Poster Wall

ICPE 2020

Virtual , Virtual Event, Maryland

September 16 – 17, 2020 (Wednesday – Thursday)

Home
Event Info
Smaller
Bigger

MY-OP01 Perinatal Outcomes Associated with Marijuana Use During Pregnancy 1432295 Alexandra Sosinsky Harvard T.H. Chan School of Public Health Perinatal Outcomes Associated with Marijuana Use During Pregnancy Background: Increases in marijuana use during pregnancy raise concerns about its reproductive safety. Objectives: We explored the associations between marijuana use during pregnancy and perinatal outcomes. Methods: This study used 2000-2014 nationwide Medicaid Analytic eXtract data to identify a cohort of pregnancies in women enrolled in Medicaid from at least three months before their last menstrual period through at least one month post-delivery. Infants were enrolled in Medicaid for at least three months after birth. Logistic regression was used to adjust for confounders including demographic characteristics, possible indications for marijuana use, other illicit drug use, and psychiatric comorbidities. Marijuana exposure was defined by the presence of an ICD-9 code (305.2x or 304.3x) during pregnancy.Results: Among 1,882,133 pregnancies, 26,882 (1.45%) were exposed to marijuana. Marijuana exposure during pregnancy was associated with an odds ratio above 1.15 for all perinatal outcomes considered except gestational diabetes, gestational hypertension, and C-section. The signals observed for neonatal abstinence syndrome, postpartum hemorrhage, preeclampsia and NICU admission disappeared upon confounding adjustment. The odds ratios moved towards the null but remained elevated for preterm delivery (1.15; 95% CI, 1.11-1.19), placental abruption (1.27, 1.16-1.40), and small for gestational age (1.31; 95% CI, 1.20-1.45).Conclusion: Marijuana exposure in pregnancy was associated with multiple adverse perinatal outcomes. Many but not all of the observed associations were explained by measured confounders. Given the underestimation of real use, unknown timing of marijuana consumption, and potential for residual confounding, the results from this study should be seen as exploratory and warrant further evaluation.

MY-OP05 Influenza Vaccination Among Frail Older Adults Receiving Post-Acute and Long-Term Care. 1432299 Elliott Bosco Brown University Influenza Vaccination Among Frail Older Adults Receiving Post-Acute and Long-Term Care. Background: Influenza vaccination is recommended for older adults receiving post-acute care (PAC) and long-term care (LTC) in long-term care facilities (LTCFs). Frailty may reduce vaccination in these populations, though little is known.Objectives: To quantify the association of various frailty measures with influenza vaccination in PAC and LTC residents.Methods: We conducted a retrospective cohort study using Medicare claims linked to Minimum Data Set (MDS) clinical assessments and LTCF data. PAC (<100 days) and LTC (100+ days) residents were identified during the 2013-2014 influenza season. Exposure was MDS-based measures of resident frailty including functional dependence (28-point Morris Activities of Daily Living), cognitive dysfunction (Cognitive Function Scale), and health instability (Changes in Health, End-stage Disease, and Symptoms and Signs). The outcome was MDS-based influenza vaccination. To assess each measures' association with vaccination, we fit generalized estimating equations accounting for LTCF clustering and adjusted for resident and LTCF characteristics. Results: We included 357,854 PAC and 162,778 LTC residents. Compared to baseline, highest levels of functional dependence [OR (95% CI): 0.95 (0.91, 0.98)], cognitive dysfunction [0.93 (0.86, 0.99)], and health instability [0.84 (0.77, 0.91)] were associated with decreased influenza vaccination among PAC residents. Highest levels of functional dependence [0.88 (0.84, 0.92)] and health instability [0.68 (0.62, 0.75)] were associated with decreased vaccination among LTC residents, however increased vaccination occurred in the highest levels of cognitive dysfunction [1.70 (1.57, 1.82)].Conclusion: Functional dependence and health instability were associated with reduced influenza vaccination in PAC and LTC residents, though cognitive dysfunction varied. Funding: Sanofi Pasteur.

MY-OP06 The association between medication adherence patterns in older community-dwelling people with multimorbidity and health-related quality of life 1432300 Caroline Walsh The association between medication adherence patterns in older community-dwelling people with multimorbidity and health-related quality of life Background: Previous evidence on medication non-adherence measurement in older people with multimorbidity is limited and the association between adherence and health-related quality of life (HRQoL) inconclusive.Objectives:To examine adherence patterns to multiple medications in older community-dwelling adults and the association between adherence patterns and HRQoL.Methods: Community-dwelling adults aged ≥70 years were recruited from 15 general practices in Ireland in 2010 (wave one) and followed two years later (wave two).Participants had ≥2 RxRisk-V conditions at wave one and had ≥2 dispensings of RxRisk-V medications (wave1-wave2). RxRisk-V index is a co-morbidity pharmacy claims-based measure used in older populations. Group-based trajectory models (GBTM) were used to group participants’ adherence patterns for RxRisk-V medications. Multilevel regression was used to examine the association between adherence patterns and (i) EQ-5D utility (linear) and (ii) vulnerability, using the Vulnerable Elders Survey (VES≥3 defined as vulnerable; logistic) at wave two, controlling for socio-demographic, clinical and social-support covariates.Results: 501 participants were included. GBTM identified five adherence groups: (1) initial low adherers, gradual increase; (2) high adherers, sharp decline ;(3) steady adherers, gradual decline;(4) consistent high adherers and (5) consistent non-adherers. After adjustment, there was no significant association between adherence groupings and EQ-5D utility. Group 5 had a significantly increased vulnerability risk compared to Group 4 (adjusted OR 1.88, 95% CI 1.01, 3.50).Conclusion: While the association between adherence trajectories and EQ-5D utility was not significant, suboptimal adherence to multiple medications in older adults with multimorbidity may be associated with increased risk of vulnerability.

MY-OP08 Repeating greedy nearest neighbor caliper propensity score matching in the same cohort can yield highly inconsistent estimates of the exposure-outcome association - a simulation study 1432302 Joris Komen University of Utrecht Repeating greedy nearest neighbor caliper propensity score matching in the same cohort can yield highly inconsistent estimates of the exposure-outcome association - a simulation study Background: Greedy nearest neighbor caliper propensity score (PS) matching is often used to account for baseline differences in observational research. This matching algorithm depends on the order of treated patients before the matching procedure, which will ultimately affect treatment effects.Objective: To assess how study results can be affected by different ordering of treated patients before PS-matching.Methods: We performed a simulation study, creating cohorts for 36 different scenarios with different sample sizes, treatment prevalence, outcome prevalence, and treatment-outcome associations. Scenarios were based on a literature search. We performed greedy nearest neighbor caliper PS-matching and repeated this 1000 times in the same cohort with a different random seed each match, to guarantee different ordering of subjects. We used a conditional logistic regression to test the association between treatment and outcome in all 1000 matched cohorts.Results: Repeating PS-matching 1000 times in the same cohort yielded large variations, in all 36 scenarios. The largest variation was found in the small cohort of 500 individuals, where the odds ratio ranged from 0.33 to 7.00. Even in larger cohorts of 10 000 individuals, odds ratios could range from 0.81 to 1.33. Interestingly, the association after repeated matching in the same cohort could be significant and non-significant after different repetitions of the matching procedure with different random seeds. We confirmed this finding in the Stockholm-Healthcare database.Conclusion: Replication of PS-matching in the same cohort can yield highly unstable estimates of the treatment-outcome association. This finding warrants careful interpretation of studies using this matching procedure.

MY-OP10 The weight-lowering effects of glucagon-like peptide-1 receptor agonists and the detection of breast cancer among women with type 2 diabetes. 1432304 Christina Santella Lady Davis Institute The weight-lowering effects of glucagon-like peptide-1 receptor agonists and the detection of breast cancer among women with type 2 diabetes. Background: There are concerns that use of glucagon-like peptide-1 receptor agonists (GLP-1 RAs) may be associated with an increased risk of breast cancer. However, one hypothesis suggests that the significant weight loss associated with these drugs may be improving the detection of prevalent breast cancer masses.Objective: We assessed whether the weight-lowering effects of GLP-1 RAs is associated with an increased incidence of breast cancer among obese women with type 2 diabetes.Methods: Using the United Kingdom Clinical Practice Research Datalink, we conducted a propensity score-matched cohort study among female obese patients with type 2 diabetes newly-treated with antidiabetic drugs between 1 January 2007 and 31 January 2018. New users of GLP-1 RAs (n=2,561) were propensity score-matched, in a 1:1 ratio, to new users of second- to third-line non-insulin antidiabetic drugs (n=2,561). Time-dependent Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) of breast cancer associated with different GLP-1 RA maximal weight loss categories (10%).Results: Breast cancer incidence gradually increased with GLP-RA maximal weight loss categories, with the highest HR observed for patients achieving at least 10% weight loss (HR: 2.15, 95% CI: 1.12-4.16). In secondary analyses, the HR for >10% weight loss was highest in the 2 to 3 years since treatment initiation (HR: 3.91, 95% CI: 1.05-14.58).Conclusions: The results of this study suggest that significant weight loss (>10%) with GLP-1 RAs may improve the detection of breast cancer among obese women with type 2 diabetes.

MY-OP11 Effect of enzyme-inducing antiepileptic drugs on failure of oral low-dose hormonal contraceptives 1432305 Amir Sarayani Effect of enzyme-inducing antiepileptic drugs on failure of oral low-dose hormonal contraceptives Background: Certain antiepileptic drugs (AED) may reduce hormonal contraceptive (HC) effectiveness via CYP3A4 enzyme induction and thus, low-dose oral estrogen HCs are not recommended for AED users. However, they remain the most common HC choice and their failure rate is unknown.Objectives: To compare contraception failure during concomitant use of oral low-dose estrogen HC with enzyme-inducing AED (carbamazepine or oxcarbazepine) or enzyme-neutral AED (lamotrigine or levetiracetam).Methods: We conducted a retrospective cohort study using IBM MarketScan Commercial Claims Data (2005-2017). We identified females aged 12-48 with epilepsy or bipolar disorder and no indication of infertility during 6-month look-back. Patients entered or re-entered the study at first day of concomitant use and were followed up to 3 years. The outcome was conception, measured with a validated algorithm that estimates gestational age for all pregnancy endpoints. We used generalized estimating equation models to calculate relative and absolute risk after standardized mortality ratio weighting via propensity score.Results: We identified 18,704 enzyme-inducing concomitant use episodes and 88,152 enzyme-neutral episodes. Bipolar diagnosis accounted for 68.7% vs. 72.5% episodes, respectively, and weighting yielded balanced cohorts. The adjusted contraception failure rate was 22.1 (95% CI 17.9-27.1) vs. 16.2 (14.6-18.0) per 1000 person-years for the enzyme-inducing and enzyme-neutral groups, respectively. The relative risk was 1.36 (1.08-1.71) and the rate difference was 5.8 (1.0-10.7).Conclusion: Use of CYP3A4-inducing AEDs increases the risk of oral low-dose estrogen HC failure. To ensure selection of more effective options, further research is needed to quantify the AED effect on other hormonal contraceptives.

MY-OP12 Prevalence of time-related biases in pharmacoepidemiology studies of anti-emetics, antifungal, and antibiotic medications during pregnancy: A systematic review 1432306 UGOCHINYERE vivian UKAH Prevalence of time-related biases in pharmacoepidemiology studies of anti-emetics, antifungal, and antibiotic medications during pregnancy: A systematic review Background: The occurrence of time related biases, such as immortal time and time-window bias, have been demonstrated in drug effectiveness observational studies. These biases often result in an observation of false protective associations between a treatment or exposure and the outcomes of interest. However, the prevalence of these biases in perinatal pharmaco-epidemiology has not been described.Objective: To systematically review and estimate the prevalence of time-related biases in observational studies of classes of medications commonly used during pregnancy (antibiotic, antifungal and antiemetic drugs) via systematic review.Methods: We systematically searched Medline and EMBASE databases between January 2013 and April 2019 for observational studies reporting the association between the antibiotic, antifungal, and antiemetic drugs and adverse pregnancy outcomes related to gestational age. Articles were reviewed independently for eligibility and method by 2 reviewers. Results: Sixteen studies were included in our systematic review (12 cohort, 3 nested case-control studies and 1 case-control study), including 10 on antibiotics, 4 on anti-emetic, and 2 on anti-fungal medications. Eight studies (50%) had time-related bias, specifically immortal time bias due to the misclassification of person-time between cohort entry and initiation of drug exposure. None of the included studies had time-window bias. Medications in studies with time-related biases appeared to be more protective for similar outcomes and comparators, compared with studies without time-related biases. Conclusion: Immortal time bias occurs frequently in epidemiological studies of medications during pregnancy. There is need for more appropriate analyses in these studies, such as time-dependent analyses, to avoid time-related biases.

MY-PO02 Empirical assessment of case-based methods for identification of drugs associated with upper gastrointestinal bleeding in the French National Healthcare System database (SNDS) 1432308 Nicolas Thurin Empirical assessment of case-based methods for identification of drugs associated with upper gastrointestinal bleeding in the French National Healthcare System database (SNDS) Introduction: Upper gastrointestinal bleeding (UGIB) is a severe event, which is frequently drug-related.Objectives: To empirically assess and calibrate case-based designs in the French National Healthcare System database (SNDS) to efficiently generate safety alerts regarding drugs associated with UGIB.Methods: All cases of UGIB were extracted from SNDS (2009-2014). Positive and negative drug controls were used to compare 196 self-controlled case series (SCCS), case-control (CC), and case-population (CP) design variants. Each variant was evaluated in a 1/10th population sample using area under the receiver operating curve (AUC) and mean square error (MSE). Parameters that had major impacts on results were identified through logistic regression. Optimum variants were replicated in the unsampled population, and negative controls estimates were used to model the distribution of the residual bias under the null.Results: AUCs for SCCS, CC and CP, respectively, ranged from 0.64 to 0.80, 0.44 to 0.61 and 0.50 to 0.67. MSEs ranged from 0.07 to 0.39, 0.83 to 1.33 and 1.96 to 4.6, respectively. Univariate regressions showed that high AUCs were achieved using SCCS with multiple drug adjustment and a 30-day risk window starting at exposure. The top-performing SCCS variant in the unsampled population yielded an AUC=0.84 and MSE=0.14, with 10/36 negative controls presenting significant estimates.Conclusion: SCCS adjusting for multiple drugs and using a 30-day risk window showed good performances for the identification of drugs associated with UGIB in the SNDS. Calibration process highlighted that low systematic error was generated but that protopathic bias and confounding by indication remained unaddressed.

MY-PO04 Use of traditional and complementary medicine in people infected with HIV in Côte d'Ivoire: cross-sectional study 1432310 Mariam Mama Djima Use of traditional and complementary medicine in people infected with HIV in Côte d'Ivoire: cross-sectional study Background: In Côte d'Ivoire, people living with HIV (PLHIV) have free access to antiretroviral therapy (ART) and cotrimoxazole. However, they may use other types of treatment including those derived from traditional and complementary medicine (TCM), for the treatment of HIV or related infections. There is limited data on the use of TCM by PLHIV on ART in Africa. Objective: The aim of this study is to describe the use of TCM and to identify the factors associated with this use in PLHIV under ART in Côte d'Ivoire.Methods: A cross-sectional study was conducted in six HIV clinics in 2016. Adults infected with HIV-1 and receiving ART for at least one year were eligible. A standardized questionnaire was used to collect demographic data, HIV characteristics, medications use and TCM use. Associated factors were identified using multivariate Poisson regression.Results: A total of 1458 PLHIV (73.4% female) with a median age of 81 months on ART were enrolled. The TCM use in the 12 months and 30 days prior to the participant interview was 64.5% and 53.2%, respectively. Factors associated with the TCM use during the last 12 months were: being female; have a poor health perception; to report the use of at least one non-HIV medication. Being unemployed or students and working in an informal sector has been inversely associated with the TCM use.Conclusion: The use of TCM affects more than 50% of PLHIV on ART regardless of the estimated period of TCM use.

MY-PO05 Determinants of Stigma among Patients with Hepatitis C Virus (HCV) Infection 1432311 McKayla Saine Determinants of Stigma among Patients with Hepatitis C Virus (HCV) Infection Background: Stigma around hepatitis C virus (HCV) infection is an important and understudied barrier to HCV treatment and elimination. However, the determinants of HCV-related stigma remain unknown.Objectives: We sought to identify the determinants of HCV-related stigma, including the impact of HCV treatment stage (i.e., spontaneously-cleared; diagnosed, untreated; previously treated, not cured; currently being treated; treated, cured) and human immunodeficiency virus (HIV) coinfection.Methods: We conducted a cross-sectional study among patients with a history of HCV infection (n=270) in Philadelphia, Pennsylvania. Stigma was evaluated using the validated HCV Stigma Scale, adapted from the Berger HIV Stigma Scale. Associations between HCV-related stigma and hypothesized demographic, behavioral, and clinical risk factors were evaluated by multivariable linear regression.Results: Most participants (95.5%) experienced HCV-related stigma. Mean stigma scores did not differ significantly between HCV-monoinfected and HIV/HCV-coinfected participants (P=0.574). However, we observed significant interactions between HIV status and multiple determinants, and therefore, stratified analyses by HIV status. Among HIV/HCV-coinfected participants, previous HCV treatment without cure, female gender, Hispanic/Latino ethnicity, and some college education were significantly associated with higher HCV-stigma scores. Annual income of $10,000-$40,000 was associated with significantly lower stigma scores. No significant associations were observed among HCV-monoinfected participants.Conclusions: We found that most participants experienced stigma associated with HCV. While stigma scores were similar between HCV-monoinfected and HIV/HCV-coinfected participants, the determinants associated with HCV-related stigma differed by HIV status. Understanding how experiences of stigma differ between HCV-monoinfected and HIV/HCV-coinfected patients may aid in the development of targeted interventions to address the HCV epidemic.

MY-PO07 Linkage of Primary Care Prescribing Records and Pharmacy Dispensing Records in the Salford Lung Study: Application in Asthma 1432313 Holly Tibble University of Edinburgh Linkage of Primary Care Prescribing Records and Pharmacy Dispensing Records in the Salford Lung Study: Application in Asthma Background: Clinical data can be repurposed to measure non-adherence – not observing the treatment regimen agreed upon with their clinician. Using prescribing records alone disregards medications not collected from the dispensary, while using dispensing records alone overlooks resolved conditions and/or discontinued treatments. Linkage between prescribing and dispensing data is not routinely conducted, and, without a common event identifier, is non-trivial. Objectives: To develop a methodology for probabilistically linking asthma pharmacy dispensing records to primary care prescribing records.Methods: A novel probabilistic methodology was developed in the Salford Lung Study dataset, matching asthma medication pharmacy dispensing records and primary care prescribing records, using semantic and syntactic harmonization, domain knowledge integration, and natural language feature extraction. Factors associated with failure or delay to collect prescriptions were assessed. Finally, we used a naïve deterministic record linkage algorithm to compare against the results of our proposed methodology.Results: We matched 83% of pharmacy dispensing records to primary care prescribing records, and approximated that 30% of prescriptions issued were not collected. Medication strength and quantity were each missing in approximately 60% of the unmatched dispensing records. The naïve algorithm identified 75% fewer matching prescribing records. Factors associated with delay (or failure) in prescription collection included season, quantity of medication supply prescribed, previous dispensing history and class of medication. Conclusions: We have developed an accurate probabilistic record linkage methodology allowing researchers to match a large percentage of pharmacy dispensing records with primary care prescribing records for asthma medications, and subsequently extract information about asthma medication non-adherence.

MY-PO11 Antiviral Chemoprophylaxis Policies and Use during Influenza Outbreaks in Long-term Care Facilities 1432317 Joe Silva Antiviral Chemoprophylaxis Policies and Use during Influenza Outbreaks in Long-term Care Facilities Background: Older adults living in long-term care facilities (LTCFs) are at high risk for influenza infection. Current guidelines include utilization of antiviral medication chemoprophylaxis for influenza outbreak management in LTCFs. Though facilities may develop their own policies, there is little empirical evidence describing the alignment of these policies with national evidence-based guidelines, or that either local policies or national guidelines are adhered to in the event of an outbreak.Objectives: To identify LTCF antiviral chemoprophylaxis utilization and adherence to both national guidelines and local facility policies in instances of influenza outbreaks.Methods: We analyzed influenza outbreaks in Rhode Island LTCFs during the 2017-2018 and 2018-2019 influenza seasons. Outbreaks were derived from the Rhode Island Department of Health influenza surveillance system, and reporting facilities were interviewed for existence of chemoprophylaxis policies. LTCF adherence to facility policies and national guidelines was assessed for each outbreak. Results: Seventy-four outbreaks were observed, of which 68% (n=50) occurred in facilities that were nonadherent to national guidelines. Observed outbreaks occurred in 53 unique facilities and 52 were interviewed. Thirty-seven percent (n=19) of facilities reported no written policy on chemoprophylaxis and 44% (n=23) reported a policy that was not concurrent. For outbreaks that occurred in facilities that reported policies (n=42), 29% (n=12) were not adherent to the facility policy. Conclusions: Use of antiviral chemoprophylaxis often deviated from both national guidelines and local LTCF policies. Adoption of antiviral chemoprophylaxis policies and harmonizing existing policies to national guidelines could reduce non-adherence to guideline-recommended practices to effectively control influenza outbreaks.

MY-PO13 Determinants of Adverse Drug Reaction (ADR) Reporting among Healthcare Professionals in Ghanaian Hospitals: A Presentation of Preliminary Qualitative Findings. 1432319 Walter-Rodney Nagumo University of Sheffield Determinants of Adverse Drug Reaction (ADR) Reporting among Healthcare Professionals in Ghanaian Hospitals: A Presentation of Preliminary Qualitative Findings. Background: Adverse drug reactions (ADRs) are associated with significant morbidity, mortality, and increased healthcare cost. Estimates show that 712 ADR mortalities occur annually, with an associated cost of about £98.5m in the UK1 while the analysis in the USA shows that, 2,216,000 hospitalised patients experienced serious ADRs and 106,000 fatal cases2. Available studies have raised a concern about under-reporting globally and lack of awareness particularly, in Africa 3. Objective: To explore the factors influencing ADR reporting among Ghanaian healthcare professionals (HCPs).Method: Qualitative data (in-depth interviews and FGDs) were collected from fifty-one HCPs (28 females and 23 males) comprising nursing, medical and pharmacy staff. Five hospitals were purposively sampled in the Tamale metropolis between September and November 2017. Interviews were audio-recorded, transcribed verbatim and thematically analysed.Results: There were significant variations in experiences and beliefs about ADRs and associated reporting. HCPs recognised human and system factors as key issues affecting ADR reporting. Most importantly, patients were apprehensive, there was perverse self-medication, hospital shopping and surrogate/carer reporting. HCPs knowledge about what to report, the reporting procedures and fear of consequences also directly affected decisions to report. Key identified system factors were funding, leadership and administrative bureaucracies.Conclusion: The pharmacist's role was viewed as significant in the ADR reporting process, but current practices and beliefs appeared to limit the degree to which they could comprehensively enhance ADR reporting. Specific interventions may have significant implications on knowledge, attitude and practice around ADR reporting.

MY-PO14 Statin use and skin cancer risk: A French prospective cohort study 1432320 Marie Al Rahmoun Paris Sud University Statin use and skin cancer risk: A French prospective cohort study Background: The relationship between statin use and incidence of skin cancers is unclear, epidemiological studies being scarce and yielding conflicting results. Objective: To explore the associations between statin use and risk of skin cancers. Methods: E3N is a prospective cohort of 98,995 French women aged 40-65 years in 1990. Health and lifestyle data were self-reported in biennial questionnaires and were matched with a drug reimbursement database allowing the identification of statin use since 2004. We used Cox models adjusted for age and skin cancer risk factors to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). Results: Over 2004-2014, 455 cutaneous melanoma, 1741 basal cell carcinoma (BCC), and 268 squamous cell carcinoma (SCC) cases were ascertained among the 62,473 women included in the study. There was no statistically significant association between ever vs. never use of statins and risks of melanoma (HR=1.16, 95%CI=0.94-1.44) or SCC (HR=0.89, 95% CI=0.66-1.19). However, ever use of statins was associated with a decreased BCC risk (HR=0.89, 95%CI=0.79-1.00). Compared with never use, this decreased BCC risk was observed from the 3rd year of cumulative use (HR=0.79, 95%CI=0.65-0.96), when time since first exposure was ≥4 years (HR=0.80, 95%CI=0.67-0.95), or when the cumulative dose reached at least 900 defined daily doses (HR=0.79, 95%CI=0.64-0.97). Conversely, women with <4 years of statin exposure had a higher melanoma risk (HR=1.37, 95%CI=1.01-1.86). Conclusion: long-term statin use was associated with a decreased BCC risk. More research is warranted to understand the mechanisms underlying this association.

MY-PO15 Coexisting conditions in children and adolescents with cancer in a section of the South African private health sector 1432321 Marianne Otoo North-West University Coexisting conditions in children and adolescents with cancer in a section of the South African private health sector Background: The immunosuppression associated with cancer, its treatment, and its resultant effect on the quality of life of patients makes the presence of coexisting conditions relatively common.Objectives: This study aimed at investigating the common coexisting conditions in children and adolescents younger than 19 years being treated for cancers in a section of the South African private health sector.Method: A cross-sectional study design was followed. Medicine claims data from 1 January 2008 to 31 December 2017 were queried to identify coexisting conditions using the International Classification of Diseases, Tenth Revision (ICD-10) codes indicated on reimbursed claims. The pharmacological drug classes of medications associated with non-specific ICD-10 codes were analyzed using the drug utilization 90% (DU90%) principle. Pharmacological classes were categorized using the Monthly Index of Medical Specialties (MIMS) classification system. Data were analyzed descriptively.Results: A total of 173 participants were included in the study. ICD-10 codes were available for 13.65% of medicine claims. Diseases of the respiratory system (J00-J99, 7.15%), gastrointestinal tract (K00-K95, 1.60%), and skin disorders (L00-L99, 0.95%) were the most prevalent specific diagnoses identified. Non-specific ICD-10 codes were recorded on 86.35% (n = 2 272) of non-cytotoxic medicine claims. Antimicrobial agents (17.40%), respiratory system agents (13.91%) and analgesics (10.64%) were the most frequently utilized pharmacological classes of medications in our patient population.Conclusion: As determined from ICD-10 codes and medication claimed on reimbursed claims, children and adolescents being treated for cancers mostly suffered from acute conditions, in particular, microbial infections and diseases of the respiratory system.

MY-PO16 Associations between Antidiabetic Medications Use and Breast Cancer Survival Outcomes among Medicare Beneficiaries 1432322 Shahariar Mohammed Fahim Associations between Antidiabetic Medications Use and Breast Cancer Survival Outcomes among Medicare Beneficiaries Background: The knowledge of prognosis variability between cancer patients using different antidiabetic medications before breast cancer diagnosis remains limited.Objective: This study examined the associations between survival outcomes, all-cause mortality (ACM) and breast cancer-specific mortality (BCSM), and antidiabetic medications that cancer patients used before their breast cancer diagnosis overall and by patient characteristics.Methods: Design: Longitudinal, retrospective cohort design. Settings: Linked Surveillance, Epidemiology and End Results (SEER)-Medicare databases from 2007 to 2013 were used to identify patients who had continual Medicare enrollment and a new breast cancer diagnosis between 2008 and 2013 (n=49,151). Exposure: Antidiabetic medication use during one year before their breast cancer diagnosis. Outcomes: ACM and BCSM. Statistical Analysis: Cox proportional hazard models were used to estimate hazard ratios (HR) and 95% confidence intervals for associations between prior use of antidiabetic medications and survival outcomes. Results: A total of 1,719 patients used antidiabetic medications during one year prior to their breast cancer diagnosis. Compared to metformin, patients using insulin, sulfonylurea, and combination drugs had higher ACM (HR: 1.62, P0.05); however, single or widowed patients had a higher likelihood of BCSM compared to married patients (HR: 1.74, and HR: 1.57, P=0.01, respectively). Diabetes severity was not significantly associated with either ACM or BCSM. Conclusions: Different antidiabetic medications used by patients before their cancer diagnosis likely affect ACM but not BCSM.

MY-PO18 CORTICOSTEROID USE AND RISK OF DEEP VEIN THROMBOSIS AND PULMONARY EMBOLISM: A SYSTEMATIC REVIEW AND META-ANALYSIS 1432324 Ananth Kashyap Rajiv Gandhi University of Health Sciences, Bangalore, India CORTICOSTEROID USE AND RISK OF DEEP VEIN THROMBOSIS AND PULMONARY EMBOLISM: A SYSTEMATIC REVIEW AND META-ANALYSIS Background: Deep vein thrombosis (DVT) and Pulmonary embolism (PE) are some of the major preventable causes of morbidity and mortality across the globe, associated with various risk factors in which the use of corticosteroids is one among them.Objective: We aimed to assess the risk of DVT and PE in those patients who used corticosteroids for any of its indications.Methods: The search was conducted in MEDLINE and Cochrane Library. The references of included studies and grey literature databases were also searched for relevant studies without any language or date restrictions. Observational studies examining the risk of DVT and PE in those patients who used corticosteroids were included for the review. The outcomes of concerns were proportion and risk of DVT and PE. New Castle Ottawa (NCO) was used to assess study quality. Three authors were independently involved in the study selection, data extraction, and quality assessment.Results: A total of five out of 3582 studies including 2,01,516 participants for DVT and 2,22,225 participants for PE were included in this meta-analysis. There was a significantly increased risk of DVT (Odds Ratio: 2.26; 95% Confidence Interval: 2.08 to 2.46; P<0.0001) and PE (Odds ratio: 3.45; 95% Confidence Interval: 3.21 to 3.71; P<0.0001) among the steroid group when compared to the control group. Four studies were of high quality and one study was moderate-quality according to NCO scale.Conclusion:The study infers that the use of corticosteroids was associated with significantly increased risk of DVT and PE when compared to the control group.

MY-PO20 Prescription Opioid Patterns Following Supply Restriction Policy among Patients Receiving Chronic Opioid Therapy: An Interrupted Time Series Analysis 1432326 Yun Shen Prescription Opioid Patterns Following Supply Restriction Policy among Patients Receiving Chronic Opioid Therapy: An Interrupted Time Series Analysis Background: In Florida, House Bill 21 (HB21) was implemented in July 2018 to limit prescriptions of Schedule II opioids for patients with acute pain to a 3-day supply. However, little is known about unintended impacts of supply restriction policies on opioid prescription access among chronic pain patients. Objective: The purpose of this study was to assess opioid prescribing patterns among patients receiving COT following HB21.Methods: We obtained pharmacy claims for opioids dispensed from 1/1/2015 to 3/31/2019 from a private health plan serving a large Florida employer. COT episodes were defined as ≥70 days’ supply in the prior 90 days. Interrupted time series analyses were conducted to compare monthly opioid prescription indicators (number of COT patients, daily Morphine Milligram Equivalents [MMEs], days’ supply) among COT patients before and after the enactment of HB21 on 7/1/2018.Results: There were significant reductions in the trend (pre-HB21 RR: 0.95, 95% CI:0.94, 0.97; post-HB21 RR: 0.65, 95% CI: 0.55, 0.78) and immediate change (RR: 0.95, 95% CI: 0.89, 1.00) for the monthly prevalence of COT patients. No significant changes in trend or immediate change were observed for monthly MME per COT patient-day. A significant immediate increase was observed for monthly number of days’ supply per prescription (RR: 1.01, 95% CI: 1.00, 1.02) and total MME per prescription (RR: 1.04, 95% CI: 1.01, 1.08), with no significant change in trend.Conclusion: Following HB21, fewer patients received COT. However, patients continuing to receive COT experienced no significant changes in days’ supply or MME per patient-day.

MY-PO22 Primary Care Chronic and High Dose Opioid Prescribing in 2011-2018: An Observational Study Using Electronic Medical Record Data 1432328 Isaac Bai Dalhousie University Primary Care Chronic and High Dose Opioid Prescribing in 2011-2018: An Observational Study Using Electronic Medical Record Data Background: Chronic opioid therapy (COT) prescribing for non-cancer pain has increased dramatically in Canada, despite limited evidence. Many risks are associated with chronic and high dose opioid therapy, which may outweigh the potential benefits.Objective: To describe opioid prescribing patterns among patients in family practices using electronic medical record (EMR) data. Methods: A retrospective secondary analysis of adults (≥18 years old) with at least one opioid prescription in 2011-2018 is being conducted using EMR data from the Maritime Family Practice Research Network, consisting of 27 rural and urban family practices in Nova Scotia. Opioid prescriptions were identified using World Health Organization Anatomical Therapeutic Chemical codes. Exclusions were cancer patients and palliative patients, identified using ICD-9-CM diagnostic codes. Primary outcomes are prevalence of patients prescribed any opioid, chronic opioid therapy (COT, ≥84 days of therapy), and high dose COT (HD-COT, >90 morphine milligram equivalents per day on average); calculated for each study year separately.Results: Results indicate the prevalence of patients prescribed any opioid decreased from 3.66% in 2011 to 3.05% in 2018. A decrease was also seen with both COT (1.39% to 1.25%) and HD-COT prescribing (0.51% to 0.29%). The highest prevalence of patients prescribed any opioid, COT, and HD-COT was in 2013, whereas the lowest prevalence was in 2018.Conclusions: Opioid prescribing declined over the study time frame in participating primary care practices in Nova Scotia. Future research will examine the association of patient and prescriber characteristics, the conduct of urine drug screening, and concurrent prescribing of sedative/hypnotic therapy.

MY-PO23 Utilization and Adherence among Infliximab Biosimilar Initiators in a U.S. National Commercial Insurance database. 1432329 Sujith Sarvesh University of Alabama Birmingham Utilization and Adherence among Infliximab Biosimilar Initiators in a U.S. National Commercial Insurance database. Background: Infliximab-dyyb, first biosimilar to infliximab, was approved for multiple indications in U.S. in 2016. Since the utilization of biosimilar in the U.S. has not been studied, it is unclear if patients on Infliximab-dyyb are adherent to their treatment.Objective: Characterization of infliximab-dyyb initiators and their adherence using national commercial administrative data.Methods: We identified Infliximab-dyyb initiators during 2016-2017 based on procedure and pharmacy codes. Using available data, we classified infliximab-dyyb initiators as: naïve Infliximab-dyyb users (never used infliximab), infliximab early switchers with 2yr) use of infliximab. Eligible patients were ≥18 years of age and continuously enrolled with medical and pharmacy coverage in 2014-2017. We evaluated baseline characteristics in all patients and in subgroup analysis of people with >=12m of coverage, calculated the proportion of days covered (PDC) >=80% at 12. Baseline was defined as 12 months prior to Infliximab-dyyb initiation date..Results: We identified 98 Naïve Infliximab-dyyb users, 114 early switchers, and 113 late switchers. Compared to (early or late) switchers, naive users were younger and more likely to be female. The highest proportion of autoimmune disease was inflammatory bowel disease (40.9-45.4%), followed by rheumatoid arthritis (15.4-29.6%) and psoriasis (9.7-13.9%). Among patients who had >=12m of follow-up, 53.6% naive users, 44.3% early switchers, and 56.1% late switchers continued to be adherent (>80%) at 12 months.Conclusion: Approximately half of the Infliximab-dyyb initiators were highly adherent at 12 months. Further studies are needed to evaluate the long-term adherence among infliximab-dyyb users.

MY-PO24 Long-term adherence and persistence with non-vitamin K oral anticoagulants in patients with atrial fibrillation and their associations with stroke risk 1432330 Joris Komen University of Utrecht Long-term adherence and persistence with non-vitamin K oral anticoagulants in patients with atrial fibrillation and their associations with stroke risk Background: Studies on adherence and persistence with non-vitamin K oral anticoagulant (NOAC) treatment rely solely on data from the early years of NOAC availability. Objective: To study long-term adherence and persistence with NOACs and their associations with stroke risk.Methods: From the Stockholm Healthcare database, we included all atrial fibrillation (AF) patients claiming a NOAC prescription until October 2018. We defined persistence as continuing to claim NOAC prescriptions after a 90-day gap. In persistent patients, we calculated the mean possession rate (MPR) to represent adherence. Using a case-control design, we calculated associations of persistence and adherence with stroke risk. All associations were adjusted for co-medications and co-morbidities as risk factors for stroke. The outcome was a composite of ischemic or unspecified stroke and TIA. Fractures and respiratory infections were falsification endpoints.Results: In 21 028 patients with a maximum follow-up of 7.4 years (median: 2.0, IQR: 2.2), persistence steadily decreased to 70% at the end of follow-up. In persistent users, the MPR remained stable at 90%, with 75% of patients having an MPR >95% throughout the study period. Non-persistence and poor adherence were both associated with increased stroke risk (non-persistence adjusted odds ratio (aOR): 2.05; 95% confidence interval (CI): 1.49–2.82, 1 percent reduction MPR aOR: 1.03; CI: 1.01–1.05). There was no association between non-persistence or poor adherence and the falsification endpoints. Conclusion: Persistence rates decreased slowly over time, but persistent patients had high adherence rates. Both non-persistence and poor adherence were associated with an increased stroke risk.

PO-0027 Why Not Biosimilars: Biologic And Biosimilar Utilization Trends In Japanese R A Patients 1400532 Lu Zhang Panalgo Tiffany Woodworth Panalgo Spotlight Poster Session Spotlight Poster Biologics & Biosimilars Background: Lower-costing biosimilars offer opportunities to increase access to biologic treatment for patients with rheumatoid arthritis (RA) and to lower overall health care expenditures in these patients. Since the 2014 market entry in Japan, the impact of biosimilars is largely unknown.Objectives: To assess the impact of biosimilars and describe utilization trends and patient characteristics in Japanese RA patients.Methods: We identified patients with ≥1 RA treatment, prior RA diagnosis and 1 year of continuous enrollment in the JMDC Inc. database between January 1, 2005 and May 31, 2019. We calculated annual prevalence of RA treatments: biologics, including biosimilars for etanercept, infliximab, and rituximab; non-biologic DMARDs; glucocorticoids; and methotrexate (MTX). Among new users of biologics, defined as no prior biologic use in prior year, we examined treatment and switching patterns between reference biologics and associated biosimilars. Finally, we examined patient characteristics among those initiating reference biologics vs biosimilars.Results: Among treated Japanese RA patients, prevalent biologic use steadily increased from 2.7% in 2005 to 18.5% in 2019. MTX use also increased (24.0 to 46.7%). Use of glucocorticoids and other DMARDs decreased (78.1 to 57.9% and 40.2 to 26.6%, respectively). Biosimilars had low uptake (0.1% in 2015 to 1.0% in 2019) despite overall increase in biologics. While prevalence of newer biologics without associated biosimilars increased (0.6% in 2008 to 11.7% in 2019), prevalence of reference biologics peaked to 10.5% in 2011 and have been steadily declining to 6.3% in 2019. Further, new initiators of reference biologics (N=536) stayed on reference for 541 days on average, and 7.3% switched to biosimilars, while new biosimilar initiators (N=37) stayed on biosimilars for 95.5 days on average, and 10.8% switched to reference. Biosimilar initiators tended to be older (50.8 [11.7] vs 47.4 [13.1] years), male (32.4 vs 27.1%), had less outpatient visits in the 12 month baseline period (19.6 vs. 24.6), and were more likely to use MTX at baseline (73.0 vs. 63.1%).Conclusions: While biosimilars are presumed to increase utilization of biologics due to better affordability, preliminary analysis of biosimilar and biologic uptake and utilization patterns indicate hesitancy to use biosimilars in Japan. Early results seem to indicate that new initiators of biosimilars may have different clinical and demographic attributes compared to initiators of reference biosimilars. Further research should be conducted examine these differences and their impact on biosimilar uptake.

PO-0029 Adherence To Laboratory Monitoring Among People Taking Oral Drugs For Multiple Sclerosis: A Population-based Study 1400534 Huah Shin Ng University of British Columbia Adherence To Laboratory Monitoring Among People Taking Oral Drugs For Multiple Sclerosis: A Population-based Study Pharmacovigilance -> Neurological/Mental Health Background: The oral disease-modifying therapies (DMTs) for multiple sclerosis (MS) are associated with improved drug adherence relative to the injectable DMTs. However, it is not known how well MS patients adhere to the blood and urine tests that are required for safety monitoring.Objectives: We examined adherence to recommended laboratory testing by persons who initiated an oral DMT for MS.Methods: Population-based health administrative and laboratory data were accessed. All people in the province of British Columbia, Canada, filling their first prescription for dimethyl fumarate, fingolimod or teriflunomide (2011-2015) were identified. The proportion of people adherent to each drug monograph’s recommended laboratory monitoring schedule, pre- and on-DMT, was estimated for lymphocytes, liver enzymes, and urine protein. The association between adherence and sex, age, socioeconomic status, comorbidity and previous exposure to non-oral DMT were examined using the generalized estimating equation and multivariable logistic regression.Results: Overall, 1016 people were included (dimethyl fumarate n=567; fingolimod n=253; and teriflunomide n=196). The proportions of people who were adherent to the recommended laboratory monitoring schedule prior to oral DMT initiation was 88-91% for liver tests and 91-94% for lymphocyte tests (dimethyl fumarate, fingolimod and teriflunomide), and 77% for urinalysis (dimethyl fumarate only). Adherence to the first on-DMT liver test was 89% for dimethyl fumarate (within 6 months); 61% for fingolimod (within 3 months); and 40% for teriflunomide (within 1 month). For dimethyl fumarate, adherence was 90% for the first lymphocyte test and 75% for urinalysis (each within 6 months). Men were less likely than women to have a urinalysis (adjusted odds ratio [aOR]=0.61; 95%CI:0.40-0.95) before initiating dimethyl fumarate, or liver (aOR=0.46; 95%CI:0.23-0.95) or lymphocyte (aOR=0.47; 95%CI:0.22-0.98) tests while taking dimethyl fumarate.Conclusions: Adherence to recommended laboratory testing was high (>77%) before oral DMT initiation, but lower once on drug. There is a need to understand the long-term consequences of suboptimal laboratory monitoring in the DMT-treated MS population. The sex differences also warrant further investigation.

PO-0030 Androgen Therapy And Longitudinal Outcomes For Hypospadias Repair Procedures In The United States 1444339 Julia Ward Androgen Therapy And Longitudinal Outcomes For Hypospadias Repair Procedures In The United States Pediatric Pharmacoepidemiology Background: Hypospadias is the most common variation of urethral development among male infants and is often treated with a surgical repair procedure. Some patients receive neo-adjuvant androgen therapy prior to surgery in order to enlarge the surgical area. However, nation-wide trends in prevalence of hormonal therapy prior to surgery remain understudied. Further, longitudinal assessments of surgical outcomes and the influence of androgen therapy on these outcomes are virtually non-existent.Objectives: As part of the Urologic Diseases in America project, we assessed 12-year trends in hypospadias repair procedures, use of preoperative androgen stimulation (PAS), and surgical outcomes among a large, national pediatric population.Methods: We utilized 2004-2016 Optum© Clinformatics® Data Mart, a de-identified adjudicated insurance claims database of 15-18 million individuals covered annually by commercial insurance. Among 316,152 males continuously enrolled in private insurance from birth through two years of life, we utilized medical and pharmacy claims to estimate the annual prevalence of hypospadias, hypospadias-related repair procedures, and use of PAS. Among 2,356 hypospadias patients with a repair procedure, we used the Kaplan-Meier estimator to calculate the 18-month cumulative probability of repeat procedures and postoperative complications (urethral fistula, stricture, or diverticulum). Analyses were conducted overall and stratified by use of PAS and procedure location.Results: Average annual hypospadias prevalence was 1%. Overall, 57% of hypospadias patients had a repair procedure by two years of age, most of which occurred in ambulatory settings. PAS was prescribed for 4% of patients with a repair procedure. Among those with an initial procedure, an average of 16% had a postoperative complication and 12% required a repeat procedure, most within 9 months of the initial procedure. The prevalence of repeat procedures decreased by 15% over the 12-year study period. Surgical complications and repeat procedures were both more common among patients receiving PAS and among those requiring an inpatient procedure.Conclusions: The majority of male infants with hypospadias underwent a repair procedure early in life. Those with an inpatient procedure or PAS were more likely to have an adverse surgical outcome; however, the prevalence of repeat procedures decreased over the study period regardless of PAS status. Although PAS and inpatient procedures may be markers of hypospadias severity, awareness of these risk factors may allow urologists to better counsel and treat hypospadias patients.

PO-0035 Potential Association of Aromatase Inhibitors and Osteonecrosis of Jaw: Signal Refining to Identify False Positive / Pseudo Signal 1400535 Subeesh Kulangara Viswam Faculty of Pharmacy, M.S. Ramaiah University of Applied Sciences Potential Association of Aromatase Inhibitors and Osteonecrosis of Jaw: Signal Refining to Identify False Positive / Pseudo Signal Pharmacovigilance -> Other Background: Signal generation through data mining algorithms is an innovative and emerging field in pharmacovigilance. However, it is important to consider confounding factors that significantly influence the signal. Aromatase inhibitors are being reported for possible association with osteonecrosis of jaw. However, this signal was calculated without adjusting the possible confounding factors.Objectives: Our study aimed to identify potential signals of Aromatase inhibitors associated osteonecrosis of jaw and assess the possibilities of the safety signal to be a pseudo signal/false positive in FDA Adverse Event Reporting System (FAERS).Methods: A systematic data mining was performed in the FAERS database using OpenVigil 2.1, a pharmacovigilance analytical tool. Drugs included in the analysis were Letrozole, Anastrozole and Exemestane and event of interest was osteonecrosis of jaw. Reporting Odds Ratio (ROR) was used to assess the relation between the drug and adverse event. A value of ROR-1.96SE >1, (SE- standard error) was considered positive. The confounding factor considered in this study was other drugs that are known to cause osteonecrosis of jaw, co-administered with the drug of interest. Through data mining and literature search, bisphosphonates and denosumab were identified as possible confounding factors. The signal strength of drug of interest were calculated with and without the presence of confounding factors.Results: FAERS database had a total of 15,178 reports for osteonecrosis of jaw. Amongst which 1,061 (6.99%) reports were associated with aromatase inhibitors. The number of reports for letrozole, anastrozole and exemestane were 448 (42.22%), 251 (23.65%) and 362 (34.11%) respectively. Signal strength (ROR-1.96SE) for letrozole, anastrozole and exemestane associated osteonecrosis of jaw without any background correction was 8.34, 6.64 and 15.14 respectively. Upon removing the reports of bisphosphonates and denosumab, signal strength drastically decreased to 0.03, 0.36 and 0.47 for letrozole, anastrozole and exemestane respectively. The signal strength of bisphosphonates and denosumab associated osteonecrosis of jaw did not change significantly (104.06 to 93.2 and 37.1 to 36.1, respectively) upon removal of aromatase inhibitors.Conclusions: Our study suggested that the signal generated for aromatase inhibitors associated osteonecrosis of jaw in FAERS database can be false positive. Careful background corrections with the identification of those confounding factors are imperative to exclude false-positive results.

PO-0037 Comparative Safety Of High-dose Versus Standard-dose Influenza Vaccination In Patients With End-stage Renal Disease 1400537 J Bradley Layton RTI Health Solutions Spotlight Poster Session Spotlight Poster Vaccines Background: High-dose influenza vaccine (HDV) is an option for patients with end-stage renal disease (ESRD), though the safety of HDV has not been evaluated in this population.Objectives: The objective of this study was to estimate the relative occurrence of adverse reactions in patients with ESRD following vaccination with HDV compared with standard-dose influenza vaccine (SDV).Methods: Using data from the United States Renal Data System, patients with ESRD aged ≥ 65 years were identified on the date of influenza vaccination during yearly influenza seasons (2010-2016). In each outcome-specific analysis, patients were followed from vaccination to observe serious (anaphylaxis, angioedema, seizure, encephalopathy, Guillain-Barré syndrome, and short-term mortality) and milder (urticaria/hives, rash, pain in limb, cellulitis, myalgia/myositis, fever, nausea and vomiting, diarrhea, and syncope) events. Propensity score-weighted hazard ratios (HR) and 95% confidence intervals (CI) comparing recipients of HDV to SDV were estimated with Cox proportional hazards models.Results: Of 520,876 vaccinated patients (mean age = 74.7, standard deviation = 7.0; 63% white race), 7.4% received HDV. Crude incidence rates of serious events were low. HRs were inestimable for anaphylaxis, angioedema, and Guillain-Barré syndrome due to too few cases, and HRs for seizure, encephalopathy, and mortality were null. HRs for some minor events were increased in the HDV group: rash (HR = 1.86; 95% CI: 1.34-2.57), diarrhea (HR = 1.26; 95% CI: 1.07-1.50), pain in limb (HR = 1.23; 95% CI: 1.12-1.34), and myalgia/myositis (HR = 1.16; 95% CI: 1.04-1.30). This pattern was consistent in subgroup and sensitivity analyses.Conclusions: HDV recipients had increased risks of several minor events, while the risks of severe events were generally low and similar between the treatment groups, consistent with clinical trials findings in the general population of older adults. The risk-benefit tradeoff of HDV versus SDV use in patients with ESRD should be carefully considered given these safety data, increases in national use of HDV, and high comorbidity burden in this population.

PO-0039 “I Know My Body Better Than You”: Patient Focus Groups To Inform A Shared Decision Aid On Oral Corticosteroid (ocs) Use During Pregnancy 1400538 Kristin Palmsten HealthPartners Institute "I Know My Body Better Than You": Patient Focus Groups To Inform A Shared Decision Aid On Oral Corticosteroid (ocs) Use During Pregnancy Pregnancy and Lactation Background: Improved dissemination of evidence based information on medication safety in pregnancy to patients is needed. OCS use during pregnancy has been associated with adverse perinatal outcomes.Objectives: Inform the development of a shared decision making aid on OCS use in pregnancy with patient focus groups.Methods: Using EHR data, we identified patients in our Midwestern (USA) health care system who had a recent live birth and a condition for which OCSs may be prescribed (i.e., asthma, other autoimmune disease). We recruited women in 2019 with invitation letters and phone calls to participate in 1 of 5 focus groups. Participants completed a survey on demographics and health and responded to open-ended discussion questions. We conducted a conventional content analysis with inductive coding on verbatim transcripts.Results: Among the 30 participants, the average age was 33 years, 66% were White, 77% had at least a college degree, and 66% had ever taken an OCS. Women consulted a variety of sources of information on medication safety in pregnancy (e.g., internet, health care providers, family, pharmacists) and expressed frustration with conflicting information. Women voiced support for patient-provider discussions about medication use during pregnancy in which the provider shares risks and benefits and the patient makes her decision. Furthermore, women generally expressed support for a patient-centered handout about OCS use during pregnancy that the provider discusses with the patient. Obstetrician/midwives were the preferred provider type for advising on OCS use during pregnancy because women trusted them to know the information, believing it was their job, and because of convenience at routine prenatal visits. Women wanted their non-obstetric specialists to collaborate with obstetric providers on medication treatment in pregnancy. When trying to decide whether to take OCSs in pregnancy, women had concerns about: the medication’s impact on their baby (e.g., miscarriage, development, birth defects), themselves (e.g., side effects including mood, sleep, weight gain), pregnancy complications (e.g., preterm birth, increased blood pressure), and lactation. Women wanted information on OCSs (e.g., indications, length of treatment, and cost), alternative treatments, and risks of not taking OCSs.Conclusions: The need for a decision aid handout on OCS use during pregnancy that obstetric providers can discuss with their patients was confirmed. The handout should at a minimum contain risks/benefits of the medication to mother, baby, pregnancy complications, and lactation.

PO-0048 Risk Of Cataract In Primary Immune Thrombocytopenia Adult Patients Treated With Eltrombopag 1400540 Margaux Lafaurie Service de Pharmacologie M�dicale et Clinique, Centre Hospitalier Universitaire de Toulouse, France Spotlight Poster Session Spotlight Poster Rare Disease Background: In preclinical studies, eltrombopag (ELT), a thrombopoietin receptor agonist, has been associated to a dose and time-dependent increased incidence of cataract in mice and rats. In the extension study, 9,3% of the patients developed or worsened cataract over a median duration of exposure of 2.4 years.However, about 80% had at least an additional risk factor. Real-life studies are lacking.Objectives: We aimed to assess the risk of cataract with ELT in the real-life practice in a nationwide cohort of primary ITP adults.Methods: The population was the FAITH cohort which is the cohort of all incident primary ITP adult patients in France from June 2010 to June 2017. This cohort was identified within the national health insurance database using a validated algorithm. A nested case-control study was conducted. Cases were patients who had a surgery for cataract after ITP onset. Up to 5 controls for each case were matched on age, sex, and duration of disease. Exposure to ELT was searched within out-hospital dispensing data between ITP onset to index date. We used 2 definitions: i)“ever exposed” compared with “never exposed”; ii) cumulative exposure quantified by the numbers of Defined Daily Doses DDDs, categorized as follows “never exposed”, 1-364 DDDs and ≥365 DDDs. Covariables were the presence of diabetes mellitus, cumulative exposure to corticosteroids, and the presence of ophthalmological risk factors for cataract. Conditional logistic regression models were used to compute adjusted odds ratios (aOR) and their 95% confidence intervals (CI).Results: The cohort included 8,464 incident primary ITP adults. During the follow-up (31,448 patient-years), 1,089 patients were exposed to ELT, including 310 with a cumulative exposure ≥365 DDDs. Overall, 573 patients had a surgery of cataract; incidence: 1.91/100 person-years (95% CI: 1.76-2.07) and 1.51/100 person-years (95%CI: 1.17-1.96) in the "ever exposed" to ELT subgroup. The 573 cases were matched with 2,638 controls. Fifty-seven (9.9%) cases and 285 (10.8%) controls were “ever exposed” to ELT; 14 (2.4%) a,d 64 (2.4%) had a cumulative exposure ≥ 365 DDDs, respectively. Cases were more frequently exposed to corticosteroids, with a higher cumulative exposure, and had more ophthalmological risk factors. In the "ever/never" exposed analysis, the aOR for ELT was 0.82 (95% CI: 0.60-1.12). The aOR was 0.81 (95% CI: 0.57-1.14) in the 1-364 DDDs group, and 0.86 (95% CI: 0.48-1.56) in the 365 DDDs group, as compared with the "never exposed" group, respectively.Conclusions: This nationwide study did not find an increased risk of cataract in primary ITP patients treated with ELT.

PO-0049 Characteristics Of New Users Of Newer Antidiabetic Drugs In Canada And The United Kingdom 1400541 Vanessa Brunetti McGill Dept of Epidemiology, Biostatistics and Occupational Health Characteristics Of New Users Of Newer Antidiabetic Drugs In Canada And The United Kingdom Drug Utilization Research -> Trends and comparisons Background: The characteristics of patients using newer antidiabetic drugs (sodium-glucose co-transporter 2 inhibitors [SGLT-2i], dipeptidyl peptidase 4 inhibitors [DPP-4i], and glucagon-like peptide 1 [GLP-1] receptor agonists) in a real-world setting are poorly understood.Objectives: To describe the characteristics of new users of SGLT-2i, DPP-4i, and GLP-1 receptor agonists in Canada and the United Kingdom (UK).Methods: The Canadian Network for Observational Drug Effect Studies (CNODES) conducted a multi-center cohort study using administrative health databases from 7 Canadian provinces and the UK Clinical Practice Research Datalink. We assembled a base cohort of new users of antidiabetic drugs between 2006 and 2018. From this base cohort, we constructed 3 separate cohorts of new users of SGLT-2i, DPP-4i, or GLP-1 receptor agonists between 2016 and 2018. Descriptive statistics were used to characterize the cohorts.Results: Our base cohort of 2,175,815 new users of antidiabetic drugs included 166,722, 194,070 and 27,719 new users of SGLT-2i, DPP-4i and GLP-1 receptor agonists respectively, across jurisdictions. New users of GLP-1 receptor agonists were more likely to be female (54%) than new users of SGLT-2i (41%) or DPP-4i (44%), while new users of DPP-4i were more likely to be older (mean ± SD: 67.8 ± 12.3 yrs) than new users of SGLT-2i (64.4 ± 11.1 yrs) and GLP-1 receptor agonists (56.7 ± 12.2 yrs). The prevalence of cerebrovascular disease (SGLT-2i: 4%, DPP-4i: 7%, GLP-1 receptor agonists: 2%) and coronary artery disease (SGLT-2i: 20%, DPP-4i: 22%, GLP-1 receptor agonists: 16%) was similar among groups. In the UK, prevalence of obesity (body mass index ≥30kg/m2; SGLT-2i: 71%, DPP-4i: 52%, GLP-1i: 88%) and impaired renal function (estimated glomerular filtration rate 2; SGLT-2i: 6%, DPP-4i: 25%, GLP-1 receptor agonists: 15%) differed greatly between groups. New use of SGLT-2i or GLP-1 receptor agonists, but not DPP-4i, was commonly preceded by insulin use in Canada (SGLT-2i: 26%, GLP-1 receptor agonists: 33%) and the UK (SGLT-2i: 17%, GLP-1 receptor agonists: 29%).Conclusions: Although all 3 classes are recommended as 2nd or 3rd line therapy, important differences in patient characteristics of new users of newer antidiabetic drugs exist. GLP-1 receptor agonists appear to have been prescribed off-label for weight loss, and SGLT-2i are less frequently used among patients with renal insufficiency. Although large trials have demonstrated benefits with SGLT2i and GLP-1 receptor agonists, these drugs do not appear to be preferentially prescribed to patients at elevated cardiovascular risk.

PO-0052 Identifying Patients With Chronic Kidney Disease In A Population-based Laboratory Database 1400542 Uffe Heide-J�rgensen Identifying Patients With Chronic Kidney Disease In A Population-based Laboratory Database Disease Epidemiology/Clinical Course -> Other Background: Hospital diagnoses of chronic kidney disease (CKD) are likely incomplete and do not reflect the time of onset, as most patients with CKD are managed in general practice. Population-based laboratory databases are attractive data sources for patient identification in pharmacoepidemiological studies, as CKD is defined by laboratory findings in the current guidelines.Objectives: To apply different algorithms for identifying CKD from hospital and laboratory data, and to describe the resulting patient cohorts.Methods: We used population-based laboratory-, hospital-, and administrative databases in Northern Denmark to identify and follow patients with CKD as defined by six algorithms in 2009-2016. Five algorithms were based on laboratory data, requiring A) one measurement of estimated glomerular filtration rate (eGFR) 2, B) two such measurements at least 90 days apart, C) two such measurements at least 90 days apart with no eGFR > 60 mL/min/1.73m2 observed in-between, D) two such measurements 90-365 days apart, E) two such measurements or two measurements of urine albumin creatinine ratio > 30 mg/g at least 90 days apart. Algorithm F included patients with a hospital diagnosis of CKD. We estimated prevalence and incidence of CKD in the adult population, described baseline characteristics, and estimated 1-year mortality in the resulting cohorts after applying each algorithm.Results: Laboratory based algorithms A-E yielded CKD prevalences of 8.3%, 5.6%, 4.8%, 4.6%, and 5.9% as of 1 January 2009, and incidences of 8.9, 7.3, 5.9, 8.8, and 7.3 per 1000 person-years. The resulting cohorts A-E included 103,435; 84,688; 68,994; 75,031; and 100,957 patients, with corresponding median ages of 71, 74, 76, 75, and 73 years. The median eGFR levels at inclusion was 56, 55, 53, 54, and 56 mL/min/1.73m2, and the median time since first ever observed eGFR 2 was 0, 13, 17, 11, and 13 months. The 1-year mortalities were 8.1% (95% confidence interval 8.0-8.3), 8.1% (8.0-8.3), 7.9% (7.7-8.1), 8.6% (8.5-9.0), and 7.2% (7.1-7.4). Algorithm F resulted in a CKD prevalence of only 0.8%, and an incidence of 2.4 per 1000 person-years. The resulting cohort included 27,947 patients with a median age of 72 years, median eGFR of 47 mL/min/1.73m2, and median time since first ever observed eGFR 2 of 70 months. The 1-year mortality was 21.6% (21.2-22.1).Conclusions: The choice of algorithm used to identify CKD patients from databases affected estimated occurrence and cohort composition. The greatest difference was seen between algorithms based on laboratory data compared to hospital diagnoses.

PO-0059 The Availability Of Laboratory Data In Real-world Data And Association Of Temporal And Patient Factors 1400543 Paul Dillon F.Hoffmann-LaRoche AG The Availability Of Laboratory Data In Real-world Data And Association Of Temporal And Patient Factors Informatics Background: In clinical trials labs are ordered following a protocol, whereas the availability of lab data in RWD may vary according to timing of treatment and patient characteristics. Using RWD lab results requires caution when selecting cohorts or for use in statistical modelling to address biases.Objectives: To evaluate the availability of lab data and the association of temporal and patient factors within 3 electronic health record RWD cohorts.Methods: A retrospective study of patients diagnosed with diffuse large B-cell lymphoma (DLBCL), advanced non-small cell lung cancer (aNSCLC) or small-cell lung cancer (SCLC), on or after January 1, 2011, with at least 2 documented clinical encounters in the Flatiron Health EHR-derived database, aged ≥18 years, and who received a standard of care 1st line treatment regimen between January 1, 2011 and January 31, 2018. Index date was the start date for the 1st line regimen of interest. Patients were excluded if they had other malignancies before the index date or a gap in visits of >90 days between diagnosis and index date. The availability of lab data for 33 lab tests (blood counts, hepatic enzymes, inflammatory markers, renal function, and electrolytes) was evaluated 1) before treatment initiation from diagnosis date to index date and 2) during treatment between index date to end of treatment date. For each period, the proportion of patients within each cohort with at least 1 lab result available for any of the labs was determined. Differences in patient factors (gender, age, race, US geographical region, insurance type, ECOG score, disease staging, diagnosis year, and follow-up time) according to lab data availability were quantified by the standardized mean difference (SMD).Results: In total there were 5182, 12,351 and 3736 eligible DLBCL, aNSCLC and SCLC patients, respectively. For each cohort, patients were more likely to have at least one lab result available during the treatment period (93-98%) than before treatment initiation (63-83%). Across the three cohorts and two periods of interest, no patient factor was consistently observed to have an SMD<0.1 according to lab availability. Higher variation in lab availability existed in patient characteristics during the treatment period, reflected in higher SMDs. In particular aNSCLC and SCLC patients with shorter follow-up time were less likely to have lab data available during the treatment period.Conclusions: Lab availability varied according to timing of treatment and patient factors in 3 RWD cohorts. Evaluating variation in lab availability in RWD is important to consider regarding risk of biases such as selection bias.

PO-0061 Population Based Prevalence Of Selected High Priority Prescribing Cascades Among Patients With Alzheimers Disease And Related Dementia 1444305 Sonal Singh Population Based Prevalence Of Selected High Priority Prescribing Cascades Among Patients With Alzheimers Disease And Related Dementia Geriatric Pharmacoepidemiology Background: Prescribing cascades (PCs) occur when a healthcare provider prescribes a second, potentially unnecessary drug therapy to address the side effect of a drug. Population based estimates of PCs among patients with Alzheimer’s disease and related dementias (ADRD) are limited.Objectives: We estimated the incidence of Calcium channel blocker-diuretic (CCB-diuretic) and antipsychotic/antiemetic-antiparkinsonism drug PC among older adults treated for ADRD in two large US health plansMethods: We identified members aged 50+ who had been treated with an ADRD drug in the 365 days prior to or on Jan 1, 2017. Participants were enrolled in medical and pharmacy coverage for 1 year before and through cohort entry date, bridging enrollment gaps up to 45-days. We excluded participants with an institutional stay encounter and censored members based on first of: disenrollment from medical or pharmacy coverage, death, or end of data. We identified incident and prevalent CCB use and identified incident diuretic use separately among both incident and prevalent CCB users. Incident diuretic use was defined as diuretic use during day 8 through day 365 of CCB use after a washout period of 183 days. Similar methods were used to identify the antipsychotic/antiemetic-antiparkinsonism PC.Results: There were 121,538 participants with ADRD eligible for inclusion. Approximately 62% were female, with a mean age of 79.5 (SD ± 8.6). Their Combined comorbidity index, a measure of comorbidity was 3 (SD ± 2.7). The mean number of generic drugs used was 9.2 (SD ± 4.9). Overall 2.1% of the ADRD cohort experienced the CCB-diuretic PC with 1586 incident diuretic users among 36,462 prevalent CCB users (4.3%) and 161 incident diuretic users among 3304 incident CCB users (4.8%). Overall 0.3 % of the ADRD cohort experienced the anti-psychotic/antiemetic-antiparkinsonism PC. There were 306 incident antiparkinsonism drug users among 20,746 prevalent antipsychotic/antiemetic drug users (1.4%); and 36 incident antiparkinsonism drugs users among 4534 incident antiparkinsonism/antiemetic users (0.8%).Conclusions: Among two large nationally representative health plans in the United States, the burden of CCB-diuretic and anti-psychotic/antiemetic-antiparkinsonism PC was low. Administrative databases can be used to determine the burden of prescribing cascades.

PO-0063 Treatment Duration Estimation Based On Claims Vs Primary Care Prescriptions Data: An Analysis Of Opioid Use In The Sidiap Database 1400544 Junqing Xie CSM, NDORMS, University of Oxford Treatment Duration Estimation Based On Claims Vs Primary Care Prescriptions Data: An Analysis Of Opioid Use In The Sidiap Database Methods in Pharmacoepidemiology -> Measurement Methods - Validation methods to confirm data (including Patient Reported Outcomes development and validation); methods to deal with missing data Background: Treatment duration assigned to each prescription is typically estimated either from claims/dispensations or electronic medical records (EMR) data, with no opportunities for comparison between the former and the latter. The System for the Development of Research in Primary Care (SIDIAP) database is one of the few data sources where both prescriptions and dispensations can be linked and can therefore be used to compare both and hence inform future drug utilisation methods.Objectives: To compare the treatment duration estimated using the fixed-time period method based on dispensation data to those recorded on EMR in primary care.Methods: Data were obtained from the SIDIAP database (Catalonia, Spain), which comprises both EMR and pharmacy dispensation information, covering about 6 million people (>80% of the Catalan population). All users of opioids (fentanyl, morphine, tramadol, codeine) during 2007-2017, aged 18 years or older, and with linked prescription and dispensation data were included. Treatment duration for each dispensation was estimated using the fixed-time period method (number of packages * fixed pre-specified and clinically informed window) and compared to those recorded on EMR. Three fixed windows were tested based on clinical assumptions, namely 7, 15, and 30 days. The consistency/agreement between both was quantified based on the metric of the median difference.Results: The fixed window of 15-days showed a median difference of -4 and 0 days, with 23 and 14 days underestimation for the 7-day window, and 22 and 20 days overestimation for the 30-day window for fentanyl and morphine respectively. Similarly, the assumption of 15-day window for tramadol performed better than others, with a 5-day median difference. Codeine showed different results, and the assumption of 7-day window led to the lowest median difference of 1 day.Conclusions: The treatment duration estimated by dispensation data using fixed-time period method is comparable to that directly recorded on EMR in aggregate level. Clinical insights should be carefully considered to pre-set a reliable time window for different drugs when using claims data.

PO-0068 Changes In Prescribing Of Sodium-glucose Co-transporter 2 Inhibitors And Dipeptidyl Peptidase 4 Inhibitors In Singapore 1400545 Louise Goh Ministry of Health, Singapore Changes In Prescribing Of Sodium-glucose Co-transporter 2 Inhibitors And Dipeptidyl Peptidase 4 Inhibitors In Singapore Drug Utilization Research -> Changing drug utilization (interventions and implementation research) Background: The Agency for Care Effectiveness issued subsidy guidance and an Appropriate Care Guide in mid-2017 to encourage use of sodium-glucose co-transporter 2 inhibitors (SGLT2i) over dipeptidyl peptidase 4 inhibitors (DPP4i) as dual or triple therapy based on clinical and cost-effectiveness assessment. It was unclear if these efforts were associated with desired changes in prescribing for managing type 2 diabetes mellitus (T2DM).Objectives: To assess changes in prescribing of SGLT2i and DPP4i in T2DM patients managed at public healthcare institutions in Singapore post-subsidy.Methods: A retrospective cohort study was conducted using drug utilisation data and electronic health records. Two random samples (n = 10,000 T2DM patients each) stratified by care setting (i.e. 25% and 75% of patients managed at public hospitals and polyclinics or primary care clinics where most patients and milder cases are managed respectively) were selected from two periods, pre-subsidy (Sep-Dec 2016) and post-subsidy (Sep-Dec 2017). Changes in prescribing rates (PR = number of patients prescribed with studied drug among all T2DM patients in a given period) from pre- to post-subsidy were assessed using logistic regression, adjusted for covariates. Effect size was estimated using odds ratios (ORs) with 95% confidence intervals (CIs).Results: The two samples were similar in age and gender distribution. Overall, PR of SGLT2i was significantly increased by four times from 1.5% to 6.1% after subsidy listing (OR 3.9, 95%CI 3.2-4.8). The increase was more pronounced when SGLT2i was used with metformin (OR 5.4, 95%CI 4.1-7.2) or insulin (OR 2.7, 95%CI 1.9-3.9) compared with other non-recommended therapies (OR 2.0, 95%CI 1.4-3.1). Metformin monotherapy use also increased significantly by 3.5 times (OR 3.5, 95%CI 3.2-3.9). Conversely, PR of DPP4i was significantly reduced by 50% from 22.4% to 18.3% (OR 0.5, 95%CI 0.5-0.6), particularly for DPP4i monotherapy (OR 0.2, 95%CI 0.2-0.3) or when used with insulin (OR 0.6, 95%CI 0.5-0.7). The increase in overall PR of SGLT2i was primarily driven by increased use in polyclinics (0.01% to 4.8%), despite a significant increase in hospitals (5.9% to 10.0%). The decrease in DPP4i PR occurred only at polyclinics (23.3% to 17.0%, OR 0.4, 95%CI 0.4-0.4) but not at public hospitals (19.7% to 22.0%, OR 1.0, 95%CI 0.9-1.2).Conclusions: Changes in prescribing suggest early success in driving appropriate use of diabetic drugs, to increase SGLT2i use as second-line dual or triple therapy and reinforce metformin monotherapy use. Differential effect by care setting warrants investigation.

PO-0070 The Impact Of Socioeconomic Status On The Risk Of Total Hip Arthroplasty During 1995 To 2017 In Denmark 1400546 Nina Edwards Department of Clinical Epidemiology, Institute of Clinical Medicine, Aarhus University Hospital, Den The Impact Of Socioeconomic Status On The Risk Of Total Hip Arthroplasty During 1995 To 2017 In Denmark Disease Epidemiology/Clinical Course -> Other Background: People are widely affected by their socioeconomic status (SES) even in health care, where low SES is associated with higher risk of developing osteoarthritis (OA). Total hip arthroplasty (THA) is an effective procedure for reducing pain and improving function and quality of life in patients suffering from disabling OA. Thus, it is plausible to expect an increased need of THA among individuals with low SES. One previous study examined the use of THA in relation to SES in four European countries finding pro-rich-area inequality. However, the study has several limitations including missing data on the individual level measures of SES. Our hypothesis is therefore, that there is a socioeconomic inequality in THA utilization in Denmark.Objectives: Examine the association between SES and the risk of THA and whether the association is age- or time-dependent.Methods: This is a population based case-control study, where we included all patients undergoing primary THA due to OA (cases) in Denmark from the 1st of January 1995 to the 31st of December 2017 using the Danish Hip Arthroplasty Register. Using the Danish Civil Registration System each case was matched on sex, region of residence and surgery date with 5 controls from the background population. We retrieved individual-level data on SES markers (education, cohabitation, income and liquid assets). We performed logistic regression to estimate adjusted odds ratios (aOR) with 95% confidence intervals for THA, adjusting for SES markers and comorbidity according to Charlson.Results: A total of 108,946 THA cases and 544,730 population controls were identified. The aOR for the risk of THA with the lowest education vs. highest education was 1.42 (1.32-1.53) for the age group 45-55. The association between education and THA decreased with increasing age. The aOR for the risk of THA for patients who lived alone vs. lived with a partner was OR 0.93 (CI: 0.91-0.96). The aOR for the age group 45-55 with the lowest vs. highest income was 1.12 (1.02-1.23) and was 0.75 (0.70-0.80) for those with the lowest vs. highest liquid assets. The aOR for the risk of THA with the lowest education vs. highest was 1.23 (1.15-1.31) in the years 1995-2000 and decreased to 1.01 (0.97-1.05) in the years 2013-2017. For the lowest income vs. the highest the aOR was 1.22 (1.12-1.32) in 1995-2000 and decreased to 0.84 (0.80-0.89) in 2013-2017.Conclusions: The association between lower level of education, lower level of income, and living alone and the high risk of THA was age dependent. The inequality seen in the risk of THA by education decreased over time suggesting a time trend towards less socioeconomic inequality.

PO-0071 The Safety Of Antibody Drug Conjugates: A Literature Review 1400547 Joan Forns RTI Health Solutions The Safety Of Antibody Drug Conjugates: A Literature Review Safety End Points -> Cancer Background: Antibody-drug conjugates (ADCs) are a class of treatment agents for cancer combining the relative selectivity of targeted treatment with the cytotoxicity of chemotherapy drugs. Over 100 ADCs are under evaluation in clinical trials worldwide, but only 5 had been approved by both the US Food and Drug Administration and the European Medicines Agency at the time of study conduct: brentuximab vedotin (BV), ado-trastuzumab emtansine (T-DM1), gemtuzumab ozogamicin (GO), inotuzumab ozogamicin (IO), and polatuzumab vedotin (PV).Objectives: To identify the most common toxicities reported from phase 2-4 clinical trials or observational studies for the five mentioned ADCs.Methods: We searched PubMed and Embase in July 2019 for peer-reviewed, English-language articles published since 2015 that reported the toxicity of ADCs. We first identified the most commonly reported adverse events (AEs) (reported in at least 10% of patients in each study) and then calculated the median reported incidence for each AE and drug.Results: The literature search identified 357 unique publications. After reviewing titles and abstracts, we selected 58 articles for full-text review; 48 of these qualified for data extraction. Among the 48 selected studies, 18 (38%) reported AEs for BV, 15 (31%) for T-DM1, 7 (15%) for GO, 7 (15%) for IO, and 1 for PV (1%). Most included articles reported on both efficacy and safety; 40 reported on clinical trials (22 phase 2, 17 phase 3, and 1 phase 4), and 8 reported on observational studies. Of the 2 ADCs with more included studies, the median incidences of AEs of any grade were higher for BV than for T-DM1 for many AEs: anaemia (31% vs. 24%), neutropenia (32% vs. 13%), thrombocytopenia (57% vs. 29%), peripheral sensory neuropathy (23% vs. 13%), fatigue (40% vs 30%), nausea (40% vs. 36%), vomiting (26% vs. 18%), and diarrhoea (29% vs. 18%). On the other hand, median incidences were lower for BV than for T-DM1 for anorexia (18% vs. 21%), pyrexia (19% vs. 25%), myalgia (12% vs. 16%), and headache (14% vs. 27%). IO had higher median AE incidences than BV and T-DM1 for neutropenia (42%), thrombocytopenia (72%), fatigue (45%), nausea (49%), and pyrexia (32%). For GO, safety information was scarce; the median incidence was 45% for thrombocytopenia, 45% for increased aspartate aminotransferase, and 50% for fatigue. Finally, PV had the highest incidence of fatigue (60%) and diarrhoea (47%) based on a single study.Conclusions: The five approved ADCs were found to have somewhat distinct safety profiles. Most studies, especially most observational studies, reported on BV and T-DM1, which have been in the market longer than the other ADCs.

PO-0074 Classification And Disproportionality Analysis Of Medication Errors Reported To The Drug Commission Of German Pharmacists 1400548 Natalie Parrau Arzneimittelkommission Deutscher Apotheker Classification And Disproportionality Analysis Of Medication Errors Reported To The Drug Commission Of German Pharmacists Pharmacovigilance -> Other Background: Medication errors (ME) may occur in every step of the medication process and can lead to preventable harm. As a national pharmacovigilance center, the Drug Commission of German Pharmacists (AMK) collects and evaluates risks associated with medicinal products, including ME.Objectives: We classified spontaneously reported ME and used disproportionality analysis to detect safety signals.Methods: Spontaneous reports from January 2013 to December 2018 were considered as (potential) ME when reporters clearly pointed this out, or when the follow up of reports revealed evidence for ME. We used the European Medicines Agency’s (EMA) classification of ME reports, supplemented by coding the affected medication process: prescribing, storing, dispensing, preparation, and administration (with or without use of a drug delivery device). In addition, we analyzed ME in children. The reporting odds ratio (ROR) and its 95% confidence interval (95% CI) were determined on the 5th level of the Anatomical Therapeutic Chemical classification system, German version. We calculated ROR if at least four cases were reported and regarded a ROR as a signal if the lower 95% CI was greater than 1.Results: Within six years, 1398 (potential) ME reports were obtained. 395 ME (28.3%) were reported as potential or intercepted errors. In 759 cases (54.3%), no harm or adverse events (AEs) were reported. However, 224 ME (16.0%) led to harm or AEs, with n=58 being serious. Regarding the medication process, ME mostly occurred while administering medicinal products by using a drug delivery device (46.8%, n=654). In total, 123 reports (8.8%) were related to the treatment of children; administering (37.4%, n=46) or preparing errors (35.0%, n=43) were mostly reported. ME affecting children mainly referred to the systemic use of anti-infectives (46.3%, n=57). Disproportionality analysis generated 55 signals. Signal validation yielded that a majority of the signals (n=34) involved ME using a drug delivery device, e.g. an inhaler device for respiratory diseases or devices for insulin injection.Conclusions: Classification and evaluation of ME with respect to the medication process is a prerequisite for effective risk minimization measures. The use of anti-infective suspensions for children and of certain drug delivery devices seems to be error-prone. Thus, continuous patient and caregiver education provides a reasonable measure for health-care professionals to prevent ME.

PO-0076 Incidence Rates Of Acute Exacerbations Of Copd In Us Claims Data Using Icd-9 And Icd-10 1444151 Ashley Cole Incidence Rates Of Acute Exacerbations Of Copd In Us Claims Data Using Icd-9 And Icd-10 Methods in Pharmacoepidemiology -> Measurement Methods - Validation methods to confirm data (including Patient Reported Outcomes development and validation); methods to deal with missing data Background: Acute exacerbations of chronic obstructive pulmonary disease (AECOPD) are important endpoints in research leveraging administrative claims data. Prior work developed and validated a US claims-based AECOPD algorithm using ICD-9 diagnosis codes, but it is unclear whether this algorithm will perform consistently following transition to ICD-10.Objectives: To describe incidence rates (IR) by calendar year for moderate and severe AECOPD using a US claims-based algorithm applied before, during, and after transition to ICD-10 coding.Methods: Using Optum’s de-identified Clinformatics® Data Mart Database, we identified person time at risk for AECOPD during 2013-2018 from the earliest observed COPD diagnosis or study start (2012) until the earliest of death, disenrollment, or end of data. Incident events included all AECOPD (moderate and severe calculated separately) observed during the at risk period. We used a previously validated algorithm to define AECOPD related diagnostic codes. Severe AECOPD was defined as hospitalizations with a primary diagnostic code for AECOPD; moderate AECOPD was defined as outpatient/emergency visits with a primary diagnostic code for AECOPD and a fill for systemic steroids, antibiotics, or methylxanthines ≤14 days later. We used ICD-9 codes from the validated definition and performed a crosswalk exercise to translate these to equivalent ICD-10 codes. ICD-9 codes were applied to all time prior to 1 Oct 2015 and ICD-10 codes were applied afterwards. We calculated IR for each calendar year before (2013-2014), during (2015-2016), and after (2017-2018) ICD-10 transition.Results: We identified 1,979,470 moderate and 280,804 severe AECOPD in 2,134,298 subjects. For both moderate and severe AECOPD, the lowest IR were observed in 2015 and the highest IR in 2016, with similar rates in periods before and after ICD-10 transition (IR/person year [95% CI], Severe AECOPD: 2013, 0.077 [0.076-0.077]; 2014, 0.063 [0.063-0.064]; 2015, 0.052 [0.051-0.052]; 2016, 0.093 [0.093-0.094]; 2017, 0.086 [0.085-0.086]; 2018, 0.061 [0.061-0.062]; Moderate AECOPD: 2013, 0.492 [0.490-0.493]; 2014, 0.434 [0.432-0.435]; 2015, 0.433 [0.431-0.434]; 2016, 0.635 [0.633-0.637]; 2017, 0.540 [0.539-0.542]; 2018, 0.485 [0.484-0.487]).Conclusions: The ICD-9 to ICD-10 transition period was associated with higher variability in IR and researchers may want to use caution when ascertaining AECOPD during this time. Similar rates of AECOPD ascertained before and after the ICD-9 and ICD-10 period suggests that our ICD-10 algorithm performs in a similar way to the validated ICD-9 algorithm. Funding: GSK (213478).

PO-0077 Caliper Considerations For Propensity Score Matching 1400549 Elizabeth Garry Aetion, Inc Caliper Considerations For Propensity Score Matching Methods in Pharmacoepidemiology -> Confounding/Bias - Methods to avoid bias and confounding, including confounder summary scores (e.g., propensity scores, disease risk scores) Background: When using propensity scores (PS) to match initiators of a referent treatment to those initiating the exposure, a maximum caliper should be specified to allow for variation of the PS within each match. Although recommendations have been made previously for calipers 0.20-0.25 the standard deviation of the logit of the PS, the methods used to make these recommendations may have limited generalizability in regulatory settings. As tighter calipers yield closer matches, a conservative absolute caliper of 0.01 is frequently used in practice without strong precedent, but may exclude more exposed patients who are unable to be matched.Objectives: Evaluate changes to covariate balance and effectiveness estimates when the absolute caliper for PS is systematically varied.Methods: We used IBM Marketscan commercial claims data 2014-2018 to create cohorts 1:1 PS-matched without replacement to evaluate comparative effectiveness using a previously established example comparing systemic stroke or embolism among initiators of rivaroxaban versus warfarin. We started with the frequently used caliper of 0.01 and then increased the caliper by increments of 0.05. At each interval, we confirmed covariate balance (no standardized difference values ≥0.1 for any dichotomous covariates included in the PS model) and whether the regulatory decision would likely be the same based on the hazard ratio [HR] falling within the confidence interval of the previously published ROCKET trial that concluded that rivaroxaban was noninferior to warfarin (0.88 [0.74-1.03]).Results: Among the 3102 rivaroxaban and 4128 warfarin initiators, the unadjusted HR was 0.71 [0.53-0.95]. When a caliper of 0.01 was used for PS matching, 81.5% of exposed patients were matched in the analytic cohort, covariate balance was achieved, and the HR was 0.90 (0.63, 1.27). With calipers of 0.10, 0.15, 0.20, 0.25, and 0.30, covariate balance continued, the number of matched exposed patients increased to 93.2%, and the HR moved further from the null (0.89 [0.63-1.25], 0.87 [0.62-1.23], 0.85 [0.61-1.19], 0.83 [0.59-1.15], and 0.80 [0.58-1.11], respectively). At a caliper of 0.35, the HR was 0.80 (0.58-1.10) and we began to see covariate imbalance.Conclusions: In this example, covariate balance was achieved using a caliper as high as 0.30, which allowed for most exposed patients to be matched and yielded an estimate that would lead to the same non-inferiority regulatory decision. These findings suggest a need to further explore increases to the caliper used for PS matching to minimize the exclusion of exposed patients who might otherwise be unable to match without impacting the overall findings.

PO-0078 Comparative Effect Of Four Antimalarial Treatments On Haematocrit In Children In Southwest Nigeria 1400550 Zacchaeus Olofin University of Ibadan Comparative Effect Of Four Antimalarial Treatments On Haematocrit In Children In Southwest Nigeria Pediatric Pharmacoepidemiology Background: Anaemia in malaria has both central (dyserythropoiesis) and peripheral causes (phagocytosis of both infected and uninfected erythrocytes and haemolysis). However, it is often difficult to disentangle the anemia effect of malaria from its treatments.Objectives: To compare the change in haematocrit following four antimalarial treatments among children of microscopically-confirmed Plasmodium falciparum infection.Methods: Data were extracted from 313 case record forms of children that met the eligibility criteria aged 3-119 months enrolled in antimalarial clinical trials in Southwest Nigeria between 1998 and 2014. Study participants were followed up over a 28 day period according to the World Health Organization recommendation for treatment of malaria research participants. Enrollment criteria included symptoms compatible with acute uncomplicated malaria, including parasite density of at least 1000/µL and absence of chronic illness or danger signs of severe malaria. Change in haematocrit level from baseline through the treatment period and 28 days post treatment were compared among children treated with artemether-lumefantrine (82), artovaquone-proguanil (41), artesunate-amodiaquine (156) and chloroquine (34). Repeated measures analysis was done by fitting a general linear model (GLM).Results: The median age of the study population was 25 months and 54% were males. The mean differences (95% CI) in haematocrit from baseline were 4.7 (95% CI = 3.6, 5.8), 4.4 (95% CI = 2.7, 6.0), 3.8 (95% CI = 3.0, 4.7) and 2.4 (95% CI = 0.5, 4.4), for artemether-lumefantrine, artovaquone-proguanil and artesunate-amodiaquine and chloroquine, respectively. Using the general lineal model, repeated measure analysis showed that there were significant differences in the mean haematocrit level over the 28-day follow-up among the four treatment groups (p<0.05).Conclusions: All children experienced increases in haematocrit after treatment, with artesunate-amodiaquine appearing to result in a greater increase in haematocrit than other antimalarial drugs although not statistically significant, but might be significant with a larger sample size.

PO-0079 Prevalence Of Time-related Biases In Pharmacoepidemiology Studies Of Anti-emetics, Antifungal, And Antibiotic Medications During Pregnancy: A Systematic Review 1400551 Ugochinyere Vivian Ukah McGill Dept of Epidemiology, Biostatistics and Occupational Health Prevalence Of Time-related Biases In Pharmacoepidemiology Studies Of Anti-emetics, Antifungal, And Antibiotic Medications During Pregnancy: A Systematic Review Methods in Pharmacoepidemiology -> Study Design - Studies assessing study design (e.g., choice of comparator, self-controlled methods, avoiding immortal time bias) Background: The occurrence of time related biases, such as immortal time and time-window bias, have been demonstrated in observational studies including drug effectiveness studies. These biases often result in an observation of false protective associations between a treatment or exposure and the outcomes of interest. However, the prevalence of these biases in perinatal pharmaco-epidemiology has not been described.Objectives: To systematically review and estimate the prevalence of time-related biases (immortal time bias and time-window bias) in observational studies of classes of medications commonly used during pregnancy (antibiotic, antifungal and antiemetic drugs) via systematic review.Methods: We systematically searched Medline and EMBASE databases between January 2013 and December 2019 for observational studies reporting the association between the antibiotic, antifungal, and antiemetic drugs and adverse pregnancy outcomes related to gestational age. Pregnancy outcomes of interest were those defined by gestational age such as pre-eclampsia, gestational hypertension, gestational diabetes, spontaneous abortion, stillbirth, preterm delivery and small for gestational age. Articles were reviewed independently for eligibility and for time-related biases by 2 reviewers.Results: Of the 1,450 articles identified during our search, 17 studies were included in our systematic review (13 cohort, 3 nested case-control studies and 1 case-control study). Among these, 11 studies reported on antibiotics, 4 on anti-emetic, and 2 on anti-fungal medications. Spontaneous abortion (N=10 studies), and stillbirth and preterm birth (N=7 studies each) were the most reported outcomes. Eight studies (50%) had the potential to have time-related bias, specifically immortal time bias due to the misclassification of person-time between cohort entry and initiation of drug exposure. None of the included studies had time-window bias. Medications in studies with time-related biases appeared to be more protective for similar outcomes and comparators, compared with studies without time-related biases.Conclusions: The potential for immortal time bias occurs frequently in epidemiological studies of medications during pregnancy. There is need for more appropriate analyses in these studies, such as time-dependent analyses, to avoid time-related biases.

PO-0082 Ssri/snri Discontinuation In Pregnancy And Obstetrical Outcomes: A Nordic Population Register-based Study 1400552 Carolyn Cesta Ssri/snri Discontinuation In Pregnancy And Obstetrical Outcomes: A Nordic Population Register-based Study Pregnancy and Lactation Background: A significant proportion of reproductive aged women use selective serotonin reuptake inhibitors (SSRI) or serotonin-norepinephrine reuptake inhibitors (SNRI) to treat mild-to-moderate depression or anxiety, yet the guidelines for use during pregnancy are inadequate. While use of most of SSRI/SNRIs during pregnancy are considered safe for the fetus, women often discontinue use during pregnancy.Objectives: To test our hypothesis that women who discontinue SSRI/SNRI use during pregnancy have higher rates of obstetrical complications compared with women who continue use throughout pregnancy.Methods: Using the Nordic population health registers, an observational cohort study was conducted including all pregnancies to women who dispensed an SSRI or SNRI in the 90 days before the start of pregnancy from 2006-2016 (Sweden), 2004-2017 (Iceland), and 2005-2017 (Norway). Pregnancies in women with a history of inpatient psychiatric care, outpatient diagnoses of psychiatric disorders other than depression or anxiety, or co-medication to treat psychiatric disorders other than depression or anxiety were excluded. To identify discontinuation, k-means longitudinal modeling was used to cluster individual trajectories of SSRI/SNRI use based on the number of days of medication coverage for nine 30-day periods beginning from the start of pregnancy. Pregnancies in the trajectory with the highest continuous use over time were classified as continuers, and all others were classified as discontinuers. Rates of obstetrical complications in pregnancies with SSRI/SNRI discontinuation and continuation were calculated.Results: In total, 36,783 (3.3%) pregnancies were identified in Sweden, 3525 (5.9%) in Iceland, and 4391 (0.8%) in Norway. Discontinuation of SSRI/SNRIs occurred in 93.6%, 92.9%, and 91.4% of pregnancies, respectively. Preliminary results show no differences in the rates of delivery by caesarian section, placental abruption, preterm delivery (<37 gestational weeks), breech position anytime during pregnancy, delivery of a small-for-gestational-age infant, and perinatal death.Conclusions: The large majority of women discontinued SSRI/SNRI use during pregnancy; however, rates of obstetrical complications were not different between continuers and discontinuers. Next steps are to include data from Finland and Denmark, to conduct multivariate regression model analyses, and to investigate short and long-term maternal psychiatric health outcomes.

PO-0088 Liver Enzymes During and After Antimalarial Therapy in Nigerian Children With Uncomplicated Plasmodium Falciparum Infection 1400553 Zacchaeus Olofin University of Ibadan Liver Enzymes During and After Antimalarial Therapy in Nigerian Children With Uncomplicated Plasmodium Falciparum Infection Pediatric Pharmacoepidemiology Background: Derangement of liver enzymes is a known adverse event associated with antimalarial due to drug-induced liver toxicity. However, it remains unclear whether these changes in liver enzyme levels persist following the completion of antimalarial therapy.Objectives: To determine the effect of artemether-lumefantrine on plasma levels of four liver enzymes, namely; alanine aminotransferase [ALT], aspartate aminotransferase [AST], alkaline phosphatase [ALP] and gamma glutamyl transpeptidase [GGT] in children with uncomplicated Plasmodium falciparum infection during and after completion of antimalarial treatment.Methods: We reviewed the records of all children who participated in a clinical trial of antimalarial drug in Ibadan, Nigeria and 102 children who met the eligibility criteria for this analysis (i.e microscopically-proven Plasmodium falciparum infection treated with artemether-lumefantrine at the recommended age-specific dose for 3 days). Study participants were followed up on days 3, 7, 14, 21, and 28 according to the World Health Organization recommendation for treatment of malaria research participants. Inclusion criteria for the clinical trial included symptoms attuned with acute uncomplicated malaria, parasite density of at least 1000/µL, absence of chronic illness or signs of severe malaria. The results of ALT (U/L), AST (U/L), ALP (U/L) and GGT (U/L) at baseline (day 0), on day 3, and day 28 post-treatment were extracted and compared using Friedman two-way ANOVA at p = 0.05. Normal range for ALT 5-45U/L, AST 15-63U/L, ALP 145-420U/L and GGT 5-90U/LResults: The median age of participants was 25 months (range = 3 to 119 months), and 49% were male. The mean ALT decreased from 25.8 (95% CI=15.5-36.0) at baseline to 23.1 (95% CI=16.4-29.8) on day-3 and 19.1 (95% CI=15.8-22.4) on day-28 (p=0.219), p = 0.219. The mean AST increased from 50.3 (95% CI=36.7-64.2) at baseline to 51.0 (95% CI=36.9-65.1) on day-3 and 55.2 (95% CI=37.5-72.9) on day-28; p=0.198. ALP decreased from 319.2 (95% CI=287.6, 250.7) at baseline to 308.3 (95% CI=279.7-336.8) on day-3 and increased to 373.8 (95% CI=327.9-419.8) on day-28; p=0.042. The GGT increased from 21.1 (95% CI=17.5-24.8) at baseline to 22.6 (95% CI=19.3-25.8) on day-3 and decreased to 18.7 (95% CI=1.9-14.9) on day-28; p=0.001.Conclusions: Substantial rise in ALP and GGT suggesting liver injury during malaria treatment among Nigerian children. Further research is needed to identify the underlying mechanism responsible for this drug-induced liver toxicity.

PO-0090 Use Of Hydrochlorothiazide In Denmark Following Publication Of Skin Cancer Risk Findings 1400554 Anton Pottegård University of Southern Denmark Use Of Hydrochlorothiazide In Denmark Following Publication Of Skin Cancer Risk Findings Drug Utilization Research -> Changing drug utilization (interventions and implementation research) Background: The antihypertensive agent hydrochlorothiazide has recently been linked to an increased the risk of non-melanoma skin cancer, in particular squamous cell carcinoma (SCC). The three key papers were published in rapid succession on June 6th 2017, December 4th 2017, and May 29th 2018, respectively. Based on an evaluation by the Pharmacovigilance Risk Assessment Committee (PRAC) under, Summary of Product Characteristics (SmPC) were updated and ‘dear health care professionel’ (DHCP) letters were distributed to prescribers across Europe October 17th 2018.Objectives: In this study, we sought to describe the impact of the dissemination of these findings on the use of hydrochlorothiazide and health care utilization among antihypertensive users in Denmark.Methods: In this nationwide observational study, we performed descriptive analyses of a cohort comprising all Danish antihypertensive treatment users 2016-2018 (n=1,253,886) with special focus on hydrochlorothiazide users (n=296,597). Data was retrieved from the Danish nationwide health registries, including the Danish National Prescription Registry.Results: The proportion of antihypertensive fills constituted by hydrochlorothiazide dropped from 12.7% to 9.3% during 2016-2018. This drop was more pronounced among younger patients and patients with a history of skin cancer. Simultaneously, the monthly rate of new hydrochlorothiazide users in Denmark dropped from 2350 to 742. The publication of an increased risk of non-melanoma skin cancer led to an estimated excess of up to 11510 physical and 228700 e-mail/phone consultations to general practitioners. No evidence for increased risk of adverse outcomes was found.Conclusions: The publication of increased risk of skin cancer with hydrochlorothiazide use have led to a decline in the use of hydrochlorothiazide in Denmark, in particular among younger patients and patients with a skin cancer history.

PO-0092 Association Between Dietary Diversity And All-cause Mortality Among Chinese Adults 65 Years Of Age Or Older, A Community-based Cohort Study 1400555 Liyuan Tao Peking University Third Hospital Association Between Dietary Diversity And All-cause Mortality Among Chinese Adults 65 Years Of Age Or Older, A Community-based Cohort Study Disease Epidemiology/Clinical Course -> Other Background: Diet is a well-established major modifiable risk factor for multiple chronic disease and mortality, but few studies have evaluated the association of dietary diversity and mortality in older adults.Objectives: We aimed to examine the prospective association between dietary diversity and the risk of all-cause mortality in a community-based cohort study of older Chinese adults.Methods: In this prospective cohort study, we invsetigated the association between dietary diversity and all-cause mortality among 17949 community-based participants aged 65 years or old in China. The consumption on the frequencies of food groups was collected at baseline, and a dietary diversity score (DDS) was calculated. Survival status and date of death were collected during the follow-up survey. We used univariate and multivariable Cox proportional hazards models to estimate hazard ratios (HRs) and 95% confidence intervals for all-cause mortality with higher DDS (score: 8-9) versus lowest DDS (score 0-1).Results: We identified 8445 death events over 57,685 person-years of follow up. Compared with participants in the lowest DDS group (score 0-1), higher DDS was associated with lower mortality risk in the univariate models. After adjusting for potential confounders, participants in the higher DDS group had a 9%-30% lower risk in all-cause mortality (p trend<0.001) compared to participants in the lowest DDS group. The inverse relationship between diet diversity score and all-cause mortality was also statistically significant in four food groups (vegetables, fish, fruits, and nuts). We observed similar results after including different confounders in model 2 and model 3 in the sensitivity analyses.Conclusions: Dietary diversity is inversely associated with all-cause mortality among the Chinese elderly, especially in the oldest old and men. These findings suggest that increasing dietary diversity may lower mortality rates in older population and that tailored interventions on improving dietary diversity may improve health and increase survival in this population.

PO-0094 Drug Survival Of Secukinumab, Ustekinumab, And Adalimumab In Patients With Psoriasis 1400556 Zenas Yiu The University of Manchester Drug Survival Of Secukinumab, Ustekinumab, And Adalimumab In Patients With Psoriasis Drug Effectiveness Background: Biologic drug survival in the real world is an important proxy measure for effectiveness. Predictors of drug survival may help patients with psoriasis choose the right biologic first time.Objectives: 1. To assess the relative drug survival of adalimumab, secukinumab and ustekinumab in patients with psoriasis; 2. To look for predictors for biologic drug survival in the same cohort.Methods: The prospective cohort study was performed using data from the British Association of Dermatologists Biologics and Immunomodulators Register (BADBIR), a multicenter longitudinal pharmacovigilance registry of patients with moderate-to-severe psoriasis in the United Kingdom and Republic of Ireland between November 2007 and August 2019. The outcomes were survival functions after 1 and 2 years, separated by discontinuation reason; adjusted hazard ratios (adjHR) for adalimumab and secukinumab compared with ustekinumab; and adjHRs for the effect of the potential predictors of drug discontinuation due to ineffectiveness in a flexible parametric survival model.Results: 9,652 participants were included in the study, with 5,542 starting on adalimumab (57.4%); 991 on secukinumab (10.3%); and 3118 on ustekinumab (32.3%) after registration. The overall drug survival of adalimumab, secukinumab and ustekinumab was 0.78 (95% CI 0.77-0.79), 0.88 (0.86-0.91) and 0.88 (0.87-0.89) in Year 1 respectively, dropping to 0.66 (0.64-0.67), 0.77 (0.73-0.80), and 0.77 (0.76-0.79) respectively in Year 2. Fourteen covariates were included. The adjHR for discontinuation of adalimumab and secukinumab compared with ustekinumab were 2.11 (1.76-2.54) and 0.67 (0.40-1.11) respectively. Psoriatic arthritis predicted for survival in the adalimumab and secukinumab cohorts (adjHR 0.67 (0.51-0.88); 0.70 (0.40-1.24) respectively) but for discontinuation in the ustekinumab cohort (adjHR 1.42 (1.12-1.81)), while having previous exposure to biologic therapies predicted for discontinuation in the ustekinumab and secukinumab cohorts (adjHR 1.54 (1.26-1.89); 1.49 (0.91-2.45) respectively) and for survival in the adalimumab cohort (adjHR 0.71 (0.55-0.92)).Conclusions: Secukinumab and ustekinumab have similar sustained drug survival, while adalimumab has a lower drug survival in patients with psoriasis. We identified psoriatic arthritis and previous biologic experience as predictors of drug survival with differential effects between the biologic therapies.

PO-0100 Readmissions And Postoperative Mortality Of Coronary Artery Revascularization In People With And Without Alzheimer’S Disease: A Nationwide Cohort Study 1400558 Mai Vu University of Eastern Finland Readmissions And Postoperative Mortality Of Coronary Artery Revascularization In People With And Without Alzheimer’S Disease: A Nationwide Cohort Study Health Economics/Outcomes Research Background: Coronary artery disease is common in older persons, and one of the management strategies is revascularization procedure (coronary artery bypass grafting and percutaneous coronary intervention). However, effectiveness as well as survival rate in persons with Alzheimer’s disease (AD) after these procedures have not been investigated before.Objectives: To compare the incidence of coronary artery revascularization, readmission, and mortality after procedure among persons with and without AD.Methods: Our study is based on a retrospective register-based Medication Use and Alzheimer’s disease (MEDALZ) cohort which includes all community dwellers with a clinically verified AD diagnosis during 2005 - 2011 in Finland (N=70 719). For each incident revascularized persons with AD, four comparison persons without AD were was matched by age, sex, and hospital district. The follow-up for this study began on the date of AD diagnosis (matching date for comparison persons). Altogether 448 persons with AD and 5999 without AD underwent revascularization during the follow-up. The outcomes were 30-day and 90-day re-admission rate after discharge, and all-cause 1-year and 3-year mortality. Analyses were conducted with Cox proportional hazard models adjusted with age, sex, comorbidities, and statin use.Results: Revascularization were less often performed in persons with AD (adjusted hazard ratio HR = 0.23, 95% CI 0.21 - 0.26). Persons with AD tended to have lower risk of 30-day readmission (adjusted HR: 0.77, 95% CI 0.57 - 1.03), and 90-days readmissions (adjusted HR: 0.87, 95% CI 0.70 - 1.08). Higher proportion of persons with AD (7.4%) died in the operative unit compared with persons without AD (4.8%) There was no difference in 1-year mortality (adjusted HR:1.20, 95%CI 0.89 - 1.62) but people with AD had higher 3-year mortality (adjusted HR: 1.73, 95%CI 1.48 - 2.01).Conclusions: The higher three-year mortality in people with AD after revascularization is in line with the overall increase in risk of death in persons with AD. Persons with AD had higher in-hospital mortality but there were no differences in the short-term outcomes regarding rehospitalizations or death. Lower rate of revascularization, especially given the higher prevalence of ischaemic heart disease in people with AD raises questions about possible undertreatment as this group also needs these operations to improve functioning and quality of life.

PO-0101 Sex Differences In Hospital Admissions Related To Adverse Drug Reactions 1400559 Linda Hendriksen Tergooi Hospital Sex Differences In Hospital Admissions Related To Adverse Drug Reactions Safety End Points -> Other Background: Adverse drug reactions (ADRs) are a major health concern and are responsible for approximately 5% of all acute hospital admissions. Women are 1.5-1.7 times more likely to develop ADRs than men. Differences in ADR-related admissions between women and men have been described before in studies on spontaneous ADR reports and in an earlier study where adjusting for potential confounders was not possible due to its ecological design.Objectives: The main objective of this study was to investigate whether there are sex differences in ADR-related hospital admissions and examine whether women or men are more prone to develop particular drug-ADR combinations. Methods: Patients were selected from the PHARMO Database Network, consisting of over four million residents of the Netherlands with an average follow-up of ten years. ADR-related hospital admissions between 2005 and 2017 were identified using hospital discharge codes indicating an ADR of a drug group used at a therapeutic dose, conform the ICD 9 and 10 coding system. Patients aged ≥ 16 years who had a dispensing of the relevant drug within three months before the hospital admission were included. Drug-ADR combinations with at least 50 hospital admissions for women or men were examined, excluding sex-specific ADRs. Age-adjusted odds ratios (OR) with 95% CIs for drug-ADR combinations for women versus men were calculated with respect to the number of female and male users. Results: There were a total of 18,469 hospital admissions involving women (0.35% of the total number of admissions) and 14,678 admissions involving men (0.35% of the total number of admissions) due to an ADR. ORs were calculated for 48 drug-ADR combinations. There were 10 combinations with a statistically significantly higher risk in women versus men and 8 in men versus women. The most substantial differences were seen in ADRs due to anticoagulants and diuretics. Anticoagulants showed a lower risk of hospital admissions with persistent haematuria (OR 0.31; 95% CI 0.21, 0.45) haemoptysis (OR 0.47, 95% CI 0.30, 0.74) or subdural haemorrhage (OR 0.61; 95% CI 0.42, 0.88) in women than in men and a higher risk of rectal bleeding in women (OR 1.48; 95% CI 1.04, 2.11) than in men. Also, there was a higher risk of hospital admission in women using thiazide diuretics causing hypokalaemia (OR 3.03; 95% CI 1.58, 5.79) and hyponatraemia (OR 3.33, 95% CI 2.31, 4.81) than in men.Conclusions: Sex differences exist for ADR-related hospital admissions due to anticoagulants and diuretics. Further research into the mechanisms of and potential confounders for these observed differences can contribute to the development of sex-specific guidelines.

PO-0107 A Standardized And Reusable Method To Link Multiple Distributed Health Plan Databases To The National Death Index 1400561 Candace Fuller Harvard Pilgrim Health Care Institute A Standardized And Reusable Method To Link Multiple Distributed Health Plan Databases To The National Death Index Informatics Background: In the US, health plan (HP) databases typically capture medically attended deaths during enrollment but many do not capture out-of-hospital deaths. Linkage to the National Death Index (NDI) can augment HP based studies with death and cause of death information, but linkage for multi-site studies can be resource intensive.Objectives: To develop and test a reusable distributed NDI linkage approach within a use case comprising of antiarrhythmic (AR) medication users and non-users.Methods: As a test case, we identified a cohort of AR users, known to be at increased risk of death, in 6 HPs in years 2000-2017 and followed them through dispensing data. For reference, we matched AR users 1:1 on age, sex, and time with non-users. We submitted patients with HP-recorded death or potential death to the NDI for linkage. We defined potential death as HP disenrollment between cohort entry and cohort exit plus 365 days, without reenrollment or medical utilization >60 days post-disenrollment.At 5 HPs that ascertain death with state death records and other sources we examined concordance, sensitivity (Sn), and positive predictive value (PPV) of death identified through NDI linkage compared to HP-recorded death (reference).We developed and tested reusable distributed NDI linkage programs which avoid sharing identifiable patient information between HPs or the coordinating center. Programs identify cohorts, select potential deaths for NDI submission, quality check NDI records flagged for submission, and select the single best match from returned NDI data.Results: We identified 60,785 AR users and 60,785 matched non-users at 6 HPs. HPs submitted 33,703 patients (AR=57%, non-users=43%) to the NDI. The NDI returned at least one potential match for 22,542 (67%) patients. Our program for selecting the best NDI match classified 13,498 (40%) patients as dead (AR=51%, non-users=26%). In analyses utilizing HP-recorded death for reference (n=5 HPs), we observed excellent concordance between HP-recorded deaths and NDI identified death (96% agreement; Cohen’s Kappa=0.919). The NDI captured 7,402 of 7,671 HP-recorded deaths (Sn=97%, PPV=97%). The NDI identified 238 deaths not identified in HP data; review of discordant records is ongoing.Conclusions: The developed approach standardized the NDI linkage across HPs. We observed high concordance between the best match chosen by the NDI linkage program and HP-recorded death. Future work will further test and refine the reusable distributed NDI linkage approach, with the goal of providing NDI linkage methods to future studies based in distributed data networks.

PO-0108 Prescribing and Non-Medical Use of Tramadol and Other Opioid Analgesics: U.S., 2009-2018 1400562 Saranrat Wittayanukorn US Food and Drug Administration Prescribing and Non-Medical Use of Tramadol and Other Opioid Analgesics: U.S., 2009-2018 Pharmacovigilance -> Other Background: Whereas most commonly prescribed opioid analgesics (OAs) are Schedule II controlled substances in the U.S., tramadol is Schedule IV. Given the evolving U.S. opioid crisis, it is important to surveil U.S. tramadol dispensing and nonmedical use trends.Objectives: 1) To estimate patterns of dispensing of tramadol and the commonly prescribed Schedule II oral OAs: hydrocodone, morphine, and oxycodone; 2) to describe the frequency, trends, and clinical characteristics of poison control center (PCC) calls and emergency department (ED) visits involving nonmedical use of these OAs.Methods: We analyzed IQVIA data to examine outpatient utilization patterns from 2009-2018. We conducted descriptive analyses of the National Poison Data System (NPDS; 2014-2018) and National Electronic Injury Surveillance System-Cooperative Adverse Drug Event Surveillance (NEISS-CADES; 2016-2017) to examine the frequency of PCC calls and ED visits associated with nonmedical use of tramadol and the other selected OAs, and to describe the medical severity and route of nonmedical use.Results: Since 2013, immediate-release (IR) tramadol single-entity (SE) has been the second most dispensed OA. Tramadol prescriptions declined from an estimated 42 million in 2014 to 33 million in 2018; while its market share percentage among OA prescriptions increased (13.6% in 2012 to 18.7% in 2018). Prescriptions dispensed for IR formulations of hydrocodone, morphine, and oxycodone also decreased from 2014 to 2018. The annual number of U.S. PCC calls involving tramadol nonmedical use declined from 1,713 calls in 2014 to 929 in 2018, which remained higher than that of morphine (n=292) but lower than that of hydrocodone (n=1,674) or oxycodone (n=2,054) in 2018. The oral route was the most commonly reported for nonmedical use calls involving tramadol (>98% of reported routes), hydrocodone (98.5%), morphine (82.8%), and oxycodone (88.8%). Among 3,983 calls involving tramadol nonmedical use, approximately 90% of medical outcomes were categorized as minor or moderate effect (symptoms that were non-life-threatening). Of U.S. ED visits attributed to nonmedical use of any prescription OAs, an estimated 3.3% (95% CI: 2.4-4.2) involved tramadol, the lowest of the OAs examined.Conclusions: The number of dispensed prescriptions of IR tramadol SE increased from 2009-2014, then gradually declined through 2018. PCC calls involving nonmedical use of tramadol or selected Schedule II OAs all declined from 2014-2018. Tramadol was less frequently implicated in PCC calls and ED visits involving nonmedical use than were the Schedule II OAs hydrocodone and oxycodone.

PO-0120 Changes In Early Life Exposure To Antibiotics: A Comparison Of Birth Cohorts Based On German Claims Data 1400564 Oliver Scholle Leibniz Institute for Prevention Research and Epidemiology - BIPS Spotlight Poster Session Spotlight Poster Pediatrics Background: Antibiotic exposure in early life has been associated with an increased risk of chronic conditions such as childhood obesity as well as with negative effects on the gut microbiota. To avoid inappropriate antibiotic prescribing, antimicrobial stewardship programs have been implemented. However, detailed information on potential changes, e.g., regarding first antibiotic use in life, is lacking.Objectives: To describe the time span between birth and first antibiotic prescription as well as the type of the first prescribed antibiotic agent in different birth cohorts in Germany (2004-2015).Methods: Using the German Pharmacoepidemiological Research Database (GePaRD), covering ~17% of the German population, we identified children born between 2004 and 2015. Each newborn was followed until 2016 or end of enrolment in the database. Prescriptions of systemic antibiotics were identified based on outpatient drug dispensations. We estimated Kaplan-Meier curves to determine time to first antibiotic prescription in life and assessed the type of the first prescribed antibiotic agent.Results: Among ~1.5 million newborns included overall, 55% received an antibiotic prescription before the age of two years. Children in more recent birth cohorts received the first antibiotic prescription later than those in less recent birth cohorts: The median age at first antibiotic prescription was 18 months for children born in 2004 and 24 months for those born in 2014. Across all birth cohorts, amoxicillin and the broad-spectrum antibiotic cefaclor were the most frequently prescribed first-in-life antibiotics, and their proportions rose between 2004 and 2014 (amoxicillin: from 31% to 37%; cefaclor: from 20% to 29%). An increasing proportion was also observed for the broad-spectrum antibiotic cefuroxime (2% in 2004 and 6% in 2014), whereas it decreased for the narrow-spectrum antibiotics erythromycin (13% in 2004 and 6% in 2014) and phenoxymethylpenicillin (7% in 2004 and 4% in 2014).Conclusions: Our study showed that the age at first antibiotic prescription has increased over the years, but the proportion of broad-spectrum antibiotics prescribed as first-in-life antibiotic agents continues to be at a high and further increasing level. The appropriateness of this prescribing behavior requires further monitoring to ensure rational antibiotic prescribing in early life.

PO-0121 Real-world Safety Of Palbociclib In Breast Cancer Patients In The United States 1400565 Daniel Beachler HealthCore, Inc. Real-world Safety Of Palbociclib In Breast Cancer Patients In The United States Safety End Points -> Other Background: Palbociclib (palb) received FDA approval in 2015 for advanced stage HR+/HER2- breast cancer and is now a standard of care.Objectives: Provide real-world safety information on palb.Methods: We conducted a new user cohort study of breast cancer patients initiating palb and/or fulvestrant (fulv) from 02/01/2015 to 09/31/2017 using data from the HealthCore Integrated Research Database (HIRD), a longitudinal claims database from commercial health plan members in the US. The fulv monotherapy cohort was historical, comprising patients initiating fulv from 01/01/2011 to 01/31/2015. Propensity score matching and Cox proportional hazard regression were used to compare incidence rates between groups during therapy for 42 safety codes of interest (including acute liver injury (ALI)) which were defined via ICD-9/10 code algorithms. For ALI, additional analyses, medical record validation and preparation of case narratives for cases confirmed by hepatology experts were conducted, and a contemporaneous comparator group of new users of fulv monotherapy examined from 02/01/2015 to 09/31/2017 was added.Results: There were 2,445 patients initiating palb including 566 new users of palb + fulv. ICD-9/10 codes of interest were identified for the palb + fulv cohort, including neutropenia, anemia, interstitial lung disease/pneumonitis, and serious infections (incidence rates >20 per 100 person-years). Compared to historical fulv new users, new users of palb + fulv had more claims for neutropenia, leukopenia, anemia, stomatitis and mucositis, and ALI (Hazard Ratios (HRs)>1.6), whereas other claims, e.g., serious infections, and QT prolongation were similar between groups (HRs<1.4). Using several algorithms to identify ALI, an elevated hazard ratio of ALI in new users of palb + fulv was observed compared to the historical fulv new users after control for additional ALI risk factors (e.g., primary ALI algorithm: HR=3.0, 95% CI=1.1-8.4). However when new users of palb + fulv were compared to the contemporaneous fulv comparator group, no increased hazard ratio was observed, although imprecise (HR=0.5 (95% CI=0.1-2.2)). In addition, available ALI cases reviewed by a blinded hepatology expert, did not directly attribute any of these cases to palb or fulv, instead attributing the majority of them to be a consequence of liver metastases progression. Residual confounding effects cannot be excluded.Conclusions: A large real-world study of new users of palb + fulv, suggests that the observed safety profile is generally consistent with the known safety profile of palb + fulv.

PO-0122 The Impact Of Refill Gaps On The Estimation Of Persistence With Opioids Use 1400566 Junqing Xie CSM, NDORMS, University of Oxford The Impact Of Refill Gaps On The Estimation Of Persistence With Opioids Use Methods in Pharmacoepidemiology -> Measurement Methods - Validation methods to confirm data (including Patient Reported Outcomes development and validation); methods to deal with missing data Background: Long-term opioid use for the treatment of common pain syndromes is increasing worldwide, with little evidence on their long-term effectiveness and safety. Refill gaps are used in drug utilisation research to define treatment discontinuation, but the choice of length of these gaps is usually subjective. There is little understanding of how different refill gaps might influence the estimation of persistence in studies of long-term opioid therapy.Objectives: To explore the impact of different refill gaps on estimates of 1-year persistence with continued opioid use. Secondly, we compared the results obtained based on dispensation vs prescription data.Methods: Data were obtained from the SIDIAP (The System for the Development of Research in Primary Care) database, which comprises both electronic medical records and pharmacy dispensation information, covering about 6 million people (>80% of the Catalan population). All incident users of opioids (fentanyl, morphine, tramadol, codeine) based on ATC codes during 2007-2017, aged 18 years or older, and with linked prescription and dispensation data were included. For each of the prescription and dispensation data, persistence was calculated as the time from initiation to discontinuation of therapy. Refill gaps of 1, 2, and 3 months were used to identify discontinuation. 1-year persistence was defined as the proportion of individuals who had one-year or more continuous opioid use.Results: 1-year persistence with different opioids estimated using electronic prescription data increased slightly with refill gap length from 1 to 3 months, ranging from 0.6% to 0.7% (codeine), 5.6% to 7.2% (tramadol), 4.4% to 5.7% (morphine) and 14.7% to 17.8% (fentanyl). Estimates were similar when based on dispensation data, eg 0.1% to 0.3%, 2.4% to 5.4%, 3.9% to 6.3% and 14.9% to 19.1% accordingly.Conclusions: Although the longer refill gap can result in a higher estimate of 1-year persistence with different opioids, the variation seems to be acceptable when refill gaps ranging from 1 to 3 months. Also, the results are similar when dispensation vs prescription data are used to estimate persistence. Sensitivity analyses using variable refill gaps are needed for drug utilisation and safety/effectiveness studies.

PO-0128 Impact Of The Choice Of Clinical Code Set On The Outcomes Of A Real-world Database Study 1400567 Amy Buchanan-Hughes Costello Medical Consulting Ltd Impact Of The Choice Of Clinical Code Set On The Outcomes Of A Real-world Database Study Methods in Pharmacoepidemiology -> Study Design - Studies assessing study design (e.g., choice of comparator, self-controlled methods, avoiding immortal time bias) Background: Analysis of real-world data commonly requires the development of clinical code sets to define particular conditions used in determining patient eligibility, outcomes and covariates of interest. However, little guidance exists on developing code sets and these can show wide discrepancies across studies of the same condition.Objectives: Using the example of chronic obstructive pulmonary disease (COPD), we evaluated the impact of the choice of code set on study outcomes including demographics, costs and resource use.Methods: Thirteen unique code sets for COPD using the International Classification of Diseases, 10th Revision (ICD-10) were drawn from a previous literature review. These were mapped to ICD-10 Clinical Modification (ICD-10-CM) and used to select patients based on 2016 weighted national estimates from the US National Inpatient Sample (NIS). Demographics, costs and resource use were compared between the populations identified by each code set, both in full and after excluding codes common to all sets (J44*, "Other COPD").Results: Although code sets comprised 3-56 codes (identifying n=626,019-918,314 patients), the three J44* codes accounted for ≥98% of principal diagnoses in the majority (n=10/13) of analyses (68-74% in the other 3 analyses); therefore, age, sex, length-of-stay (LOS) and costs were similar across all code sets. Compared to patients with a principal diagnosis of J44* (n=626,019), patients with other principal diagnoses were typically younger (lowest proportion of patients ≥65 years was 17.9% vs 59.8% for J44*), had higher costs (highest mean $29,593 vs $8,445 for J44*) and longer LOS (highest mean 14.1 vs 4.2 days for J44*).Conclusions: The dominance of J44* as a principal diagnosis led to large overlaps between the 13 cohorts, so overall results were broadly similar among code sets. However, the additional patients identified by broader code sets typically had different characteristics to the "core" cohort. These results highlight the importance of validation studies to identify appropriate code sets, including careful consideration of the context in which specific codes are used in clinical practice and claims billing.

PO-0129 Screening For Potential Prescription Drug Effects On Prognosis Of Localized Colorectal Cancer 1444176 Anne Katrine Duun-Henriksen Screening For Potential Prescription Drug Effects On Prognosis Of Localized Colorectal Cancer Disease Epidemiology/Clinical Course -> Cancer Background: Identification of harmful or beneficial effects of drugs on cancer prognosis has inspired efforts in drug safety and drug repurposing for cancer treatment. Large registries are excellent data resources for screening for potential candidate drugs.Objectives: In this study, we present a framework for screening the potential effects of prescription drugs on cancer prognosis based on the Danish nationwide health registries, illustrated by localized colorectal cancer (CRC) patients.Methods: We used the National Cancer Registry to identify all one-year survivors of localized CRC in the period 2002-2015 and retrieved their prescription history from the Danish Prescription Registry. We followed the patients from one year after diagnosis up to 5 years after diagnosis. The framework consists of two estimation steps and three evaluation criteria for each unique drug class (4th ATC level). First, inverse probability (IP)-weighted hazard ratios (HRs) for overall and cancer-specific death from Cox models were used to compare drug users vs. nonusers. Stabilized IP weights were estimated from a logistic regression model with sex, age at diagnosis, period of diagnosis, marital status, income, and cancer treatment included as covariates. Secondly, five-year survival probabilities for users and nonusers were estimated using the IP-weighted Kaplan-Meier (KM) estimator. For each drug class, use was defined as redeeming at least one prescription within the first year after diagnosis. Drug classes with less than 100 users were not considered. For each drug class, the analyses were restricted to individuals with no prescriptions within the last year prior to diagnosis. Drug classes fulfilling all of the following three criteria were defined as potential signals: Criteria 1: The HR for overall death was >1.5 or <0.67 or the confidence interval did not overlap 0.83-1.2. Criteria 2: The HR for cancer-specific death was as least the magnitude as the HR for overall death. Criteria 3: The IP-weighted KM five-year survival estimates for users and nonusers was statistically significantly different at a 5% significance level.Results: In total, 15,274 one-year survivors of CRC filled prescriptions of 941 unique drugs in 412 unique drug classes during the first year after cancer diagnosis. Out of the 412 drug classes, 94 were eligible for screening. From the evaluation, we identified seven drug classes with a potential effect on the prognosis of one-year survivors of CRC.Conclusions: Applied on localized CRC patients, our proposed screening framework identified seven potential candidates among 94 drug classes for further investigation.

PO-0130 Temporal Trends And Changes In Indication For Opioid Use In Catalonia, 2007-2017: A Population-based Study Of 6 Million Individuals 1400568 Junqing Xie CSM, NDORMS, University of Oxford Temporal Trends And Changes In Indication For Opioid Use In Catalonia, 2007-2017: A Population-based Study Of 6 Million Individuals Drug Utilization Research -> Trends and comparisons Background: North America is currently suffering from an opioid epidemic, and recent data suggesting an increase in opioid use is also seen in northern and central Europe, but little data have been reported for the situation in southern Europe.Objectives: To evaluate secular trends of opioid and their subtypes use in primary care and changes in indication associated with opioid initiation from 2007 to 2017 in Catalonia, Spain.Methods: Data were obtained from the SIDIAP (The System for the Development of Research in Primary Care) database, which contains electronic medical records and dispensation data for ~ 6 million people (>80% of the Catalan population). The incidence proportion of opioid use (codeine, tramadol, fentanyl, and morphine, with the latter three classified as strong opioids) in each calendar year was calculated. The denominators were registered individuals meeting following criteria at the first day of each calendar year: 1) ≥18 years old, 2) ≥1 year of valid data, 3) no registered use of the same opioid in previous one year. The numerators were all new opioid users during each calendar year. Age-sex standardization was performed using the population at the year of 2007 as the reference. Indications were obtained based on the related diagnosis closet prior to therapy initiation from a pre-specified list including musculoskeletal pains, bone fracture/s, post-surgery, and cancer.Results: Between 2007 and 2017, the incidence proportion of total opioid use decreased firstly and then increased, with the lowest rate in 2013 (2.0%) and the highest rate in 2017 (3.3%). Specifically, the use codeine followed a similar trend with total opioid, but tramadol, fentanyl and morphine increased continuously from 0.61% to 1.71%, 0.03% to 0.12%, and 0.01% to 0.05% respectively. In the same period, musculoskeletal pains (back pain, neck/shoulder pain, and osteoarthritis pain) consistently dominated the indications for total opioid, codeine, tramadol treatment, with an attributable 90% of total use over years. But the proportion of strong opioid/s use related to musculoskeletal pain increased from 75% to 80% for fentanyl and 40% to 72% for morphine, which is companied by cancer-related opioid use decreasing from 17% to 8% and 55% to 13% respectively.Conclusions: Use of strong opioids in primary care has increased substantially during the study period (2007-2017) in Catalonia, replacing the use of weaker opioids (codeine). Musculoskeletal pain has gradually become the most common indication for strong opioid prescriptions, which is an unexpected sign of chronic pain management.

PO-0131 Potentially Inappropriate Use Of Non-steroidal Anti-inflammatory Drugs In The General Older German Population 1400569 Thi Ngoc Mai Nguyen R. Karl Universitat Heidelberg Potentially Inappropriate Use Of Non-steroidal Anti-inflammatory Drugs In The General Older German Population Geriatric Pharmacoepidemiology Background: Non-steroidal Anti-inflammatory Drugs (NSAIDs) are listed as potentially inappropriate drugs (PIM) for adults aged 65 years and older in all frequently used PIM lists due to their negative risk-benefit profile in this population. However, these lists strongly differ in listed active substances and circumstances, in which NSAIDs could be indicated and thus should not be considered as PIM. So far, few studies compared different PIM lists, and none has focused on NSAIDs.Objectives: The aims of this study were to compare the differences in the prevalences of NSAIDs - PIM use in a sample of the general older German population according to the recommendations of 5 PIM lists: Beers criteria, PRISCUS list, the EU(7)-PIM list, the STOPP criteria and the FORTA list and to identify the determinants of NSAIDs - PIM use.Methods: We conducted a cross-sectional survey using the 14-year follow-up data from a population-based cohort study from Germany. Participants aged 65 years or older who reported the use of NSAIDs were included (n=287). Drug data were obtained according to the “brown bag method,” which ensures the inclusion of over-the-counter preparations. To detect whether an NSAID is also a PIM, the 2019 Beers criteria, the 2015 STOPP criteria, the EU(7)-PIM list, the PRISCUS list, and the FORTA list were used. Multivariate logistic regression models were used to assess the determinants of NSAIDs-PIM use according to the 5 different PIM lists. Two-sided p-values < 0.05 were considered to be significant.Results: Prevalence of NSAIDs -PIM use varied greatly among the PIM definitions used: STOPP criteria (54.7%), Beers criteria (45.6%), EU(7)-PIM list (29.6%), FORTA list (21.3%) and PRISCUS list (3.5%). 27% of all NSAIDs -PIM users would be only identified by one list, whereas none of them would be identified by all five lists. Determinants of NSAIDs -PIM use also varied by the PIM list used. For example, for the STOPP criteria, cardiovascular disease (CVD), chronic kidney disease (CKD), gout and peptic ulcer disease (PUD) were statistically significantly associated with the use of NSAIDs - PIM among NSAIDs users (Odds ratios [95% confidence intervals]: 4.91 [2.46, 10.24], 12.51 [5.28, 34.11], 2.22 [1.01, 5.04] and 5.46 [2.16, 15.47], respectively).Conclusions: The recommendations of currently available PIM lists for NSAIDs highly vary and there is an urgent need for harmonization in order to provide better guidance to physicians for the prescription of NSAIDs for older adults.

PO-0133 Prescription Drug Use In Testicular Cancer Patients: A Danish Nationwide Registry Study 1444177 Anne Katrine Duun-Henriksen Prescription Drug Use In Testicular Cancer Patients: A Danish Nationwide Registry Study Drug Utilization Research -> Trends and comparisons Background: In developed countries testicular cancer is the most common solid tumour in men between 14 and 44 years of age. Testicular cancer is a highly curable disease with a 5-year survival rate around 95%, which gives rise to a growing population of long-term survivors.Objectives: To describe the use of prescription drugs in testicular cancer survivors compared to a cohort matched on age.Methods: From the Danish nationwide registries, we identified males diagnosed with incident testicular cancer during 2000-2015 and a randomly selected male comparison cohort matched on age at date of diagnosis. Eligible comparisons had to be cancer free and alive on the date of diagnosis (index date). Prescription history was obtained from the Danish National Prescription Registry. The men were followed from time of diagnosis/index date until date of death, date of second (testicular cancer patients)/incident cancer (comparisons) or end of study (31.12.2016) whichever occurred first. Drug use was assessed for 0-5, 5-10 and 10-15 years post diagnosis by estimating prescription rates with 95% confidence intervals (CI) for each time interval.Results: In total, 5,043 patients and 22,197 individuals in a comparison cohort were included in the study with a mean age at time of diagnosis/index date of 38 years. Median follow up from diagnosis was 8.8 years (interquartile range: 4.9-13.6 years). Testicular cancer patients had a higher prescription rate in all three intervals compared to the comparison cohort. Within the first 5 years after diagnosis/index date the rates among cancer patients and comparisons were 5.80 (95% CI: 5.76-5.83) prescriptions/person-year and 4.81 (95% CI: 4.80-4.83) prescriptions/person-year, respectively. Further, 5-10 years after diagnosis/index date, the rate among cancer patients and comparisons were 6.99 (95% CI: 6.95-7.03) prescriptions/person-year and 6.00 (95% CI: 5.99-6.02) prescriptions/person-year. The rates among cancer patients and comparisons 10-15 years after diagnosis/index date were 8.23 (95% CI: 8.16-8.29) prescriptions/person-year and 7.11 (95% CI: 7.09-7.14) prescriptions/person-year, respectively. We will assess and present which drug classes that contributed to this elevated prescription rate in testicular cancer survivors.Conclusions: We confirmed a larger prescription drug burden in testicular cancer survivors during the first 15 years after diagnosis compared to a comparison cohort.

PO-2317 Maternal Thyroid Dysfunction During Pregnancy And The Risk Of Adverse Outcomes In The Offspring: A Systematic Review 1437230 Mengqin Ge The University of Hong Kong Maternal Thyroid Dysfunction During Pregnancy And The Risk Of Adverse Outcomes In The Offspring: A Systematic Review Pediatric Pharmacoepidemiology Background: Maternal thyroid function plays an important role in optimal fetal development. However, thyroid dysfunction is among the most prevalent disorders in pregnant women. Previous studies suggested that maternal thyroid diseases may be associated with adverse neurocognitive outcomes and impair fetal other organ development. Objectives: To review the available evidence for an association between maternal thyroid dysfunction during pregnancy and the risk of various adverse outcomes in offspring Methods: We performed systematic literature searches in PubMed, EMBASE, and Cochrane. Studies were included if they investigated the association between maternal thyroid hormone function during pregnancy and the risk of adverse outcomes in their children. The Proffered Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline was followed. The quality of the design and reporting of included studies was assessed using the Newcastle-Ottawa Scale (NOS). Results: We identified 30 eligible articles: 22 studies of neurodevelopmental disorders and 8 studies of other disorders. All studies were based on a population-based prospective birth cohort or a national register-based screening cohort. The review found an association between maternal hypothyroidism and an increased risk of autism spectrum disorder (HR: 1.31, 95% CI: 1.17 - 1.47, p < 0.00001), epilepsy (HR: 1.22, 95% CI: 1.06 - 1.39, p = 0.004) and schizophrenia (OR: 1.70, 95% CI: 1.13 - 2.55, p = 0.010). Children born from mothers with higher free T4 levels are more likely to be diagnosed with depression (OR: 1.21, 95% CI: 1.00 - 1.14, p = 0.05), but the result did not attain a statistically significant level. The review also suggested a higher risk of other disorders in the offspring including pediatric endocrine morbidity (HR: 1.92, 95% CI: 1.08 - 3.4, p = 0.0263), congenital heart disease (OR: 1.68, 95% CI: 1.02 - 2.78, p = 0.0416), and asthma (HR: 1.37, 95% CI: 1.04 - 1.80, p = 0.0252) born from mothers with abnormal thyroid hormone levels. Overall, the quality of the studies is satisfied (60% “Good”, 10% “Fair” and 30% “Poor”); however, most of the studies did not adjust for pregnancy-related complications and birth outcomes which may have an intermediary role in the association. Conclusions: This systematic review suggests a possible association between maternal thyroid function and various disorders in the offspring. Future studies are needed to investigate the association with consideration of more confounding factors (e.g. thyroxine or anti-thyroid treatment).

PO-2320 The Potential Beneficial Effects Of Low-dose Aspirin For The Primary Prevention Of Gastrointestinal Cancers: A Population-based Cohort Study In Hong Kong 1400570 Jessica Shami University of Hong Kong The Potential Beneficial Effects Of Low-dose Aspirin For The Primary Prevention Of Gastrointestinal Cancers: A Population-based Cohort Study In Hong Kong Drug Effectiveness Background: The effects of low-dose aspirin use in gastrointestinal cancers remain unclear in the general Chinese population.Objectives: To evaluate associations between the use and duration of low-dose aspirin and the risk of colorectal cancer (CRC), gastric cancer (GC), and esophageal cancer (EC) in Hong Kong.Methods: We conducted a retrospective population-based cohort study, using the Clinical Data Analysis and Reporting System (CDARS), an electronic clinical database managed by the Hong Kong Hospital Authority. Patients prescribed either low-dose (75-300mg/daily) aspirin or paracetamol (reference exposure) between January 2004 and December 2008 were matched (without replacement) by propensity score at a 1:1 ratio with follow-up until December 31, 2017. Potential confounders included age, sex, a list of comorbidities and concurrent medications. The primary outcome was an incident diagnosis of gastrointestinal cancer (either CRC, GC, or EC) identified using ICD-9 codes (International classification of diseases, 9th revision). The follow-up period started from the first prescription of either low-dose aspirin or paracetamol (i.e. index date) until the occurrence of outcome, any cancer diagnosis, death, or end of study period (December 31, 2017) whichever came first. We used competing risk Cox regression with death as the competing risk to generate hazard ratios (HRs) with 95% confidence intervals (CIs).Results: We included 166,466 low-dose aspirin or paracetamol users. The mean (standard deviation) follow-up was 8.7 (4.2) years for low-dose aspirin users and 9.0 (4.3) years for paracetamol users. The incidence rates of CRC, GC, and EC were 20.4, 5.4, and 2.2 per 10,000 person-years respectively for low-dose aspirin users and 22.9, 6.8, and 2.9 respectively for paracetamol users. Low-dose aspirin (vs. paracetamol) use was associated with a reduced risk of CRC, GC, and EC (CRC: HR = 0.88 [0.82-0.94]; GC: 0.77 [0.68-0.88]; EC: 0.72 [0.59-0.89]). Low-dose aspirin use for ≥3 years was associated with a reduced risk of CRC, GC, and EC (CRC: HR = 0.74 [0.67-0.82]; GC: 0.43 [0.35-0.52]; EC: 0.43 [0.32-0.59]) compared to <3 years low-dose aspirin use.Conclusions: The use of low-dose aspirin was associated with a significant reduced risk of CRC, GC and EC among the study cohort. The use of low-dose aspirin for more than 3 years was more protective of CRC, GC, and EC compared to less than 3 years low-dose aspirin use.

PO-2323 A Comparison Study Based On Multi-view Latent Dirichlet Allocation 1400571 Lin Zhuo Peking University Third Hospital A Comparison Study Based On Multi-view Latent Dirichlet Allocation Informatics Background: There are several methods for multinomial distribution similarity calculation, which work in different scenarios, and the jury is still out on which is the best.Objectives: To compare the inappropriate prescription detecting performance of topic mapping with three traditional multinomial distribution similarity calculation methods based on multi-view Latent Dirichlet Allocation.Methods: A multi-view extension of the latent Dirichlet allocation topic modeling algorithm was chosen to generate the diagnosis-medication topic models indicating the underlying health status of patients using diagnosis and medication as variables, with data from the Chinese Monitoring Network for Rational Use of Drugs (CMNRUD) database. Each topic consisted of a set of diagnoses, medications that were highly related to each topic, and the percentages for each of these. On the basis of the topics, topic mapping(TM), a newly proposed multinomial distribution similarity calculation method was used to calculate the similarities of distribution of the diagnoses and medications on the topics built in the previous step and find the inappropriate medication use prescriptions by setting a threshold. The same process was undertaken for Kullback-Leibler(KL), Cosine Similarity(COS), and Vector’s Dot Product(DOT) methods to find the anomaly detection thresholds in the scene of multi-view topic models. The prescription manual review result by experts in the Beijing Regional Prescription Review (BRPR) database was used as the gold standard to assess the validity of the model.Results: A total of 44 million prescriptions were used to generate topics using the diagnoses and medications from the CMNRUD database. A 15,000 random samples of the BRPR database were used for validation. Compared with the other three methods, the DOT had the highest specificity of 63.05% and a positive predictive value of 15.05%. However, TM had the highest sensitivity of 81.84% and a negative predictive value of 96.00% and was superior in terms of area under the curve statistically significantly.Conclusions: Overall, topic mapping methods can find more inappropriate medication use prescriptions and perform well in terms of model validity assessments. The multi-view inappropriate medication use prescription detection model based on it can be used as a primary screening tool and will likely complement and improve the manual review.

PO-2324 Monoclonal Antibody Induced Progressive Multifocal Leukoencephalopathy And Cytokine Release Syndrome: Analysis Of Eudravigilance And Faers Databases And Risk Minimisation Measures 1400572 Kalindi Hapani APCER Lifesciences Monoclonal Antibody Induced Progressive Multifocal Leukoencephalopathy And Cytokine Release Syndrome: Analysis Of Eudravigilance And Faers Databases And Risk Minimisation Measures Biologics/Biosimilars Background: Monoclonal antibodies (MABs) are used in different therapeutic areas. There has been increased concern for spontaneous reports of progressive multifocal leukoencephalopathy (PML) and cytokine release syndrome (CRS) attributed to MABs that modulate the immune system.Objectives: 1. To detect the cases of PML & CRS with MABs from FAERS and EudraVigilance safety database and to apply signal detection algorithms to identify significant signals for PML & CRS. 2. To identify the risk minimization measures (RMMs) prevailing in EU & US for PML and CRS with MABs and to discuss the need of awareness in healthcare professionals (HCP) regarding RMMs.Methods: EudraVigilance and FAERS database were searched using each MAB name approved by EMA & USFDA as suspect drug and adverse reaction of PML and CRS. Proportional reporting ratios (PRR) and 95% confidence intervals (CI) were calculated for MABs with ≥2 case reports. European public assessment reports (EPARs) and REMS database were further explored for the type of RMMs implemented in EU and US.Results: 72 and 81 MABs were identified by searching the EPARs & Purple Book respectively. About 70% of PML cases and 34% of CRS cases retrieved from both databases were found with MABs. MABs with highest number of PML cases (n=EVD, FAERS) were natalizumab (n=1223, 1658), rituximab (n=847, 1374) and alemtuzumab (n=50, 86). Similarly, MABs with highest number of CRS cases were blinatumomab (n=195, 186), rituximab (n=107, 114) and obinutuzumab (n=6, 45). The percentages of PML & CRS versus all EVD cases for each drug ranged from 0.034% (adalimumab) to 5.28% (natalizumab) and 0.025% (alemtuzumab) to 13.14% (blinatumomab). Similarly, percentages of PML & CRS versus all FAERS cases for each drug ranged from 0.008% (adalimumab) to 1.77% (rituximab) and 0.12% (nivolumab) to 6.06% (blinatumomab). Both EU & US product labels of natalizumab describes PML with guidance on detection and mitigation of PML. Additional RMMs (aRMMs) and REMS are implemented in EU & US respectively for natalizumab. However, no aRMMs or REMS exist for PML for other 4 MABs. For blinatumomab, one of the goals included in REMS is to inform HCPs about the risk of CRS which may be life-threatening or fatal. However, there are no aRMMs in place for blinatumomab in EU.Conclusions: Our results expand the knowledge of the relationship of MABs with PML and CRS. Calculated statistics and further analysis may reveal a strong signal which requires further study so that patients and HCPs can be communicated regarding these ADRs and its minimisation strategies.

PO-2328 Time to Onset Analysis of Letrozole Associated Osteonecrosis of Jaw in FDA Adverse Event Reporting System (FAERS) Database 1400573 Subeesh Kulangara Viswam Faculty of Pharmacy, M.S. Ramaiah University of Applied Sciences Time to Onset Analysis of Letrozole Associated Osteonecrosis of Jaw in FDA Adverse Event Reporting System (FAERS) Database Pharmacovigilance -> Cancer Background: Our study aimed to refine the potential signal of letrozole associated osteonecrosis of jaw which we identified in a previous study using disproportionality analysis (proportional reporting ratio of 13.1). As the signal identified from spontaneous reporting database known to have confounding factors, it is important to refine and reconfirm the signal using an alternate analytical method.Objectives: To analyze the association between letrozole and osteonecrosis of jaw using time to onset analysisMethods: A systematic data mining was performed in FAERS Database (Q4/2003-Q3/2019) using OpenVigil 2.1, a pharmacovigilance analytical tool. Cumulative incidences of osteonecrosis of jaw was assessed using the Kaplan‐Meier method and Log‐rank test was used to compare the cumulative incidences of osteonecrosis of jaw between patients treated with letrozole and those treated with all the other drugs. Time‐to‐onset profiles were analyzed using the Weibull shape parameter test. A larger value of scale parameter stretches the distribution, whereas a smaller value shrinks the distribution. If the drug and event are not associated, the distribution tend to have a larger value of scale parameter. shape parameter 1, the hazard was considered to have early failure‐type profile, random failure type profile and wear‐out failure‐type profile respectively. Shape and scale parameter of letrozole were also compared with all the other drugs reported for osteonecrosis of jaw.Results: We identified 15,178 reports for osteonecrosis of jaw, of which 448 (3%) were attributed to letrozole. The median time‐to‐onset of osteonecrosis of jaw associated with letrozole was 266(range: 88-850) days. Reports of patients treated with other drugs showed a shorter time‐to‐onset, 122.5 days (range: 28.25-470.5). Letrozole associated osteonecrosis of jaw indicated a wear‐out failure‐type profile and the time to onset data are more widely distributed which is unlikely for an onset time data (shape:1.2, scale:50.2). In contrast, all other drugs associated osteonecrosis of jaw reports indicated an early failure‐type profile (shape: 0.4, scale: 5.5). The cumulative incidence of osteonecrosis of jaw did not significantly differ between patients treated with letrozole vs other drugs (log‐rank test, P < 0.08).Conclusions: Our study suggests that the initial signal generated for letrozole associated osteonecrosis of jaw in FAERS database was a false positive. Time to onset analysis can be used as a measure to identify false signals.

PO-2330 Real - World Data Analysis Of Potential Drug - Drug Interactions In Greek Community Pharmacies 1400574 Georgios Poulentzas Democritus University of Thrace Spotlight Poster Session Spotlight Poster Drug-Drug Interaction Background: Drug-drug interactions may lead to severe clinical outcomes concerning patients’ health, thus posing a serious threat to various fields of Public Health. To date, they have been significantly understudied in Greece.Objectives: The goal of this study was to determine the prevalence of potential drug-drug interactions (pDDIs) and shed light on the factors responsible for the growth of this problem.Methods: For this cross-sectional study, anonymous e-prescription data for the years 2012 -2017 were randomly obtained from 5 community pharmacies in Thessaloniki, the second-largest city in Greece. The sampling process was as follows: for each year patients were chosen by increasing identification number based on the coding of the pharmacies, ensuring randomized selection, and their e-prescription history was found. This process was repeated until 1.000 patients using 2 or more medicines concurrently for 3 or more months were gathered for each year. The study group assumed concurrent use of the medications if they were dispensed within a time period of 30 days. Using the online Drug Interaction Checker (www.drugs.com/drug_interactions.html), pDDIs were categorized depending on their clinical significance into major, moderate, and minor. Prevalence was calculated as “Number of patients with at least one pDDI/Total number of patients” and logistic regression was used in order to identify the factors associated with increased risk of potential drug-drug interaction existence.Results: During the study period, we collected e-prescription data of 8.320 patients, distributed in all 6 years. The overall prevalence of major pDDIs was 12.5% (45.4% for moderate pDDIs). The most common major pDDIs were those between calcium-channel blockers and statins (29.5% of pDDIs) followed by anticoagulant and PPIs (10.7% of pDDIs). The most common potential adverse effect of the assessed pDDIs (8.2%) was major myopathy. After adjustment, statistically significant associations seem to exist between risk of major pDDIs and number of administered drugs (Odds ratio 5.72, 95% CI: 4.87 - 6.72).Conclusions: The prevalence of pDDIs in this study is higher than what has been reported in other European countries, with polypharmacy being the main contributing factor. Therefore, given that pDDI prevalence is considered a quality marker of prescribing practices, our data argue for a need for improvement in that area in Greece.

PO-2334 Concomitant Use Of Benzodiazepines And Opioids And Risk Of Fractures In Older Adults With Chronic Non-cancer Pain In South Korea, 2011-2015: A Nested Case-control Study 1400575 Ye-Jin Kang Spotlight Poster Session Spotlight Poster Geriatrics Background: Although benzodiazepines (BZDs) and opioids are each associated with an increased risk of fractures, the level of the risk associated with the combined use of these drugs is uncertain.Objectives: To investigate the relationship between the combined use of BZDs and opioids with the risk of fractures in older patients with chronic non-cancer pain.Methods: In this nested case-control study, we used the National Health Insurance Service-National Sample Cohort database of Korea from 2002 to 2015. Among osteoarthritis or low back pain patients (>64 years), cases were defined as the incident diagnosis of hip, humerus or forearm fracture between 2011 and 2015. For each case, up to four controls were matched using age (within 5 years), sex, and cohort entry date. We categorized BZDs exposure windows as current, recent and past use and defined current BZDs exposure as use within 30 days before index date, recent BZDs exposure as use within 31-90 days before index date and past BZDs exposure as use within 91-365 days before index date. We categorized each exposure to BZDs with and without opioids. We estimated adjusted odds ratios (aORs) and 95% confidence intervals (CIs) of combined use of BZDs and opioids using conditional logistic regression analysis, adjusting for concomitant diseases and concomitant medications. We also performed subgroup analysis according to age groups, type of fracture, the presence of osteoporosis and an antidepressants prescription.Results: We identified 3,487 cases and 12,350 controls. Among them, 81.8% of cases and 82.0% of controls were female, and mean ages were 75.2 (±6.5) and 74.7 (±6.0) years respectively. The aORs (95% CIs) for current use of BZDs + opioids and BZDs alone were 1.34 (1.15-1.57) and 1.04 (0.91-1.17), each compared to non-use. In subgroup analysis of hip fracture, aORs for current use of BZDs + opioids and only BZDs were 1.55 (1.16-2.06) and 1.04 (0.82-1.33) respectively. The aORs for current use of BZDs + opioids were 1.25 (1.03-1.51) in patients with osteoporosis, and 1.06 (0.89-1.25) in patients without osteoporosis. Finally, the aORs for current and past use of BZDs + opioids in patients had a prescription of antidepressants were each 1.44 (1.21-1.72) and 1.31 (1.05-1.64). There was no association with these outcomes in patients without a record of antidepressants.Conclusions: Concomitant use of BZDs and opioids was associated with an increased risk of fractures in older adults with chronic non-cancer pain.

PO-2339 Equality Analysis On Three Doses Coverage Of Dpt Vaccine Among Children Aged 12 ~ 59 Months In China 1400576 Xiaoli Liu Peking Universtiy Third Hospital Equality Analysis On Three Doses Coverage Of Dpt Vaccine Among Children Aged 12 ~ 59 Months In China Vaccines Background: Diphtheria pertussis tetanus (DPT) vaccine is one of the important vaccines in children's routine immunization. Three full doses of DPT vaccine are the best way to control pertussis, diphtheria and neonatal tetanus. There are relatively few studies on the vaccination coverage of DPT vaccine in China, and research on disparities in coverage is scarce.Objectives: To understand three doses coverage of DPT vaccine and socio-economic inequality distribution among children aged 12~59 months in 20 counties in 10 provinces of China.Methods: In China, using three-stage probability proportion to size sampling, 2,892 children aged 12-59 months with an immunization card were selected from 20 counties across 10 counties in 2016. Some socio-economic characteristics were obtained from face-to-face questionnaire survey and copies of vaccination certificates. We performed two-level logistic regression models to identify the determinants and computed the concentration index to evaluate socioeconomic inequality of three doses coverage of DPT vaccine.Results: In 2016, the overall three doses coverage of DPT vaccine among children aged 12-59 months in the surveyed areas of China was 77.6%. The concentration index value was 0.0376 (P<0.001). In 2016, having a second or higher birth order (adjust OR: 1.38; 95%CI: 1.12-1.70), being born at home (adjust OR: 1.61; 95%CI: 1.04-2.49), having a minority caregiver (adjust OR: 1.36; 95%CI: 1.01-1.84) were significantly related with higher risk of unvaccinated three doses coverage of DPT vaccine.Conclusions: The socioeconomic inequality in three doses coverage of DPT vaccine still existed in 2016. To improve inequality in three doses coverage, the government should pay more attention to the disadvantaged groups, especially the children who were ethnic minorities, had a second or higher birth order and were delivered at home.

PO-2341 Vaccination Coverage And Its Determinants Of Japanese Encephalitis Vaccine Among Children Aged 12-59 Months In Rural Counties Of China In 2016 1400577 Xiaoli Liu Peking Universtiy Third Hospital Vaccination Coverage And Its Determinants Of Japanese Encephalitis Vaccine Among Children Aged 12-59 Months In Rural Counties Of China In 2016 Vaccines Background: Japanese encephalitis (JE) is caused by the JE virus, a flavivirus transmitted by mosquitoes. The majority of JE virus infections are asymptomatic, with about 0.2-0.4% of the infected individuals developing encephalitis. The vaccination of children with JE vaccine is considered to be the most effective means for prevention of JE virus infection. The Chinese Ministry of Health integrated JE vaccine into the National Immunization Program (NIP), providing free vaccinations for children under 7 years old in the whole country. However, there are still some outbreaks of JE in some parts of China, especially in rural areas. This study attempted to estimate the first dose coverage of JE vaccine in 2016 among children aged 12-59 months in rural areas of China and identify the factors hindering high vaccination coverage in those areas.Objectives: To understand the first dose coverage of JE Vaccine and its determinants of vaccination coverage among children aged 12-59 months in rural counties of China.Methods: In 2016, using three-stage probability proportion to size sampling, 2235 children aged 12-59 months with an immunization card, which is the record of children's vaccination and every child should be in accordance with national regulations to establish the immunization card and receive vaccination in China, were selected from 14 rural counties of China in 2016. Some socio-demographic and socio-economic characteristics were acquired from face-to-face questionnaire survey and copies of the immunization cards. We used multivariate logistic regression models to identify the determinants of the first dose coverage of JE Vaccine.Results: In 2016, the first dose coverage of JE Vaccine among children aged 12-59 months in rural areas of China was 88.5%. The adjusted analysis showed that the children who were born at home (adjusted OR: 2.44; 95%CI: 1.12-5.30) or in households with the lowest per capita income (adjusted OR: 2.70; 95%CI: 1.35-5.39) were significantly related with unvaccinated coverage of the first dose of JE Vaccine.Conclusions: The first dose coverage of JE Vaccine was still at a relatively low level in 14 rural counties of China. To improve the coverage of Japanese Encephalitis Vaccine, the government should pay more attention to the disadvantaged groups, especially the children who were delivered at home, or from households with the lowest per capita income.

PO-2342 Variation In Influenza Vaccination Among Residents Of Predominantly White And Black Long-term Care Facilities 1400578 Elliott Bosco Brown University Spotlight Poster Session Spotlight Poster Vaccines Background: Annual assessment and provision of influenza vaccination is both guideline-recommended and legally required for older adults in long-term care facilities (LTCFs). Disparities in vaccination between white and black LTCF residents have been documented, but data are older, not nationally representative, and do not differentiate within versus between LTCF disparities or short-stay versus long-stay residents.Objectives: To compare the vaccination coverage of white and black residents in LTCFs serving predominantly white or black residents, stratified by short-stay and long-stay status. We hypothesized that vaccination coverage would be lower in black-predominant LTCFs, while differences between white and black residents would be largest in white-predominant LTCFs.Methods: We conducted a retrospective cohort study using 2013-2014 Medicare claims linked to Minimum Data Set (MDS) assessments and LTCF data from LTCFocus. Short-stay (<100 days) and long-stay (100+ days) LTCF residents were identified during influenza season (October-March). Residents were stratified by race and residence in a predominantly white or black LTCF (≥70% residents white or black). The outcome was LTCF influenza vaccination coverage, the proportion of residents vaccinated as defined by the MDS.Results: We included 805,321 short-stay (14,252 LTCFs) and 750,404 long-stay residents (14,488 LTCFs). Overall, average influenza vaccination coverage was higher for white short-stay residents [Mean (95% CI): 61.4 (61.0, 61.8)] than black residents [52.6 (51.8, 53.4)], and higher for white long-stay residents [78.0 (77.8, 78.3)] than black residents [72.9 (72.2, 73.5)]. Average short-stay vaccination coverage was higher for white residents [64.1 (63.7, 64.5)] than black residents [55.3 (54.2, 56.4)] in predominantly-white LTCFs, but similar among white [43.6 (39.1, 48.2)] and black residents [43.8 (41.0, 46.6)] in predominantly-black LTCFs. Average long-stay vaccination coverage was higher for white residents [79.8 (79.5, 80.0)] than black residents [74.8 (74.0, 75.7)] in predominantly-white LTCFs, but similar among white [67.9 (64.5, 71.3)] and black residents [66.6 (64.6, 68.7)] in predominantly-black LTCFs.Conclusions: Vaccination coverage for white and black residents was lowest in black-predominant LTCFs. Differences between white and black residents were largest in white-predominant LTCFs. Efforts to improve influenza vaccination in LTCFs should further investigate reasons (e.g., LTCF quality) for differences in vaccination across and within LTCFs. Funding: Sanofi Pasteur.

PO-2343 Women Are Started On A Lower Daily Dose Of Metoprolol Than Men Irrespective Of Dose Recommendations 1400579 Linda Hendriksen Tergooi Hospital Women Are Started On A Lower Daily Dose Of Metoprolol Than Men Irrespective Of Dose Recommendations Drug Utilization Research -> Other Background: In a pharmacokinetic study of metoprolol, women had a two times higher maximal concentration in the blood and a significantly higher area under the curve than men after 100 milligrams (mg) twice daily. However, current guidelines do not distinguish dosage advices for women and men.Objectives: This study was conducted to evaluate whether physicians prescribe the same starting dose of metoprolol for women and men under real-life circumstances. To see if differences occurred after the start of metoprolol, the dosing regime was followed up to 10 prescriptions.Methods: A cohort analysis was performed in the Rotterdam Study (RS) and results were replicated in the Integrated Primary Care Information (IPCI) database - A Dutch GP database. The study population comprised all patients aged ≥18 years starting on metoprolol. The primary outcome was the mean daily dose in mg of the first metoprolol prescription. The secondary outcome was the dose regimen, expressed as the mean daily dose in mg per prescription up to 10 prescriptions. Data were analysed using an independent-samples t-test to compare the mean daily dose of metoprolol for both sexes per prescription. A linear regression analysis on the first prescription was used to adjust for age and in the RS also for indications.Results: In total, 1,263 women and 870 men were included from the RS. The mean daily dose of the first prescription was significantly lower in women than in men; 61.1 mg and 65.9 mg respectively. Women received a 4.8 mg (95% CI -7.8, -1.8) lower dose. For both sexes, dosages increased over time and the difference remained statistically significant for the first three prescriptions. In total 23,074 women and 19,562 men were included from the IPCI database. Also in this database, the mean daily starting dose of metoprolol was significantly lower in women than in men; 57.3 mg and 62.0 mg respectively. The dose of the first prescription was 4.6 mg (95% CI -5.3, -4.0) lower in women. Dosages increased over the 10 prescriptions for both women and men and the difference in dose remained statistically significant. The dose of the first prescription was statistically significantly lower in women after adjustment for age in the RS and the IPCI database and after adjustment for indications in the RS.Conclusions: In real life, women received a lower starting dose of metoprolol than men despite the absence of sex-adjusted dosage recommendations in guidelines. Furthermore, both women and men received lower dosages than recommended. This might mean that prescribers start at a lower dose because they fear that women are more prone to adverse reactions.

PO-2347 Hospitalization For Thromboembolic Events Does Not Improve Adherence To Oral Anticoagulants In High-risk Atrial Fibrillation Patients 1400580 Thanh Phuong Pham Nguyen Perelman School of Medicine at the University of Pennsylvania Hospitalization For Thromboembolic Events Does Not Improve Adherence To Oral Anticoagulants In High-risk Atrial Fibrillation Patients Drug Utilization Research -> Changing drug utilization (interventions and implementation research) Background: Adherence to oral anticoagulants (OACs) in atrial fibrillation (AF) patients is poor in practice. It is not known whether and how adherence changes after hospitalization for a thromboembolic event (i.e., ischemic stroke (IS) or thromboembolism (TE)), a sentinel health event that may represent an opportunity to enhance medication adherence.Objectives: The purpose of this study is to examine the impact of hospitalization for an IS/TE on post-discharge adherence to OACs in high-risk AF patients.Methods: We conducted a quasi-experimental pre-post observational study design using population-based U.S. commercial insurance healthcare claims from the 2009-2016 OptumTM ClinformaticsTM Data Mart. Adult AF patients taking OACs with a random hospitalization for a nonbleeding-related reason occurring after the first observed OAC prescription fill, with no other admissions within the preceding and following six months, were identified. Adherence was estimated by the proportion of days covered (PDC) within six months before and after hospitalization. Difference-in-difference (DID) analysis via a generalized linear model was employed to compare pre- and post-hospitalization PDCs by reasons for hospitalization (i.e., IS/TE versus other nonbleeding-related reasons), adjusting for imbalanced baseline characteristics between groups.Results: We identified 21,400 individuals with AF on OACs who were subsequently hospitalized for a nonbleeding-related reason. Of these, 1,147 (5.4%) were hospitalized for IS/TE, and 20,253 (94.6%) for other reasons. Baseline characteristics were similar between groups. After covariate adjustment, 6-month adherence declined by 1.1% less in individuals hospitalized for IS/TE, compared to other nonbleeding reasons, even though this difference was not statistically significant (p=0.17). Individuals with minority race/ethnicity had a larger reduction by 0.7% in PDC as compared to white patients after discharge (p=0.006).Conclusions: This real-world study suggests that more effective collaborative strategies across care settings are needed to improve adherence to OACs, particularly after a thromboembolic event and during transition of care.

PO-2352 Elevated Risk Of Hospitalization For Infection In Anemic Patients In Japan At The Beginning Of Impaired Renal Function 1400581 Tomomi Kimura Astellas Elevated Risk Of Hospitalization For Infection In Anemic Patients In Japan At The Beginning Of Impaired Renal Function Disease Epidemiology/Clinical Course -> Other Background: Anemia in chronic kidney disease (CKD) is known to be associated with an increased risk of renal and/or cardiovascular outcomes, as well as higher mortality. Anemia is also associated with chronic infection, but the association with acute, severe infection at early stage CKD has not yet been elucidated in the Japanese population.Objectives: To descriptively compare the incidence and hazard ratio of the hospitalization for infection between anemic and non-anemic patients with early stages of impaired renal function.Methods: This was a retrospective cohort study using annual health checkup data and insurance claims between 2008 and 2019 for employees and their dependents aged 74 years or younger, provided by JMDC. Subjects aged 18 years or older who had an estimated glomerular filtration rate (eGFR) of ≥60 ml/min/1.73m2 at the previous checkup, and first experienced an eGFR <60, were enrolled and followed until hospitalization for infection or end of data period (including death). Anemia was defined with baseline hemoglobin value, age, and sex according to Japanese guidelines for Renal Anemia in Chronic Kidney Disease, regardless of treatment for anemia. Hospitalization for infection was defined as those with an anti-infective prescription on the admission date and infectious disease-related diagnosis codes on the inpatient claims. Propensity score for anemia (or non-anemia) at baseline was estimated and the stabilized inverse-probability-weighted Kaplan-Meier estimates were calculated. Hazard ratio was estimated using the Cox proportional hazard model adjusted for baseline age, sex, eGFR, proteinuria, Charlson comorbidity index score, HbA1c, systemic blood pressure, and current smoking status.Results: In total, 32,870 subjects were enrolled and followed-up in an average of 4.1 years. At baseline, 1,396 (4.2%) of the subjects were anemic. The crude incidence rates per 1000 person-years (95% confidence interval [CI]) were 7.64 (5.55-10.26) in anemic group vs 2.30 (2.05-2.58) in the non-anemic group. Unweighted and weighted 5-year Kaplan-Meier survival estimates (95% CI) of time to first hospitalization for infection in the anemic and non-anemic groups were 96.5 (95.2-97.5) vs 98.9 (98.8-99.1) and 97.6 (95.7-98.7) vs. 98.9 (98.8-99.0), respectively. The adjusted hazard ratio (95% CI) was 2.349 (1.672-3.301).Conclusions: Anemia in very early stage CKD was associated with elevated risk of hospitalization for infection, independent of the level of renal impairment.

PO-2354 Over - The - Counter Drugs: A Pharmacoepidemiological Study In Greece 1400582 Theodora Tatsiou Aristotle University of Thessaloniki Georgios Poulentzas Democritus University of Thrace Over - The - Counter Drugs: A Pharmacoepidemiological Study In Greece Drug Utilization Research -> Trends and comparisons Background: Some medicines are available over the counter (OTC) in Greece without a prescription from physicians. There is a limited understanding of patients’ knowledge level regarding OTC medicines, how they get informed about their proper use and whether there are any associated potential health issues.Objectives: The goal of this study is to determine the knowledge level, habits, sources of information and selection criteria of Greek residents regarding OTC medicines, as well as their opinion about the dispensing of these medicines in non-pharmacy settings.Methods: During 5 months of 2017, 920 people aged over 17, who were approached in community pharmacies and gathering points, were interviewed anonymously. 782 of those individuals participated (85% response rate) by answering a predesigned, cross-sectional questionnaire consisting of 12 multiple choice questions. Statistical analysis was performed with IBM SPSS Statistics v.15.0 program using descriptive statistics and chi-square while crude odds ratios were calculated with logistic regression.Results: 507 respondents (64.8%) exhibited good knowledge about which medicines are categorized as OTC and which are not. Healthcare professionals (such as pharmacists and physicians) are the main source of information about the proper use, dosage and adverse effects of OTC medications for 648 (84%) of the respondents. When asked about the selection criteria regarding the purchase of OTC medication, physicians play the most critical role for 510 respondents (65.6%), followed by pharmacists for 468 (60.2%) and friends and family for 173 (22.3%). Product price is a more important factor for the selection of OTC medicines when it comes to elderly people (OR 1.73, 95% CI 1.10-2.72) Furthermore, 61.2% of the respondents stated that they are reluctant to buy OTC medicines from a non-pharmacy setting, while 69.6% believes that retail of OTC medicines in super markets will increase their abuse and misuse. However, younger people seem to be more positive about purchasing them elsewhere than pharmacies than older people (OR 2.20, 95% CI 1.37-3.54).Conclusions: Healthcare professionals play a crucial role in people’s information about OTC use and they have the biggest impact in the final selection. Finally, Greek society presents reluctance to accept OTC disposal elsewhere than community pharmacies.

PO-2355 Effect Of Tenofovir Containing First-line Art In Patients With Moderate/severe Reduced Creatinine Clearance At Baseline: A Retrospective Study At Two Referral Hospitals In Namibia. 1400583 Francis Kalemeera Spotlight Poster Session Spotlight Poster BRACE Background: Caution is advised against the use of tenofovir disoproxil fumarate (TDF) in patients with baseline creatinine clearances (CrCl) 3ml/min/year, were observed in 40.1% (n=1,146) and 60.4% (n=1728), respectively. Decline occurred in 2.3% (n=65). Improvement was associated with severely low CrCl (adjusted hazard ratio [aHR]=13.4 [95%CI: 6.7 - 26.9, pp=0.033) and a longer duration of ART (aHR=1.1 [95%CI: 1.05 - 1.2, ppp<0.001). The cumulative hazard rate was higher in gp I.Conclusions: Improvement was more likely than decline in gp I pts. Significantly more decline events were observed amongst pts in gp II. Greater proportions migrated to a lower CrCl stage in gp II. Although, gp I pts were more likely to experience new onset severe reduced CrCl than gp II pts, the absolute proportions were low. TDF should not be withheld from HIV-positive pts with a baseline CrCl <60ml/min.

PO-2362 The Application Of Mixed Methods Research In Pharmacoepedemiology: A Synthesis Review 1400584 hongling chu Peking University Third Hospital The Application Of Mixed Methods Research In Pharmacoepedemiology: A Synthesis Review Methods in Pharmacoepidemiology -> Study Design - Studies assessing study design (e.g., choice of comparator, self-controlled methods, avoiding immortal time bias) Background: Mixed methods research (MMR) combined with research on the basis of quantitative and qualitative data, can answer research questions comprehensively and deeply by integrating qualitative and quantitative study. Specifically, qualitative research can be used to generate research hypotheses before analyzing data, answer the question like “what, how, why”, further explain the research results; or used to test the completeness of the data.Objectives: In this synthesis review we aimed to answer 4 key questions: What kinds of questions in the field of pharmacoepidemiology can be solved by using MMR; how the MMR solve the problems of pharmacoepidemiologic studies; What are the observed benefits of using MMR; What are the limitations and barriers of using MMR.Methods: We searched MEDLINE, EMBASE, Cochrane, SAGE, Web of Science, Google Scholar, CNKI, WANGFANG TADA. Included studies were published in English or Chinese, of any using MMR in the field of pharmacoepidemiologic. Data were analyzed using framework analysis and meta-synthesis approach.Results: We included 56 studies using MMR in the field of pharmacoepidemiologic. In general, MMR by integrating qualitative and quantitative data can be used to: comprehensively and deeply analyze the factors affecting the description of a medication; evaluate the feasibility of translating the finds from study to real world practice; deeply explain the quantitative result from the big data; establish or select the meaningful outcomes; and solve problem related to medication management. The mainly limitation is mainly about shortage of an optimal design before implementing the MMR.Conclusions: Mixed methods research can be used in the field of pharmacoepidemiology to overcome some disadvantages generated by only using quantitative study, to further improve the quality of research.

PO-2366 Linezolid Induced Skin Reactions In A Multidrug-resistant Infective Endocarditis Patient: The First Case With Systematic Review 1400586 Jose Kochuparambil Mary Queens Mission Hospital, Kanjirappally, India Linezolid Induced Skin Reactions In A Multidrug-resistant Infective Endocarditis Patient: The First Case With Systematic Review Pharmacovigilance -> Other Background: There are no cutaneous adverse reactions (CARs) reports available in the literature on linezolid (LNZ) induced reactions in multidrug-resistant (MDR) infective endocarditis (IE) patient. So, there is a lack of information on the management of such incidents. Here we report the first case of CARs in an MDR IE patient by LNZ administration with a systematic review.Objectives: To report a case on linezolid induced cutaneous adverse reactions (CARs) and to critically evaluate the literature.Methods: Descriptive studies addressing LNZ induced CARs in MDR patient were considered for review retrieved from PubMed database using keywords based search. Study selection, data abstraction and quality assessment were performed.Results: Case report: A 24-year-old female patient admitted to the hospital with complaints of fever (101o F)associated with chills, rigors, and shortness of breath for 4 days. She was an MDR IE patient with a history of rheumatic heart disease, currently, on LNZ 600mg IV BD against Staphylococcus coagulase-negative as that was the only sensitive option. The patient experienced flares of cutaneous reactions with multiple hyper pigmented maculopapular lesions all over the body after one week of LNZ therapy. LNZ dose was tapered gradually and discontinued by ensuring that the patient receives appropriate corticosteroids. Her reactions were completely subsided and an improvement in infection was observed after 1 month of therapy. Systematic Review: A total of 4 reports from the PubMed database were included in the final review with 2 male and 2 female patients. LNZ was immediately stopped in all the cases due to spontaneous reporting of CARs and the reaction was progressively resolved.Conclusions: Healthcare professional should be aware of the CARs especially on MDR patients as they are exposed to multiple drugs. More spontaneous monitoring and reporting required for better quantification.

PO-2367 Disposal Of Leftover Medicines: Prevalence And Practices Among The Rural Population Of Kerala 1400587 Jose Kochuparambil Mary Queens Mission Hospital, Kanjirappally, India Disposal Of Leftover Medicines: Prevalence And Practices Among The Rural Population Of Kerala Drug Utilization Research -> Quality use of medicines (including indicators and guidelines) Background: Leftover medicines in the household is an evolving problem. There are no regulations from the government to safely dispose medicines kept as left over home. Along with this there is a greater number of people accessing medical care and the prescription of medicines also increased resulting in an increased per person consumption of medicines.Objectives: To determine prevalence of households with leftover medicines and its disposal practices among the rural population of Kerala, India.Methods: A cross-sectional study was conducted among the rural population of three districts of Kerala. The pre-tested data collection tool used for collecting information on medicine usage and medicine wastage. The houses numbered with multiples of 5 were selected for the study resulting in a total household of 181.Results: Among the 181 houses surveyed, 93 (51%) households were reported to have left over medicines kept at home. The major reason for medicine left over was found to be symptom relief [42] (45%) and change of prescriber [18] (20%). Among the total families surveyed, 121 (67%) of the families responded that they are disposing their solid medicines along with household waste and in the case of liquid preparations, 98 (54%) were disposing their medicines along with household waste. Some of the responders disposed the medicines as fertilizers for plants.Conclusions: There is no established system for safe medicine disposal in the community. Even though there are some good practices for collecting the unused medicines by non-profit organisations to re-use it among the needy, they are not standardized. The majority of respondents would support the implementation of a proper system for re-collecting the medicines. Hence, a regulated drug take-back program may result in a reduction in the household medicine wastage.

PO-2370 Doxycycline Adherence Scale: The Impact On The Incidence Of Leptospirosis Among Flood Affected South Indian Population. 1400588 Jose Kochuparambil Mary Queens Mission Hospital, Kanjirappally, India Doxycycline Adherence Scale: The Impact On The Incidence Of Leptospirosis Among Flood Affected South Indian Population. Disease Epidemiology/Clinical Course -> Other Background: Leptospirosis is a zoonotic disease caused by Leptospira after flood or heavy rain. Doxycycline is the most commonly used prophylactic antibiotic for the prevention of leptospirosis. Adherence to doxycycline is crucial in case of moderate to severe exposure to contaminated water. Non-adherence and unawareness towards chemoprophylaxis after the flooding in the Indian state of Kerala on the year 2018 contributed towards an epidemic. This incident pointed towards the need of developing and implementing a scale to assess adherence towards doxycycline.Objectives: To develop and implement a scale to measure adherence to the administration of doxycycline.Methods: A five-item doxycycline adherence scale was prepared based on the leptospirosis policy statement of the health ministry and insights from the past incidence of leptospirosis. This prospective, interventional study included 200 flood victims aged 18 years and older, displaced from the home and staying in the rescue homes. They were interviewed for basic demographic details and the doxycycline adherence scale was given to them for the next 5 weeks on a weekly basis by 5 different interviewers. Half of the study sample was provided with counseling on rational use of doxycycline. The impact of counseling and adherence to therapy in reducing the incidence of leptospirosis was analyzed by comparing the response from infected to the non-infected responders with chi-square method with p value < 0.05 to be significant.Results: 161 (80.5%) people were adherent to the doxycycline chemoprophylaxis and 17 cases of leptospirosis (9.5%) were reported among the cohort by the health workers. The number of cases of leptospirosis reported among the adherents vs non-adherents (3% vs 31%, p < 0.001) and among the population that received counseling vs no counseling (2% vs 15%, p < 0.001) showed statistically significant results.Conclusions: Proper adherence and providing counseling regarding the rational use of medicines and health habits can decrease the incidence of leptospirosis. Implementation of doxycycline adherence scale can help to monitor the patients and provide corrective intervention regarding non-adherence.

PO-2371 Venous Thromboembolism And Risk Of Cancer In Users Of Low-dose Aspirin: A Danish Population-based Cohort Study 1400589 Frederikke Troelsen Department of Clinical Epidemiology, Aarhus University Hospital Venous Thromboembolism And Risk Of Cancer In Users Of Low-dose Aspirin: A Danish Population-based Cohort Study Disease Epidemiology/Clinical Course -> Cancer Background: Long-term use of low-dose aspirin has been associated with a reduced risk of several cancers. Venous thromboembolism (VTE) is a marker of increased cancer risk; however, it remains unknown whether VTE is a marker of occult cancer in users of low-dose aspirin.Objectives: To investigate the risk of cancer subsequent to VTE among users of low-dose aspirin.Methods: We performed a population-based cohort study in Denmark, linking prospectively collected, individual-level data from existing Danish health registries during 2006-2017. We identified all patients with a first-time hospital-based diagnosis of VTE who also redeemed a prescription for low-dose aspirin within 90 days prior to the date of their first-time VTE. We defined new use of aspirin as first-time prescription within one year prior to the VTE diagnosis and current use as first-time prescription more than one year prior to the VTE diagnosis. We computed absolute risks of cancer, treating death as a competing risk, and age-, sex-, and calendar-period standardized incidence ratios (SIRs) by comparing the observed cancer incidence among aspirin users with the expected risk of cancer based on national cancer incidence during the study period.Results: We followed 8,891 users of aspirin with VTE for a median of 2,5 years (interquartile range (IQR): 0.6-5.3 years). Current users comprised 87% of the study population. The median age at VTE diagnosis was 77 years (IQR: 68-84 years) for current and 68 years (IQR: 58-78) for new users, respectively. We observed 1,159 cancers among current users of aspirin. The one-year absolute risk of cancer was 7.0% (95% confidence interval [CI]: 6.5-7.6) with a corresponding SIR of 3.0 (95% CI: 2.8-3.3). The SIR decreased to 1.1 (95% CI: 1.0-1.2) during the subsequent years of follow-up. We observed 182 cancers among new aspirin users yielding an absolute risk of 6.1% (95% CI: 4.8-7.6) during the first year of follow-up. The SIRs were 3.4 (95% CI: 2.6-4.3) and 1.2 (95% CI: 1.0-1.5) after one and more than one year of follow-up, respectively.Conclusions: VTE might be a harbinger of cancer even in users of low-dose aspirin regardless of treatment duration.

PO-2376 Antidepressant Dispensing Among Women With Breast Cancer, United States, 2001-2018 1400591 Jihye Park UNC Gillings School of Global Public Health Antidepressant Dispensing Among Women With Breast Cancer, United States, 2001-2018 Disease Epidemiology/Clinical Course -> Cancer Background: Breast cancer is the most common cancer among women in the US. In women with early breast cancer, prevalence of clinically significant depression or anxiety symptoms is twice as many than seen in the general female population, and antidepressant (AD) use is common in this population.Objectives: To provide an updated summary of patterns of AD dispensing among women with breast cancer diagnosis using a sample of US commercial health insurance claims data, 2001-2018.Methods: Using the MarketScan claims databases from 2001 to 2018, we identified women aged 18-64 with at least one breast cancer diagnosis code. Women were eligible if they had at least one-year of continuous enrollment pre- and post-breast cancer diagnosis date (index date). A woman was classified as an AD user if she filled at least one prescription at any point within one year of the index date. Descriptive statistics were used to summarize the distribution and trend of AD dispensing. The prevalence of psychiatric diagnosis was determined using the diagnosis codes of depression, anxiety disorder, and adjustment disorder.Results: Overall, 176,867 women with a diagnosis of breast cancer with a minimum one-year post-diagnosis were identified (median=2.9, IQR=1.8-3.0 years). During the first year after breast cancer diagnosis, 58,205 women (32.9%) had at least one prescription fill for an AD (median time to first fill=45 [IQR=17-121] days). While there was some fluctuation, an overall decrease from 33.8% to 31.1% was observed in the proportion of one-year post-diagnosis AD users between 2001 and 2009, whereas there was a steady increase between 2010 and 2018 from 32.0% to 33.4%. Of all ADs, selective serotonin reuptake inhibitors (SSRIs) were the most common class that was first filled after diagnosis date (49.1%), followed by serotonin and norepinephrine reuptake inhibitors (SNRIs; 26.6%), and tricyclic antidepressants (6.9%). When assessing the trend of first fill after diagnosis by drug class among AD users, there was an observed decline in SSRI prescription fills from 51.7% to 46.3%, whereas an observed increase in SNRI prescription fills from 18.5% to 37.0%. Among all women in the cohort, 47,021 (26.6%) had at least one psychiatric diagnosis at any point prior to and during the first year after index date.Conclusions: In a population of commercially-insured women with breast cancer diagnosis, nearly 33% had at least one prescription fill for an AD during one-year post-diagnosis. Understanding patterns of AD discontinuation and co-prescription with cancer therapy can inform pharmacologic management of depression and anxiety disorders in women with breast cancer.

PO-2377 Feasibility Of Integrating A Web-based Tool To Monitor Adherence To Respiratory Medications Using Pharmacy Claims Data In Primary Care 1400592 Alia Yousif Faculty of Pharmacy, Universit� de Montr�al Feasibility Of Integrating A Web-based Tool To Monitor Adherence To Respiratory Medications Using Pharmacy Claims Data In Primary Care Informatics Background: Medication nonadherence in patients with asthma or chronic obstructive pulmonary disease (COPD) is associated with poor treatment outcomes. As a large majority of these patients are treated in primary care, family physicians need to assess adherence accurately to propose effective interventions, which is a challenging task. As a potential solution, we previously developed e-MEDRESP, a novel web-based tool that allows physicians to monitor adherence to respiratory medications using pharmacy claims data.Objectives: 1) To test the feasibility of implementing e-MEDRESP in family physicians’ electronic medical records (EMR); 2) To evaluate its use in clinical practice; 3) To assess patients’ and physicians’ satisfaction with the tool.Methods: A 12-month feasibility study was initiated in July 2019 in 13 family medicine clinics in Quebec (Canada). e-MEDRESP is available for 16 family physicians and 296 of their patients with asthma or COPD were recruited. Eligible patients had a physician-diagnosed asthma or COPD and must have been prescribed a respiratory controller medication in the prior year. Counters were embedded in the tool to track physician use of e-MEDRESP. Patient and physician satisfaction with the tool were collected, respectively, via phone interviews and online questionnaires. Data were analyzed using descriptive statistics.Results: To date, 137 patients had at least one medical visit (all-cause) with one of the participating physicians. e-MEDRESP was consulted by 12 physicians for 66 (48%) of these patients. Fifty-seven patients underwent a phone interview; 91% reported discussing their medication use with their physician; 32% confirmed seeing their e-MEDRESP report on the physician’s computer and indicated that it was easy to interpret. Using a Likert scale (5=Completely agree; 1=Completely disagree), physicians reported that the tool helped to better evaluate their patients’ medication use, with a mean rating of 4.8 ± 0.7. They also felt that the tool facilitated communication on medication adherence and helped them adjust the prescribed therapy, with mean ratings of 4.8 ± 0.5 and 4.3 ± 0.9, respectively. All physicians reported that they intend to continue to use e-MEDRESP and appreciated its seamless integration in the EMR.Conclusions: Preliminary data suggest that the integration of e-MEDRESP within physician workflow is feasible. Complete results are expected in July 2020. The capacity of e-MEDRESP to enhance adherence and the proportion of patients whose prescription changed following a physician consultation of the tool will be assessed.

PO-2378 Deriving The European Leukemia Net ( E L N ) Risk Categorisationin Acute Myeloid Leukemia ( A M L ) Patients In A Real-world Data Source 1400593 Tao Xu F. Hoffmann-La Roche Ltd. Deriving The European Leukemia Net ( E L N ) Risk Categorisationin Acute Myeloid Leukemia ( A M L ) Patients In A Real-world Data Source Disease Epidemiology/Clinical Course -> Cancer Background: ELN risk categorization is key for diagnosis and management of AML. It is based on cytogenetics in combination with specific recurrent molecular mutations. However, the assignment of patients to ELN risk groups using real-world data (RWD) can be limited by data availability and timing of test results.Objectives: 1) To assess availability of ELN data elements captured in routine clinical practice in an electronic health record (EHR) database, 2) to derive a modified ELN stratification in AML patients based on minimum available variables and 3) describe overall survival (OS) by assigned risk group.Methods: This retrospective cohort study included patients from the Flatiron Health EHR-derived de-identified database, diagnosed with AML between 2014 and 2019 in the Flatiron Health network which is predominantly community oncology clinics. The database contains demographic and clinical data curated via technology-enabled abstraction. Cytogenetic and molecular testing results were abstracted from unstructured data. We implemented an algorithm based on the 2017 ELN definition, while accounting for data completeness. Median OS (mOS) from the time of 1L treatment (tx) was evaluated across risk groups.Results: Of 815 AML patients, 700 patients had 1st line (1L) tx information available. Of these, 589 had ELN markers tested pre-1L tx. Cytogenetic markers such as t(8;21), were tested in 552 (78.9%) of the patients pre-1L tx. Molecular markers, such as RUNX1 or TP53, were more commonly tested in recent years e.g. TP53 was tested in 7.2% of the patients in 2014 and 38.6% in 2019. Overall, the test prevalence was <50% for these markers. Cytogenetic tests still dominated the genetic profiling techniques used (~ 80%), however, increased use of next generation sequencing (NGS) was observed: 7.2% in 2014, 50% in 2019.The algorithm identified AML patients with favorable N=94, mOS 35.6 months (95% CI 26.9, NA), intermediate N=225, mOS 12.5 (95% CI 10.9, 18.4) and adverse N=179, mOS 9.1 (95% CI 7.9, 10.5) risk. Ninety-one patients had insufficient markers for risk categorization and were assigned to “unknown” with mOS 13.9 months (95% CI 12.2, NA), and had slightly better mOS than the patients with no ELN markers tested N=111, mOS 10.2 months (95% CI 5.8, 18.2). Results were robust when restricted to patients receiving intensive tx.Conclusions: Increased use of NGS testing was observed in patients with AML. EHR-derived data can be used to define risk groups based on the 2017 ELN criteria. We demonstrated the stratification of patients and outcomes with the derived ELN algorithm.

PO-2387 Fertility Treatment And Oral Contraceptive Discontinuation For Identification Of Pregnancy Planning In Routinely Collected Health Data - An Application To Analgesic And Antibiotic Utilisation 1400595 Sarah Hjorth Fertility Treatment And Oral Contraceptive Discontinuation For Identification Of Pregnancy Planning In Routinely Collected Health Data - An Application To Analgesic And Antibiotic Utilisation Pregnancy and Lactation Background: Women with unplanned pregnancies use folic acid less frequently, and more often smoke and use potentially teratogenic medications in the first trimester. Yet most perinatal pharmacoepidemiological studies based on routinely collected data lack information on pregnancy planning. Further, for pregnancies to appear in routinely collected data, they need to proceed beyond a certain point, for example week 12, or end in live birth. If pregnancy intention could be identified, it would be possible to ascertain pregnancies prospectively from an earlier time.Objectives: To evaluate fertility treatment or discontinuation of oral contraceptives (OC) as proxies for pregnancy planning in routinely collected data. We hypothesised that (1) women on fertility treatment, and women who discontinue OC before pregnancy, would be less likely to use prescription analgesics in the three months before pregnancy and in the first trimester than women who discontinue OC during pregnancy, and (2) there would be no difference in use of prescription antibiotics.Methods: This registry-based cohort study used data on pregnancies in the Medical Birth Registry of Norway from 2006 to 2017 linked to data from the Norwegian Prescription Database. We included women with either filled prescriptions for clomiphene in the year before pregnancy, or OC dispensed that covered the date one year before pregnancy start. Participant characteristics and proportion of pregnancies with filled prescriptions for antibiotics and analgesics were displayed for women on fertility treatment, women who discontinued OC before pregnancy, and women who discontinued during pregnancy.Results: Of 175 628 included pregnancies, fertility treatment was identified in 20 580, and OC discontinuation in 155 048. As hypothesised, women who discontinued OC during pregnancy were slightly more likely to fill prescriptions for analgesics before pregnancy (10.5%) than women who discontinued before pregnancy (9.4%); however, they also filled more prescriptions for antibiotics (13.7% compared to 11.9% in women who discontinued OC before pregnancy). In the first trimester, patterns were similar. Contrary to our hypothesis, women on fertility treatment had the highest proportion of analgesic prescription fills before pregnancy (11.5%).Conclusions: Fertility treatment and OC discontinuation from routinely collected health data may be markers of pregnancy planning, but cannot stand alone in identifying unplanned pregnancies.

PO-2392 Uncontrolled Blood Pressure And Therapeutic Inertia In Treated Hypertensive Patients: A Descriptive Cohort Study In Clinical Practice Research Datalink (cprd) 1400596 Celine DARRICARRERE Servier Uncontrolled Blood Pressure And Therapeutic Inertia In Treated Hypertensive Patients: A Descriptive Cohort Study In Clinical Practice Research Datalink (cprd) Disease Epidemiology/Clinical Course -> Cardiovascular Background: Hypertension is a major and fast-growing cause of cardiovascular (CV) diseases and deaths. Treatment and control of this condition are crucial to reduce CV events and related burden. One of the identified factors contributing to uncontrolled hypertension is therapeutic inertia, which is the failure of health-care provider to modify therapy when blood pressure (BP) targets are not reached.Objectives: To quantify uncontrolled BP over time and therapeutic inertia in UK patients treated for hypertension in Primary Care.Methods: A retrospective cohort of patients diagnosed with primary hypertension and receiving antihypertensive drugs from January 2015 to June 2017, was created from the CPRD (GOLD). Patients with secondary hypertension were excluded. Index date (ID) is the first prescription of hypertensive drug during study period. Patients were followed for 18 months. Primary outcomes were proportions of patients with uncontrolled BP (one BP record, Systolic ≥ 140 mmHg or Diastolic ≥ 90 mmHg) at ID and over time; time to first BP control (one BP record Systolic < 140 mmHg and Diastolic < 90 mmHg) in patients uncontrolled at ID and proportion of therapeutic inertia defined as 2 consecutive uncontrolled BP records with no treatment change.Results: 581 260 treated hypertensive patients (mean age 67.8 years, 51.3% women) were included, among whom 37.2% (N=216 014) had an uncontrolled BP record at ID, 32.6% had a controlled BP record (N= 189 291) and 30.3% (N=175 955) did not have any BP record. At ID, 516 231 patients (88.8%) were prevalent users of HT drugs and mean duration since first antihypertensive medication prescribed was 8.3 years. The mean number of BP records per patients during follow-up was 2.1 at 1 year, with 1.1 and 1.0 for uncontrolled and controlled values respectively. Over a median follow-up of 18 months, 59.2% of the patients had at least one uncontrolled BP record and 12.8% of patients did not have any BP record. Focusing on uncontrolled population at ID (N=216 014), 72.9% had at least one uncontrolled BP record during the follow-up and 28.3% had only uncontrolled BP records. Therapeutic inertia occurred at least once for 55.7% of the uncontrolled population. However BP control was reached for 61.5% of the uncontrolled population at ID, within 185 days in median.Conclusions: This cohort study showed that recording of BP measures in hypertensive patients’ medical record is poor. Based on this low number of measures, uncontrolled BP and therapeutic inertia remain common.

PO-2395 Assessment Of Clinical Outcome Of Twice Daily Dosing Schedule Of Deferasirox In Transfusion Dependent Paediatric Beta Thalassemia Patients A Randomized Controlled Study 1400597 Juny Sebastian JSS College of Pharmacy Assessment Of Clinical Outcome Of Twice Daily Dosing Schedule Of Deferasirox In Transfusion Dependent Paediatric Beta Thalassemia Patients A Randomized Controlled Study Drug Effectiveness Background: Deferasirox has proven good efficacy and acceptable safety for the management of thalassemia patients. However, some patients are unresponsive or intolerant to once daily administration of deferasirox even at a high dose.Objectives: To evaluate the efficacy and tolerability of twice daily dosing of deferasirox among transfusion dependent paediatric beta thalassemia patients.Methods: This prospective randomized single blinded parallel study was conducted at the Pediatric Hematology clinic of a tertiary care hospital over a period of one year. Study enrolled transfusion dependent paediatric beta thalassemia patients prescribed with deferasirox who visited the study site for their regular blood transfusions and follow up. All the eligible subjects were randomized by block randomization method to intervention and control group. Once daily dosing of deferasirox was changed to twice daily dosing with the same total daily dose for the intervention group.Whereas, in the control group, the patients were continued with the once daily deferasirox dosing. Serum ferritin levels were tested at baseline and after 6 months of enrollment to assess the clinical outcome. Statistical analysis was performed using independent t test with the help of SPSS software verson 22.Results: Forty one patients were randomized into intervention (n=20) and control group(n=21). A statistically significant mean decrease in serum ferritin levels was detected in the intervention group (P value 0.011), while, the serum ferritin levels of the control group was significantly increased from baseline (P 0.000). The number of drug related problems was also significantly reduced in intervention group (P 0.031) when compared to control group ( P 0.062).Conclusions: This study concludes that twice daily dosing of deferasirox with the same total daily dose significantly enhances the iron chelation efficacy and tolerability among transfusion dependent paediatric beta thalassemia patients when compared to once daily regimen.

PO-2396 Pre-existing Autoimmune Disease And Immune-related Adverse Events (irae) With Checkpoint Inhibitors In Metastatic Melanoma 1400598 Marina Machado Research Institute of the McGill University Health Centre Pre-existing Autoimmune Disease And Immune-related Adverse Events (irae) With Checkpoint Inhibitors In Metastatic Melanoma Safety End Points -> Cancer Background: Melanoma patients with pre-existing autoimmune disease were excluded from clinical trials testing checkpoint inhibitors because of concerns of irAE.Objectives: We used real-world data to evaluate irAE in metastatic melanoma patients with and without pre-existing autoimmune diseases using checkpoint inhibitors.Methods: We assembled a cohort of adults with metastatic melanoma using US MarketScan® Databases. All patients were assigned a time-zero as the date of first exposure to ipilimumab (IPI), pembrolizumab (PEM), nivolumab (NIV) or NIV/IPI over Jan 2012-July 2017. We required health/drug plan coverage for 12 months before time-zero and identified in this period all autoimmune disease diagnostic codes from physician claims/hospitalizations. Outcomes (incident irAE) following time-zero were identified from physician claims/hospitalizations primary diagnostic codes. Patients were censored due to in-hospital death, loss of health/drug plan, end of study period (Dec 2017), gap between doses ≥ 60 days or therapy switch. We estimated irAEs rates, stratified by pre-existing autoimmune disease and sex, and hazard ratios (HR) with 95% confidence intervals (CI), using Cox model, adjusted for age, sex, calendar year, comorbidities, past health care use, past and current cancer therapy, and pre-existing autoimmune disease.Results: We studied 2315 patients initiating IPI (62.2%), PEM (16.8%), NIV/IPI (11.8%), and NIV (9.3%). The median age was 60 years (interquartile range 52-67), 62.1% were male, and 27.7% had pre-existing autoimmune disease including hypothyroidism (12.5% in men, 22.8% in women), interstitial lung disease (2.7% in men, 2.3% in women), and myositis (2.3% in men, 4.9% in women). Most frequent irAEs were hypothyroidism (new cases) and inflammatory bowel disease. The rate (per 100 person-years) of irAEs in patients with pre-existing autoimmune diseases was 7.3 (95% CI 4.2-12.6) versus 4.1 (95% CI 2.5-6.6) in those without autoimmune disease. The rate was 5.4 (95% CI 3.5-8.3) in men and 4.4 (95% CI 2.4-8.3) in women. In multivariate analyses, presence of pre-existing autoimmunity was associated with higher irAE risk vs absence (HR 2.17, 95% CI 1.01-4.66); irAE risk was lower with IPI vs NIV/IPI (HR 0.19 95% CI 0.04-0.86); we did not to see difference in women vs men (HR 0.77, 95% CI 0.35-1.70).Conclusions: Over a quarter of metastatic melanoma patients starting checkpoint inhibitors had pre-existing autoimmune disease. We observed higher irAE risk in patients with pre-existing autoimmunity and lower irAE risk in IPI, although 95% CIs around estimates were wide.

PO-2398 Trends In Acid Suppressant Drug Prescriptions In Primary Care In The United Kingdom 1400599 Devin Abrahami Trends In Acid Suppressant Drug Prescriptions In Primary Care In The United Kingdom Drug Utilization Research -> Quality use of medicines (including indicators and guidelines) Background: Proton pump inhibitors (PPIs) and histamine-2 receptor antagonists (H2RAs) are among the most commonly prescribed drugs worldwide. However, there is some evidence that they are being used inappropriately, either in individuals without underlying disease, or for a longer duration than necessary. In light of emerging safety concerns, it has become increasingly important to understand how these drugs are being prescribed in clinical practice.Objectives: To examine the prescribing patterns of PPIs and H2RAs over time by quantifying prevalence of use in the general population, intensity of use among drug users, and describing persistence patterns among new users.Methods: Using the United Kingdom Clinical Practice Research Datalink, we defined a cohort of patients prescribed PPIs and H2RAs between January 1990 and December 2018. Annual prescription rates were estimated by dividing the number of patients with at least one prescription by the total population registered in the data. Yearly intensity of use was estimated by dividing the number of prescriptions by the person-years of follow-up, and change in intensity was calculated per decade using negative binomial regression. Finally, Kaplan-Meier curves were used to examine persistence to original treatment course among new users.Results: PPIs were prescribed to 2,714,785 patients, while H2RAs were prescribed to 855,248 patients. Among new PPI and H2RA users, only 48.4 and 54.1%, respectively, had an approved indication for use. The prevalence of PPI use significantly increased from 1990 to 2018 (0.2 to 14.2%), while the prevalence of H2RA use remained low over the study period (2.5 to 1.9%). Prevalence of use was highest among females and the elderly (at least 60 years old). For every increase in year from 1990 to 1999, the yearly intensity of PPI use increased by 16% (IRR: 1.16, 95% CI: (1.16 to 1.16)), while the intensity of use remained constant from 2000 to 2009 and 2010 to 2018. In contrast, intensity of H2RA use decreased by 11% per year from 1990 to 1999 (IRR: 0.89, 95% CI: 0.89 to 0.90), decreased by 10% per year from 2000 to 2009 (IRR: 0.90, 95% CI: 0.90 to 0.90), and increased by 5% per year from 2010 to 2018 (IRR: 1.05, 95% CI: 1.05 to 1.05). Overall, 38% of PPI and 47% of H2RA users continued using these drugs beyond the one-year duration recommended in treatment guidelines.Conclusions: This study demonstrates that use of PPIs has been steadily increasing over time, while use of H2RAs remains low. Given the high prevalence of patients who do not have an approved indication for use and the high persistence to treatment, programs which target medication overuse should be considered.

PO-2406 Drug-drug Interaction Surveillance Study: Comparing Self-controlled Designs In Five Empirical Examples In Real-world Data 1400601 Katsiaryna Bykov BWH Div of Pharmacoepidemiology and Pharmacoeconomics Spotlight Poster Session Spotlight Poster Drug-Drug Interaction Background: Two self-controlled designs, the case-crossover (CCO) and the self-controlled case series (SSCS), have been utilized for studying drug-drug interactions (DDI) in real-world data (RWD). While sharing the strength of controlling for time-invariant confounding, these designs rely on different assumptions and may vary in their susceptibility to bias in RWD analyses.Objectives: To compare the CCO and the SCCS for identifying DDIs in electronic healthcare data.Methods: Using a large US commercial healthcare insurance database, we implemented the CCO and the SCCS designs to evaluate the following 5 DDI-outcome scenarios: (1) simvastatin-clarithromycin and muscle-related toxicity (expected, harmful DDI); (2) atorvastatin-valsartan and muscle-related toxicity; and (3-5) dabigatran-P-glycoprotein inhibitor (clarithromycin, amiodarone, verapamil) and bleeding. For both designs and all scenarios, analyses were conducted within person-time exposed to the object drug (statins and dabigatran). In the CCO, the odds of exposure to inhibiting drugs during a hazard window immediately preceding the outcome were compared to odds during a referent window 30 days prior. In the SCCS, outcome rates during person-time exposed to inhibiting drugs were compared to rates during unexposed person-time. We adjusted for possible bias associated with inhibiting drug by comparing DDI estimates to estimates from patients unexposed to the object drug (controls).Results: The designs yielded similar estimates for simvastatin-clarithromycin DDI: odds ratios (ORs) 2.61 (95% CI 1.52 - 4.50) before and 1.29 (95% CI 0.73 - 2.28) after adjustment in CCO; incidence ratios (IRs) 2.67 (95% CI 1.82 - 3.91) and 1.32 (95% CI 0.89 - 1.96) in SCCS. Both designs showed no association between atorvastatin-valsartan exposure and muscle-related toxicity, either before or after adjustment. Due to small numbers of patients exposed to dabigatran and clarithromycin, we could not evaluate it using the CCO; the SCCS yielded IRs of 2.48 (95% CI 1.06 - 5.78) before and 1.20 (95% CI 0.51 - 2.82) after adjustment. Both designs showed no association between amiodarone or verapamil and bleeding while on dabigatran but revealed positive association in patients unexposed to oral anticoagulants (controls), leading to protective adjusted estimates.Conclusions: We observed similar estimates from CCO and SCCS analyses of DDIs in real-world data, with SCCS displaying better efficiency. The use of control had a larger impact on the results than the choice of a design, highlighting the importance of appropriate control group selection.

PO-2411 The Positive Predictive Value Of Hip Fracture Diagnoses And Surgical Procedure Codes In The Danish Multidisciplinary Hip Fracture Registry And The Danish National Patient Registry 1400602 Thomas Hjelholt Department of Clinical Epidemiology, Institute of Clinical Medicine, Aarhus University Hospital, Den The Positive Predictive Value Of Hip Fracture Diagnoses And Surgical Procedure Codes In The Danish Multidisciplinary Hip Fracture Registry And The Danish National Patient Registry Methods in Pharmacoepidemiology -> Measurement Methods - Validation methods to confirm data (including Patient Reported Outcomes development and validation); methods to deal with missing data Background: The Danish Multidisciplinary Hip Fracture Registry (DMHFR) may be a valuable source for epidemiological research, if the data are valid. Previous hip fracture studies based on DMHFR have relied on few validation studies of hip fracture diagnosis registered in Norway, Finland, USA and Canada, with coding practices and healthcare systems different from Denmark.Objectives: To examine the validity of hip fracture diagnoses and surgical procedure codes in the DMHFR and the Danish National Patient Registry (DNPR) by calculating the positive predictive value (PPV).Methods: We identified a random sample of 750 hip fracture patients registered in the DMHFR between 2014 and 2017, treated at five specific hospitals in the Central or Northern region of Denmark. The hospitals were representative for the entire region, which is comparable to the national population in terms of sociodemographic and health-related characteristics. Diagnoses were coded by the 10th revision of the International Classification of Diseases, while procedures were coded by the Nordic Medico-Statistical Committee classification in the DNPR and directly transferred to the DMHFR. Using the surgical procedure description from the medical record as gold standard, we estimated the PPV of the hip fracture diagnoses and surgical procedure codes in the DMHFR and the DNPR with 95% confidence interval (CI).Results: The PPV was 90% (95% CI: 86%-93%) for fracture of the neck of femur, 92% (95% CI: 87%-95%) for trochanteric fracture, and 83% (95% CI: 78%-88%) for subtrochanteric fracture. Joining trochanteric and subtrochanteric fracture resulted in a PPV of 97% (95% CI: 95%-98%). Procedure codes had a PPV of 100% for primary prosthetic replacement and internal fixation with intramedullary nail, 96% (95% CI: 85%-99%) for internal fixation using screws alone, 91% (95% CI: 84%-96%) for internal fixation using plates and screws, and 89% (95% CI: 83%-94%) for internal fixation with other or combined methods. Stratifying by age group, gender, hospital type and calendar year of surgery were comparable to the overall PPV estimates.Conclusions: Our findings indicate a high quality of the hip fracture diagnoses and corresponding procedure codes in the DMHFR and the DNPR, with a majority of PPVs above 90%. Thus, the DMHFR and the DNPR are a valuable data source on hip fracture for epidemiological research.

PO-2412 In Utero Exposure To Antiretroviral Drugs And Pregnancy Outcomes: Secondary Analysis Of The Anrs Pharmacovigilance Database From The Anrs-promoted Hiv Research Studies 1400603 Laura Saint-Lary Unite Mixte de Recherche INSERM-Universit� 1027 In Utero Exposure To Antiretroviral Drugs And Pregnancy Outcomes: Secondary Analysis Of The Anrs Pharmacovigilance Database From The Anrs-promoted Hiv Research Studies Pregnancy and Lactation Background: More than 1.5 million pregnant women lived with HIV in 2016 and were eligible to receive antiretroviral drugs, raising the question of whether this treatment affects perinatal outcomes. Some studies have reported increased risks of adverse pregnancy outcomes in case of prenatal exposure to antiretroviral drugs, depending on class used. In 2018, the WHO has warned of an increased number of neural tube defects related to pre-conceptional exposure to Dolutegravir.Objectives: We assessed the adverse pregnancy outcomes associated with antiretroviral drugs exposure notified to the French ANRS.Methods: A systematic review has been performed using the ANRS pharmacovigilance database implemented for each ANRS-promoted HIV research from 2004 to 2019 with a routine notification pregnancy and its outcome: abortion, ectopic pregnancy, stillbirth, prematurity (50). Odd ratios (OR) and corresponding 95% CIs have been computed to assess association between outcomes and prenatal exposure to antiretroviral combinations.Results: Of 1 148 pregnancies identified, our systematic review included 924 pregnancies reported in 34 studies. There was no significant association between spontaneous abortion, prematurity and low-birth weight risks exposure to non-NNRTI combinations compared to non-nucleoside reverse transcriptase inhibitor (NNRTI) combinations. An increased risk of low-birth weight was reported in new-born exposed to exclusive nucleoside reverse transcriptase inhibitor (NRTI) combinations (N=4; OR 7.50 [1.49-37.83]).Conclusions: We did not present adverse pregnancy outcomes in case of exposure to different antiretroviral classes compared to NNRTI-based combinations, which is reassuring although not statistically powered for that purpose. Exposure to NRTI combinations seems to be associated with an increased risk of low-birth weight, but this result should be interpreted with caution considering the small size group and indication bias to use of exclusive NRTI combinations.

PO-2414 Nonselective Beta-blocker Eyedrops and the Risk of Fall-related Injuries: A Danish Nationwide Active Comparator, New User Propensity Score Matched Cohort Study 1400604 Sidsel Pedersen University of Southern Denmark Spotlight Poster Session Spotlight Poster Geriatrics Background: Falls in older patients represent a major public health issue. Risk factors for falls are multifactorial and include the use of commonly prescribed medication such as betablockers. Timolol is a non-selective beta-blocking eye drop agent used in the treatment of glaucoma. Due to high systemic absorption, the use of timolol eye drops may cause alterations in heart rate and rhythm potentially increasing falls risk. However, this has not been investigated in older adults.Objectives: To investigate the risk of injurious falls in older patients using topical non-selective betablockers compared to prostaglandin analogues.Methods: We conducted an active comparator, new user cohort study using information on redeemed prescriptions and hospital discharge diagnoses from the Danish nationwide registries. We identified all patients ≥65 years, who initiated either timolol eye drops or topical prostaglandin analogues and matched them in a 1:1 ratio, based on a high-dimensional propensity score. The primary outcome was injurious falls (defined as contusion, wounds and fractures other than hand and feet) within 90 days after a first prescription of either drug. Fall-related fractures was used as a secondary outcome.We obtained incidence rate ratios (IRR), incidence rate differences (IRD), and the number needed to treat to harm one additional patient (NNTH) for both unmatched and matched cohorts.Results: In total, 40,420 older adults were identified as users of timolol and 55,754 as users of prostaglandin analogues. Of these, 43% were male and the median [IQR] age was 75 [70-80] years. For the unmatched cohort, the IRR of injurious falls was 1.20 (95% CI 0.98-1.49), IRD was 4 (0-8) cases per 1000 person years, and NNTH was 1,695 for one year of treatment with timolol compared to prostaglandin analogues.In the propensity score matched cohort we identified 28,431 first users of timolol eye drops and 28,431 users of prostaglandin analogues. The IRR of injurious falls was 0.96 (95%CI 0.73-1.27), IRD -1 (95% CI -6 to 5) per 1000 person years, and a number needed to treat for one year to prevent one additional case of 12,500. For fall-related fractures, the IRR was 1.30 (95% CI 0.96-1.76), IRD of 4 cases per 1000 person years and a NNTH of 252.Conclusions: We found no increased risk of injurious falls after initiation of timolol eyedrops in an older Danish population.

PO-2415 Decreasing Adherence To Antiretroviral Therapy Over 4 Years Of Follow-up In A Commercially-insured Us Population Of Patients With Hiv 1400605 Rachel Bhak Spotlight Poster Session Spotlight Poster Adherence Background: Research has shown single-tablet regimens (STRs) of antiretrovirals (ARVs) lead to better medication adherence and viral suppression than multi-tablet regimens (MTRs); however literature in the commercially-insured US population is limited.Objectives: This study compared yearly and longer-term ARV adherence among HIV patients overall and by STRs vs. MTRs.Methods: A retrospective study using Optum Clinformatics US-based commercial claims data was conducted. Patients with an HIV-1 diagnosis during 2011-2017, age ≥ 18 years at index (date of first complete ARV regimen during the study period), and continuous enrollment for ≥ 3 months before index (baseline) and ≥ 12 months after index (observation) were included. MTRs were required to be comprised of 3 or more agents across at least 2 classes. Adherence was measured as the proportion of days covered (PDC) and compared using a Chi-square test. PDC was examined in the 1-year observation period for the overall analysis, and each year following index among patients with at least 4 years of continuous data. A subgroup analysis was conducted among patients with index during 2014-2016 to evaluate modern ARV adherence.Results: Among the 15,153 included patients, median age was 45 years, the majority were male (88%), and 53% were in the Southern region of the US. At index, 58% (n = 8,715) were receiving an STR and 43% (n = 6,438) an MTR. During year 1, the proportion of patients with PDC ≥ 0.90 was 63% overall, and greater for STR than MTR (67% vs. 58%, P P < .001). Among patients with at least 3 years of observation (STR: n = 489, MTR: n = 322), PDC ≥ 0.90 also decreased over time (59% in year 1 to 42% in year 3), and was consistently significantly higher for STR than MTR (P < .001).Conclusions: Adherence in this population of patients with HIV showed room for improvement in the first year of observation overall and in the modern ARV era, with those receiving STRs having higher adherence compared to those receiving MTRs. For the patients with 4 years of follow-up, adherence tended to decrease year on year. Maintaining high rates of ARV adherence is a critically important aspect of therapy for patients with HIV.

PO-2418 Trends Over Time In The Implementation Of Opioid Risk Minimization Measures And In Opioid-related Harms In Canada: A Mixed-methods Study 1400606 Camille Goyer YolaRx Consultants Trends Over Time In The Implementation Of Opioid Risk Minimization Measures And In Opioid-related Harms In Canada: A Mixed-methods Study Drug Utilization Research -> Changing drug utilization (interventions and implementation research) Background: Canada is experiencing a major opioid epidemic. Multiple guidance’s, policies and risk minimization measures (RMMs) have so far been implemented without an integrated approach, which makes the evaluation of their effectiveness challenging.Objectives: To develop a timeline of policies and interventions in Canada over the past five years and assess trends over time in pharmacovigilance metrics, including the reporting of opioid-related harms, including deaths and, social media coverage.Methods: The study involved four components: 1) Scoping review of opioid RMMs between 1 January 2016 to 15 November 2019 in Canada; 2) Descriptive analysis of opioid-related harms reported to Canada Vigilance, the national post-market spontaneous reporting database; 3) Qualitative review of opioid-related deaths using narratives of Quebec coroner’s reports; 4) Social media coverage on opioids and opioid-related harms (Facebook, Twitter and Reddit) using web-scrapping. Time period for the last three components covered 1 January 2009 to 31 August 2019. Trends over time in RMMs implementation was compared to trends over time in the reporting and social media coverage, without implying causality.Results: There were 414 Canadian RMMs and policies identified over the study period with a 4-fold increase as of 2016. The most common types were harm reduction (32.0%) (e.g., supervised injection site) and opioid awareness/education (20.0%) (e.g., educational webinars). In Canada Vigilance, there were 6,727 opioid-related harms (4,970 patients) with oxycodone being the most common suspected opioid (30.3%). Most cases were reported by consumers (63.8%) and 98.2% were considered serious. The number of reported cases remained constant over the years 2009 and 2018 (range: 53 - 177/year) while it increased dramatically during the fourth quarter of 2018 (n=661). There were 1,320 opioid-related deaths in Quebec; based on the Coroners’ reports, 69.8% were caused by opioids other than heroin, methadone and opium. Compared to 2015, the number of posts increased by 35.4% (Reddit), 329.0% (Facebook) and 381.5% (Twitter) in 2016. Trends plateaued for Reddit and Twitter as of 2017 and increased by 92.5% in 2019 for Facebook, compared to 2016.Conclusions: A large number of RMMs have recently been implemented in Canada. Trends in the reporting of opioid-related harms increased dramatically as of 2018 and opioid-related deaths in Quebec continue to rise. Media coverage and public awareness have dramatically increased since 2016. Further research is needed to identify which RMM has the greatest impact.

PO-2419 Endostar Continuous Versus Intermittent Intravenous Infusion Combined With Chemotherapy In Treating Patients With Advanced Non-small Cell Lung Cancer: A Meta-analysis 1400607 Lu Xu School of Public Health, Peking University Endostar Continuous Versus Intermittent Intravenous Infusion Combined With Chemotherapy In Treating Patients With Advanced Non-small Cell Lung Cancer: A Meta-analysis Drug Effectiveness Background: Endostar is a novel recombinant human endostatin approved to treat non-small cell lung cancer (NSCLC) in combination with first-line chemotherapy in China. Both intermittent intravenous (II) infusion and continuous intravenous (CI) infusion of Endostar are widely used currently in clinical practice.Objectives: To compare the efficacy and safety of CI of Endostar with II.Methods: We did a systematic review and meta-analysis by searching relevant databases from inception until June 25, 2019. Randomized controlled studies (RCTs), non-randomized controlled trials (NRCTs) and cohort studies compared CI of Endostar with II in combination with chemotherapy in stage III or IV NSCLC patients and reported efficacy or safety outcomes were eligible. Two reviewers independently screened records, extracted data and assessed risk of bias. Pooled risk ratios (RRs) with 95% confidence intervals were calculated using random effects meta-analysis for short-term efficacy and safety outcomes, and hazard ratios (HRs) for survival outcomes. GRADE approach was used to evaluate the quality of evidence in each outcome.Results: Finally nine studies involving 597 patients were included, including two RCTs, three NRCTs and four cohort studies. For short-term efficacy, moderate quality of evidence showed that there were no significant differences between CI of Endostar and II in objective response rate (ORR; RR 1.34, 95% CI 0.91-1.98, P = 0.14) and disease control rate (DCR; RR 1.11, 95% CI 0.94-1.30, P = 0.21). Low quality of evidence indicated that CI of Endostar significantly improved both overall survival (OS; HR 0.69, 95% CI 0.48-0.99, P = 0.046) and progression-free survival (PFS; HR 0.71, 95% CI 0.55-0.93, P = 0.01) compared with II. As for safety outcomes, high quality evidence found that CI of Endostar significantly substantially reduced the risk of myelosuppression (RR 0.55, 95% CI 0.32-0.96, P = 0.03) and cardiovascular toxicity (RR 0.21, 95% CI 0.06-0.78, P = 0.02) in comparison with II. Low quality of evidence also supported reduced risk of laryngeal hemorrhage in CI compared with II. No significant differences were found for the other adverse events.Conclusions: In NSCLC patients, compared with II, CI of Endostar had similar short-term efficacy, substantially lower risk of myelosuppression and cardiovascular toxicity. Although insufficient evidence supported the survival benefit of CI compared with II, large RCTs with long follow-up are needed to demonstrate these survival benefits.

PO-2420 Effects Of Incretin-based Therapies On Weight-related Indicators Among Patients With Type 2 Diabetes: A Network Meta-analysis 1400608 Lu Xu School of Public Health, Peking University Effects Of Incretin-based Therapies On Weight-related Indicators Among Patients With Type 2 Diabetes: A Network Meta-analysis Drug Effectiveness Background: Nowadays, a new class of anti-diabetic drugs based on incretin including glucagon-like peptide-1 receptor agonists (GLP-1 RAs) and dipeptidyl-peptidase IV inhibitors (DPP-4Is) have been introduced into clinical practice. Several studies reported that they can help control body weight.Objectives: To evaluate the effects of incretin-based therapies on body weight, body mass index (BMI) and waist circumference (WC).Methods: Databases including Medline, Embase, the Cochrane library and clinicaltrials.gov were searched for randomized controlled trials concerning the effects of incretin-based therapies on body weight, BMI and WC from inception to June 23, 2017. Risk of bias tool recommended by Cochrane handbook was used. Subgroup analysis, sensitivity analysis, and GRADE evaluation were performed as well.Results: A total of 292 RCTs (262 for weight, 191 for BMI and 56 for WC) were included in this study. Nine treatments were involved in this study, including GLP-1 RAs, DPP-4Is, insulin, metformin (Met), sodium-dependent glucose transporters 2 (SGLT-2), sulfonylurea (SU), thiazolidinedione (TZD), α-glycosidase inhibitor (a-Glu) and placebo. A total of 227, 118, 183, 269 and 260 trials were at low risk in terms of random sequence generation, allocation concealment, blinding of outcome assessment, complete outcome data, and selective reporting. As for weight, compared with placebo, DPP-4Is increased weight slightly by 0.31kg (95% CI: 0.05, 0.58) and GLP-1 RAs decreased weight by -1.34kg (95% CI: -1.60, -1.09). As for BMI, GLP-1 RAs decreased BMI by -1.10 (95% CI: -1.42, -0.78) and DPP-4Is had negligible effects on BMI compared with placebo. As for WC, compared with placebo, GLP-1 RAs decreased WC by -1.28cm (95% CI: -1.69, -0.86) and DPP-4Is represented a neutral impact. Subgroup analysis showed that DPP-4Is have no statistically significant effect on reduction of all three indicators compared with placebo, but GLP-1 RAs lowered all three indicators. According to the sensitivity analysis, the main results were robust. According to the ranking results, GLP-1 RAs and DPP-4Is ranked second and sixth respectively in terms of effects on weight loss. Regarding BMI deduction, GLP-1 RAs and DPP-4Is ranked first and fourth respectively. Speaking of WC decrease, GLP-1 RAs and DPP-4Is ranked second and fourth respectively. The evidence quality of ranking of treatments were low, very low and moderate for weight, BMI and WC, respectively.Conclusions: GLP-1 RAs are better at lowering all three indicators than DPP-4Is. Overall, the effects of GLP-1 RAs on weight, BMI and WC are good.

PO-2423 Psychotropic Medication Use In Former Icu Patients With Mental Health Problems: A Prospective Observational Follow-up Study 1400609 Debbie Janssen Utrecht University Psychotropic Medication Use In Former Icu Patients With Mental Health Problems: A Prospective Observational Follow-up Study Drug Utilization Research -> Other Background: An increasing number of patients with critical illness survive an admission to the intensive care unit (ICU) because of advancements in critical care. However, approximately 50-70% of these survivors develop mental health problems some time after ICU discharge. Little is known on 'whether' and 'how' these patients are being treated.Objectives: To describe the extent to which patients with mental health problems after admission to an Intensive Care Unit (ICU) initiate and use psychotropic medication.Methods: All adult patients who stayed in the ICU of the University Medical Center Utrecht for 48 hours or more between 2013 and 2017, alive after 1 year and not admitted to the ICU with brain injury, were eligible. Questionnaires that included the Hospital Anxiety and Depression Scale (HADS), the Impact of Event Scale (IES/IES-R), and psychotropic medication use, were used to identify components of mental health problems, i.e. depression, anxiety and posttraumatic stress disorder (PTSD).Results: Of the 1328 former ICU patients, 24.3% (n=323) had developed any of the mental health problems mentioned above. Of this group, 29.7% (n=96) used psychotropic medication one year after ICU discharge versus the 10.6% (n=107) of patients without these problems (OR 3.17, 95% CI 2.29-4.38). They were further 3.31 (95% CI 2.03-5.40) times more likely to initiate psychotropic medication (12.1% vs 3.7%) after ICU admission. Similar patterns were observed for individual groups of psychotropics: antidepressants, antipsychotics and benzodiazepines.Conclusions: Former ICU patients with mental health problems were almost three times more likely to use and initiate psychotropic medication than former ICU patients without these problems, although only one third of the patients were treated with psychotropic medication. Future research should investigated whether mental health problems are properly diagnosed and treated in former ICU patients.

PO-2425 Serum Lipid Trajectories In The Years Before A Lymphoma Diagnosis 1400610 Joan Fortuny RTI Health Solutions Serum Lipid Trajectories In The Years Before A Lymphoma Diagnosis Disease Epidemiology/Clinical Course -> Cancer Background: Serum cholesterol levels are known to be low in patients newly diagnosed with a lymphoma. Only one study has examined the trajectory of cholesterol levels in the years before a lymphoma diagnosis.Objectives: To describe the trajectories of lipid serum levels (total, low and high-density lipoprotein cholesterol and triglycerides) in the years before a diagnosis of lymphoma in a European population.Methods: We conducted a case-control study on patients with a lymphoma diagnosed between 2000 and 2017. Up to 5 noncancer controls per case were matched on practice, sex, calendar year, and age at lymphoma diagnosis date or an equivalent date among the controls (index date). Study participants were aged ≥ 40 years at index date; had no evidence of another cancer, HIV infection, or organ transplantation before or on the index date; and had ≥ 1 cholesterol measurement before the index date and before the first prescription of a hypolipidemic drug. General practitioner and hospital data from the Clinical Practice Research Datalink in the United Kingdom were used. Multilevel, multivariable linear longitudinal models were fit to examine the trajectories of serum lipid levels in the years before the index date with time since index date, case status, and confounders in the model.Results: During the study period, 3,161,492 patients had a valid cholesterol measurement. Among these, 11,969 cases of non-Hodgkin lymphoma and 473 of Hodgkin lymphoma met the study’s inclusion and exclusion criteria. Controls were selected for non-Hodgkin lymphoma and Hodgkin lymphoma cases (59,537 and 2,357, respectively). The most common subtypes of lymphoma were plasma cell neoplasm, chronic lymphocytic leukemia, and diffuse large B cell lymphoma. For all lymphoma types, crude mean cholesterol levels in the years before index date was lower in cases than in controls. For most lymphoma subtypes, the adjusted cholesterol levels showed a more pronounced decrease in the 4 years before diagnosis. On the other hand, triglycerides levels were not clearly related to case status. Results remained similar when restricting to patients who had never used any hypolipidemic drug and had ≥ 1 measure of cholesterol within 1 year before index date.Conclusions: This research is the first to replicate a similar study conducted in the United States, while more thoroughly adjusting for potential confounders. The different behavior of cholesterol and triglycerides, described here for the first time, offers insight into the role of lipids in lymphomagenesis. The study suggests the potential role of cholesterol as a biomarker for lymphoma in the years prior to its diagnosis.

PO-2426 Incidence Of Severe Abnormal Uterine Bleeding Following Oral Anticoagulant Use 1400611 Ting-Ying Huang Harvard Medical School Incidence Of Severe Abnormal Uterine Bleeding Following Oral Anticoagulant Use Safety End Points -> Cardiovascular Background: Recent reports have led to concerns over clinically relevant gynecologic adverse events among women of reproductive age and exposed to oral anticoagulants (OAs).Objectives: To determine Severe Abnormal Uterine Bleeding (SAUB, defined as vaginal bleeding (VB) requiring medical, transfusion or surgical intervention) incidence rate among OA users in US, overall and by menopausal age group.Methods: We used a new user cohort design and followed women who had newly initiated an OA, including non-vitamin K antagonist oral anticoagulants and warfarin, in the Sentinel Distributed Database between October 19, 2010 and September 30, 2015. Eligible members had an indicated condition (venous thromboembolism or atrial fibrillation/flutter), no evidence of index exposure, and no history of a qualifying VB management or joint replacement, in the 6-month lookback period. We defined three different SAUB outcomes - VB diagnosis followed by: 1) same-day medical intervention (insertion of intrauterine device, vaginal packing, initiation of an oral contraceptive or antifibrinolytic), 2) same-day transfusion, or 3) gynecologic surgery (hysterectomy, hysteroscopy, hysteroscopic polypectomy, myomectomy, dilation and curettage, endometrial ablation, or uterine artery embolization) within 30 days. Individual OA new users contributed at-risk time from their first valid index exposure date until the earliest of: SAUB occurrence, disenrollment, end of exposed time or study period, or recorded death.Results: In 1,050,192 OA new users, the overall SAUB incidence rates were 0.6, 1.7, and 5.0 per 1,000 person-years for outcomes with medical, transfusion, and surgical management respectively. When stratified by menopausal age group, the SAUB incidence rates in reproductive-aged (<50 years) women elevated to 11.8, 13.7, and 33.0 per 1,000 person-years; all estimates were consistently higher than those observed in menopause-aged (50+ years) women. SAUB with surgical management was notably more common than SAUB with medical or transfusion management in all ages.Conclusions: Based on real-world evidence, our findings suggest that patients on OA therapy may experience serious adverse outcomes related to gynecologic conditions such as SAUB while on OA therapy. As clinically significant VB is likely underreported since these women are often referred to gynecologic practices or end up in urgent/emergent care settings, we alert medical practitioners of potential SAUB cases. Further studies are warranted to facilitate the understanding of true SAUB risk in the OA user population and whether the class-wide association exists.

PO-2431 Digital Breast Tomosynthesis In Diagnosis And Screening Of Breast Cancer: A Systematic Review 1400613 Lu Xu School of Public Health, Peking University Digital Breast Tomosynthesis In Diagnosis And Screening Of Breast Cancer: A Systematic Review Medical Devices Background: Digital mammography (DM) will be affected by the breast parenchymal overlap, and only one image can be collected for each position in breast cancer screening. As a new digital photography method, digital breast tomosynthesis (DBT) enables imaging of the breast, and reduces parenchymal overlap.Objectives: To compare digital breast tomosynthesis (DBT) combined with digital mammography (DM) versus DM in screening and diagnosis of breast cancer.Methods: We did a systematic review and meta-analysis by searching English databases (Medline, Embase, Cochrane Library and Web of Science) and Chinese databases (Sinomed, CNKI, VIP and Wanfang) from inception until December 2, 2019. For using DBT in diagnosis, meta-analysis and SORC curve were completed via the bivariate model and HSROC model, respectively. For using DBT in screening, we calculated relative risk (RR) and 95% CI by random effects meta-analysis. Quality assessment was based on the Revised Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) checklist.Results: Fifteen diagnostic studies and 21 screening studies were included. In diagnostic setting, sensitivity was 0.90 (95% CI 0.86-0.94) for DBT plus DM versus 0.81 (95% CI 0.74-0.86) for DM. Specificity was 0.77 (95% CI 0.70-0.83) versus 0.71 (95% CI 0.63-0.79). Area under curve (AUC) was 0.91 (95% CI 0.88-0.93) versus 0.83 (95% CI 0.80-0.86). Positive likelihood ratio was 3.9 (95% CI 3.0-5.2) versus 2.8 (95% CI 2.1-3.8). Negative likelihood ratio was 0.13 (95% CI 0.08-0.19) versus 0.27 (95% CI 0.19-0.37). Diagnostic ratio was 31 (95% CI 17-57) versus 11 (95% CI 6-18). In screening setting, compared with DM, DBT plus DM can increase the detection rate of breast cancer (RR = 1.24, 95% CI: 1.17-1.32, I2 = 22.9%), increase the detection rate of invasive breast cancer (RR = 1.26, 95% CI: 1.16-1.33, I2 = 37.2%), increase the detection rate of in situ cancer in breast duct (RR = 1.24, 95% CI: 1.06-1.46, I2 = 58.3%), and reduce the recall rate (RR = 0.83, 95% CI: 0.78-0.87, I2 = 95.7%), increase the positive predictive value of recall (RR = 1.40, 95% CI: 1.22-1.61, I2 = 86.1%), and increase the biopsy rate (RR = 1.12, 95% CI: 1.03-1.22, I2 = 87.9%). For the positive predictive value of biopsy, there was no statistically significant difference between the two groups (RR = 1.01, 95% CI: 0.87 ~ 1.19, I2 = 91.2%).Conclusions: DBT plus DM is better than DM in the diagnosis of breast cancer. Compared with DM, DBT plus DM can increase the detection rate of breast cancer and reduce the recall rate.

PO-2432 Comparative Effectiveness Of Statins And Fibrates For Primary Prevention Of Cardiovascular Disease And Mortality: A Cohort Study 1400614 Joseph Blais University of Hong Kong Comparative Effectiveness Of Statins And Fibrates For Primary Prevention Of Cardiovascular Disease And Mortality: A Cohort Study Drug Effectiveness Background: The role of fibrates and their comparative effectiveness to statins for primary prevention of cardiovascular disease and mortality remains uncertain.Objectives: To determine whether statins versus fibrates are associated with a reduced risk of mortality and cardiovascular events in individuals without a history of cardiovascular disease.Methods: Observational, active comparator, new user, cohort study, using the Clinical Data Analysis and Reporting System of the Hong Kong Hospital Authority. Adults without a history of cardiovascular disease who initiated a statin or a fibrate were enrolled from 2003-2012 and followed-up until the first occurrence of an outcome of interest, death, or 31 December 2017. Outcome measures were all-cause mortality, cardiovascular mortality, and major cardiovascular events, defined as a composite of cardiovascular mortality, myocardial infarction, stroke and transient ischemic attack. Cox-proportional hazards models, with inverse probability of treatment weighting on the estimated propensity score were used to estimate hazard ratios (HR). We performed subgroup analyses stratified by baseline low-density lipoprotein cholesterol concentration, high-density lipoprotein cholesterol concentration, triglyceride level, and diabetes status.Results: A total of 271 491 participants (mean age 62; 46% men; median follow-up 7.3 years) were treated with a statin (n=213 218) or a fibrate (n=58 273). Compared with fibrates, statins were associated with a reduction in all-cause mortality (HR 0.96, 95% confidence interval (CI) 0.92-1.00), cardiovascular mortality (0.84, 95% CI 0.77-0.92), and major cardiovascular events (0.87, 95% CI 0.83-0.91). With the exception of participants with low high-density lipoprotein cholesterol concentrations (<1.0 mmol/L), statin use was consistently associated with a decreased risk of cardiovascular mortality and major cardiovascular events across subgroups.Conclusions: Statins were more effective than fibrates at reducing major cardiovascular events and cardiovascular mortality in participants without cardiovascular disease. Our results suggest that statins are preferred as initial treatment in individuals with elevated triglyceride levels or diabetes. However, individuals with low HDL cholesterol concentrations appeared to derive similar benefits from statins and fibrates.

PO-2448 Applying Sufficient-cause Principle To Estimate Risks Of Adverse Events In The General Indicated Patient Population Should A New Drug Is Used 1400616 Chuntao Wu Alexion Applying Sufficient-cause Principle To Estimate Risks Of Adverse Events In The General Indicated Patient Population Should A New Drug Is Used Methods in Pharmacoepidemiology -> Analytical Methods - e.g., marginal structural models, instrumental variables, sensitivity analyses Background: In drug development, it is of interest to assess the risk of an adverse event (AE) in the general indicated patient population using the data from clinical development. Suppose that in the randomized placebo-controlled clinical trials of Drug A, the incidence proportions of an AE of interest are IP(1) in the treatment arm and IP(0) in the placebo arm, and IP(1) is greater than IP(0). We denote the background incidence proportion of the AE of interest in the general indicated patients as IP(b0). The incidence ratio (IR=IP(1)/IP(0)) and incidence difference (ID=IP(1) - IP(0)) based on the data in the clinical trials are commonly used to estimate the incidence proportion of the AE of interest in the general indicated patients if Drug A is used as IP(b1)=IP(b0)*IR and IP(b1)=IP(b0) + ID, respectively. These two approaches usually yield different results, which cause confusion.Objectives: Using the sufficient-cause principle, we propose a new approach to estimate the risk of an AE in the general indicated patients should a new drug is used.Methods: In clinical trial data, we define IP(0) as the proportion of patients with sufficient causes resulting in AE of interest, absent of drug A . Thus, 1 - IP(0) is the proportion of patients without any sufficient causes for the AE, i.e., free of event. We define ID/(1 - IP(0)) as the additional sufficient-cause completion rate when Drug A is used, and denote it as CR (completion rate). Suppose Drug A yields the same additional sufficient-cause completion rate in the general indicated patients, the estimated incidence proportion of the AE when Drug A is used in the general indicated patients is calculated as IP(b0) + CR * (1 - IP(b0)), in which IP(b0) is the proportion of patients with sufficient causes for the AE in the background, without drug A; and 1 - IP(b0) is the proportion of patients without any sufficient causes for the AE.Results: In a hypothetical example of clinical trial, IP(1) and IP(0) of an AE are 0.3 and 0.1, and the background incidence proportion in the general indicated patients is IP(b0) = 0.25. Using IR (0.3/0.1=3) and ID (0.3-0.1=0.2), the expected incidence proportions in the indicated patients using drug A are estimated as 0.75 (0.25 * 3) and 0.45 (0.25+0.2), respectively. Using our approach, the estimated incidence proportion is calculated as 0.25 + [0.2/(1-0.1)]* (1-0.25)=0.42.Conclusions: The proposed approach avoids conflicting results from applying IR and ID, and the estimated incidence proportion is bounded between 0 and 1. This approach can be expanded to incidence rate measures and to that a study drug reduces the risk of an AE.

PO-2450 Prescription Patterns Of Outpatients And The Potential Of Multiplexed Pharmacogenomic Testing 1437470 Sze Ling Chan Singapore Health Services Prescription Patterns Of Outpatients And The Potential Of Multiplexed Pharmacogenomic Testing Rare Disease Background: Pre-emptive pharmacogenomic (PGx) testing is potentially an efficient approach to improve safety and efficacy of drug treatment. However, it is unclear who and when to test in initial implementation efforts. Objectives: The aim of this study was to describe the prescription pattern of PGx drugs among medical outpatients at a large tertiary-care hospital in Singapore to ascertain the potential of multiplexed PGx testing. Methods: This is a retrospective cohort study of adult medical outpatients. The 5-year cumulative incidence (CI) for receiving new PGx drugs was estimated from dispensing records using competing risks analysis. We considered 3 groups of drugs in our analysis: (i) all PGx drugs (defined as Clinical Pharmacogenetics Implementation Consortium (CPIC) level A or B), (ii) PGx drugs with a CPIC or Dutch Pharmacogenetics Working Group (DPWG) guideline and (iii) PGx drugs with a clinical effect of class D-F based on DPWG guidelines (termed ‘high risk’ drug). Comparisons of CIs were also done by demographic and clinical characteristics using Gray’s test. Results: The 5-year CIs of receiving any new PGx drug, PGx drug with guidelines, and a ‘high risk’ drug were 42.6%, 37.3% and 13.7%, respectively. The most common newly prescribed PGx drugs were omeprazole, codeine, simvastatin, dextromethorphan and clopidogrel (5-year CIs: 17.4%, 8.8%, 8.1%, 4.9% and 3.7%, respectively). The 5-year CI of receiving at least 1 PGx drug was higher for patients >40 years old at index visit (43.6% vs. ≤40 years old 36.0%, p -22), Malays and Indians (50.3% and 49.8% vs. Chinese 31.1%, p -22) and those who attended one of the following 4 specialties at the index visit compared to other specialties: infectious diseases (46.2% vs. 42.6%, p = 2.9 x 10-4), psychiatry (48.3% vs. 42.3%, p = 7.4 x 10-13), renal (49.8% vs. 40.9%, p -22) and rheumatology & immunology (54.8% vs. 41.7%, p -22). Patients prescribed ≥5 drugs at index visit had a 5-year CI of 51.7% compared to those prescribed 0-4 drugs (41.7%, p -22). Conclusions: Medical outpatients have a substantial probability of benefiting from pre-emptive PGx testing. Certain subgroups of patients have higher probability of receiving PGx drugs, but predictive modelling may be needed for yielding insights into risk stratification for pre-emptive genotyping implementation programs.

PO-2452 End-of-life Anticancer Treatment. A Nationwide Registry-based Study 1400618 Mette Bliddal University of Southern Denmark (Department of Clinical Research) End-of-life Anticancer Treatment. A Nationwide Registry-based Study Disease Epidemiology/Clinical Course -> Cancer Background: Anticancer treatments near the end of a patient’s life should generally be avoided, as it leaves the patient with no significant anticancer effect but increases the risk of severe side effects. Reports of the frequency of end-of-life treatment including chemotherapy vary markedly.Objectives: To describe the magnitude of all anticancer end-of-life treatment in Denmark and to identify subpopulations of patients more likely to receive anticancer treatment at the end of their lives.Methods: Using the Danish national health registers, we performed a nationwide cohort study identifying all deceased cancer patients (≥18 year of age) during 2010-2015 with 1) a diagnosis of a solid tumor defined by ICD10 codes, 2) a clinical oncology contact within 12 months prior to death, and 3) and cancer listed as their primary cause of death. Use of any anticancer treatment in the last 30 days of life was categorised as end-of-life treatment. We used logistic regression models to estimate odds ratios (OR) with 95% confidence intervals (CI) for receiving end-of-life treatment according to sex, age group, cancer group, Charlson Comorbidity Index (CCI), and time since diagnosis.Results: We identified 42,277 patients (mean age 70 years), of whom 15.9% received end-of-life anticancer treatment. The probability of receiving end-of-life treatment did not change over the study period (p-value 0.09). Chemotherapy alone was the most frequent treatment accounting for 78% of all end-of-life treatments in 2010 decreasing to 71% in 2015. In contrast, immunotherapy and hormone/anti hormone both used as single and combination therapy increased during the study period. Patient dying with breast cancer had the highest frequency of end-of-life treatment (23%), followed malignant melanoma (21%) and prostate cancer (18%). Protective factors for end-of-life treatment were female sex (OR 0.91, 95% CI 0.86-0.97), older age (OR 90+ years 0.10, 95% CI 0.07-0.16 and 75-89 years 0.32, 95% CI 0.27-0.38 compared to 18-44 years), high comorbidity according to CCI (OR CCI 3+ 0.82, 95% CI 0.76-0.88 compared to 0), and being diagnosed >6 months prior to death (OR diagnosed 1-5 years prior to death 0.53, 95% CI 0.50-0.57).Conclusions: In this nationwide Danish sample of cancer deaths, we found a stable overall rate at 15.9% of receiving anticancer treatment one month prior to death. During the period, a decreasing use of chemotherapy was out-balanced with an increasing use of immunotherapies and targeted treatments. Although, some end-of-life treatment is unavoidable, further research is needed to assess if the frequency of treatment during this period of life reflects inappropriate use.

PO-2457 Drug Use Variability By Community Sampling Place And Setting For Rural Women 1400619 Megan Dickson University of Kentucky College of Medicine Drug Use Variability By Community Sampling Place And Setting For Rural Women Health Economics/Outcomes Research Background: Rural drug use across the US has increased, as has the number of studies examining those increases. However, there is a paucity of studies focused specifically on rural women.Objectives: This study examines self-reported drug use in two rural county settings among women on public assistance (Welfare to Work) and in community confinement (jails) to: 1) determine how drug use patterns compare across sampling place, and 2) describe how county setting relates to self-reported drug use. It was hypothesized that rural women in jails would report more drug-involvement than those in Welfare to Work settings, but that there would also be differences related to county of recruitment.Methods: The current study design includes data from two datasets of drug-using rural women who were consented during the same time period (between December 2012 and August 2015), in the same two rural Appalachian Kentucky counties. The first dataset includes random selection and interviews with 279 women recruited from two county jails. The second dataset is 254 low-income women on Welfare to Work (public assistance). The current study includes baseline measures from both datasets such as self-reported lifetime drug use from the Addiction Severity Index. Participation rates exceeded 80% in the both samples. Logistic regressions controlling for demographic differences and county were used to compare drug use patterns across samples.Results: Women recruited from jail were significantly older (pp=.022), married (p=.010), and have less than a high school education (p=.011). When controlling for recruitment county and demographic differences, women recruited from jail were more likely to self-report having used of 8 of the 9 drugs examined, as well as injecting (AOR=12.66; p) and multiple drug use (AOR=14.31; p.001). Examining individual samples across county sites found similarities in drug use patterns by county of recruitment. Specifically the prevalence of methamphetamine use was significantly higher (p.001) in both samples from the same site, while illicit methadone use was significantly lower (p.05).Conclusions: Results support study hypotheses, suggesting that although jail participants are more drug-involved, drug use patterns may also be associated with county of recruitment. Implications include the need to: 1) consider both the community sampling place and setting when selecting subjects, 2) screen rural women for treatment needs and refer to interventions or treatment, as appropriate, and 3) improve service availability for rural women to reduce the likelihood of future drug use and related behaviors.

PO-2461 Real-world Outcomes Of Patients With Locally Advanced Or Metastatic Epithelioid Sarcoma 1444125 Priyanka Bobbili Real-world Outcomes Of Patients With Locally Advanced Or Metastatic Epithelioid Sarcoma Health Economics/Outcomes Research Background: Epithelioid sarcoma (ES) is an ultra-rare and aggressive subtype of sarcoma that typically affects young adults. Limited data are available on effectiveness and safety of conventional systemic therapies for treatment of advanced (unresectable and/or metastatic) ES.Objectives: This study aimed to collect real-world (rw) data on outcomes of ES patients receiving at least first-line (1L) or 2 or more lines (2L+) of systemic therapies.Methods: Retrospective chart review was conducted in patients with advanced ES who initiated systemic therapy between 2000-2017 at 5 US cancer centers. RECIST assessment was not performed in clinical practice, thus rw overall response rate (rwORR), duration of response (rwDOR), disease control rate (rwDCR; stable disease ≥32 weeks or response of any duration) and progression-free survival (rwPFS) were assessed by review of radiology reports. Overall survival (OS), rwDOR and rwPFS were estimated from start of therapy by Kaplan-Meier method. Endpoints were compared among patients treated with monotherapy vs combination therapy. Adverse events (AEs) resulting in treatment modification, discontinuation, hospitalization, permanent sequelae or death, were abstracted from medical records.Results: Of 74 eligible patients, 53 (72%) were male, and 63 (85%) had metastatic ES. INI-1/BAF47 was undetectable in 90% of 41 tumors tested. Median (range) age at diagnosis was 33 years (11-76), and number of lines of therapy received was 2 (1-7). Anthracycline-based (54%) and gemcitabine-based (24%) regimens were used most often in 1L. 1L rwORR was 15%, rwDCR was 20%, median rwDOR was 3.3 months, median rwPFS was 2.5 months, and median OS was 15.2 months. Among 46 (62%) patients with evaluable data in 2L+, rwORR was 9%, rwDCR, was 20%, median rwDOR was 4.5 months, median rwPFS was 6.0 months, and median OS was 10.0 months. In 1L, the 20 (27%) patients who received monotherapy had a lower rwORR (5% vs 19%) and rwDCR (16% vs 22%) compared with patients who received combination therapy; a similar trend was observed in 2L+. Over 50% of patients had at least one AE; most frequently febrile neutropenia (14%), pain (9%), anemia, dyspnea, fever, thrombocytopenia and transaminitis (5% each).Conclusions: Cytotoxic chemotherapies and tyrosine kinase inhibitors have limited efficacy in advanced ES and are associated with standard toxicities. This US-based rw study provides benchmarking treatment efficacy and safety data to understand the medical need of patients with advanced ES.

PO-2462 Concordance Of Race Information Derived From Electronic Health Records And Imputed Using Health Plan Administrative Data 1400620 Nancy Lin Health Catalyst Concordance Of Race Information Derived From Electronic Health Records And Imputed Using Health Plan Administrative Data Informatics Background: Real world evidence on potential heterogeneity in the safety or effectiveness profiles of selected medications by race/ethnicity is needed, however directly collected race information is typically unavailable in claims data.Objectives: To assess completeness and agreement of race coming from health insurer claims data (imputed) compared with race from electronic health records (EHR).Methods: We identified 3 cohorts in a US health insurer population linked to EHR data: members born between 10/2015-8/2017 (children); female members with a live birth delivery between 10/2015 - 8/2017 (mothers), and members with a heart failure diagnosis between 10/2015 - 8/2019 (HF patients). For each cohort, we calculated measures of concordance, sensitivity, specificity, and positive predictive value (PPV) comparing race/ethnicity from claims data to race/ethnicity available in EHR.Results: Of 19,600 children, 17,763 mothers, and 33,640 HF patients with linked data, race/ethnicity was classified as “other, multiple, unknown or missing” based on EHR for 16%, 12%, and 28% respectively. When restricted to members with a single recorded race in EHR, all cohorts were predominantly White (children: 84%; mothers: 77%; HF patients: 80%), with 5%, 8%, and 14% classified as African American among children, mothers, and HF patients respectively. The PPV for imputed White race was 94-95% across the 3 populations while PPVs for imputed African American race were lower and more variable (children: 40%; mothers: 66%; HF patients: 74%), likely influenced by differences in the racial composition of the cohorts. Imputed African American race had low to moderate sensitivity (children: 42%; mothers: 47%; HF patients: 70%) but was highly specific (96% - 98%).Conclusions: Completeness of race information in the EHR, and level of agreement between imputed race vs EHR-derived race, can vary depending on the population studied. Researchers are advised to consider the potential impact that the anticipated racial composition of the study population and specific data environment may exert on the performance of a selected imputation approach.

PO-2464 Identifying Valid Algorithms For Number Of Chemotherapy Treatment Lines In The Danish National Patient Registry 1437217 Signe S�rup Department of Clinical Epidemiology, Institute of Clinical Medicine, Aarhus University Hospital, Den Identifying Valid Algorithms For Number Of Chemotherapy Treatment Lines In The Danish National Patient Registry Methods in Pharmacoepidemiology -> Measurement Methods - Validation methods to confirm data (including Patient Reported Outcomes development and validation); methods to deal with missing data Background: Information on treatment with chemotherapy in all Danish hospitals is registered in the Danish National Patient Registry (DNPR). Yet, the number of lines of chemotherapy, which is important when evaluating cancer treatment, is not directly captured. Objectives: To identify which algorithms for number of lines of chemotherapy per patient based on treatment codes and dates in the DNPR have the highest percentage agreement with a reference standard of documentation in medical records. Methods: We included patients diagnosed between January 1, 2012, and December 31, 2016, with American Joint Committee on Cancer stage III or IV epithelial ovarian cancer, stage III or IV gastric adenocarcinoma, stage III or IV renal cell carcinoma, stage III or IV non-small cell lung cancer, or stage II, III, or IV urothelial cell carcinoma, who received chemotherapy at the Department of Oncology, Aarhus University Hospital, Denmark. We developed 2 algorithms for number of chemotherapy lines based on dates and treatment codes (eg, “treatment with cisplatin” or “cytostatic treatment”) in the DNPR. Algorithm I used the number of days between consecutive chemotherapy administrations (any chemotherapy treatment code). Algorithm II also used information on chemotherapy drugs if available or the number of days between consecutive chemotherapy administrations if chemotherapy drugs were unspecified. We evaluated the algorithms, setting the number of allowed days between administrations from 28 to 50. We calculated the percentage agreement between the number of chemotherapy lines identified from the algorithms and the reference standard—information on chemotherapy drugs abstracted from medical records and subsequently coded according to lines of chemotherapy. Results: In total, 179 patients were included (approximately 36 patients with each cancer type). For algorithm I, the highest percentage agreement with the reference standard was found when using <45 days between consecutive dates of chemotherapy (67.6% [95% CI: 60.1%-73.8%]). However, the percentage agreement was higher for algorithm II using <45 days between consecutive dates of chemotherapy administration for registrations with unspecified chemotherapy drugs (90.5% [95% CI: 85.0%-93.7%]). Conclusions: The highest percentage agreement between the data from medical records on number of lines of chemotherapy (reference standard) was found for an algorithm based on routinely recorded data from DNPR that incorporated both registration of specific chemotherapy drugs and <45 days between consecutive administrations if the drug name was unspecified.

PO-2465 Identification Of An Appropriate Time Window For The Exclusion Of Prevalent Cases Recorded Following Registration With The Clinical Practice Research Datalink 1400621 Thomas Leahy PHMR Ltd Spotlight Poster Session Spotlight Poster Databases Background: Defining incidence cases of a disease in primary electronic health records such as the Clinical Practice Research Datalink (CPRD) is complicated by the lack of patient history prior to registration with a CPRD-registered practice. In the CPRD it has been shown that prevalent conditions are often recorded shortly after registration with a GP practice and therefore to avoid the inappropriate inclusion of these individuals as incident cases, first diagnoses recorded in a specified period post registration are typically excluded. However, CPRD publications rarely report how the time period post-registration is chosen, with many appearing to use a default choice of 12 months.Objectives: To compare approaches to determine the length of the period post-patient registration required to avoid the inclusion of non-incident cases in a study population.Methods: As a case study, all patients with a first recorded non-valvular atrial fibrillation (NVAF) diagnosis in the CPRD between 2006 and 2015 were identified. Five different approaches were then used to choose a period post-registration to exclude patients (i) a standard 12-month period (ii) visual inspection of a plot of the incidence of the condition over time since registration, (iii) visual inspection of a plot of the cumulative sum (CUMSUM) over time since registration, (iv) a variance changepoint method to identify the point at which the incidence variance changes post-registration, and (v) a reverse time series trend approach. The impact of each approach on the size of the incident study population was assessed.Results: In the period 2006 to 2015 there were 48638 patients with a NVAF diagnosis. The use of a 12-month period under approach (i) resulted in a study population of 43992. Approaches (ii), (iii), (iv) and (v) resulted in the identification of 7 months, 3 months, 3 months and 5 months as being the suitable exclusion periods, respectively. Application of these periods resulted in study populations of 45439, 46825, 46825 and 46080 respectively. These reflected relative increases in the study population of 3.29%, 6.44%, 6.44% and 4.75% respectively.Conclusions: Using the standard 12-month period post-patient registration may unnecessarily reduce the study population size. We suggest alternative approaches are used which at a minimum involve visual inspection of relevant plots of the incidence over time since registration and potentially use objective, data driven approaches to choose the cut-off. Sensitivity analyses could be used to illustrate any impact of using periods of differing length.

PO-2466 Risk Of Skin Cancer In New Users Of Thiazides And Thiazide-like Diuretics: A Cohort Study Using An Active Comparator Group 1400622 Rahel Schneider University of Basel Spotlight Poster Session Spotlight Poster Cancer Background: Long-term use of hydrochlorothiazide (HCTZ) has been associated with an increased risk of skin cancer. It remains unclear whether the risk of skin cancer is increased for all thiazides and thiazide-like diuretics (TZs).Objectives: To examine whether the use of different TZs increases the risk of skin cancer compared to calcium channel blockers (CCBs).Methods: We conducted a cohort study using the United Kingdom Clinical Practice Research Datalink (CPRD). We established a comparative cohort including patients aged 18 to 85 years with a first-time prescription for a TZ or a CCB between 1998 and 2017. We excluded patients with recorded use of TZs and CCBs. The three co-primary outcomes of interest were basal cell carcinoma (BCC), squamous cell carcinoma (SCC), and cutaneous malignant melanoma (CMM). We estimated incidence rates (IR) per 100,000 person-years and IR ratios (IRRs) in short-term users (< 20 prescriptions) and long-term users (≥ 20 prescriptions) of TZs and CCBs using negative binomial regression. We used fine stratification on the propensity score (PS) to control for 23 baseline covariates. We conducted analyses for TZs overall and for bendroflumethiazide (BFT), indapamide, and HCTZ separately.Results: We identified 271,154 TZ users (87.6% BFT, 5.8% indapamide, and 3.6% HCTZ) vs 275,263 CCB users. Crude IRs for BCC were 460 (95% CI, 446-475) for short-term and 398 (381-415) for long-term TZ use. After PS-weighting, we observed IRRs for BCC of 1.24 (1.17-1.31) for short-term and 1.17 (1.10-1.23) for long-term TZ use. Results were not meaningfully different when we evaluated BFT, indapamide, and HCTZ separately. Crude IRs for SCC were 65 (60-70) for short-term and 72 (65-80) for long-term TZ use, which resulted in weighted overall IRRs of 1.07 (1.05-1.09) for short-term and 1.15 (1.12-1.18) for long-term TZ use. HCTZ showed a weighted IRR for SCC of 1.29 (1.24-1.35) in short-term users, which increased to 1.95 (1.87-2.02) in long-term users. BFT and indapamide revealed null results. Crude IRs for CMM were 72 (66-77) for short-term and 61 (54-68) for long-term TZ use. We observed overall IRRs of 1.11 (1.09-1.13) for short-term and 1.02 (0.99-1.05) for long-term TZ use. Long-term use of indapamide was associated with an increased risk of CMM (IRR, 1.43 [1.35-1.50]), whereas HCTZ and BFT showed null results.Conclusions: Our results corroborate the previously reported increased risk of SCC in association with long-term use of HCTZ and do not suggest a class effect for all TZs. Long-term use of indapamide may be a risk factor for CMM, but further evidence is needed.

PO-2467 Clinical Characteristics Of Sepsis Patients By Survival Status: An Analysis Of Electronic Medical Records 1400623 Brian Buysse Syneos Health Clinical Characteristics Of Sepsis Patients By Survival Status: An Analysis Of Electronic Medical Records Disease Epidemiology/Clinical Course -> Other Background: Sepsis-related mortality is estimated to be 20% globally. There is lack of understanding characteristics of patients with fatal sepsis.Objectives: To describe clinical characteristics of sepsis deaths and of those who survived sepsis using a US database of electronic medical records (EMRs).Methods: EMR data of patients with newly diagnosed sepsis within 12 months (ICD-10 codes included A40 and A41) in the past 5 years was analyzed in TriNetX Analytics (TriNetX, Cambridge, US; January 2020), a live database of ~60M US patients. Disease diagnoses, microbiological investigations and antimicrobial therapy were assessed in two cohorts, one of patients who had died within 1 month of sepsis diagnosis (deceased) and one of patients who had lived for at least 6 months (survived).Results: Included were 21,241 deceased and 90,304 surviving patients. Deceased patients were older at sepsis diagnosis than surviving patients (mean 69.3 and 57.4 yrs, respectively) and had greater prevalence of chronic conditions recorded in the year before sepsis. For instance, metastatic cancer was 4 times more frequent in deceased than in surviving patients. Microbiological investigations within 7 days before and including day of sepsis were done in only 34% of deceased and 38% of surviving patients. About 25% of both cohorts received antimicrobials within 30 days before sepsis. Antibiotics were given to ~80% of both cohorts in the 7 days following and including day of sepsis diagnosis, and this was higher in patients with microbiological investigations (87% in both cohorts). In deceased patients, vancomycin plus piperacillin-tazobactam was the most common first-line (15% of patients) and second-line (10%) therapy. In contrast, ceftriaxone monotherapy was the most common first-line antibiotic therapy in surviving patients (13%)—only 7% received vancomycin plus piperacillin-tazobactam as first-line therapy. The two most common second-line therapies in surviving patients were levofloxacin (7%) and amoxicillin (7%); only 2% of deceased patients took levofloxacin and only 1% amoxicillin. Antifungal therapy following sepsis diagnosis was higher in deceased (16%) than in surviving patients (3%).Conclusions: These results lay bare differences between patients who survive and succumb to sepsis in the US. The apparent low use of blood culture tests indicates that most patients receive empirical antimicrobial treatment. Differences in antimicrobial treatment between deceased and surviving patients may be explained by sepsis cause and severity. Sepsis and death are known to undercoded in EMRs, so results are approximate.

PO-2468 Drug-induced Vitamin B12 Deficiency - A Population-based Hypothesis-free Symmetry Analysis Screening 1400624 Daniel Henriksen Drug-induced Vitamin B12 Deficiency - A Population-based Hypothesis-free Symmetry Analysis Screening Pharmacovigilance -> Other Background: Vitamin B12 is a water-soluble vitamin required for formation of hematopoietic cells. Several drug-classes have been associated with vitamin B12 deficiency, but the evidence is typically based on spontaneous adverse event reportings, or case reports.Objectives: To perform a population-based hypothesis-free symmetry analysis screening of drug-classes associated with laboratory confirmed vitamin B12 deficiency.Methods: The study population consisted of all subjects living on the island of Funen (pop: 493 000 in 2016) who had at least one creatinine measurement between 2000-2016 as well as a vitamin B12 measurement. Subjects were included at the first occurrence of vitamin B12 deficiency (=130 pmol/L) within 365 days up to the date of their initial measurement indicating de novo vitamin B12 deficiency. We then performed a symmetry analysis screening of all newly initiated drug classes (ATC fourth level) one year prior to and after the date of the vitamin B12 deficiency measurement, anchoring the date to the date of the initial prescription of the given drug class. A symmetrical distribution of the number of patients with vitamin B12 deficiency before and after date of the prescription of the given drug class is expected if no association exist (i.e. a sequence ratio (SR) of 1.0). The symmetry analysis design is robust to confounders that are stable over time. We took only SRs into account if the difference between groups (difpresc,before/after), that is persons who filled a prescription before versus after vitamin B12 deficiency date, was above 10.Results: A total of 1738 persons were included for further analysis, 1014 (58%) were female, and the median age was 68 years (47-79 years). The drug-class with the largest difpresc,before/after was proton pump inhibitors (Npresc->B12def: 173/94, SR 1.84, 95%CI 1.44-2.39), followed by loop-diuretics (Npresc->B12def: 111/63, SR 1.76, 95%CI 1.31-2.44), paracetamol (Npresc->B12def: 144/113, SR 1.27, 95%CI 1.00-1.64), and angiotensin converting enzyme (ACE) inhibitors (Npresc->B12def: 63/34, SR 1.85, 95%CI 1.26-2.90). Metformin yielded an SR of 3.78 (95%CI 1.99-8.96), but the number of persons were low (Npresc->B12def: 34/9).Conclusions: Using the symmetry analysis to screen a large population-based cohort for laboratory-confirmed vitamin B12 deficiency confirmed several well-known adverse drug reactions, but also identified unknown but biologically plausible signals. The next step is to explore these more in depth.

PO-2469 Biliary Tract Cancer (btc) Epidemiology In The United States 1400625 Sandra Batten PAREXEL Biliary Tract Cancer (btc) Epidemiology In The United States Disease Epidemiology/Clinical Course -> Cancer Background: BTC is a rare diagnosis in the US (~8000 people annually) but represents a high disease burden due to late diagnosis, poor prognosis, and limited treatment options. Monitoring the epidemiology of BTC will inform public health management and development of clinical programs and research.Objectives: Evaluation of BTC incidence, prevalence, mortality and survival rates (SR) using the latest data from SEER and NPCR databases.Methods: Data (2001-2015) were assessed by age, gender, race/ethnicity (White [W]; White, Hispanics [H]; Black [B]; Asian/Pacific Islander [A/PI]; American Indian/Alaska Native [AI/AN]); and anatomic site (extrahepatic cholangiocarcinoma [ECC], intrahepatic cholangiocarcinoma [ICC], gall bladder cancer [GBC], and ampulla of Vater cancer [AVC]). Incidence was reported as age-standardized rate (ASR; per 100,000 person-yrs) using the 2000 US standard population. Estimates of annual percent changes (APC) were calculated using the weighted least squares method.Results: ASR for BTC in the US was 5.04. Incidence was slightly higher in males vs females (M, ASR = 5.31; F, ASR = 4.85), and was consistent across most subtypes; although females had a higher incidence of GBC (ASR = 1.92; M, ASR = 1.14). Using incidence of W ethnicity as a reference value (ASR = 4.64), H and A/PI have higher incidence rates (incidence rate ratio = 1.65 and 1.45, respectively). During the survey period, total incidence of BTC increased significantly (APC = 1.76). The highest increase was reported for ICC (APC = 6.65). Overall mortality ASR for patients (pts) with BTC was 3.40. H, A/PI, and AI/AN had higher mortality rates (mortality rate ratio = 1.36, 1.22, and 1.32, respectively) when using W as reference (ASR = 3.25). The 5-yr SR for BTC is 15.2%; all subtypes had a consistent rate (GBC = 18%; ICC = 8.5%; ECC = 9.1%), except AVC (34.5%). SRs among ethnic groups were also largely consistent with the overall, with B having the lowest SR (13.8%) and H the highest (16.6%). While the 5-yr SR is low for BTC overall, it is extremely low for pts with distant disease (3.0%), compared with regional (19.1%) or localized disease (31.5%). The 10-yr prevalence for BTC is 10.8 per 100,000.Conclusions: Incidence of BTC, particularly ICC, has increased over the last decade and the 5-yr SR for BTC remains low. Discrepancies among race/ethnicity groups may reflect socio-economic and health care access differences. These data underscore the serious unmet need among pts with BTC. Public health efforts should be applied to raising awareness of the growing BTC population and integrating novel therapy options for BTC pts.

PO-2470 Effect Of Several Optimization Strategies In A Clinical Decision Support System For Drug-drug Interactions On Alert Burden And Alert Acceptance 1400626 Katoo Muylle Vrije Universiteit Brussel Effect Of Several Optimization Strategies In A Clinical Decision Support System For Drug-drug Interactions On Alert Burden And Alert Acceptance Medical Devices Background: Clinical decision support systems (CDSS) are associated with a high alert burden and too many alerts of low clinical importance leading to alert fatigue and high override rates compromising the primary objective of patient safety.Objectives: To evaluate the effect of the implementation of an optimized CDSS for drug-drug interaction (DDI) screening on alert burden and alert acceptance.Methods: This retrospective pre-post study was conducted at a 721-bed university hospital. Only inpatients were included in the analysis. The intervention comprised of several optimizations in the CDSS including the customization of the knowledge base and alert severity classification together with different required actions in order to override the alerts according to DDI severity level, the creation of individual screening intervals, patient-specific alerting, a new alert design, and the creation of a follow-up system. This pre-post study analyzed in detail the alert burden in the pre- and post-intervention period. The alert acceptance was evaluated not only based on in-dialog actions (i.e. prescription discontinued directly from the alert screen), but also based on the actual administration of the drugs involved in the DDI.Results: In the pre-intervention period, 1087 alerts (92.0% level 1 alerts) were triggered, accounting for 19 different class-based DDIs. In the post-intervention period, 2630 alerts (38.4% level 1 alerts) were triggered, representing 86 different class-based DDIs. The relative risk for alert acceptance based on in-dialog analysis in the post-intervention period compared to the pre-intervention period was 4.019 (95% confidence interval (CI) 3.170 - 5.097; 25.5% versus 6.3%). This relative risk was 1.164 (95% CI 1.083 - 1.252; 54.4% versus 46.7%) when alert acceptance was measured based on administration rates. Analyzing only level 1 alerts, the relative risk for prescription acceptance was 5.504 (95% CI 4.262 - 7.109; 34.1% versus 6.2%) and for administration acceptance was 1.78 (95% CI 1.364 - 1.602; 67.3% versus 45.5%).Conclusions: The implementation of the new CDSS was successful given the significant higher alert acceptance. This increase in alert acceptance was mainly due to the higher acceptance of level 1 alerts, which is probably related to the optimization strategies leading to more specific alerts. Alert acceptance based on administration rates was about twice as high as the alert acceptance based on prescription rates.

PO-2472 An FDA Quality Assessment Tool For Evaluation Of Real-world Evidence Observational Studies 1400627 Marie Bradley US Food and Drug Administration An FDA Quality Assessment Tool For Evaluation Of Real-world Evidence Observational Studies Methods in Pharmacoepidemiology -> Study Design - Studies assessing study design (e.g., choice of comparator, self-controlled methods, avoiding immortal time bias) Background: Recent legislation encourages the US Food and Drug Administration to evaluate the use of real-world evidence (RWE) in support of regulatory decision making. Assessing the quality of RWE observational studies to ensure they meet the FDA requirements for adequate and well-controlled studies is critical.Objectives: To build an FDA specific quality assessment tool for observational studies based on a previous FDA quality assessment framework.Methods: The 2012 FDA quality assessment framework (FDA framework) was evaluated against published quality assessment tools commonly used for observational cohort studies, including the Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I) and the GRACE checklist (GRACE), and in light of evolving observational study designs and methods for quality improvement. Important features of and gaps in the quality elements of the FDA framework were identified. These were used to inform the development of a more comprehensive FDA specific quality assessment tool.Results: The main gaps identified in the FDA framework were: Study design: Use of a new user design and selection of participants based on variables assessed after index exposure were not addressed. Exposure: Adherence to treatment was not assessed. Analytic approach: Quantitative bias analysis or the appropriateness of methods to combine data from different sources was not addressed. Missing data for important confounders was not assessed. Important elements of the FDA framework that were not prioritized in the other tools included appropriateness of aims, study design and base population (denominator), exposure definition and measurement, extent of loss to follow-up, generalizability of findings and the inclusion of a statement on biologic plausibility or alternative explanations.Conclusions: We identified several important features and gaps in the existing FDA quality assessment framework for observational cohort studies. To address these gaps, we built a comprehensive quality assessment tool for use in the evaluation of observational studies at FDA .

PO-2473 Missing Data Analysis In Multi-center Studies When Access To Person-level Data Is Restricted 1400628 Qoua Her University of North Carolina at Chapel Hill Missing Data Analysis In Multi-center Studies When Access To Person-level Data Is Restricted Methods in Pharmacoepidemiology -> Measurement Methods - Validation methods to confirm data (including Patient Reported Outcomes development and validation); methods to deal with missing data Background: Distributed regression analysis (DRA) is a privacy-protecting analytical method that performs multivariable-adjusted regression analysis in multi-database settings using only summary-level information. However, data quality issues such as missing data continue to challenge the utility of DRA. Few method recommendations are available to guide analyses with missing data in multi-center studies where access to person-level data is restricted.Objectives: To evaluate the performance of DRA after handling missing data with multiple imputation by chain equations (MICE) at each data-contributing site.Methods: We created simulated cohort studies (n=10,000) of bariatric surgery patients based on data from the OptumLabs® Data Warehouse, which includes de-identified administrative claims and electronic health record data, to evaluate the comparative safety and effectiveness of two common bariatric procedures. Three types of outcomes were evaluated: percent weight change by the first post-surgery year (continuous), rehospitalization within the first 30 post-surgery days (binary), and time to first rehospitalization during the first post-surgery year (time-to-event). We induced missing data to the pre-surgery BMI variable under different missing mechanisms (missing completely at random, missing at random, and missing not at random) and levels of missingness (10%, 50%, and 80%). We partitioned the data into 3 datasets of approximately equal size to simulate a 3-site multi-center study with missing data, where access to person-level data is restricted. We performed site-specific MICE to create 5 imputed datasets and then conducted DRA (linear, logistic, and Cox proportional hazards) to estimate the global regression parameters and standard errors using summary-level information. We compared the results obtained from this approach to the results obtained from the benchmark analysis of the pooled person-level data where MICE was conducted by site. A difference of -6 between estimates from the two approaches was deemed to demonstrate the validity of DRA.Results: The use of MICE and DRA produced regression parameter estimates and standard errors that were precise to those computed from the benchmark analyses under all missing data scenarios considered. The difference in the regression results from the DRA compared to those the from benchmark ranged from 10-15 to 10-17.Conclusions: MICE plus DRA can be used to handle missing data without sharing any person-level data in multi-center studies.

PO-2477 Long-term Safety And Effectiveness Of Etanercept In Pediatric Psoriasis Patients: Results From A Prospective Observational Cohort Study 1400629 Yun Gu Long-term Safety And Effectiveness Of Etanercept In Pediatric Psoriasis Patients: Results From A Prospective Observational Cohort Study Biologics/Biosimilars Background: Etanercept is a soluble TNF receptor fusion protein indicated for treatment of severe plaque psoriasis, and can be prescribed from 6 years of age.Objectives: The study aimed to evaluate long-term safety and effectiveness of etanercept in pediatric plaque psoriasis patients.Methods: It was a long-term, prospective, observational study of pediatric psoriasis patients prescribed etanercept and followed for up to 5 years in 8 EU countries. Patients were classified as prospective (starting etanercept within 30 days pre-enrolment or any time after) or retrospective (starting etanercept > 30 days pre-enrolment). Incidence rate and relative risk were estimated for defined outcomes, with patients classified by timing of AE relative to etanercept: current/recent or previous treatment.Results: There were 72 patients (32 prospective and 40 retrospective) enrolled at 28 sites, with 51% males. Mean age was 14.5 years with average disease duration of 7.1 years. Patients were followed for a mean of 3. 8 years, with 29 (40%) completing 5 years follow-up. No serious or opportunistic infections or malignancies were reported. There were 25 serious adverse events (SAEs), including 15 SAEs in 11 (15.3%) patients with current/recent etanercept treatment and 10 SAEs in 7 (13.0%) previously exposed, with an incidence rate of 6.21 (95% CI: 3.48 - 10.24) and 8.54 (95% CI: 4.09 - 15.70) per 100 person-years, respectively. Among them, 7 (28.0%) were considered possibly related to etanercept (1 Cholesteatoma, 1 hepatic steatosis, 3 psoriasis and 2 drug ineffective). Infections and infestations were most frequently reported treatment-emergent AE (TEAEs) with 49 events in 19 (26.4%) and 7 events in 5 (9.3%) patients with current/recent exposure and previous exposure, respectively. Among 32 prospective patients, 28 (87.5%) completed at least 24 weeks of therapy, and 5 (15.6%) required ≥ 1 subsequent course of etanercept. Sixteen (50%) required > 1 treatment period with other systemic treatments. After treatment with etanercept, 93.8% (30/32) had a decrease in disease severity. Twenty six (81.3%) prospective patients had ≥ 1 dose change/discontinuation during initial treatment period, with drug ineffective being the most common reason (32.4%).Conclusions: In this one of the largest studies with the longest follow-up of pediatric psoriasis patients treated with etanercept in the real-world setting, no serious or opportunistic infections or malignancies were observed, and the majority of prospective patients had a decrease in disease severity after etanercept treatment.

PO-2480 Effect Of Domperidone, Ondansetron, And Olanzapine Containing Antiemetic Regimen On Qtc Interval In Patients With Malignancy 1400630 Ashwin Kamath Kasturba Medical College, Mangalore, Manipal Academy of Higher Education, Manipal Effect Of Domperidone, Ondansetron, And Olanzapine Containing Antiemetic Regimen On Qtc Interval In Patients With Malignancy Safety End Points -> Cardiovascular Background: Domperidone, ondansetron, and olanzapine can prolong the QT interval. The use of combinations of these drugs in various clinical situations is not uncommon; however, there is no data regarding the effect of the drug combination on the QTc interval.Objectives: To determine the presence of any QTc prolonging effect of the drug combination when used as antiemetic in patients receiving cancer chemotherapy.Methods: We carried out a prospective, observational study of patients with malignancy, aged 18−80 years of either gender with Eastern Cooperative Oncology Group status ≤ 2, who were to receive domperidone (Day 2 and 3), ondansetron (Day 2 and 3), and olanzapine (Day 1-3) containing antiemetic regimen. The study was approved by the institutional ethics committee, and informed consent was obtained from the study participants. Electrocardiograms were recorded before (Day 0) and during (Day 1-3) the administration of antiemetics, for three consecutive days. The absolute value and the change in the QTc interval compared to baseline value was recorded and analyzed. The QT interval was corrected for heart rate using Bazett’s method. Freidman’s test was used to determine the presence of any significant difference in the QTc interval before and after the administration of the drugs. If a significant difference was seen, Wilcoxon signed-rank test was used to identify the time points with a significant difference, using Bonferroni correction. A p-value 500 ms. More female patients had ∆QTc >30 ms compared with males (56.52% versus 46.15%; p = 0.8018).Conclusions: QTc prolongation is a concern with olanzapine alone and in combination with domperidone and ondansetron. However, no clinically significant adverse events occurred. The clinical relevance of the observed QTc prolongation needs to be investigated further.

PO-2481 Drug-related Hospital Admissions Following Emergency Department Visits: Preliminary Results Of A Cross-sectional Study 1400631 Zuzana Ocovska Charles University, Faculty of Pharmacy in Hradec Kralove Drug-related Hospital Admissions Following Emergency Department Visits: Preliminary Results Of A Cross-sectional Study Safety End Points -> Other Background: Drug-related hospital admissions (DRA) have attracted much research attention worldwide. And yet, this topic has not been sufficiently investigated in the Czech Republic. Objectives: To determine the prevalence and preventability of DRA, to identify the implicated medications in DRA and to examine the preventability aspects of DRA. Methods: This cross-sectional study examines hospital admissions following emergency department visits in the University Hospital Hradec Králové in 2018. Planned hospitalisations and visits to emergency room without inpatient hospitalisation are not included. The data are retrieved retrospectively using electronic medical records. The data are entered into an Access database. The DRA identification process draws upon the drug-related admissions adjudication guide developed by Thevelin et al.1 Medical Dictionary for Regulatory Activities and the Anatomical Therapeutic Chemical classification system are used for classification of DRA and implicated medications. Descriptive and inferential statistics are used for statistical analyses. Results: Out of 951 eligible hospital admissions analysed so far, 926 have been included and 154 DRA have been identified. The prevalence of DRA was 16.6% (95% CI 14.2-19.0). 118 DRA were related to treatment safety and 36 DRA were related to treatment effectiveness. Antithrombotic agents (21.2%), psycholeptics (9.3%), antiinflammatory and antirheumatic products (5.9%) and antineoplastic agents (5.9%) represented the most common medication classes involved in DRA related to treatment safety. Diuretics (25.0%), antihypertensives (22.2%), antithrombotic agents (16.7%) and drugs used in diabetes (13.9%) represented the most common medication classes involved in DRA related to treatment effectiveness. Gastrointestinal disorders and metabolism and nutrition disorders were the leading causes of DRA. The number of drugs in medication history was a predictor of DRA (OR 1.11, 95% CI 1.07-1.16). 56.5% of identified DRA were potentially preventable. Conclusions: Measuring the scope and nature of DRA is essential for the development of risk minimization measures. The relatively high preventability observed suggests the research on preventability aspects of DRA deserves attention. The study was supported by Charles University (Project SVV 260 551, Project GAUK 14120). 1 Thevelin S et al. Development of a standardized chart review method to identify drug-related hospital admissions in older people. Br J Clin Pharmacol. 2018;84(11):2600-14.

PO-2484 Comorbidities Analysis Of Patients With Atopic Dermatitis On Non-biologic Systemic Drug Therapy Using Optum Electronic Health Records Databases 1439060 Huifang Liang Comorbidities Analysis Of Patients With Atopic Dermatitis On Non-biologic Systemic Drug Therapy Using Optum Electronic Health Records Databases Disease Epidemiology/Clinical Course -> Other Background: Managing comorbid conditions may improve treatment outcomes in patients with atopic dermatitis (AD). Limited data exist on comorbidities in AD patients who are on non-biologic systemic drug therapy. Objectives: To describe comorbidities in AD patients who are on non-biologic systemic drug therapy. Methods: A retrospective cohort study was performed to identify patients with AD in the Optum™ Electronic Health Records Databases from 2008-2018 based on the algorithm of Silverberg et al. The algorithm uses two diagnoses of AD, or the combination of (i) one diagnosis of eczema or rash and nonspecific skin eruption and (ii) ≥2 historic codes for asthma, food allergies or allergic rhinitis. Patients were eligible if they were in the integrated data network and had ≥1 year of activity before the index date (or the first date of receiving non-biologic systemic drug therapy after the diagnosis date). Prespecified thirteen comorbidities of interest were assessed using data within one year before the index date. Descriptive statistics were generated for overall and pediatric and adult populations separately. Results: A total of 286,941 patients were included. Their mean age was 37.1 ± 26.9 years, with children and adults being 34.4% and 65.6%, respectively. Overall, females were predominant (73.3%) among adults; however, males were predominant among children (56.4%). The majority of non-biologic systemic immunosuppressants were corticosteroids (97.6%). The use of the index systemic immunosuppressants was higher in adults than children (3.3% vs 0.64%). Among patients with AD who were on non-biologic systemic drug therapy, over 75% of adults and over 20% of children had at least one comorbidity. More females than males had comorbidities (73.8% vs 63.4%). Asthma was the most common allergic comorbidity in both adults (33.3%) and children (39.1%). Several neuropsychiatric disorders were also common in all patients: sleep disturbances (adults vs children: 20.2% vs 5.2%), anxiety (24.3% vs 2.6%), depression (24.0% vs 1.1%), and attention-deficit/hyperactivity disorder (ADHD) (2.6% vs 4.0% in children). In addition, common non-allergic comorbidities in adults included hypertension 47.2%, hyperlipidemia 42.1%, and diabetes 18.8%. Conclusions: In patients with AD on non-biologic systemic drug therapy, asthma is the most common allergic comorbidity; common non-allergic comorbidities include sleep disturbances, depression and anxiety. Screening and intervening on atopic and nonatopic comorbidities are essential to alleviating the burden for patients with AD.

PO-2487 Risk Of Pregnancy Loss After In Utero Exposure To Triptans 1444212 Caroline Hurault-Delarue Risk Of Pregnancy Loss After In Utero Exposure To Triptans Pregnancy and Lactation Background: Triptans are serotoninergic agonists with vasoconstrictive properties, indicated in migraine and vascular facial algia. Data on the effects of triptans during pregnancy mainly concern sumatriptan and are reassuring concerning the teratogenic risk. However, we can wonder about the potential risks of these drugs on the course of pregnancy.Objectives: The objective was to study risks of pregnancy loss after in utero exposure to triptans.Methods: We performed a comparative study in EFEMERIS database, including 135,576 pregnant women who delivered between 2004 and 2017, in a French department. Exposure to triptans (ATC code N02CC: Selective serotonin (5-HT1) agonists) during pregnancy was described. The comparisons of pregnancy loss prevalences between exposed and non-exposed pregnant women were performed using logistic regressions.Results: During the study period, 802 pregnant women received at least one prescription and dispensation of triptans (0.6%), mainly during the first trimester of pregnancy (78.2%). Zolmitriptan, sumatriptan and eletriptan were the most prescribed. Among exposed women, 10.6% of pregnancies ended in a pregnancy loss (spontaneous abortions and fetal deaths) against 4.9% in the comparison group. After adjustment for confounding factors, a slight increased risk was observed for the triptan-exposed group (adjusted OR=1.3 [1.1-1.7]). An additional analysis taking into account NSAID exposure confirmed the slight increased risk of pregnancy loss after triptan exposure (3.9% in women neither exposed to triptans nor to NSAIDs, 6.3% in women exposed to triptans and not exposed to NSAIDs, 13.6% in women not exposed to triptans and exposed to NSAIDs, 17.2% in women exposed to both triptans and NSAIDs).Conclusions: This study showed a slightly significant association between dispensation of triptans during pregnancy and pregnancy loss. The results of this study should be interpreted with caution due to bias, particularly the frequent concomitant exposure to NSAIDs known to increase the risk of fetal death. Anyway, the role of triptans cannot be ruled out because of a plausible pharmacological mechanism. In conclusion, looking forward to further pharmacoepidemiological investigations, the use of triptans should not be trivialized in pregnant women, and requires a careful benefit/risk evaluation.

PO-2488 Drug Prescriptions In French Pregnant Women Between 2015 And 2017, A Study In The Egb Database 1444213 Caroline Hurault-Delarue Drug Prescriptions In French Pregnant Women Between 2015 And 2017, A Study In The Egb Database Pregnancy and Lactation Background: Studies performed in France before 2014, consistently showed that a large number of drugs was prescribed to pregnant women.Objectives: The aim of this study was to describe French drug prescription practices in pregnant women between 2015 and 2017. The secondary objective was to identify teratogenic and fetotoxic drug prescriptions during pregnancy.Methods: This study was carried out using data from a Permanent Beneficiaries Sample (Echantillon Généraliste des Bénéficiaires - EGB), a French national health database which includes 1/97th of the French population. Our study population included all pregnant women, aged 10 to 60, who were registered in the EGB and had a pregnancy outcome between 2015 and 2017. Drugs prescribed and dispensed to women during pregnancy and the 3 months before, were collected and described for each year and according to pregnancy trimesters. Prescriptions of major teratogen or fetotoxic drugs were described.Results: We identified 27,596 pregnant women. Among them, 93% received drug prescriptions and dispensations during pregnancy with an average of 7.4±5.4 different drugs. "Alimentary tract and metabolism (75.3%)", "nervous system (63.9%)" and "blood and blood forming organs (58.9%)" classes were the most frequently prescribed to pregnant women. The 5 most frequently prescribed drugs were paracetamol (60.3%), iron (49.2%), folic acid (45.3%), phloroglucinol (43.7%) and colecalciferol (41.7%). The most commonly prescribed drugs included some that have not yet been well evaluated in pregnancy as tixocortol, sertaconazole, helicidine, metopimazine and domperidone. Prescriptions and dispensations of teratogenic or fetotoxic drugs, as Non-Steroidal Anti-Inflammatory Drugs after the 6th month of pregnancy and retinoids were observed. Valproic acid prescriptions to pregnant women have become extremely rare.Conclusions: This descriptive study demonstrates that several drugs are prescribed and dispensed to pregnant women in France. These include drugs with a proven teratogenic or fetotoxic effect and many drugs that have not yet been well evaluated in pregnancy.

PO-2489 Benefits And Limitations Of Using Individual And Linked English Routine Data Sources To Identify Incident Cancers In Cancer Epidemiology Studies 1400632 Helen Strongman London School of Hygiene and Tropical Medicine Benefits And Limitations Of Using Individual And Linked English Routine Data Sources To Identify Incident Cancers In Cancer Epidemiology Studies Informatics Background: Pharmacoepidemiologists commonly use individual and linked English primary care, hospitalisation, cancer registration and death registration data in cancer studies. We do not know how this choice of data source(s) affects selection of incident cancer populations and measurement of cancer exposures and outcomes for a wide range of cancers.Objectives: To describe the benefits and limitations of using different individual and combinations of English electronic health data to identify incident cancers.Methods: We used linked Clinical Practice Research Datalink (CPRD) GOLD primary care, cancer registration, hospitalisation, and death registration data for our descriptive study. We implemented alternative case definitions to identify first site-specific cancers at the 20 most common cancer sites, based on the first ever cancer diagnosis recorded in each individual or combination of data sources between 2000 and 2014. We calculated positive predictive values and sensitivities for each case definition, compared to a gold standard algorithm that used information from all linked datasets to identify first cancers. We described completeness of grade and stage information in the cancer registration dataset.Results: 168,634 gold standard cancers were identified. Positive predictive values of all case definitions were &ge94% for the four most common cancers (breast, lung, colorectal, prostate) and &ge80% across cancer sites (oral cavity, oesophageal, stomach, liver, pancreas, malignant melanoma, cervix, uterus, ovaries, kidney, bladder, brain, thyroid, Non-Hodgkin lymphoma, multiple myeloma, leukaemia). Sensitivity for case definitions that used cancer registration alone or in combination was &ge92% for the four most common cancers and &ge80% across all cancer sites except bladder cancer, for which sensitivity was 65% using cancer registration alone. For case definitions using linked primary care, hospitalisation and death registration data, sensitivity was &ge89% for the four most common cancers, and &ge80% for all cancer sites except kidney (69%), oral cavity (76%) and ovarian cancer (78%). Sensitivities were generally lower when primary care or hospitalisation data were used alone. Completeness of staging data in cancer registration data was high for most cancers from 2012.Conclusions: Ascertainment of incident cancers was good when using cancer registration data alone or in combination with other datasets, and when using a combination of primary care, hospitalisation and death registration data, with variation between cancer sites.

PO-2490 Comparison Of A New Three-item Self-reported Measure Of Adherence To Medication With Pharmacy Claims Data 1400633 Julie Lauffenburger BWH Div of Pharmacoepidemiology and Pharmacoeconomics Spotlight Poster Session Spotlight Poster Adherence Background: Less than half of patients with cardiometabolic disease consistently take prescribed medications. While health insurers and some delivery organizations use insurer claims to measure adherence, most do not have rapid access to these data during routine interactions given administrative lag. Self-reported scales exist, but their practical utility is often limited by costs or length. By contrast, the accuracy of a new, free 3-item self-reported measure has been demonstrated in individuals with HIV but to our knowledge has not yet been compared with insurer claims or evaluated in other conditions.Objectives: To evaluate the concordance between this self-reported tool and claims-based adherence measures in patients with cardiometabolic disease.Methods: We used a linked pharmacy claims and electronic health record (EHR) dataset from a recently-completed pragmatic trial of patients with cardiometabolic conditions. After 12 months of follow-up, intervention subjects were mailed a survey with the 3-item measure that queries about medication use in the prior 30 days. As in prior research, responses were linearly transformed and averaged. Adherence was also measured in pharmacy claims in Month 12 and Months 1-12 of the trial using proportion of days covered (PDC) metrics. We compared sensitivity, specificity, and likelihood ratios (LR) for non-adherence between self-report (average<0.80) and claims (PDC<0.80) as the referent and varied the survey cutpoint for "non-adherence" as a sensitivity analysis. Results: Of 459 patients returning the survey (response rate: 43.5%), 50% were non- adherent in claims in Month 12, while 21% were non-adherent based on the survey. Compared with claims, the specificity of the 3-item metric was high (Month 12: 0.93), and sensitivity was relatively poor (Month 12: 0.25). Positive and negative LRs were 1.46 (95%CI: 1.01-2.09) and 0.91 (95%CI: 0.82-1.00), respectively. Survey measures were similarly accurate for longer (12 month) claims measurement windows (e.g., Sensitivity: 0.24 and Specificity: 0.87 in Months 1-12). Varying the cutpoint for non-adherence to <0.90 and <1.0 for the survey improved accuracy.Conclusions: A new 3-item self-reported adherence measure has high specificity but poor sensitivity for non-adherence compared to pharmacy claims in cardiometabolic disease. This tool may be useful for clinicians or healthcare organizations to identify those needing adherence support in the absence of rapid access to claims data, as any admission of non-perfect adherence could potentially prove actionable.

PO-2493 Comparative Long-term Cardiovascular Outcomes Between Ranibizumab Versus Aflibercept In Patients With Maculopathy 1444242 Wan-Ju Lee Comparative Long-term Cardiovascular Outcomes Between Ranibizumab Versus Aflibercept In Patients With Maculopathy Safety End Points -> Cardiovascular Background: Intravitreal injection of anti-vascular endothelial growth factors (anti-VEGFs) were effective for patients with maculopathy to maintain their vision. However, few reports indicated anti-VEGFs may associate with higher risk of cardiovascular (CV) events, but little is known about the risk profiles between different anti-VEGFs.Objectives: To compare the risk of CV events between anti-VEGFs (ranibizumab vs. aflibercept) in patients with maculopathy.Methods: We conducted a retrospective cohort study using Taiwan’s National Health Insurance Database from 2010 to 2017. We included adult patients (aged 20+ years) diagnosed with maculopathy newly receiving Intravitreal anti-VEGFs. Study primary outcomes were hospitalization due to ischemic stroke, myocardial infarction (MI) and transient ischemic attack (TIA). The secondary outcomes included hospitalization due to hemorrhagic stroke and all-cause mortality. We followed up patients for 2 years from initiation of ranibizumab or aflibercept. We performed Cox proportional hazard regression with adjustment of patient’s age, sex, comorbidities and co-medications to compare the risk of outcome events between ranibizumab vs. aflibercept.Results: We included 13,258 and 1972 patients receiving ranibizumab and aflibercept patients, respectively. The average age was 67.6 (SD 11.1) and 71.2 (SD 9.5) years old for ranibizumab and aflibercept users, respectively. We found the incidence rate of ischemic stroke, MI and TIA were 11.2, 7.7, 21.76 per 1000 person-year for ranibizumab users and 8.07, 0.78, 16.69 per 1000 person-year for aflibercept users. Compare to ranibizumab, the adjusted hazard ratio (AHR) of ischemic stroke, MI and TIA were 0.8 (95%CI, 0.5-1.0), 0.3 (95%CI, 0.1-0.8), and 0.8 (95%CI, 0.6-1.0) respectively. The risks of hemorrhagic stroke and all-cause mortality were similar between ranibizumab vs. aflibercept users. The incidence rate was 4.12 vs. 3.37 and 21.97 vs. 23.02 per 1000 person-year for hemorrhagic stroke and all-cause mortality for ranibizumab vs. aflibercept users respectively.Conclusions: We found patients receiving aflibercept posed lower risks of ischemic stroke, MI and TIA compared to ranibizumab patients. The finding warrants clinical considerations on patients’ cardiovascular outcomes while selecting intravitreal anti-VEGFs.

PO-2496 Using Natural Language Processing To Improve Patient Age Ascertainment From Adverse Event Reports 1400634 Phuong Pham University of Florida Using Natural Language Processing To Improve Patient Age Ascertainment From Adverse Event Reports Pharmacovigilance -> Other Background: Missing information in structured data fields is a significant challenge when conducting analyses of adverse event reports submitted to the FDA Adverse Event Reporting System (FAERS). Age is a critical data element because age-related physiological changes can affect the pharmacokinetics and pharmacodynamics of drugs. While demographic data is often missing in the report’s structured data fields, it may be found in the report’s free-text narrative. Natural language processing (NLP) has been leveraged to extract patient demographic information from unstructured clinical text in electronic health records. However, the potential for NLP to improve the identification of patient age from FAERS reports has not been previously evaluated.Objectives: Evaluate the performance of a rule-based NLP tool to extract patients’ age from the report narrative and its potential impact on addressing missing age in the structured data field.Methods: Using all FAERS reports received from January 1, 2002 through December 31, 2018, we evaluated the annual proportion of reports missing an age in the structured field before and after the application of the NLP tool. A random sample of 1500 reports was selected from the dataset for reviewers to manually identify patients’ age from the report narrative (gold standard). Reviewers were blinded to the structured field and NLP output. Each report was read by two different reviewers and disagreements were adjudicated by a third reviewer. The gold standard of manually extracted ages was then compared to the NLP output. Sensitivity, specificity, positive predictive value (PPV), and F-measure (a metric accounting for both sensitivity and PPV) were calculated.Results: Of 10,300,609 reports received in the study period, the overall percentage of reports missing age in the structured data field was 32%. The percentage of reports missing age in the structured field increased over time, from 21.9% in 2002 to 43.8% in 2018. The NLP tool achieved high performance, with sensitivity of 98.5%, specificity of 92.9%, PPV of 94.9%, and F-measure of 96.7%. When the NLP algorithm was applied to reports without age in the structured data field from 2002 to 2018, the tool identified age for an additional one million reports (10% of the total number of reports).Conclusions: NLP has potential utility to extract patients’ age from report narratives of postmarketing adverse event reports. The use of this tool would help facilitate pharmacovigilance and research using FAERS data.

PO-2498 The Efficacy And Safety Of Metformin For Insulin Resistance And Obesity In Children And Adolescents: A Systematic Review And Meta-analysis 1400635 Reem Masarwa McGill Dept of Epidemiology, Biostatistics and Occupational Health The Efficacy And Safety Of Metformin For Insulin Resistance And Obesity In Children And Adolescents: A Systematic Review And Meta-analysis Pediatric Pharmacoepidemiology Background: The worldwide prevalence of childhood obesity and its complications has markedly increased over the last several decades. Metformin is prescribed “off-label” for the treatment of obesity and insulin resistance in children and adolescents. However, the efficacy and safety of metformin remains unclear.Objectives: To compare the efficacy and safety of metformin in children and adolescents via systematic review of randomized controlled trials (RCTs) and observational studies and meta-analysis of RCTs, with a particular focus on body mass index (BMI), insulin resistance, and adverse effects.Methods: We systematically searched PubMed, Embase, the Cochrane Library, Scopus and ClincalTrials.gov from inception to November 2019. We included all published RCTs and observational studies assessing the efficacy and safety of metformin monotherapy compared to lifestyle interventions in children and adolescents (4-19 years) with obesity, prediabetes, diabetes, or non-alcoholic fatty liver disease. We used the ROB-2 and ROBINS-I tools to assess the quality of RCTs and observational studies, respectively. The primary outcomes were changes in BMI, BMI z-score, and homeostatic model assessment insulin resistance (HOMA-IR), and the occurrence of gastrointestinal (GI) adverse effects. Observational studies were included in the qualitative systematic review only. Data were pooled across RCTs using DerSimonian and Laird random-effects models with inverse variance weighting, and amount of heterogeneity present was estimated using the I2 statistic.Results: Our systematic review included 25 RCTs (1,705 participants) and 5 observational studies (301 participants). Ages ranged from 4 to 19 years, and follow-up duration ranged from 2 months to 2 years. Compared with lifestyle interventions, metformin was associated with a greater decrease in BMI (weighted mean difference [WMD]: −0.69 kg/m2, 95% confidence interval [CI] −1.28, −0.10, I2: 76%), BMI z-score (WMD: −0.15, 95% CI −0.27, −0.04, I2: 89%), and HOMA-IR (WMD: −0.92, 95% CI −1.71, −0.13, I2: 68%). Sensitivity analyses restricted to higher-quality studies produced similar results for BMI z-score (WMD: −0.16, 95% CI, −0.33, 0.01, I2: 86%), but attenuated effects for BMI (WMD: −0.25 kg/m2, 95% CI, −0.98, 0.45, I2: 77%) and HOMA-IR (WMD: −0.54, 95% CI, −1.77, 0.68, I2: 43%). Metformin was associated with an increased risk of GI adverse effects (relative risk: 1.84, 95% CI 1.37, 2.49, I2: 10%).Conclusions: Metformin therapy has modest but favorable effects on weight, and insulin resistance, and a tolerable safety profile.

PO-2504 The Magnitude Of The Warfarin-amiodarone Drug-drug Interaction Varies With Renal Function: A Propensity-matched Cohort Study. 1400636 Todd Miano University of Pennsylvania Spotlight Poster Session Spotlight Poster Drug-Drug Interaction Background: Amiodarone inhibits warfarin metabolism and is associated with major bleeding during warfarin therapy. Managing this drug-drug interaction (DDI) is challenging because of substantial interpatient variability in DDI magnitude. Because renal dysfunction induces changes in drug metabolism and protein binding that could alter cytochrome P450 inhibition mechanisms, we hypothesized that renal dysfunction alters the impact of the warfarin-amiodarone DDI.Objectives: To determine whether the warfarin-amiodarone drug-drug interaction varies with renal function in hospitalized patients with atrial fibrillation.Methods: We examined the interaction between renal function, warfarin, and amiodarone in a propensity-matched cohort study. Renal function was estimated with creatinine clearance (CrCL). Propensity score estimation and matching was stratified by baseline renal function. Warfarin response was measured with the warfarin sensitivity index (WSI), a dose-normalized international normalized ratio (INR) measure and was modeled with multilevel mixed-effects linear regression. Time to the first supratherapeutic INR (>4) was modeled using Cox regression.Results: Propensity score matching resulted in 4518 matched pairs. Renal dysfunction was observed in 54% of patients and the supratherapeutic INR rate was 17.9 per thousand-days. Amiodarone’s effect on warfarin response varied three-fold across the renal function range, increasing WSI by 36% in patients with normal renal function (CrCl 115 ml/min), but by only 11.8% in patients with severe renal dysfunction (CrCl 15 ml/min). Similarly, amiodarone had a strong effect in patients with normal renal function, HR 1.80 (1.23,2.64), but a negligible effect on supratherapeutic INR hazard in patients with severe renal dysfunction, HR 1.01 (0.75,1.37).Conclusions: Our results suggest that renal function is a novel factor that explains substantial variability in the warfarin-amiodarone drug-drug interaction. This information could inform warfarin dosage adjustment and monitoring, and may have implications for the selection of oral anticoagulation agents in patients treated with amiodarone.

PO-2505 Effect Of Renin-angiotensin System Inhibitors On The Comparative Renal Safety Of Nonsteroidal Anti-inflammatory Drugs And Opioids In Hospitalized Patients. 1400637 Todd Miano University of Pennsylvania Spotlight Poster Session Spotlight Poster Drug-Drug Interaction Background: Nonsteroidal anti-inflammatory drugs (NSAIDS) are increasingly important alternatives to opioids for acute pain control in hospitalized patients, as health systems implement opioid minimization initiatives. Increasing NSAID use may increase acute kidney injury (AKI) rates in this vulnerable population, particularly in patients with predisposing risk factors. Conflicting data in outpatient populations suggests that NSAID nephrotoxicity is magnified by renin-angiotensin system (RAS) inhibitors, but no studies have examined this question in hospitalized patients.Objectives: To determine whether the renal safety of short term treatment with NSAIDs vs. oxycodone is altered by concomitant RAS treatment in hospitalized patients.Methods: Retrospective, active-comparator cohort study of patients admitted to four hospitals in Philadelphia, PA. To minimize confounding by indication, NSAIDs were compared to oxycodone, and RAS inhibitors were compared to amlodipine. We tested synergistic interaction between NSAIDs and RAS by comparing the difference in AKI rate between NSAID vs. oxycodone in RAS exposed patients to the difference in AKI rate between NSAID vs. oxycodone in amlodipine exposed patients. In a secondary analysis, we restricted the cohort to patients with baseline diuretic treatment. AKI rates were adjusted for 78 baseline characteristics with inverse probability of treatment weighted Poisson regression.Results: The analysis included 25,571 patients who received a median of 2.4 days of analgesia treatment. The overall AKI rate was 23.6 per 1000 days. The rate difference (RD) for NSAIDs vs. oxycodone in amlodipine exposed patients was 4.1 per 1000 days (CI95 -2.8, 11.1), and the rate difference for NSAID vs. oxycodone in RAS exposed patients was 5.9 per 1000 days (CI95 1.9, 10.1), resulting in an interaction estimate consistent with negligible synergy: 1.8 excess AKI events per 1000 days (CI95 -6.2, 9.9). An analysis restricted to patients treated with baseline diuretics produced a stronger, albeit non-significant interaction estimate: 9.9 excess AKI events per 1000 days (CI95 -5.05, 24.83).Conclusions: This study did not observe synergistic nephrotoxicity with short term NSAIDs+RAS treatment in the absence of concomitant diuretics, suggesting that RAS treatment may not be a reason to choose an opioid in lieu of a NSAID. Clinically important synergistic effects of NSAID+RAS cannot be ruled out, however, in patients receiving concomitant diuretics.

PO-2506 A Weighting Method To Remove Bias From Within-subject Exposure Dependency In Case-crossover Studies 1400638 Kiyoshi Kubota NPO Drug Safety Research Unit Japan A Weighting Method To Remove Bias From Within-subject Exposure Dependency In Case-crossover Studies Methods in Pharmacoepidemiology -> Analytical Methods - e.g., marginal structural models, instrumental variables, sensitivity analyses Background: The case-crossover design with more than one control period may be biased from within-subject exposure dependency and/or time trends when analyzed by standard conditional logistic regression.Objectives: We propose a weighting method to remove bias from within-subject exposure dependency, with or without a time trend, by extending the method of Vines and Farrington in 2001.Methods: We modified Greenland's formula for case-crossover likelihood by weighting the exposure probability in the case period. This probability is estimated from cases when no time trend exists but from cases and time-controls if a time trend exists. The method was applied to real-world data from Japan and Australia. The Japanese study used claims data of corporate type health insurance plans. We selected 311 cases prescribed celecoxib who had peripheral edema (defined by new use of furosemide) and 1,555 time controls during 84, 168, and 336 day time windows. The Australian study used the Australian Department of Veterans’ Affairs administrative claims database. We re-analyzed case-control data on psychoactive medicines and hip fracture (Pharmacoepidemiol Drug Saf 2015; 24:576-82). We selected 1,316 cases prescribed selective serotonin reuptake inhibitors with hip fracture and 3,948 time-controls during a 180-day time window. Time windows were divided into 2 to 336 periods in the Japanese study and 2 to 180 periods in the Australian study. The data were analyzed by standard conditional logistic regression and our proposed weighting method, with and without time-controls.Results: The odds ratio (OR) estimated by the usual case-control design was 2.26 (95% confidence interval: 1.75-2.92) in the Japanese study and 1.32 (1.15-1.53) in the Australian study. After adjusting for time trends, standard conditional logistic regression in the case-crossover study overestimated the OR and bias increased with longer time windows. The ORs were overestimated as 4.74 (3.34-6.71) and 7.16 (5.27-9.74) with and without time-controls, respectively, with 1 day periods and a 336 day time window in the Japanese study. The ORs were overestimated as 1.75 (1.46-2.11) and 2.02 (1.71-2.37) with and without time-controls, respectively, with 1 day periods and a 180 day time window in the Australian study. Our proposed weighting method produced an OR close to the case-control OR which was stable when the exposure period or time window varied (1.88 to 2.40 Japan; 1.29 to 1.35 Australia).Conclusions: Using our method, case-crossover studies can be used for pharmacoepidemiology studies when exposure periods are autocorrelated or time trends exist.

PO-2507 Comparative Pharmacovigilance Analyses Of Pimavanserin And Mortality In Parkinson’S Disease 1400639 Joshua Brown University of Florida Comparative Pharmacovigilance Analyses Of Pimavanserin And Mortality In Parkinson’S Disease Pharmacovigilance -> Neurological/Mental Health Background: Pimavanserin is approved for treatment of Parkinson’s disease (PD)-related psychosis but has been associated with an increased risk of death. Media reports regarding this association have not taken into account limitations of the data source nor relevant comparisons to treatment alternatives.Objectives: This study aimed to conduct a comparative pharmacovigilance assessment of pimavanserin versus treatment alternatives.Methods: This was a retrospective analysis of adverse event case reports submitted to the U.S. Food and Drug Administration’s Adverse Event Reporting System (FAERS) from 2016 through Q3/2019.Reports were assessed for exposure to pimavanserin, clozapine, quetiapine, haloperidol, and other antipsychotics. PD-specific medications were used to restrict the base population to those with confirmed PD. The outcome of interest was all-cause death. A Proportional Reporting Ratio (PRR) and 95% confidence limits were calculated for each two-by-two contingency of outcome (death) and exposure (pimavanserin, etc.). For each outcome/exposure pair, the baseline population was altered to include the full FAERS sample, only reports with PD treated with levodopa, and only reports with PD treated with multiple medications. The sample was also stratified by time period before April 2018 and after September 2018 to capture periods of public knowledge and federal response. A lower confidence limit (LCL) of ≥2 was considered as the commonly accepted threshold for a drug safety signal.Results: As of Q3/2019, there were 2,287 reports of death associated with pimavanserin. In the full FAERS base population, pimavanserin yielded a PRR LCL=2.08 but was smaller when restricted to those with PD treated with levodopa (LCL=1.15) or multiple PD medications (LCL=1.63). Stratification by time showed a possible reporting bias associated with pimavanserin. Pimavanserin metrics were similar in direction and magnitude to clozapine and quetiapine.Conclusions: Compared in context to treatment alternatives, pimavanserin was not associated with excess risk of death in patients with PD. Given the severity of PD-related psychosis, treatment options are needed for shared decision-making between physicians and patients that balances both risks and benefits.

PO-2512 Comorbidities And Characteristics In A Cohort Of Patients With Nafld/nash From The Netherlands 1400640 Jetty Overbeek PHARMO Institute Comorbidities And Characteristics In A Cohort Of Patients With Nafld/nash From The Netherlands Disease Epidemiology/Clinical Course -> Other Background: The burden of non-alcoholic fatty liver disease (NAFLD) and non-alcoholic steatohepatitis (NASH) is increasing. However, limited real-world information is available regarding patients with NAFLD/NASH in Europe.Objectives: To create a dynamic cohort of patients suffering from NAFLD and/or NASH with detailed information to study the safety and effectiveness of future treatment options.Methods: All patients with a diagnosis of hepatic steatosis in the General Practitioner (GP) Database of the PHARMO Database Network in the period 2009-2018 were selected. The date of first recorded diagnosis was defined as the index date. All patients were followed from index date until end of study period, data availability or death, whichever occurred first. Patients were characterised in terms of demographics, laboratory measurements, comorbidities and medication use.Results: In total, 9,050 patients were included. Of these patients 54% was male and the mean age (±SD) was 55 (±14) years. The mean (±SD) BMI at index date was 31 (±5) kg/m2 and 43% was current or former smoker. The median (IQR) ALAT, ASAT and bilirubin value at index date were 51 (33-76) U/l, 35 (26-50) U/l, and 9 (7-12) µmol/L, respectively. The mean (±SD) lipid values at index date were 5.2 ± 1.2 mmol/l (total cholesterol), 3.0 ± 1.0 mmol/l (LDL), 1.2 ± 0.4 mmol/l (HDL), and 2.2 ± 1.5 mmol/l (triglycerides). The most common comorbidities were uncomplicated hypertension (2.9%) and type 2 diabetes (2.7%). In the year before the index date the most frequent reasons for GP visits were general complaints, such as abdominal pain (6%) and weakness/tiredness (6%). The most frequently prescribed drugs were omeprazole (26%), diclofenac (16%) and simvastatin (15%).Conclusions: This study from PHARMO’s GP database gave insight in the characteristics of a real-world cohort of NAFLD/NASH patients from the Netherlands. At index the majority of patients were obese and had elevated ALAT levels. Further criteria are needed to distinguish between NAFLD and NASH. This cohort makes it possible to study patients with NAFLD/NASH in a real-world setting.

PO-2513 Pain Severity And Analgesics Use In The Community-dwelling Older Population. A Drug Utilization Study From Germany 1400641 Ben Schoettker German cancer research center (DKFZ) Pain Severity And Analgesics Use In The Community-dwelling Older Population. A Drug Utilization Study From Germany Drug Utilization Research -> Other Background: Chronic pain is common in the older population and a significant public health concern. However, comprehensive studies on analgesics use in this age group from Germany are scarce.Objectives: This study aims to give a comprehensive overview on the prevalence of overall analgesics use and the prevalence of use of the most common therapeutic groups of analgesics in community-dwelling older adults from Germany.Methods: A cross-sectional study was carried out using data from a German cohort of 2038 community-dwelling adults aged 63-89 years. Descriptive statistics and logistic regression models were applied to assess the utilization of analgesics by age, sex, pain severity, pain duration and locations.Results: One out of four (25.0%) study participants was suffering from high intensity or disabling pain (grades II-IV). However, less than one out of four (22.9%) used any analgesic drugs, and about half (50.5%) of those still reported to suffer from high intensity or disabling pain (grades II-IV) despite the intake of analgesics. Among analgesics users, occasional non-steroid anti-inflammatory drugs (NSAIDs) use was the most frequent pain therapy (in 43.6% of user), followed by metamizole (dipyrone) use (16.1%), regular NSAIDs use (12.9%), strong opioids use (12.7%) and weak opioids use (12.0%). In multivariate logistic regression models, higher age, higher pain severity, longer pain duration, abdominal pain and back pain were statistically significantly associated with opioids use. Metamizole use was also statistically significantly associated with higher pain severity but inversely associated with pain duration.Conclusions: A significant number of older German adults are affected by high intensity and disabling chronic pain despite receiving analgesics. Long-term studies are needed to compare the effectiveness and safety of different treatments for chronic pain in older adults.

PO-2515 Using Electronic Data To Validate Acute Pancreatitis Diagnoses 1400642 Gregory Nichols Kaiser Permanente Using Electronic Data To Validate Acute Pancreatitis Diagnoses Health Economics/Outcomes Research Background: Recorded diagnoses of acute pancreatitis (AP) using ICD9/ICD10 codes only are often inaccurate in the United States resulting in limited utility for case identification.Objectives: To validate ICD9/ICD10 diagnoses of AP recorded in the electronic medical record (EMR) of an integrated health system and to determine whether other recorded data could be included as covariates in an algorithm to more accurately identify cases of AP in common US EMR and claims data.Methods: We identified random samples of potential cases of AP from inpatient diagnoses (any position) and from outpatient or emergency department diagnoses immediately preceding a hospitalization for any reason, and from hospitalizations for differential diagnoses (cholangitis and cholecystitis) in 2012-2018. We conducted an EMR review to confirm cases meeting the Atlanta Classification or cases with a clinician recorded note of AP. Using confirmed cases as the “gold standard”, we used logistic regression to develop EMR-based and claims-based algorithms using variables typically available in these data sources for identifying cases of acute pancreatitis. Algorithm performance was assessed with the C statistic, sensitivity, specificity, and positive and negative predictive value.Results: Of the 550 patients with an AP diagnosis from any source, 467 (84.9%) met the gold standard as did 5 of 150 patients (3.3%) with a differential diagnosis. An AP diagnosis from any source had high sensitivity (98.9%), modest specificity (63.6%), and a reasonably good C statistic (0.813). A simple EMR-based model attained a C-statistic of 0.913 using an AP diagnosis (from any source), body mass index > 30 kg/m2, serum lipase > 3 times upper limit of normal (ULN), and presence of diabetes. A simple claims-based model attained a C-statistic of 0.892 using an AP diagnosis (from any source) and dichotomous variables for whether a serum lipase test was performed and whether an abdominal ultrasound was performed; sensitivity was 89.8%, specificity 79.4%, and positive and negative predictive values of 90% and 79%, respectively. Use of ICD10 vs. ICD9 diagnoses appeared to only slightly improve case identification.Conclusions: Consistent with available studies, we found diagnoses of AP do not accurately identify confirmed cases in US data sources. Our results suggest that the addition of a few variables typically available in existing data sources can increase accurate identification of AP cases, producing an area-under-the-curve (C statistic) of approximately 0.90 with high levels of sensitivity and specificity, and positive and negative predictive value.

PO-2517 Almost Half Of Seniors Initiating Oral Bisphosphonate Therapy Are Exposed For 3 Or More Years - Understanding The Benefits And Harms Of Prolonged Therapy Is Needed 1400643 Kaley Hayes Almost Half Of Seniors Initiating Oral Bisphosphonate Therapy Are Exposed For 3 Or More Years - Understanding The Benefits And Harms Of Prolonged Therapy Is Needed Drug Utilization Research -> Quality use of medicines (including indicators and guidelines) Background: Oral bisphosphonates (BP, alendronate and risedronate) are the primary drugs used to treat osteoporosis in Canada. Clinical guidelines recommend at least 3 years of BP therapy to reduce fracture risk. The benefits and harms of continued therapy is uncertain. Rare but serious adverse events are associated with BP use >5 years.Objectives: To describe patterns of long-term (≥3 years) BP use.Methods: The Ontario government covers prescription drugs listed on the public formulary for residents aged ≥65 years. This includes alendronate and risedronate since 2000/11 for women and 2003/07 for men. We identified first dispensation of BP drugs among adults aged ≥66 years from 2000/11 to 2016/12, and followed them forward until death or 2019/12 to identify patients with ≥3 years of BP use (proportion of days covered ≥80%, using 3-year rolling window approach after each BP dispensation). Patterns of long-term BP use were described based on length of time between first BP dispensation and the start of long-term therapy period, and compared between men and women.Results: We identified 260,784 eligible seniors initiating BP therapy in Ontario, mean age=75.4 (SD 7.1) years, 77% women, 57% risedronate. Of these, 47% (n=121,786) met criteria for long-term therapy, mean age=75.0 (SD 6.7) years, 79% women, 55% risedronate. The mean length of BP therapy was 7.6 (SD 3.4, median 6.9, 99th percentile 16.6) years for women and 6.7 (SD 3.0, median 6.0, 99th percentile 15.1) years for men. Overall, 31% of patients (33% of women, 26% of men) were treated for ≥5 years. Eighty percent met criteria for long-term therapy after their first BP dispensation, 6% had a gap <12 months, and 14% had a gap of ≥12 months. Eleven percent of long-term therapy patients switched between alendronate and risedronate, and 35% switched between brand and generic formulations. The annual number of patients undergoing long-term BP therapy increased steadily from 2001 (n=3,079) until peaking in 2007 (n=15,649), then decreased annually thereafter, reaching n=5,783 in 2016.Conclusions: Almost half of patients initiating oral BP are exposed to long-term therapy, with 80% meeting a minimum 3 years of therapy after their first BP dispensation and 31% continuing therapy for at least 5 years. The benefits and harms of continuing BP therapy beyond 3 years are needed to inform the true effects of long-term exposure. Better understanding of switching between alendronate and risedronate and brand and generic formulations will inform studies that consider the relative effects of prolonged BP therapy.

PO-2518 Risk Of Hospitalisation Due To Infection In Patients With Psoriasis: A Population-based Cohort Study Using The United Kingdom Clinical Practice Research Datalink 1400644 Zenas Yiu The University of Manchester Risk Of Hospitalisation Due To Infection In Patients With Psoriasis: A Population-based Cohort Study Using The United Kingdom Clinical Practice Research Datalink Disease Epidemiology/Clinical Course -> Other Background: Psoriasis is associated with multiple comorbidities and treated with systemic therapies that may increase the risk of serious infections. The relationship between psoriasis and serious infection, however, is as yet unclear.Objectives: The primary objective was to determine whether patients with psoriasis have a higher risk of hospitalisation due to infection; the secondary objective was to determine whether patients with psoriasis have a higher risk of death due to infection compared with matched patients without psoriasis from primary care.Methods: We performed a cohort study of adult patients (≥18 years of age ) with psoriasis delineated from the UK Clinical Practice Research Datalink (CPRD GOLD) and linked to Hospital Episode Statistics (HES) and national mortality records between 01/04/2003 and 31/12/2016. Each patient with psoriasis was matched to up to 6 individuals without psoriasis on age, sex, and primary care practice. Hospitalisation due to infection was ascertained in the linked HES records. Unadjusted and adjusted stratified Cox proportional hazard models were estimated, with the adjusted model inclusive of potential confounders such as lifestyle factors and comorbid conditions.Results: 69,312 patients with psoriasis and 338,598 comparison patients were followed up for a median (inter-quartile range [IQR]) of 4.9 (5.9) and 5.1 (6.3) years respectively. Patients with psoriasis had a higher incidence rate of serious infection (20.5/1000 person-years, 95% CI 20.0-21.0, n=7629) compared with those without psoriasis (16.1/1000 person-years, 95% CI 15.9-16.3, n=30756). The unadjusted hazard ratio for serious infection in patients with psoriasis was 1.46 (95% CI 1.42-1.50), and the fully adjusted hazard ratio was 1.36 (95% CI 1.31-1.40). The unadjusted hazard ratio for death due to infection in patients with psoriasis was 1.30 (95% CI 1.18-1.43), and the fully adjusted hazard ratio was 1.33 (95% CI 1.08-1.63).Conclusions: Psoriasis is associated with a small increase in the risk of serious infection. Further research is needed to understand the mechanism by which psoriasis predisposes to a higher risk of infection.

PO-2520 Anti-hypertensive Drug Prescribing Pattern, Adherence To Guidelines And Blood Pressure Control At A Teaching Hospital In Ghana 1400645 Mark Amankwa Harrison Korle bu Teaching Hospital Anti-hypertensive Drug Prescribing Pattern, Adherence To Guidelines And Blood Pressure Control At A Teaching Hospital In Ghana Drug Utilization Research -> Quality use of medicines (including indicators and guidelines) Background: Blood pressure (BP) control is low in many low and middle income countries, despite various therapy options. Current guidelines can help achieve outcomes improvement. Therefore, guideline adherence needs to be studied.Objectives: To evaluate the pattern of anti-hypertensive (anti-HPT) drug prescribing, its adherence to hypertension treatment guidelines by clinicians at the Korle Bu teaching hospital (KBTH) and association with blood pressure control.Methods: Over a 3-month period, we carried out a cross-sectional study of hypertensive out-patients attending the KBTH polyclinic for treatment. After patients were selected, their medical notes were reviewed to collect data on: patient demographics and BP; the pattern of anti-hypertensive drug class and combination prescribing; level of generic and single pill combination (SPC) prescribing. Adherence to guidelines was measured as prescribing pattern conformity to recommendations by the 2018 European Society of cardiology (ESC) and the 2017 Ghana Standard Treatment Guidelines (STG).Results: Out of 310 questionnaires 304 were completed, and 31.6% were diabetic. Monotherapy and dual therapy were 18.8% (n=57) and 52.6% (n=160) respectively. The 5 leading drugs, calcium channel blockers (CCB), diuretics, angiotensin receptor blockers (ARB), angiotensin-converting enzyme inhibitors (ACE-I), and beta blockers were given to 86.8% (n=264), 45.7% (n=139), 33.2% (n=101), 27% (n=82), and 11.8% (n=36) of patients respectively. Majority (71.4%; n=217) were on amlodipine. For 2 drugs, 50% (n=80) were CCB plus a renin-angiotensin system (RAS) blocker. The most common were CCB plus diuretic (37.1%), CCB plus ACE-I (23.8%) and CCB plus ARB (26.2%). SPC and generic prescriptions were 2.8% (n=7) and 84.5% (n=256) respectively. BP control was 41.8%. The 5 leading drugs agreed with 1st line recommendations of the 2017 Ghana STG. The 5 leading drugs and pattern of combination therapy agreed with 2018 guidelines of the ESC. The number of anti-HPT drugs was negatively associated with BP control with significance. (B= -0.402; p-value=0.015). Amlodipine, indapamide and hydrochlorothiazide were more likely to result in BP control (OR=1.53, 95% CI: 0.91 - 2.58; OR=1.7, 95% CI: 1.54 - 1.89; OR=1.7, 95% CI: 1.55 - 1.87).Conclusions: Anti-hypertensive drug prescribing was based on the 5 major drug classes, and the majority of patients received 2 drugs (CCB and RAS blockers). Prescribing pattern agreed with the Ghana and ESC guidelines but SPC prescribing needs to be improved. CCB and diuretics are more likely to control BP in Ghanaians.

PO-2522 Proton Pump Inhibitors Are Not Associated With An Increased Risk Of Colorectal Cancer 1400646 Josephina Kuiper PHARMO Institute Proton Pump Inhibitors Are Not Associated With An Increased Risk Of Colorectal Cancer Safety End Points -> Cancer Background: The clinical relevance of proton pump inhibitors use as a risk factor for colorectal cancer in humans is still unclear.Objectives: To investigate the risk of colorectal cancer after the use of proton pump inhibitors within a linked Cancer Registry-General Practitioner Database cohort.Methods: Colorectal cancer patients diagnosed between 2007-2014 with at least six years of primary care data available prior to diagnosis (index date) were identified and matched to four controls on gender, birth year, general practice and period of primary care data availability. Proton pump inhibitor use was determined in the six years prior to the index date and analysed with conditional logistic regression, adjusted for potential confounders.Results: 1,041 cases (53%) and 3,751 controls (47%) ever used a proton pump inhibitor, yielding an odds ratio (OR) of 1.08 (95% confidence interval (CI) 0.97-1.21). Current use showed the highest OR (1.30 (95% CI 1.16-1.47). Long-term use of proton pump inhibitors (≥4 years) was not associated with colorectal cancer (OR 0.92 (95% CI 0.76-1.11)). A variation in OR for tumour stage and tumour subsite was observed with the highest OR for stage I tumour (1.28 (95% CI 0.87 - 1.87) and proximal colon (1.19 95% CI 0.87-1.63).Conclusions: No increased risk in colorectal cancer was seen with the use of proton pump inhibitors. Current use was associated with an increased likelihood to be diagnosed with colorectal cancer, but this is likely the result of reverse causality. No increased risk was seen for long-term use of proton pump inhibitors.

PO-2527 Treatment Patterns Of Patients With Major Depressive Disorder And Suicidal Ideation Or Attempt 1400648 David Kern Janssen Research & Development Treatment Patterns Of Patients With Major Depressive Disorder And Suicidal Ideation Or Attempt Disease Epidemiology/Clinical Course -> Neurological/Mental Health Background: Patients diagnosed with major depressive disorder (MDD) and having suicidal ideation comprise approximately one-quarter of those with MDD, yet little is known about how they are treated.Objectives: To understand the treatment patterns of medications and procedures for subjects with MDD and suicidal ideation (SI) or suicide attempts (MDD+SI/SA) 1 year following their first occurrence of SI/SA.Methods: We identified patients diagnosed with MDD and a diagnosis for conditions related to SI/SA from a large US claims database of commercially insured individuals during 1/1/2014 to 6/30/2019. The index date was the first observed medical claim in the database for SI/SA. Patients were required to have continuous enrollment in the database ≥1 year prior to the index date, with no minimum follow-up required. Patients with evidence of autism, mania, psychosis or dementia any time prior to index were excluded. Treatment patterns included procedures for electroconvulsive therapy (ECT) and transcranial magnetic stimulation (TMS), and medications of selective serotonin reuptake inhibitors (SSRI), serotonin and norepinephrine reuptake inhibitors, tricyclic antidepressants, monoamine oxidase inhibitors, other antidepressants (mainly bupropion and trazodone), anxiolytics, hypnotics/sedatives, lithium, psychostimulants, and atypical antipsychotics. Changes in class-based regimens were determined according to switching or adding of a new medication class to an existing regimen. Treatment patterns were captured up to 1 year following the index SI/SA.Results: We identified 49,575 patients diagnosed with MDD+SI/SA. Patients were followed for 265 days on average, with 51% having a full year of observation. During this time, 75% received ≥1 treatment of interest. Of the treated, 57% received ≥2 distinct regimens, while 32% received ≥3 and 17% received 4 or more. Use of multiple therapy classes simultaneously was common, occurring in 62% of patients. There was heterogeneity in treatments received, with SSRI alone the most common first treatment (17%) followed by other antidepressants (10%), anxiolytics (9%), antipsychotics (7%) and anticonvulsants (6%). Use of ECT (0.9%) and TMS (0.3%) any time during follow-up was rare.Conclusions: Of patients with MDD+SI/SA receiving pharmacotherapy or ECT/TMS, nearly two-thirds received multiple medication classes simultaneously at some point in the year following their first recorded SI/SA event. Use of medications other than antidepressants during this time was relatively common practice. Switching or adding-on classes was also frequent.

PO-2528 Selecting An Appropriate Comparator Group In Vaccine Cohort Studies Of Older Adults: Recombinant Zoster Vaccine As An Example 1400649 Jennifer Han GSK Vaccines S.r.l. Selecting An Appropriate Comparator Group In Vaccine Cohort Studies Of Older Adults: Recombinant Zoster Vaccine As An Example Vaccines Background: Observational studies are critical for assessing effectiveness and safety of vaccines in the real-world setting. However, substantive differences between individuals who do and do not receive the vaccine can lead to a number of biases. Selection of an appropriate comparator group is an important study design consideration, especially in studies of older adults with multiple comorbidities.Objectives: To identify the most appropriate comparator group for a cohort study assessing safety outcomes in recipients of the two-dose adjuvanted Recombinant Zoster Vaccine (RZV) aged ≥ 50 years.Methods: Curated administrative data in the Sentinel common data model format from 4 health plan participants in the US Food and Drug Administration’s Sentinel Initiative were used to select commercially-insured adults ≥ 50 years of age with medical and drug coverage from 1/2018-1/2019. The exposed group was individuals who received ≥ 1 RZV dose, an adjuvanted subunit vaccine approved in the US for the prevention of herpes zoster in people aged ≥ 50 years. Six different comparator selection algorithms were evaluated, using various combinations of receiving ≥ 1 healthcare utilization service: annual physical exam; age-appropriate screening; annual exam or screening; age-appropriate vaccination; annual exam and influenza vaccination; and outpatient encounter for chronic condition management. Demographics, healthcare-seeking behaviors, and the prevalence of comorbidities and immunosuppressive conditions for the RZV recipient group and the comparator groups were descriptively compared.Results: 261,050 and 133,256 individuals received ≥ 1 and two doses of RZV, respectively, during the study period. The number of individuals with ≥ 1 comparator visit ranged from 1,666,326 to 6,698,184, depending on the definition. The mean age for each of the comparator groups was similar to that of RZV recipients (68 years, standard deviation, 9). However, there was substantial variation in the prevalence of some comorbidities across the six comparator groups compared to RZV recipients. Adults who received a routine annual physical exam or an age-appropriate preventive screening measure were one of the most comparable unexposed groups compared to RZV recipients.Conclusions: Differences among potential comparator groups highlight the importance of the rigorous selection of a comparator group in cohort studies evaluating vaccine effectiveness or safety, especially in older adults who have multiple comorbidities.

PO-2530 The State Of Real-world Evidence In Comparative Effectiveness And Safety Studies Based On Healthcare Insurance Claims 1400650 Katsiaryna Bykov BWH Div of Pharmacoepidemiology and Pharmacoeconomics The State Of Real-world Evidence In Comparative Effectiveness And Safety Studies Based On Healthcare Insurance Claims Methods in Pharmacoepidemiology -> Study Design - Studies assessing study design (e.g., choice of comparator, self-controlled methods, avoiding immortal time bias) Background: There is growing interest in broader use of real-world data (RWD) to inform decision-making, however, concerns about the validity of RWD analyses continue to detract from their utility. Some of these concerns may be attributable to mistakes in study design and analysis, but the relative contribution of these avoidable problems is not known.Objectives: To evaluate the quality of observational studies on the comparative safety and effectiveness of medications conducted using US health insurance claims data.Methods: Using a systematic PubMed search (January 2010 - May 2019), we identified published observational studies (cohort or case-control design) that were conducted using US healthcare claims and evaluated the comparative effectiveness or safety of medications in the therapeutic areas of cardiovascular diseases, diabetes, asthma or chronic obstructive pulmonary disease, and osteoporosis. A random sample of 75 studies (10 case-control and 65 cohort studies) was selected for evaluation. We developed a questionnaire to identify major methodological issues in published real-world evidence (RWE) studies, focusing on time-related biases, adjustment for causal intermediaries, biases due to depletion of susceptible individuals, reverse causation, residual confounding, differential surveillance (detection bias), and informative censoring, as well as exposure and outcome misclassification. Each study was evaluated by two reviewers; discordance was resolved through consensus within the overall team.Results: We present interim results for 10 case-control and 30 cohort studies; complete results will be available by the time of presentation. Potential for time-related biases was the most common methodological issue and was identified in 60% of cohort studies (immortal person-time) and 80% of case-control studies (time-window bias). Half of the reviewed studies (52%) adjusted for variables measured during follow-up without appropriate statistical methods; 47% had the potential for bias due to depletion of susceptible individuals, 37% due to reverse causation, 25% due to detection bias, and 10% due to informative censoring. In only 45% of studies a new user design was implemented; 45% used an active comparator.Conclusions: The prevalence of well-known design and analysis flaws that may lead to biased results is high among RWE studies on medication safety and effectiveness. Recognizing and avoiding known methodological pitfalls could substantially improve the quality and validity of RWD studies, ultimately increasing the confidence in evidence generated from them.

PO-2535 Improving Identification Of Anaphylaxis 1400651 David Carrell Kaiser Permanente Washington Health Research Institute Improving Identification Of Anaphylaxis Safety End Points -> Other Background: Anaphylaxis is an acute life-threatening illness that is often misidentified by ICD diagnostic codes. This threatens the validity of claims-based epidemiologic studies of anaphylaxis as an adverse drug event.Objectives: (1) Conduct physician adjudication of potential ICD-10 based anaphylaxis events to identify genuine cases. (2) Improve the accuracy of claims-based anaphylaxis identification using natural language processing (NLP) of unstructured clinical notes and machine learning methods.Methods: We obtained medical records with diagnosis codes for anaphylaxis from inpatient (IP), emergency department (ED), and outpatient (OP) encounters in an integrated healthcare system in Washington State from October 2015 to December 2018. Eligible OP events were also required to have modifier codes suggesting presence of an acute condition. Potential drug-related events (ICD-10 code T88.6) were over-sampled. Two physicians performed adjudication using established events criteria; disagreements were resolved by discussion. We also reviewed IP and ED encounters with diagnosis codes for allergic reactions and adverse drug reactions to identify further potential anaphylaxis cases.Results: Out of 247 potential events sampled, 240 (97%) had medical records adequate for adjudication (162 ED/IP, 78 OP). The overall positive predictive value (PPV) for validated anaphylaxis events was 64.6% (95% CI 58-70%) with minimal heterogeneity by setting (ED/IP vs. OP), sex, race, or specific ICD-10 code. PPV was higher among those under age 40 (72%, 95% CI 63-79%) versus age 40 or higher (58%, 95% CI 49-66%, p=0.02). Among validated events, common causes were food (39%), medications (34%), and insect bite or sting (12%). Among false positives, 94% were considered to be another serious allergic reaction. Clinical context (timing, alternative clinical conditions, information source) was important in making the proper classifications for some cases. Adjudication of 76 encounters identified by allergy and adverse drug reaction codes yielded a single anaphylaxis event (PPV 1.3%, 95% CI 0.0-9.0%). Toward improving identification of actual anaphylaxis events, we are currently developing machine learning models using predictors from structured electronic health record (EHR) data and NLP-extracted information from the clinical notes of these 316 adjudicated cases.Conclusions: ICD-10 diagnosis codes for anaphylaxis had moderate PPV for validated events. Many of the misclassified events were serious allergic reactions. These findings have implications for pharmacoepidemiologic studies that seek to estimate treatment-related risks of anaphylaxis using EHR data.

PO-2536 Prescription Drugs Among Dementia Patients With Behavioral Disturbance 1400652 Wenjun Zhong Merck & Co., Inc. Prescription Drugs Among Dementia Patients With Behavioral Disturbance Geriatric Pharmacoepidemiology Background: Behavioral disturbance (BD) in dementia patients is common and requires many healthcare resources, but there have been no FDA approved medications for this indication.Objectives: This study examined real-world medication use in dementia patients with BD to inform future new drug investigational studies.Methods: A cross-sectional study was conducted using Truven MarketScan Commercial and Medicare Supplemental database for outpatient medication use in Oct 2015-Dec 2018. Patients with two outpatient or one inpatient dementia diagnosis were identified through ICD-10-CM codes, and the first event of BD (including agitation, aggression, psychosis, delirium, and wandering) at or after the dementia diagnosis was identified through ICD-10-CM codes as the index BD. Medications prescribed before (days’ supply should cover BD index date) and after BD index date (within 6 months) were identified through National Drug Code. All the medications used in this population were counted and the frequencies were reported both in therapeutic classes and in generic names.Results: Among 232,921 patients with dementia during Oct 2015-Dec 2018, 83,024 (36%) had at least one BD event after dementia diagnosis during the same time period. Of these BD patients, the mean (standard deviation) age was 81 (13) years at the time of index BD (90% of the population were >=65 years, and 50% were >=85 years); about 60% were female. Among them, 85% (n=70,481) received prescriptions before or after the index BD event. Top 5 prescription classes were antidepressants (60%), antihyperlipidemic drugs (45%), beta blockers (40%), parasympathomimetic drugs (39%) and tranquilizers/antipsychotic drugs (37%). Top 5 drugs prescribed before BD index event were donepezil (20%), levothyroxine (18%), atorvastatin (16%), memantine (15%), and amlodipine (13%). Top 5 drugs prescribed after BD index event (within 6 months) were donepezil (27%), levothyroxine (21%), memantine (21%), atorvastatin (20%), and quetiapine (19%). Other central nervous system drugs (top 5 CNS) prescribed after BD index event were sertraline (13%), mirtazapine (12%), trazodone (12%), lorazepam (11%) and escitalopram (10%).Conclusions: Patients with dementia and BD were an older population. A large proportion of them were exposed to prescription medications including CNS drugs such as antidepressants and antipsychotics. There is an unmet medical need in this population since this patient population is susceptible to the side effects of antidepressants and antipsychotics. This study provided real-world medication use data for future new drug investigational studies for dementia patients with BD.

PO-2541 The Influenza Vaccination Continuity In Korean Elderly With High-risk Conditions 1400653 Na-Young Jeong Ewha Womans University The Influenza Vaccination Continuity In Korean Elderly With High-risk Conditions Vaccines Background: Since different flu viruses circulate every year and change during flu season, influenza vaccination is needed every year. In particular, the continuity of influenza vaccination is essential to the people at high risk for complications in the case of influenza infection such as elderly and patients with specific diseases.Objectives: To evaluate the continuity of influenza vaccination during the 2015/16 to 2017/18 flu seasons for elderly with high-risk conditions.Methods: We used linked databases of the Korea Centers for Disease Control & Prevention’s vaccination registry, which contains all National Immunization Program vaccination records, and the National Health Insurance Service claims data. The patients with high-risk conditions were defined as those who visited ≥1 in inpatient setting or ≥3 in outpatient setting for the influenza season, Sep 30 to Apr 30, and the preceding 8 months with diseases specified by the Centers for Disease Control and Prevention Advisory Committee on Immunization Practices as high-risk groups. Those who died before Apr 30, 2018 were excluded. The proportion of vaccination for each season were calculated, and the association of vaccination continuity and the types or the number of diseases were assessed by using logistic regression.Results: A total of 1,617,546 elderly who had at least one high-risk disease and at least one flu vaccination during the three seasons were identified. Among them, 1,247,297 (77.11%) were vaccinated in all three seasons. In patients with one high-risk condition in all three seasons, people with pulmonary disease, cardiac disease, or diabetes had higher odds of vaccination continuity (Pulmonary: 1.22, 1.20-1.23; Cardiac: 1.15, 1.13-1.16; Diabetes: 1.15, 1.14-1.17), while patients with liver disease, renal disease, Immunodeficiency disease, or cancer were less likely to vaccinated successively. People with more than one high-risk condition during all flu seasons are likely to be vaccinated continuously than those with only one condition (1.06, 1.04-1.08).Conclusions: We observed that of all high-risk elderly were vaccinated in all three seasons, which means the vaccination continuity were high. Because this study used only 3 seasons from the 2015/16 flu season, further studies using more influenza seasons are needed.

PO-2543 Community Pharmacy Practice Evaluation In The United Arab Emirates Using Simulated Patients With Common Cold 1400654 Subish Palaian Ajman University Community Pharmacy Practice Evaluation In The United Arab Emirates Using Simulated Patients With Common Cold Drug Utilization Research -> Quality use of medicines (including indicators and guidelines) Background: Patients seek medical advice from pharmacists regarding minor ailments. Pharmacists on the other hand are expected to counsel patients appropriately and dispense medications rationally. Simulated client method using ‘Mystery patients’ provides opportunity to investigate and explore actual dispensing practices (including client-provider interactions).Objectives: To explore the dispensing and counseling practices of community pharmacists in United Arab Emirates.Methods: A simulated client study was conducted among a conveniently selected 34 community pharmacies in Ajman, UAE in January 2020. Three simulated clients (third year pharmacy students) visited community pharmacies and exhibited symptoms of common cold (ICD-10 code J00). Uniformity was maintained while exhibiting symptoms across pharmacies and each of them exchanged their roles routinely across encounters. Ethical approval was obtained for the research and pilot tested with three pharmacies. Filled checklists were analyzed and statistical analysis performed using Pearson correlation coefficient and independent t-test at α =0.05.Results: The Simulated clients’ interaction was with 9 female and 25 male pharmacists, five Arabs versus 29 non-Arabs. All pharmacists dispensed at least one medicine (average 3.08; n=105). Most common therapeutic category was antihistamines (21.90%; n= 23), followed by expectorants (16.19%; n=17). Commonly dispensed medicines were pseudoephedrine+Loratidine combination (n=8; 7.61%), followed by guaifenesin+pseudoephedrine+dextromethorphan+paracetamol in 6 (5.71%) cases. Only 12 pharmacists counseled on ‘treatment duration’ and none asked about ‘medical history’. The labeling ‘when to take’ was mentioned by 20 (58.82%) pharmacists with the average dispensing time of 37.20± 25.97 seconds. Dosage regimen was appropriately chosen in 22 (64.70%) pharmacies. The average cost of medicines per patient was AED 55.32±36.93 (1 USD= 3.67 AED). Only one pharmacist counseled on medication allergy, 4(11.76%) on side effects, 2(5.88%) on contraindications, 3(8.82%) on drug-drug interactions and 31(91.17%) pharmacists provided appropriate instructions on medications usage. The average counseling time was 75.29 ± 74.62 seconds. Counseling duration had a positive correlation with cost (p=0.015), dispensing duration (p=0.000) and associated with race (p=0.003) but not gender (p= 0.8).Conclusions: Pharmacists spent less time in understanding patients’ condition and counseling. They were predominantly selling medicines with little or no information on safe and effective usage.

PO-2550 Ability Of Primary Healthcare Records To Assess Medicinal Products Discussed By The Pharmacovigilance Risk Assessment Committee 1444186 Robert Flynn Ability Of Primary Healthcare Records To Assess Medicinal Products Discussed By The Pharmacovigilance Risk Assessment Committee Drug Utilization Research -> Health policy and governance Background: The EU’s Pharmacovigilance Risk Assessment Committee (PRAC) is responsible for protecting patients by ensuring the safety of marketed medicines. The European Medicines Agency (EMA) supports such evaluation by analysing routinely collected Primary care Health Records (PHRs) to assess drug utilisation, risks of medicines and effectiveness of risk minimisation measures.Objectives: To evaluate the capture of exposure to medicines discussed at PRAC in four established PHR databases.Methods: Four PHR databases were searched for substances discussed at consecutive PRAC meetings from September to November 2018: IQVIA Medical Research Data (IMRD) - France, IMRD - Germany, IMRD - United Kingdom and the Clinical Practice Research Datalink Aurum. Data extracted were: number of prescriptions, number of patients with at least one prescription, whether a Centrally (CAP) or Nationally (NAP) Authorised Product and authorisation date. Exposures were assessed up to 31 August 2018. To estimate the number of substances each database could meaningfully assess adverse events, we calculated the numbers of patient exposures required to detect a statistically significant event associated with a theoretical relative risk of 2.0, with adverse event rates taken from the Summary of Product Characteristics frequency categories. This was based on a hypothetical comparison of two proportions using a 2-sided Fisher test with α = 0.05, power = 0.90 and equal numbers of exposed / comparator patients.Results: There were a total of 128 drug substances (or classes): 83 CAPs (64.8%) and 45 NAPs (35.2%). Of CAPs, 6 (7.2%) had been authorised for 5 years. Where marketed (n = 41), all NAPs had been authorised for > 5 years. No individual database was able to detect a doubling in effect size for the majority of CAPs for common (1/100), uncommon (1/1000), rare (1/10000) or very rare (1/100000) events, or NAPs for uncommon, rare or very rare events. An analysis of pooled allowed the detection of an uncommon adverse event associated with a doubling of risk for 26.5% of CAPs and 60.0% of NAPs. For events that were rare or very rare the proportion was lower (CAPs: 4.8% and 0%; NAPs: 40.0% and 11.1%). Coverage was worse for “orphaned” substance.Conclusions: More advanced and recently authorised medicinal products had less coverage in the available databases compared to more established, nationally licensed products. To enable better informed regulatory decisions there is a need to access complementary data sources, particularly capturing exposure in specialist care settings.

PO-2551 Hypertensive Disorders Of Pregnancy And Use Of Antihypertensive Medications In The 2015 Pelotas Birth Cohort Study 1400655 Lisiane Leal McGill University, Department of Epidemiology, Biostatistics, and Occupational Health Hypertensive Disorders Of Pregnancy And Use Of Antihypertensive Medications In The 2015 Pelotas Birth Cohort Study Pregnancy and Lactation Background: Hypertensive disorders of pregnancy (HDP) affect approximately 140,000 women in Brazil. Pharmacotherapies play an important role in reducing the occurrence of adverse outcomes. However, the patterns of use of antihypertensives among women with HDP living in Brazil is unclear.Objectives: To describe the use of antihypertensive agents among women with HDP in the Pelotas Birth Cohort. To compare the patterns of use, by trimester, between women with and without HDP.Methods: We used data from the 2015 Pelotas (Brazil) Birth Cohort, one of the largest birth cohorts of women living in low and middle-income countries. HDP was classified as (i) chronic hypertension; (ii) preeclampsia superimposed on chronic hypertension; (iii) gestational hypertension and (iv) preeclampsia-eclampsia. Antihypertensive medications were defined using the 4th level of the Anatomical Therapeutic Chemical Classification System and the substance name. Descriptive statistics were used to describe this population. Patterns of use of antihypertensive medications were presented by trimester of use.Results: Of the 4,270 eligible women in the cohort, 4,262 were included. HDP was present in 1,336 (31.35%) of women in the cohort. The majority of women with HDP were white (66.47%), had between 9-11 years of education (38.25%), and were of lower-middle socioeconomic class (54.12%). Comorbidities were slightly higher in women with HDP, compared to women without an HDP and these women were more likely to be hospitalized (30.16%). Among women with HDP, 636 (47.60%) reported having gestational hypertension, 409 (30.61%) chronic hypertension, 191 (14.30%) preeclampsia, and 89 (6.66%) preeclampsia superimposed on chronic hypertension. Approximately 71% of women with HDP reported not using antihypertensive medication during pregnancy. Among all antihypertensives reported by women with HDP, methyldopa in monotherapy was the most frequent (16%), regardless of the type of HDP. Methyldopa uses increased from 11.10-18.79% over the course of pregnancy for women with HDP, while those without HDP were found to increase their use of ‘Other lipid modifying agents’ (1.03-4.27%).Conclusions: In the Pelotas Birth Cohort, women with HDP were found to have higher rates of comorbidities and hospital admissions. Patterns of use of methyldopa were in-line with the Brazilian guidelines as the first-line therapy for HDP. However, a large number of women with HDP not using antihypertensive agents require further investigation.

PO-2552 Operational Considerations And Challenges In Use Of An External Comparator Arm To Contextualize Safety Data From Maternal Immunization Trials: The South Africa Epidemiology Of Obstetric And Neonatal Outcomes (SAFE ONE) Study Example 1400656 Vera Frajzyngier Pfizer Operational Considerations And Challenges In Use Of An External Comparator Arm To Contextualize Safety Data From Maternal Immunization Trials: The South Africa Epidemiology Of Obstetric And Neonatal Outcomes (SAFE ONE) Study Example Vaccines Background: Immunization of pregnant women to protect their infants has been shown to be safe and effective, with several vaccines undergoing clinical trials. The availability of expected rates of obstetric and neonatal outcomes in an external comparator cohort helps to contextualize safety signals observed in vaccine development programs and facilitates recognition of true safety concerns.Objectives: To describe operational challenges and considerations encountered during a retrospective chart review study estimating the incidence of obstetric and neonatal outcomes among pregnant women and their infants at 3 clinical trial sites (and associated clinics/hospitals) in South Africa.Methods: An electronic composite of individual birth register files from each institution in a region served as a sampling framework, with random sampling of charts proportional to the expected regional recruitment in the clinical trial. Inclusion and exclusion criteria replicated those of the vaccine trial. Data from paper-based or electronic maternity case records were abstracted onto a shared REDCap database. Outcomes were defined using Global Alignment of Immunization Safety Assessment in Pregnancy standardized definitions, and necessitated capture of detailed clinical data spanning antenatal care to post-delivery admissions.Results: Charts for 9414 women and their newborns were abstracted. Among 10,600 files originally sampled <10% were not located initially and were replaced by subsequent files on a random number list. Potential reasons for missingness included patient referrals to other facilities, filing specific outcomes (e.g. stillbirth) in separate locations, and misfiling. Charts of patients referred to tertiary facilities and filed separately were tracked and abstracted where feasible. Neonatal charts were often systematically missing; bias mitigation necessitated obtaining data from unplanned data sources across various hospital units. Benefits of access to electronically scanned files were offset by misordered and missing pages. Complex pregnancies and/or deliveries required clinically experienced abstractors and additional quality control measures.Conclusions: Despite careful planning and pilot testing, unanticipated challenges were encountered in the review of both electronically scanned and paper-based source documents. Lessons learned from the SAFE ONE study can inform the design of future chart review-based comparator cohorts in LMIC.

PO-2554 Systematic Evaluation Of The Efficacy-effectiveness Gap Of Systemic Treatments In Extensive Disease Small Cell Lung Cancer 1400657 Christine Cramer Systematic Evaluation Of The Efficacy-effectiveness Gap Of Systemic Treatments In Extensive Disease Small Cell Lung Cancer Drug Effectiveness Background: The recommendations in treatment standards are conventionally based on clinical trial data, in which patient populations are studied that are not necessarily a reflection of the general population seen in clinical practice. Important patient characteristics predictive for treatment response are often underrepresented in clinical trial populations.Objectives: The aim of this study is to assess how clinical outcomes in real-world (effectiveness) correspond to the outcomes in clinical trials (efficacy) of systemic treatments for ED SCLC, and to search for patient factors that may explain a gap.Methods: All patients diagnosed with ED SCLC between 2008 and 2014 in six Dutch large teaching hospitals (Santeon network) were identified and followed-up from date of diagnosis until death or end of data collection. For every patient, an efficacy-effectiveness factor (EE factor) was calculated by dividing individual patients’ overall survival (OS) by the pooled median OS assessed from clinical trials with the respective first-line treatment. A multivariable linear regression analysis was applied to study the association between patient and treatment characteristics and the magnitude of the efficacy-effectiveness gap.Results: From 792 diagnosed patients, 568 (72%) started with first-line treatment. For all studied regimens (carboplatin-etoposide, cisplatin-etoposide, and cyclophosphamide-doxorubicin-etoposide), the median OS in real-world is shorter than the clinical trial reference median OS. Overall, the median EE factor was 0.79 (p<0.001 from 1.00). Poor performance status (ECOG≥2) and a higher age at diagnosis (age≥65 years) were independent predictors for a lower EE factor. The EE gap was 43% in patients with both age ≥65 years and ECOG ≥2 (EE factor 0.57). The mean age and the proportion of patients with ECOG≥2 in real-world were different from those in clinical trials (mean age of 66 versus 62 years, and ECOG≥2 25% versus 17%; both p<0.001). In patients with ECOG≥2 and aged ≥65 years in real-world, early discontinuation of treatment and no subsequent line of treatment were more prevalent.Conclusions: Overall survival of patients with ED SCLC treated with systemic therapy in real-world practice is 21% shorter than for patients included in trials. Differences in age at diagnosis and performance status explain this gap.

PO-2555 Co-occurrence Of Birth Defects Among Infants With Neonatal Abstinence Syndrome, 2012-2016 1400658 Jennita Reefhuis U.S. Centers for Disease Control and Prevention Co-occurrence Of Birth Defects Among Infants With Neonatal Abstinence Syndrome, 2012-2016 Pregnancy and Lactation Background: Prenatal opioid use has been associated with neonatal abstinence syndrome (NAS) and certain birth defects, but few data exist on their co-occurrence.Objectives: We compared the prevalence of birth defects among newborns with and without NAS.Methods: We used the 2012-2016 Agency for Healthcare Research and Quality’s Healthcare Cost and Utilization Project’s National Inpatient Sample, a nationally representative sample of de-identified hospital discharge records. Among the 30 diagnosis codes per discharge, we identified (1) newborns using International Statistical Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes V30.XX-V39.XX (1/2012-9/2015 discharges) and ICD-10 code Z38.XX (10/2015-12/2016); NAS using ICD-9 code 779.5 (excluding potential iatrogenic NAS: 765.00-765.05, 770.7, 772.1X, 779.7, 777.5X, 777.6) and ICD-10 code P96.1; and birth defects using Salemi et al. (Birth Defects Research. 2019;111:1365-1379). We calculated the prevalence of specific birth defects (for those with >10 unweighted discharges) among newborns with and without NAS and corresponding prevalence ratios (PR) and 95% confidence intervals (CI), accounting for sample weights.Results: We identified 3,795,785 discharges representing over 18.9 million newborn hospitalizations during 2012-2016. NAS and/or a birth defect were diagnosed in 6.1 and 22.5 per 1,000 newborns, respectively. The prevalence of 11 of 14 specific birth defects with sufficient sample size was significantly higher among newborns with NAS than without NAS, with the strongest associations for pulmonary valve atresia/stenosis (PR=4.2, CI: 3.2-5.4), cleft palate (PR=4.5, CI: 3.5-5.9), and cloacal exstrophy (PR=4.6, CI: 3.3-6.4).Conclusions: While these increases in prevalence of birth defects among infants with NAS might be related to prenatal opioid exposure, they could also result from common underlying causes of NAS and birth defects or variations in birth defect ascertainment among infants with and without NAS using billing records and across ICD coding transitions.

PO-2556 Trends In Gabapentinoid Prescriptions And Coprescription With Opioids In The Four Uk Nations 1437786 Alvi Rahman McGill Dept of Epidemiology, Biostatistics and Occupational Health Trends In Gabapentinoid Prescriptions And Coprescription With Opioids In The Four Uk Nations Drug Utilization Research -> Trends and comparisons Background: Gabapentinoid drugs, namely gabapentin and pregabalin, were approved in the United Kingdom in 1993 and 2004, respectively, for the treatment of epilepsy, neuropathic pain and generalised anxiety disorders (pregabalin). In the last decade, the rate of patients newly treated with gabapentinoids has tripled but the potential differences in prescription and coprescription with opioids between the four UK nations is unknown. Objectives: To estimate the annual rate of patients newly prescribed gabapentin and pregabalin in each UK nation (England, Northern Ireland, Scotland, and Wales) and the proportion of patients concomitantly prescribed opioids. Methods: We used the Clinical Practice Research Datalink, a UK primary care database, to identify patients registered between 1993 and 2017. We fitted Poisson regression models to estimate the annual rate (per 100,000 person-years (PY)) of patients newly prescribed gabapentin and pregabalin in each UK nation, and the rate ratio (RR) from 2007 to 2017. We also calculated the proportion of patients with same-day coprescription with opioids. Results: The annual rate of patients newly prescribed pregabalin increased steadily during the study period. Between 2007 and 2017, the rate rose from 118 to 351 per 100,000 PY (RR: 2.97; 95% CI: 2.85-3.11) in England, from 96 to 418 per 100,000 PY (RR: 4.37; 95% CI: 3.96-4.82) in Scotland, and from 104 to 370 per 100,000 PY (RR: 3.55; 95% CI: 3.23-3.91) in Wales. However, in Northern Ireland, we observed a rapid escalation from 546 (95% CI: 511-582) per 100,000 PY in 2007 to 1,139 (95% CI: 1,088-1,193) per 100,000 PY in 2010, at which point the rate was 5 times higher than in the other nations. This peak was followed by a decline to 532 (95% CI: 493-573) per 100,000 PY in 2017, but the rate remained higher than in the other nations. We observed a regular increase in patients newly prescribed gabapentin in England, Scotland, and Wales, which stabilized from 2015 onwards. In Northern Ireland however, following a plateau from 2005 to 2009, the rate of first gabapentin use sharply increased and surpassed rates in the other nations in 2017. The proportion of concomitant prescription of opioids remained relatively constant through the study period. In 2017, 16 to 19% of patients newly prescribed gabapentinoids in each of the four nations were concomitantly prescribed opioids. Conclusions: Since their introduction in the UK, the use of gabapentinoids has notably increased in all nations, more so in Northern Ireland with an unprecedented peak in pregabalin prescriptions. A substantial proportion of patients were concomitantly prescribed opioids in all nations.

PO-2557 Medication Error Reporting By African Countries To Vigibase 1400659 George Sabblah Food and Drugs Authority, Ghana Medication Error Reporting By African Countries To Vigibase Pharmacovigilance -> Other Background: Thirty-seven African countries are full members of the World Health Organisation (WHO) Programme for International Drug Monitoring (PIDM) and may contribute medication error (ME) reports to VigiBase, the WHO global database of individual case safety reports. Currently, there is limited information on the number of ME reports submitted by these countries, the type of medicines contributing to these errors, the characteristics of these events and the type of reporters. These pieces of information will help in formulating educational strategies and interventions to prevent harm associated with MEs.Objectives: The objective of this study is to determine the frequency of ME reports submitted by African countries to VigiBase and to describe their characteristics.Methods: ME cases from African countries in VigiBase from 1st January 1997 to 31st December 2018 were retrieved in January 2020 using the narrow Standard Medical Dictionary for Regulatory Activities (MedDRA) Queries. Descriptive statistics were used to summarize key variables of interest, namely, number of reports submitted by each country; year the reports were first submitted to VigiBase; the source of reports; reporter type; most reported events classified at the preferred term (PT) level and most reported medicine classified by the fourth level ATC codes. Spearman's ρ was used to determine the relationship between the number of reports and the year of joining PIDM.Results: A total of 3,388 ME reports were retrieved from VigiBase, which is 0.4% of all ME reports in the database. Out of the 37 full members of the PIDM, only 15 contributed ME reports with 3,338 (98.5%) of these reports submitted by three countries; Morocco (56.8%), Egypt (24.6%) and South Africa (17.1%). There was no statistically significant relationship between the year of joining PIDM and number of reports (r= 0.44, p=0.0973). The majority of reports were from spontaneous reporting system, 2,825 (83.5%) and studies 540 (15.9%). Patients submitted 1326 (30.1%) of the ME reports. Top five medicine classes with most frequently reported MEs were Insulins and analogues, Immunostimulants, Other analgesics and antipyretics, Direct acting antivirals and Immunosuppressants. The top five reported MedDRA PTs in decreasing order were Incorrect dose administered, Inappropriate schedule of product administration, Product administration error, Product dose omission and Medication error. Conclusions: There is a relatively small number of ME reports from a limited number of African countries. Low reporting puts patients at risk of experiencing preventable MEs. Low rate of reporting from patients’ needs further investigation.

PO-2558 Assessment Of Parents Knowledge Attitude And Practice Towards Vaccines That Are Not Covered Under Expanded Program On Immunization In India 1400660 Juny Sebastian JSS College of Pharmacy Assessment Of Parents Knowledge Attitude And Practice Towards Vaccines That Are Not Covered Under Expanded Program On Immunization In India Pediatric Pharmacoepidemiology Background: Vaccines are safe, simple and one of the most cost-effective way to save the lives of children worldwide. In India emphasis is given to vaccines that are covered by the Expanded Program on Immunisation (EPI) even though India accounts for a large number of childhood deaths that can be prevented by proper utilisation of vaccines that are not covered by the EPI.Objectives: To assess the knowledge, attitude and practice of parents towards Rotavirus, Pneumococcal, Influenza and Hepatitis A vaccines and to assess the factors contributing for non-vaccination.Methods: A cross- sectional observational study using a validated questionnaire was conducted on parents of children between three to five years of age visiting the paediatric department of JSS Hospital, Mysuru. Median split method was used to classify the knowledge level to adequate or inadequate. Chi square test was used for the statistical analysis of the study.Results: Out of 207 study participants, children of 97.58% parents were fully immunized according to the EPI schedule, but the non- EPI vaccine coverage was 22.34%. Among the study population, 41.88% of mothers and 63.82% fathers found to have inadequate knowledge of non-EPI vaccines and the study participants belong to upper socioeconomic class demonstrated better knowledge (78.3%) where as lower class demonstrated least (5.29%) knowledge. The higher rate of vaccination was observed with Hepatitis A (37.19%) followed by rotavirus vaccine (26.57%) among the vaccines under study. The knowledge level in the parents who have not vaccinated their children with non-EPI vaccines was found to be inadequate when compared with parents in the vaccinated group (76.36%). Lack of knowledge about non EPI vaccines (65.7%) among parents were the major constraint for not vaccinating their children. There was a significant association between the non-EPI vaccination coverage and urban residential area (p<0.01), having two or more kids (p<0.05), educational status of the parents(p<0.05), socio-economic class of upper/upper lower/upper middle (p<0.05) and adequate knowledge on non-EPI vaccines(p<0.05).Conclusions: Most of the sampled parents were unaware of the non-EPI vaccines. As the parental knowledge, attitude and practice about non-EPI vaccines are important determining factors for the immunization status of their child, future efforts by providing education on non-EPI vaccines are required to improve the parents confidence on non-EPI vaccines and to decrease the non- EPI vaccine hesitancy.

PO-2559 Second-generation Antipsychotics And The Risk Of Chronic Kidney Disease: A Population-based Case-control Study 1400661 Mikkel Hoejlund University of Southern Denmark Second-generation Antipsychotics And The Risk Of Chronic Kidney Disease: A Population-based Case-control Study Safety End Points -> Other Background: Second-generation antipsychotics (SGA) are associated with development of metabolic disturbances which can lead to cardiovascular and renal disease.Objectives: To examine the association between use of SGAs and the risk of chronic kidney disease (CKD).Methods: We conducted a population-based case-control study by identifying 21,434 cases of incident CKD from 2001 to 2016 in the Funen Laboratory Cohort containing more than 7.7 million creatinine samples. Cases were identified based on creatinine measurements and matched by risk-set sampling to four population controls using age, sex, and calendar year. We obtained information on drug exposure and comorbidities from the Danish National Prescription Register and the Danish National Patient Register. We calculated odds ratios (OR) for the association between SGA use and CKD using conditional logistic regression.Results: Use of SGAs was associated with increased risk of CKD among ever-users (OR 1.24, 95%CI: 1.12-1.37) and current users (OR 1.26, 95%CI: 1.12-1.42). We found no clear evidence of dose-response-relationship. Both short duration (1-2 antipsychotic prescriptions; OR 1.22, 95%CI: 1.01-1.48), as well as long-term use (>30 prescriptions; OR 1.45, 95%CI 1.19-1.76) were associated with an increased risk of CKD. Both use of SGAs with mild and high risk of metabolic disturbances was associated with increased risk of CKD (Mild risk: OR 1.21, 95%CI: 1.06-1.39 and high risk OR 1.36, 95%CI: 1.11-1.68). Recent use of NSAIDs, prior use of lithium, hypertension, or prior AKI were not clearly associated with development of CKD in connection to SGA exposure. All SGAs, except for aripiprazole, were associated with increased risk of CKD, and the highest risk was found for clozapine (OR 1.81, 95%CI: 1.22-2.69).Conclusions: Use of SGA is associated with a small-to-moderately increased risk of incident CKD. All investigated second-generation antipsychotics, except for aripiprazole, were associated with an increased risk of CKD.

PO-2560 Assessing The Risk Of Non-melanoma Skin Cancer Associated With Hydrochlorothiazide Use In Hypertensive Patients : A Nested Case-control Study In Korea 1400662 Bora Yeon Korea Institute of Drug Safety & Risk Management Spotlight Poster Session Spotlight Poster AsPEN Background: Hydrochlorothiazide (HCTZ) is one of the most frequently prescribed antihypertensive drugs. There have recently been safety issues associated with the use of HCTZ and skin cancer.Objectives: To evaluate the association between HCTZ use and the risk of non-melanoma skin cancer (NMSC) in patients with hypertension.Methods: We performed a nested case-control study using the National Health Insurance claim data from 2002 to 2017. As an initial cohort, we included the patients with hypertensive disorder (defined by ICD-10: I10-I15) older than 30 years, and excluded the patients with diagnosed NMSC (ICD-10: C44) or prescribed HCTZ for one year prior to first diagnosis of hypertension (entry date). We identified patients (case) who diagnosed with NMSC during 2008-2017 in the cohort. For each case, we used risk-set sampling and randomly matched 4 population controls by same sex, age and entry date. We excluded those who were diagnosed with cancer or HIV and those prescribed immunosuppressants before first diagnosis of NMSC (index date). HCTZ use considered from entry date to 2 years prior to the index date. Using conditional logistic regression models, we computed the odds ratio (OR) for the association of NMSC with HCTZ use adjusted for the potential confounders. In addition, we conducted the subgroup analysis on the risk of developing NMSC based on HCTZ cumulative dose and long-term use.Results: We identified 10,565 NMSC patients with matched 42,260 controls and among them, HCTZ users were 4,098(38.8%) and 15,984(37.8%), respectively. There was no statistically significant associated with the use of HCTZ and the risk of NMSC (OR: 1.02, 95% CI: 0.98-1.07). As a result of analyzing the cumulative dose and long-term use of HCTZ, we found no statistically differences.Conclusions: HCTZ use is not associated with an increased risk of non-melanoma skin cancer in Korean people. However, additional studies are needed to confirm the association between HCTZ and non-melanoma skin cancer, as previous studies in Asian populations are limited. Also, healthcare professionals should be aware of NMSC risk associated with HCTZ among Caucasian population and regularly monitor if suspicious skin lesion occurs.

PO-2562 Twenty Years Of Modified Prescription- Event Monitoring: Summary Of Studies Conducted Between 2000 And 2020 1400663 Vicki Osborne Drug Safety Research Unit Twenty Years Of Modified Prescription- Event Monitoring: Summary Of Studies Conducted Between 2000 And 2020 Pharmacovigilance -> Other Background: Modified Prescription-Event Monitoring (M-PEM) is the only national system in England developed solely to prospectively monitor the utilisation and safety of recently marketed medicines available to all primary care physicians (GPs). It is not a spontaneous reporting system and is a pharmacoepidemiological method based on the fundamental concept of monitoring events regardless of relatedness to drug exposure. Where drugs are of the same class, direct comparisons between drugs are possible. Alternatively, within cohort comparisons are possible using 1) different exposure time periods and 2) pre and post exposure time periods.Objectives: We aimed to summarise the studies that have been conducted using M-PEM methodology between 2000 and 2020, to quantify the contribution of this method to pharmacoepidemiology and pharmacovigilance.Methods: M-PEM uses an observational cohort design in England. All dispensed primary care National Health Service prescriptions for a target medication are obtained for the required study period. Data collection forms (DCFs) are sent to GPs for individual patients dispensed these prescriptions. These are sent after a pre-specified study observation period following the date of dispensing (e.g.12 months) and request data on drug utilisation (DU) characteristics and events during the observation period. Summary descriptive statistics were calculated for completed M-PEM studies.Results: Over the past 20 years, 20 M-PEM studies were completed on 19 different medications. These included 3 inhaled corticosteroids, 2 antipsychotics, 2 antidiabetics and 2 opioids. A variety of different therapeutic areas were covered including 7 central nervous system, 4 respiratory and 3 cardiovascular medications. Median response rate for M-PEM DCFs was 55.3% (IQR: 44.8, 62.1%). Median cohort size was 5687 patients (IQR: 1554, 11316), with a minimum cohort size of 63 patients and a maximum cohort size of 17546 patients. After stratification by sex, median number of males was 3371 (IQR: 822, 5219) and median number of females was 2390 (IQR: 602, 6834). In total, M-PEM contributed DU and safety data on 132078 patients in the post-marketing setting.Conclusions: M-PEM is a prospective cohort methodology that can collect DU and safety data on medicines prescribed in primary care, covering a wide variety of therapeutic areas. Over the past 20 years, M-PEM has collected data on large numbers of patients, with response rates higher than other GP postal surveys. Overall, M-PEM has contributed important data for the post-marketing assessment of medicines.

PO-2565 Estimating Patient Satisfaction From Disease Blog Texts On The Internet: A Pilot Study For Patient-centric Pharmacovigilance 1400664 Shinichi Matsuda Chugai Pharmaceutical Spotlight Poster Session Spotlight Poster Digital Epi Background: While adverse event reporting databases and healthcare databases have played a central role in the current pharmacovigilance system, it is faced with a challenge how to obtain feedbacks directly from patients who are the main stakeholders for healthcare services. As an example of conventional researches to receive the feedbacks, there is a survey using a QOL questionnaire. However, because most of the QOL questions are developed from the perspective of healthcare professionals, it tends to focus only on the pre-specified intentions of the questioners. Although such questionnaires partly evaluate patient satisfaction, it may overlook important patient opinions. Therefore, we thought that an analysis based on the data of patients' actual narratives can enhance the understanding of whether patients are satisfied or unsatisfied with the treatments. Using disease blog texts on the Internet, we conducted a pilot study to estimate patient satisfaction on their treatments.Objectives: To visualize and evaluate patient satisfaction by applying natural language processing and machine learning approaches to patient-written disease blog texts.Methods: We analyzed textual data collected from the Japanese disease blogs on the Internet: tōbyōki blogs, which translate literally as a diary-like account of a struggle with disease (Matsuda, et al. JMIR Public Health Surveill 2017). Using supervised machine learning approach, we not only created a model to estimate patient satisfaction index (PSI) based on the patients’ texts but also evaluated the validity of the model.Results: Our preliminary model showed moderately good result in classifying patient satisfaction (positive or negative) for each sentence. For instance, the sentence “During that time, the nurse was stroking my back all the time.” showed positive PSI (+0.4375); in contrast, the sentence “And next is muscle injection, so it hurts again.” showed negative PSI (-0.99805) as expected. These results suggested that PSI becomes a useful indicator to identify the transition of patient satisfaction before and after the critical event of the patient journey such as disease notification or initiation of the drug treatment.Conclusions: Patient satisfaction could be estimated based on the text data written by patients. Patient-generated data on the Internet should be a promising data source to implement patient-centric pharmacovigilance.

PO-2566 The Effect Of The Indicated Use Of β-blockers On The Risk Of Exacerbations In Chronic Obstructive Pulmonary Disease: The Rotterdam Study 1400665 Leila Karimi Gazafroudi Erasmus University Medical Center The Effect Of The Indicated Use Of β-blockers On The Risk Of Exacerbations In Chronic Obstructive Pulmonary Disease: The Rotterdam Study Drug Effectiveness Background: While several observational studies have reported a reduced risk of exacerbations in COPD patients treated with β-blockers (BB), physicians are reluctant to prescribe BB in patients with cardiovascular disease and concomitant COPD. Recently, the BLOCK COPD randomized clinical trial (RCT) reported an increased risk of COPD exacerbations in metoprolol treated COPD patients without an indication for the use of BB. The current discrepancy between findings of observational studies and the BLOCK COPD RCT might result in further reduction of prescribing BB in COPD patients with a cardiac indication.Objectives: We investigated whether the association between use of BB and risk of COPD exacerbations differed between patients with and without an indication for treatment with BB.Methods: Within the Rotterdam Study, a prospective population-based cohort study, we followed COPD subjects until the first COPD exacerbation or end of follow-up. COPD exacerbations were defined as acute episodes of worsening of respiratory symptoms requiring systemic corticosteroids and/or antibiotics or hospitalization. Individuals were considered to have an indication for treatment with BB in case of having hypertension, coronary heart disease, atrial fibrillation, or heart failure at baseline. The association between BB use, as a time-varying variable, and COPD exacerbations was assessed using Cox proportional hazards models adjusted for age, sex, smoking, incident cardiovascular disease (i.e., heart failure, hypertension, atrial fibrillation, and coronary heart disease) during follow-up, respiratory drugs, and nitrates.Results: In total 1,312 COPD patients with a mean age=69.7±9.2 were included. In patients with an indication for treatment with BB (n=755, mean age=70.4±8.8), current use of cardioselective BB was significantly associated with a reduced risk of COPD exacerbations (Hazard Ratio (HR)=0.69, 95% Confidence Interval (CI): 0.57-0.85). In contrast, in subjects without an indication for treatment with BB (n=557, mean age=68.8±9.7), use of cardioselective BB was not associated with an altered risk of COPD exacerbations (HR=0.94, 95% CI: 0.55-1.62).Conclusions: In the Rotterdam Study, use of cardioselective BB reduced the risk of exacerbations only in COPD patients with concomitant cardiovascular disease. Our results could explain the discordant findings between the BLOCK COPD RCT (excluding patients with a therapeutic indication for the use of BB) and previous observational studies.

PO-2567 Time to Onset Analysis of Montelukast Associated Suicidal Ideation Using Data From FDA Adverse Event Reporting System Database (FAERS) 1400666 Subeesh Kulangara Viswam Faculty of Pharmacy, M.S. Ramaiah University of Applied Sciences Time to Onset Analysis of Montelukast Associated Suicidal Ideation Using Data From FDA Adverse Event Reporting System Database (FAERS) Pharmacovigilance -> Neurological/Mental Health Background: Montelukast is used for the treatment of chronic asthma and for the symptomatic relief of allergic rhinitis. The association of montelukast use and suicidal ideation is being discussed in the recent past. However, the time-to-onset for such event is seldom investigated.Objectives: To examine the time-to-onset profile of montelukast associated suicidal ideation using Food and Drug Administration Adverse Event Reporting System (FAERS) database.Methods: A systematic data mining was performed in FAERS Database (Q4/2003-Q3/2019) using OpenVigil 2.1, a pharmacovigilance analytical tool. Cumulative incidences of suicidal ideation were assessed using the Kaplan‐Meier method and Log‐rank test was used to compare the cumulative incidences of suicidal ideation between patients treated with montelukast and those treated with all the other drugs. Time‐to‐onset profiles were analyzed using the Weibull shape parameter test. A larger value of scale parameter stretches the distribution, whereas a smaller value shrinks the distribution. shape parameter 1, the hazard was considered to have early failure‐type profile, random failure type profile and wear‐out failure‐type profile respectively. Shape and scale parameter of montelukast was also compared with all the other drugs reported for suicidal ideation.Results: We identified a total of 30,408 reports for suicidal ideation of which 1,531(5%) were attributed to montelukast. The median time‐to‐onset of the event associated with montelukast was 8 (range: 3-31) days. Patients treated with other drugs showed almost similar time‐to‐onset, 6 days (range: 3-21). The cumulative incidence of suicidal ideation significantly differed between patients treated with montelukast and other drugs (log‐rank test, P < 0.002). Montelukast associated suicidal ideation (shape:0.453, scale:37.1) and other drug‐associated suicidal ideation (shape: 0.48, scale: 17.1) indicated early failure‐type profile, which means the event is more of acute and early onset in nature. However, the time to onset data is more widely distributed in both instances.Conclusions: Montelukast associated suicidal ideation had an onset profile that was similar to that of other drugs. As montelukast associated suicidal ideation is more of acute and early onset in nature, attention should be more focused on the earlier stages of treatment especially in patients with baseline risk.

PO-2568 Ten Years Of Specialist Cohort Event Monitoring: Summary Of Studies Conducted Between 2010 And 2020 1400667 Vicki Osborne Drug Safety Research Unit Ten Years Of Specialist Cohort Event Monitoring: Summary Of Studies Conducted Between 2010 And 2020 Pharmacovigilance -> Other Background: Specialist Cohort Event Monitoring (SCEM) is a national system in the UK developed to prospectively monitor the utilisation and safety of recently marketed medicines available in specialist care settings (e.g. hospitals and specialists’ clinics). SCEM are formal pharmacoepidemiological studies; events are monitored regardless of relatedness to drug exposure. SCEM can collect information on both the study drug of interest and a reference (contextual) cohort of patients using another drug (i.e. standard of care), to allow contextualisation of drug utilisation (DU) characteristics.Objectives: We aimed to summarise the studies that have been conducted using SCEM methodology between 2010 and 2020, to quantify the contribution of this method to pharmacoepidemiology and pharmacovigilance.Methods: SCEM uses an observational cohort design in the UK. Prescribers are identified through clinical research networks and patients are subsequently identified by specialist prescribers. Informed consent is obtained for participation. Research personnel are used to aid with recruitment of prescribers, although only prescribers have any contact with patients. Data are sampled at national level and research personnel recruit in all areas of the UK. Data collection forms (DCFs) are sent to prescribers for individual patients. These are sent at baseline and after a pre-specified study observation period following the medication start date (e.g. 12 weeks). Data is collected on DU characteristics and events during the observation period. Summary descriptive statistics were calculated for completed SCEM studies.Results: Over the past 10 years, 4 SCEM studies were completed between 2010 and 2020 on 5 different medications (cohort sizes n=124, 125, 528, 845, 2067 and 2542; 1 medication was examined in 2 different studies for different indications). Medications studied were 2 antipsychotics, 2 oral anticoagulants and 1 antiplatelet. Median number of prescribers for these studies was 235 (IQR:43, 1196). Median completion rate for SCEM DCFs (completion of both forms) was 98.9% (IQR:98.1, 100.0%). After stratification by sex, median number of males was 339 (IQR:103, 1149) and median number of females was 443 (IQR:62, 918). In total, SCEM have contributed DU and safety data on 6231 patients in the post-marketing setting.Conclusions: SCEM is a prospective cohort methodology that can collect DU and safety data on medicines prescribed in specialist settings. Over the past 10 years, SCEM has collected data on large numbers of patients and has contributed important data for the post-marketing assessment of medicines.

PO-2569 Incidence Of Anemia In Patients With Chronic Kidney Disease - A Population-based Cohort Study 1400668 S�ren Vestergaard Aarhus University Hospital Psychiatry Incidence Of Anemia In Patients With Chronic Kidney Disease - A Population-based Cohort Study Disease Epidemiology/Clinical Course -> Other Background: Anemia is common in patients with advanced chronic kidney disease (CKD) and is associated with increased mortality risk.Objectives: We conducted a population-based cohort study to clarify the incidence of anemia in patients with newly progressed CKD stages 3-5.Methods: Using a population-based registry with complete laboratory test results available for the entire population of Northern Denmark (~1.9 million people) during 2009-2016, we identified cohorts of patients with newly progressed CKD (stages 3-5). CKD was defined by two creatinine tests ≥90 days apart both revealing an estimated glomerular filtration rate (eGFR, MDRD) 10 g/dL), moderate (Hb 8-10 g/dL), or severe (Hb <8 g/dL) anemia. We computed cumulative incidence proportions (risks) of anemia by CKD stage, examined anemia trajectories and calculated adjusted hazard ratios (HRs) of anemia by CKD stage, using multivariable Cox regressions with a robust sandwich estimator.Results: We identified 55,940 individuals with newly progressed CKD in stages 3-5, who over time contributed to CKD stage cohorts [n=41,958 patients in 3a, n=17,875 in 3b, n=5,182 in 4, and n=931 in 5]. Within 1 year 36.1% (95% CI: 35.7 - 36.6) of patients with CKD 3a developed anemia, increasing to 62.3% (95% CI: 61.7 - 62.9) within 5 years. Anemia risks were higher in higher CKD stages, as 80.8% (95% CI: 80.1 - 81.5) of CKD 3b patients, 90.6% (95% CI: 89.6 - 91.5) of CKD 4 patients, and 98.8% (95% CI: 97.8 - 99.4) of CKD 5 patients developed anemia within 5 years. Cox regression showed increases in anemia hazards with increasing CKD stage [adjusted HRs of 1.55 (95% CI: 1.52 - 1.58) in CKD 3b; 2.38 (95% CI: 2.30 - 2.47) in CKD 4; and 5.42 (95% CI: 5.09 - 5.77) in CKD 5, compared to CKD 3a. One in five patients with any CKD stage 3-5, and half of all patients with CKD 5 experienced severe anemia (Hb <8 g/dL) during total 8 years of follow-up.Conclusions: A substantial proportion of patients with newly progressed CKD develop anemia during the subsequent year, with increasing anemia risk at higher CKD stages. Clinicians should be aware of the high risk of anemia in patients with CKD stage 3-5 in everyday clinical care and exercise vigilance about associated complications.

PO-2574 Incidence Of Anorexia Nervosa And The Prescribing Of Psychotropic Medications From Uk Primary Care, 1996-2016: A Retrospective Database Study 1444346 Maedeh Y Beykloo Incidence Of Anorexia Nervosa And The Prescribing Of Psychotropic Medications From Uk Primary Care, 1996-2016: A Retrospective Database Study Drug Utilization Research -> Trends and comparisons Background: Anorexia nervosa (AN) is a psychiatric disorder characterised by restrictive eating and an intense fear of weight gain (1). The most recent study reporting the incidence of AN was conducted using primary care records from 2000 to 2009 and found the incidence to be stable by gender despite minor fluctuations across the years (2). There is currently a lack in evidence for the efficacy of psychotropic medications in AN, however, they are sometimes prescribed if there is a less than optimal treatment response with first line psychological treatments.Objectives: This study aimed to comprehensively assess trends in the incidence of AN in order to provide up-to-date results and to measure the use of psychotropic medications in individuals with AN over two decades.Methods: A retrospective database study was conducted using The Health Improvement Network (THIN) database which consists of anonymised medical records from general practices throughout the UK. Individuals aged 12-50 years and with recorded AN diagnosis were identified between 1996 and 2016. The incidence of AN were calculated annually and Kaplan Meier analyses were used to measure the prescribing duration and cessation of psychotropic medications in this cohort.Results: A total of 17,597 individuals were identified from 1996-2016, of which 80.28% were females. The incidence of AN remained fairly constant for females and males from an overall incidence of 15.21/100,000 person years in 1996, to 21.84/100,000 person years in 2006, followed by a decrease in 2016 to 18.25/100,000 person years. Almost half of individuals with AN (47.44%) received a prescription for psychotropic medication in the study period. Most individuals receiving medication were prescribed antidepressants (98.13%), whilst under a fifth (17.29%) were prescribed antipsychotics, which is in line with our previous study (3). The most common antidepressant prescribed was fluoxetine (59.73%) and the most common antipsychotic was olanzapine (8.90%). Under half of individuals (46%) initiated on a psychotropic medication continued with a psychotropic prescription after six months.Conclusions: This study provides the most up-to-date trends for AN incidence and highlights the use of psychotropic medications for the treatment of AN in UK primary care, despite lack of sufficient evidence. As our study is limited to patient data as recorded in primary care, we call for greater investigations into the benefits and harms associated with psychotropic medications in this population.

PO-2576 Pneumococcal Vaccination Coverage In Individuals (16 To 59 Years) With A Newly Diagnosed Risk Condition In Germany 1400669 Arijita Deb Merck & Co., Inc. Pneumococcal Vaccination Coverage In Individuals (16 To 59 Years) With A Newly Diagnosed Risk Condition In Germany Vaccines Background: The German Committee on Vaccinations (STIKO) recommends pneumococcal vaccination for individuals with certain underlying medical conditions who are at increased risk of developing pneumococcal disease such as those with congenital or acquired immune-deficiencies or immunosuppression, anatomical and foreign-material associated risks for pneumococcal meningitis and chronic conditions.Objectives: This study estimated the pneumococcal vaccine coverage rate (VCR) among individuals aged between 16 to 59 years in Germany recommended to receive pneumococcal vaccination, according to the STIKO.Methods: A retrospective cohort database analysis was conducted using the German InGef database. The study population comprised individuals aged between 16 to 59 years with at least one new at-risk (chronic conditions) or high-risk condition (immunocompromising conditions). The study cohorts were identified at time of first diagnosis in each calendar year from 2014 to 2016 and were followed up until 2017 (2 to 4 years follow-up). The underlying medical conditions were identified using International Classification of Diseases, 10th revision, German Modification (ICD-10 GM) codes, German Procedure Classification (OPS) codes and German doctor fee’s schedule (EBM) codes in the InGef research database.Results: The pneumococcal VCR among adults under 60 years with risk conditions after four years of follow-up increased slightly from 2014 to 2017 from 0.33% to 0.97%. Adults aged 50 to 59 years were more likely to be vaccinated than those aged 16 to 49 years. Adults with high-risk conditions were more likely to be vaccinated than adults with at-risk conditions (1.98% vs. 0.79% in 2017). A higher number of risk conditions increased the likelihood of vaccination. Median time to vaccination from the first diagnosis of the risk conditions was 385 days and 451 days, respectively, for adults with at-risk conditions and high-risk conditions.Conclusions: Despite recommendations from STIKO, pneumococcal vaccination coverage remains suboptimal in vulnerable populations. Further efforts are required to raise awareness and increase immunization levels among individuals with risk conditions.

PO-2578 Incidence Of Ocular Adverse Events Associated With Mometasone Implants Among Patients With Nasal Polyposis 1400670 Efe Eworuke US Food and Drug Administration Incidence Of Ocular Adverse Events Associated With Mometasone Implants Among Patients With Nasal Polyposis Medical Devices Background: Long term use of oral or intranasal steroids has been previously linked to the development of cataracts (posterior subcapsular cataracts) and increased intraocular pressure. It remains unknown whether sinus mometasone implants carry similar risk, given the route of administration (endoscopically) and location of stent placements (nasal).Objectives: To determine the incidence rates of diminished visual acuity, glaucoma, and cataracts among patients who had a single mometasone implant.Methods: We conducted a retrospective cohort study using data from 01/2016-09/2018 in the IBM® MarketScan® Commercial and Medicare Supplemental Databases. We included patients, 18 years and older with a diagnosis of nasal polyposis (ICD-9: 471.x or ICD-10: J33.x) who had a procedure code and/or national drug code for insertion of mometasone 370µg implant. Patients with a diagnosis of diminished visual acuity, glaucoma, cataracts, or nasal septal perforation during the baseline period were excluded. We also excluded patients who received treatment for glaucoma during the baseline period and those with evidence of eye laser surgery, trabeculoplasty or cataract surgery. Follow-up began on the date of mometasone implant insertion for a maximum of one year or earliest of the following events: outcome of interest, loss of medical or drug coverage, study end date or death.Results: We identified 957 patients with mometasone stent who met the study inclusion/exclusion criteria. The mean age of the cohort was 45.8 ± 13.2 years with 60% male. The majority of patients used oral (74.3%) and intranasal (42.7%) steroids before implant insertion. Only 2.9% had sinus surgery before implant insertion. The incidence of glaucoma, cataracts and diminished visual acuity was 23.4, 32.9 and 6.2 per 100,000 person-years, respectively.Conclusions: Our study explores the risk of glaucoma, cataracts and diminished visual acuity following mometasone stent implants. Additional studies are needed to characterize the risk of these adverse events.

PO-2581 Covariate Missingness, Complete Case Analysis, And Transportability, Oh My! 1400671 Michael Webster-Clark UNC Gillings School of Global Public Health Covariate Missingness, Complete Case Analysis, And Transportability, Oh My! Methods in Pharmacoepidemiology -> Analytical Methods - e.g., marginal structural models, instrumental variables, sensitivity analyses Background: Interest in transporting study results to external target populations is increasing. While work has addressed differing distributions of effect measure modifiers (EMM), little attention has been paid to missing EMM data. The appropriateness of complete case analysis may depend on whether missingness is completely at random (MCAR), at random with respect to observed covariates (MAR), or not at random with respect to EMM value (MNAR).Objectives: We sought to identify cases when complete case analysis would induce bias in estimates by adding known missingness to real trial and target population data.Methods: We assessed performance of inverse odds of sampling weighting to estimate effects of dabigatran vs warfarin on stroke in real-world patients when data on EMM were missing. Specifically, we reweighted Randomized Evaluation of Long-Term Anticoagulation trial participants to resemble new users of oral anticoagulant therapy (specifically, warfarin new users) in a random sample of US Medicare patients. We then estimated intention-to-treat hazard ratios in the weighted trial. Next, we added missingness to trial or target populations in history of transient ischemic attack (TIA), an EMM, and re-estimated treatment effects. Missingness was a logit-linear function of age and past warfarin use (if MAR) and TIA (if MNAR), evaluated in 21 scenarios with differing model coefficients. We report the medians (confidence limits: 2.5th, 97.5th) of 1000 bootstraps.Results: We used data from 15,132 trial participants over age 65 and 74,891 warfarin initiators in Medicare. We estimated a hazard ratio (HR) of 0.656 (95% confidence interval [CI] of 0.41, 1.08 and confidence limit ratio [CLR] of 2.65) after reweighting the trial data to resemble Medicare warfarin initiators. Complete case analysis in the presence of trial MCAR, MAR, and MNAR resulted in estimates with substantially increased variance (largest CLR of 4.55) but minimal bias (most biased HR = 0.635) compared to our full data estimate, even with strong associations between covariates and missingness. Complete case analysis in the presence of MAR and MNAR covariate data in Medicare with the same coefficients for missingness appeared capable of adding more bias (most biased HR=0.709) but less variance (largest CLR: 2.81).Conclusions: When transporting estimates from randomized trials to external populations, using trial complete cases may be acceptable if the study population is large even if covariate data is MAR or MNAR. Taking complete cases in the target population, however, may add some bias if data is MAR or MNAR because the distribution of EMM can differ from the full target.

PO-2588 Proton Pump Inhibitor Use And The Risk Of Hepatocellular Carcinoma: A Systematic Review Of Real-world Data 1444306 Ambrish Singh Proton Pump Inhibitor Use And The Risk Of Hepatocellular Carcinoma: A Systematic Review Of Real-world Data Benefit-Risk Assessment, Communication, and Evaluation (BRACE) Background: Proton pump inhibitors (PPIs) are one of the most frequently used prescription drugs globally. Recently, there have been concerns that PPI therapy is associated with an increased risk of cancer such as hepatocellular carcinoma (HCC). However, the results of real-world clinical studies are conflicting.Objectives: A systematic literature review was conducted to identify all the evidence and determine the effect of the use of PPI on the risk of HCC based on the published real-world data (RWD) studies.Methods: We performed a systematic review and meta-analysis of RWD studies, adhering to the PRISMA guidelines, to evaluate the risks of HCC with PPI use. All RWD studies that assessed the risk of HCC in PPI users compared to PPI nonusers were included. We used the generic inverse variance method and RevMan (5.3.5) to meta-analyses the risk ratios (RRs) using a random-effects model. We used Newcastle-Ottawa scales (NOS) to assess the study quality. The study protocol was registered at PROSPERO [CRD42019109885].Results: Of the identified studies, six (4 cohort and 2 case-control) with more than 33,102 HCC cases were included in our analysis. In PPI users, compared with PPI nonusers, the unadjusted and adjusted risk ratio (RR) for HCC (studies, n= 4) was 1.80 [95 % confidence interval (CI): 1.03 - 3.14, p2>88%) among the studies and the majority (n= 5) of the studies were of moderate to high quality, based on NOS.Conclusions: This meta-analysis found a significantly increased risk of HCC in PPI users compared to PPI nonusers; however, the results should be interpreted with caution owing to the inclusion of observational studies with significant heterogeneity.

PO-2590 Newer Second-line Glucose-lowering Drugs Versus Thiazolidinediones And Cirrhosis Risk In Type 2 Diabetes 1400672 Jeff Yang UNC Gillings School of Global Public Health Newer Second-line Glucose-lowering Drugs Versus Thiazolidinediones And Cirrhosis Risk In Type 2 Diabetes Safety End Points -> Other Background: Type 2 diabetes (T2D) accelerates progression of chronic liver disease to cirrhosis. Thiazolidinediones (TZDs) have been shown to reduce hepatic steatosis and hepatocellular injury in non-alcoholic fatty liver disease. However, the impact of second-line glucose-lowering drugs (SL-GLDs) on cirrhosis risk is unknown.Objectives: To compare cirrhosis risk with use of newer SL-GLDs vs. TZDs among patients with T2D.Methods: We conducted a retrospective cohort study using a 20% sample of Medicare Fee-for-Service. We compared initiators of TZD vs. newer SL-GLDs - sodium-glucose co-transporter 2 inhibitors (SGLT2i), dipeptidyl peptidase-4 inhibitors (DPP4i), and glucagon-like peptide receptor agonists (GLP1RA) - between January 2007 (2013 for SGLT2i) and September 2015. Eligible patients had ≥2 same-drug class prescription claims, were ≥65 years old, and had ≥1 year of continuous enrollment in Parts A/B/D, with no claims for either study drug or cirrhosis/cirrhosis-related complication, prior to index date. We defined cirrhosis using outpatient and inpatient ICD-9 codes 456.1, 571.2, or 571.5 [Lapointe et al, PLoS One 2018, specificity 61-77%, sensitivity 98-99%]. We estimated crude incidence rates (IR), as well as adjusted hazard ratios (aHR) and 95% CI using propensity-score (PS) weighted Cox proportional hazards models controlling for demographics; comorbidities; medication use; and healthcare utilization, censoring for treatment changes.Results: We identified 309,456 new users of the 4 study drug classes. Median follow-up ranged from 0.39-0.64 years. We observed 692, 293, and 89 cirrhosis events in the DPP4i, GLP1RA, and SGLT2i comparisons, respectively. Crude cirrhosis IRs ranged from 4.8 (95% CI 4.1-5.5) to 9.3 (95% CI 6.8-12.7) events per 1,000 person-years. Compared to new use of TZD, initiation of newer SL-GLDs was associated with increased risk of cirrhosis, with elevated aHR estimates for SGLT2i (1.72, 95% CI 1.05-2.82) and DPP4i (1.38, 95% CI 1.15-1.66); GLP1RA was the most comparable agent to TZD (1.18, 95% CI 0.85-1.64). Results were consistent across a number of a priori sensitivity (1% PS trimming, initial treatment, varying lag times, competing event analysis, PS matching) and subgroup analyses (age, sex, baseline metformin, baseline liver disease, baseline insulin).Conclusions: TZDs were associated with lower cirrhosis risk than newer SL-GLDs except GLP1RAs, suggesting that TZDs and GLP1RAs may reduce cirrhosis risk in patients with T2D. Further study will assess the impact of different validated cirrhosis definitions on safety results.

PO-2592 Effectiveness And Safety Of Vitamin K Antagonists In Patients With Non-valvular Atrial Fibrillation And Renal Disease: A Cohort Study 1400673 Antonios Douros Lady Davis Institute Effectiveness And Safety Of Vitamin K Antagonists In Patients With Non-valvular Atrial Fibrillation And Renal Disease: A Cohort Study Drug Effectiveness Background: Patients with non-valvular atrial fibrillation (NVAF) and renal impairment are at high risk of both thrombosis and bleeding complicating antithrombotic treatment. However, available clinical evidence on benefits and risks of antithrombotic drugs commonly used in this setting such as vitamin K antagonists (VKAs) or antiplatelet agents is scarce.Objectives: To assess the effectiveness and safety of VKAs compared to antiplatelet agents in NVAF patients with chronic kidney disease (CKD) or undergoing dialysis.Methods: We conducted a population-based cohort study using linked administrative healthcare databases from Quebec. We assembled a cohort of patients with incident NVAF from 2000 to 2014 and then constructed three subcohorts based on renal status (dialysis, CKD, no renal disease) at cohort entry. Using a nested case-control approach, we identified all cases with a first-ever diagnosis of ischemic stroke, major bleeding or all-cause mortality (depending on the outcome under study) during follow-up and matched them to 30 controls on age, sex, date of cohort entry, and duration of follow-up in each subcohort. We estimated incidence rate ratios (RRs) and 95% confidence intervals (CIs) of each outcome associated with current use of VKAs compared with current use of antiplatelet agents using conditional logistic regression. Sensitivity analyses accounted for possible exposure and outcome misclassification, restricted to patients with high baseline ischemic or bleeding risk, and explored the potential impact of reverse causality.Results: Of 231,411 patients with NVAF, 3,613 were undergoing dialysis and 30,608 had CKD. Among dialysis patients, VKAs were associated with a numerically decreased risk of ischemic stroke (RR, 0.79; 95% CI, 0.46 to 1.36), an increased risk of major bleeding (RR, 1.42; 95% CI, 1.04 to 1.93), and a similar risk of all-cause mortality (RR, 0.98; 95% CI, 0.81 to 1.20), compared with antiplatelet agents. Among CKD patients, there was a decreased risk of ischemic stroke (RR, 0.64; 95% CI, 0.56 to 0.73), an increased risk of major bleeding (RR, 1.47; 95% CI, 1.35 to 1.59), and a decreased risk of all-cause mortality (RR, 0.86; 95% CI, 0.81 to 0.90). Sensitivity analyses yielded highly consistent results.Conclusions: Our results suggest that VKAs could decrease the risk of ischemic stroke in NVAF patients with CKD or undergoing dialysis at the cost of a higher risk of major bleeding. Until data from randomized trials are available, individualized benefit-risk assessment for the antithrombotic management of these vulnerable populations is required.

PO-2594 Methodologic Considerations For Evaluating Adherence To Oral Oncologic Therapy In A High Mortality Population 1442625 Danielle Chun UNC Gillings School of Global Public Health Methodologic Considerations For Evaluating Adherence To Oral Oncologic Therapy In A High Mortality Population Disease Epidemiology/Clinical Course -> Cancer Background: Continuous and on-time medication use is critical to for drug effectiveness and improving patient outcomes. However, key methodological decisions must be made when designing and analyzing studies of medication adherence and persistence in populations with a high risk of short-term mortality, such as the setting of metastatic cancer. Objectives: To compare different measures of and approaches to quantifying adherence and persistence to oncologic therapy in a high mortality population using a motivating example of oral targeted therapy (TT) in patients diagnosed with metastatic renal cell carcinoma (mRCC). Methods: Patients aged ≥66 years old with a primary diagnosis of mRCC who initiated TT within 4 months of their diagnosis were identified in the Surveillance, Epidemiology, and End Results (SEER)-Medicare linked database from 2007-2015. We estimated 6-month adherence to oral TTs using an 80% threshold for proportion of days covered, including and excluding patients who died within 6 months. We then estimated the proportion of patients who were persistent (e.g., continuous TT users) at 6 months: (1) censoring patients at death and (2) treating death as a competing risk using nonparametric cumulative incidence functions. A gap of up to 30 days between TT fills was allowed. Results: Among 507 mRCC patients initiating an oral TT (sunitinib:63%, pazopanib: 26%, sorafenib:10%), 32% died within 6 months. After excluding those who died, the proportion of adherent patients was substantially higher (38%) than when these patients were included but assigned zero days covered after death (26%). The proportion of patients who were persistent at 6 months was also substantially higher when censoring patients at death at 82% (95% confidence interval (CI): 78, 85%) compared with treating death as a competing risk at 50% (95% CI: 46, 54%). Conclusions: Different approaches to the treatment of death result in substantially different estimates of oral oncologic adherence and persistence in high-mortality populations. This may have potential for bias when making comparisons in this setting. Because short-term mortality is so common in the metastatic cancer setting, studies of oncologic therapy adherence and persistence should explicitly report the proportion of patients who die over time and use appropriate methods to account for this competing risk.

PO-2599 Anticholinergic And Sedative Drug Load Among Medicare Beneficiaries 1400674 Shahar Shmuel UNC Gillings School of Global Public Health Anticholinergic And Sedative Drug Load Among Medicare Beneficiaries Geriatric Pharmacoepidemiology Background: Medications with anticholinergic and sedating properties are widely used among older adults despite strong evidence of their harm. The Drug Burden Index (DBI), a clinical screening tool that incorporates dose and cumulatively measures these properties across drug classes, may identify patients at high risk of physical and cognitive function-related adverse events. However, the DBI has been studied mainly outside the US, and not using US Medicare claims. Objectives: We developed an updated US-based DBI drug list and operationalized the DBI for Medicare claims, and used this framework to (1) measure the monthly distribution based on mean daily DBI, (2) identify patient-level predictors of high mean daily DBI exposure, and (3) describe temporal trends in the prevalence of several DBI thresholds. Methods: We screened medications for DBI properties and operationalized the DBI for Medicare claims. We then conducted a retrospective cohort study of a 20% random, nationwide sample of 4,137,384 fee-for-service Medicare beneficiaries aged 66+ years (134,757,039 person-months) from 2013 to 2016. We measured the monthly distribution based on mean daily DBI, categorized as (1) >0 vs. 0 (any use) and (2) 0, 02, and examined temporal trends. We described patient-level factors (e.g., demographics, healthcare use, and health status) associated with high (>2) versus low (00) DBI exposure decreased slightly over the study period, while the monthly point prevalence was stable. Predictors of high monthly DBI exposure (>2) included certain indicators of increased healthcare use (e.g., high number of drug claims), white race, younger age, low-income subsidy, frailty, and a psychosis diagnosis code. Conclusions: This study operationalized the DBI in Medicare claims and described patterns of DBI exposure over time. The predictors of high monthly DBI exposure can be used to inform discussions between patients and providers about medication appropriateness and potential de-prescribing. Future studies should assess the association between the DBI and adverse events among Medicare patients.

PO-2600 Process Indicators And Outcomes To Assess The Effectiveness Of Additional Risk Minimisation Measures In Europe 1400675 ESTHER ARTIME Autonomous University of Madrid Spotlight Poster Session Spotlight Poster BRACE Background: Good Pharmacovigilance Practice XVI recommends the evaluation of additional risk minimisation measures (aRMMs) with process indicators and outcomes.Objectives: To describe study designs, measures and results in risk minimisation evaluation plans (RM Plans) that include process indicators and outcomes in Europe.Methods: A systematic search was conducted using the EU PAS Register, PubMed, and grey literature to identify RM Plans for products involving one or multiple studies to assess the effectiveness of aRMMs via process indicators and outcomes, between 1Jan2011 and 12Oct2019. Process indicators include measures of awareness, receipt, use, knowledge, and self-reported behaviour via surveys. Outcome measures include health outcomes and behavioural outcomes via measures of prescribing, investigations and monitoring. Data were abstracted from study reports, manuscripts and abstracts.Results: A total of 129 studies involving 102 products were identified. Eighteen products (18%) with RM Plans assessed process indicators and outcomes; 10/18 plans included process indicators and outcomes within the same study. Most products belonged to the antineoplastic and nervous system groups (61%). Two thirds had HCP educational materials, while patient cards were implemented in half of the products. The 18 RM Plans included 29 studies: cross-sectional surveys (66%; 47% targeted patients and 90% HCPs), retrospective designs (59%; 47% using a pre/post approach) and prospective designs (10%). While 62% measured process indicators [receipt (n=14) or use of aRMMs (n=12), level of knowledge (n=17) and self-reported behaviour (n=15)], prescribing behaviour was assessed via drug utilisation studies in 41% of cases and health outcomes were measured in 28%. The proportion of patients receiving the target drug under on-label/off-label conditions, over time, before and after the aRMMs or at a specific data point, was evaluated by 11 of 29 studies. Monitoring of metabolic parameters, reading errors and inappropriate use were assessed by one study each. Outcomes measured were adverse events. Drug utilisation and health outcomes data were mostly obtained from existing healthcare databases (10 of 29). In four RM Plans a correlation between process indicators and outcomes was noted. Study results were available for 14/18 RM Plans.Conclusions: Only one in five RM Plans identified in this review assessed both process indicators and outcomes, with health outcomes being measured in a minority. Few within-patient correlations between process indicators and outcomes were noted.

PO-2602 Availability Of Secondary Healthcare Data Sources In Colombia: A Systematic Review 1400676 Juan Franco Bayer Healthcare Pharmaceuticals Availability Of Secondary Healthcare Data Sources In Colombia: A Systematic Review Informatics Background: Real-world evidence (RWE) is emerging as a fundamental component of the post-marketing evaluation of medicinal products. Even though the focus on RWE studies has increased in Colombia, the availability of secondary data sources to perform this type of research is not well documented.Objectives: We aimed at identifying and characterizing secondary data sources available in Colombia, which could be used for pharmacoepidemiology studies.Methods: We performed a systematic literature review on PubMed, EMBASE and VHL using a combination of controlled vocabulary and keywords for the concepts of electronic health records, epidemiologic studies and Colombia.Results: A total of 323 publications were included in this review. These comprised 123 identified secondary data sources including pharmacy dispensing databases, government datasets from vital statistics and healthcare provision data collected by the Ministry of Health, disease registries, claims databases with insurance data, and electronic heath records, among others. These data sources were particularly used for cross-sectional studies focused on portraying disease epidemiology in a particular population. Almost all of the databases (95%) contained patient demographic information, while pharmacological treatment and specific diagnostic tests were present in 44% and 39% of the databases, respectively. Even though the database owner was identifiable in 94%, access information and requirements was only available in 44% of the articles. Only the pharmacy-dispensing database, local cancer registry and the government databases included a description of quality control measures or of the overall quality of the information available.Conclusions: The diversity of databases identified shows that Colombia has a high potential to continue enhancing its RWE strategy. Greater efforts are required to improve data quality and accessibility. The linkage between databases will expand data pooling and integration to boost the translational potential of RWE.

PO-2603 Antihypertensive Polytherapy In Australia: Prevalence Of Inappropriate Combinations, 2013-2018 1400677 Michael Falster Centre for Big Data Research in Health, UNSW Sydney Antihypertensive Polytherapy In Australia: Prevalence Of Inappropriate Combinations, 2013-2018 Drug Utilization Research -> Quality use of medicines (including indicators and guidelines) Background: Antihypertensive polytherapy plays a significant role in treating hypertension and is endorsed increasingly in hypertension guidelines. There has been little population-based research to better understand how antihypertensive combinations, particularly inappropriate polytherapy, are used in routine clinical care.Objectives: To estimate the prevalence of inappropriate antihypertensive polytherapy in Australia.Methods: We conducted a retrospective cohort study using pharmaceutical claims data. We used a nationally representative 10% sample of Pharmaceutical Benefits Scheme (PBS) eligible Australians and their dispensing history to identify people aged 18+ years exposed to at least one PBS-listed antihypertensive, 2013-2018. We measured annual: prevalence of antihypertensive polypharmacy (≥40 days concomitant exposure), inappropriate antihypertensive combinations (against guideline recommendations; within-class polytherapy) and combinations to be used with caution.Results: Almost half (47.5%) of people using antihypertensives in 2018 experienced polytherapy. Among these, 2.4% had an inappropriate combination (1.5% against guidelines; 1.0% within-class polytherapy). Inappropriate combinations were more prevalent in people experiencing polytherapy with three (3.7%) or four (16.1%) antihypertensive medicines than people on dual therapy (0.7%). Inappropriate combinations occurred at a lower rate in people using fixed-dose rather than free-drug combinations for dual therapy (0% vs 0.7%; p3 antihypertensives is required.

PO-2604 Long-term Trajectories Of Medicine Use Among Older Adults Experiencing Polypharmacy In Australia 1400678 Michael Falster Centre for Big Data Research in Health, UNSW Sydney Long-term Trajectories Of Medicine Use Among Older Adults Experiencing Polypharmacy In Australia Drug Utilization Research -> Quality use of medicines (including indicators and guidelines) Background: Polypharmacy (the concurrent use of five or more medicines) is often considered a marker of potentially inappropriate care. Patients experiencing polypharmacy may follow a diverse range of clinical pathways, yet there is little understanding about how the number and nature of medicine use changes over time.Objectives: To identify common pathways of medicine use over time among older patients experiencing polypharmacy.Methods: We conducted a retrospective cohort study using administrative pharmaceutical claims data for a 10% sample of Australian Pharmaceutical Benefits Scheme (PBS) beneficiaries. We identified people aged 70 years and over who were exposed to five or more medicines on February 15, 2014. Using group-based trajectory modelling, we explored changes in the quarterly number and type of medicines people used over a five-year period (2014-2018).Results: In our cohort of 98,539 older people, we identified two groups of people experiencing polypharmacy encompassing seven distinct trajectories: sustained polypharmacy (77% of people, 4 trajectories); and decreasing medicine use (23%, 3 trajectories). Patients in sustained polypharmacy trajectory groups with a lower mean number of medicines (e.g. 6 unique medicines) had relatively stable trajectories, while those with a higher mean number of medicines (e.g. 15 unique medicines) experienced greater seasonal variation in both the number and type of medicines used. Within the sustained polypharmacy trajectories, patients retained on average two-thirds of their medicines of exposure (chemical substance level) between adjacent quarters. For patients in decreasing medicine use trajectories, the most common trajectory was a slight drop in medicine exposure within the first three months. Overall, 21% of patients were no longer experiencing polypharmacy after one year.Conclusions: Polypharmacy among older patients is largely a chronic phenomenon, reflecting the nature of multimorbidity within this population. However, there is an underlying volatility in the medicines involved, reflecting both changes in treatment and seasonal fluctuation in medicine dispensing. Care must be taken when measuring and interpreting polypharmacy, as a single measure may not reflect a patients’ sustained medicine use over time.

PO-2607 Prevalence And Incidence Of Narcolepsy In Japan: A Study Using Japanese Claims Data 1443093 Motonobu Sakaguchi Motonobu Sakaguchi Prevalence And Incidence Of Narcolepsy In Japan: A Study Using Japanese Claims Data Disease Epidemiology/Clinical Course -> Neurological/Mental Health Background: Previous studies suggest a relatively higher prevalence of narcolepsy in Japan than in Europe and North America, however these studies with Japanese data had some methodological limitations such as reliance on self-report questionnaires. More data is therefore, required to accurately estimate the epidemiology of narcolepsy in Japan. Objectives: The objective of the study was to determine the prevalence and incidence of narcolepsy overall, and by age and gender, and estimate the numbers of total narcolepsy cases across Japan. Methods: A retrospective cross-sectional study was conducted using the JMDC database. Patients who were continuously enrolled from January 2014 to December 2017 were eligible. Prevalence cohort was composed of the eligible population, and incident cohort was restricted to those without any narcolepsy diagnoses any time before 1st January 2017 from the eligible population. Considering the local practice in diagnosing narcolepsy, narcolepsy cases were identified by two definitions to obtain a range of prevalence and incidence rates within which the true rate was likely to lie. A) broad case definition (upper limit): patients having one diagnosis (ICD-10 code, G474) followed by a confirmed diagnosis; B) narrow case definition (lower limit): patients having at least one sleep test, including electroencephalogram, polysomnography, or multiple sleep latency test, and any subsequent confirmed diagnosis. Descriptive analyses were used to summarize the demographics for the patients with narcolepsy. All analyses were performed using SAS 9.4 (SAS Institute, Cary, NC). Results: Among the prevalent cohort, 1,066 cases and 571 cases were fulfilled using the broad and narrow definition respectively. Among the incident cohort, 121 cases and 76 cases were met by the broad and narrow definition respectively. The age-sex adjusted prevalence of overall narcolepsy was 22.3-45.3 per 100,000 persons, projected to 28-57 thousand narcolepsy patients across Japan. The highest prevalence occurred among those 20-29 years of age. The age-sex adjusted incidence rate was 0.37-0.64 per 100,000 person-years, suggesting 500-800 newly diagnosed patients every year Conclusions: We concluded that the prevalence of overall narcolepsy is not as high as previously reported in Japan, but may still be higher than the prevalence in Western countries. In this study, it appears that the incidence rate was the highest in the second decade, fitting the natural history. The reason for the lower prevalence in the age 10-19 years needs to be further explored.

PO-2609 Using Multiple Random Index Dates For Each Patient Improves Precision Of Prescription Durations Estimated With The Reverse Waiting Time Distribution 1400679 Katrine Nielsen Aarhus University Using Multiple Random Index Dates For Each Patient Improves Precision Of Prescription Durations Estimated With The Reverse Waiting Time Distribution Methods in Pharmacoepidemiology -> Analytical Methods - e.g., marginal structural models, instrumental variables, sensitivity analyses Background: Prescription durations can be estimated using the reverse Waiting Time Distribution (rWTD) as the time within which 80% of the current users have redeemed a new prescription. When there is seasonal variation in redemption rates, rWTD estimates vary with the chosen index date, which can be mitigated by using individual random index dates. However, using a single index date for each patient is statistically inefficient.Objectives: Using more than one random index date for each patient to efficiently utilize the information in the data and hence improve the precision of the estimates.Methods: First, we select a one-year sample window supplemented with a preceding period of equal length. Second, for each patient we uniformly sample multiple random index dates within the sample window and identify the last prescription, if any, within one year prior to the index date. Based on time from the last prescription to the index date, prescription durations are estimated using the rWTD with robust variance estimation to account for the non-independence of observations from the same patient. We conducted a simulation study with log-normal distributed prescription durations both with and without seasonal variation in redemption rates. We generated 1,000 datasets such that on average 5,000 patients had prescriptions each year, of whom 25% ended treatment. We analyzed the simulated datasets with the rWTD (log-normal) after sampling 1, 5, 10 or 20 random index dates for each patient and estimated the relative bias and coverage probability of nominal 95% confidence intervals. We finally applied the method to Danish data on warfarin redemptions from 2013-2015.Results: In the simulation study without seasonal variation, the relative bias was negligible (-0.41 to -0.20%) whether sampling 1 or 5 times. The coverage probability was slightly lower when sampling 5 times (89.0 to 93.9%) as compared to 1 (92.4 to 93.6%). When durations varied seasonally, bias was higher whether sampling 1 or 5 times (-0.73 to 6.71%). Consequently, the coverage probability was lower and decreased with the number of sampled index dates. Patterns continued with more random index dates (10 and 20). The estimated durations of warfarin prescriptions in Denmark were 93.4 (91.8; 95.0) and 93.8 (92.9; 94.8) days with 1 and 5 index dates, respectively.Conclusions: Precision improves with the number of random index dates, but it does not reduce bias due to misspecification, e.g. with seasonal variation in redemption rates. Given misspecification, the improved precision lowered coverage probability.

PO-2610 Probabilistic Linkage Of The French Cystic Fibrosis Patient Registry With The French National Health Data System 1400680 Manon Belhassen PELyon Probabilistic Linkage Of The French Cystic Fibrosis Patient Registry With The French National Health Data System Disease Epidemiology/Clinical Course -> Other Background: To efficiently assess the management of the cystic fibrosis (CF), the French Cystic Fibrosis Patient Registry was linked to the SNDS (French Claims Data System, with information on health care utilization of the national population).Objectives: The aim of this linkage was to build a complete dataset that will open up new perspectives of research on CF patients, such as the changes in healthcare consumption, and to relate it to the epidemiological transformation of CF.Methods: The records of CF patients included between 2006 and 2016 in the French CF Registry were linked by probabilistic method to their corresponding SNDS data. To perform this linkage, a scoring system was developed with variables available in both data sources, ranging from the least (sex, month and year of birth, region of residence) to the most specific ones (dates of spirometry, dates of transplant, and sweat test). Next, a decision algorithm was developed. In a last step, to allocate remaining patients, “comorbidities” (e.g. diabetes, aspergillosis or pancreatic insufficiency) were used. Sensitivity analyses were performed using other scoring systems, to test the validity of the linkage.Results: Of the 7,671 patients included in the French CF Patient Registry between 2006 and 2016, 6,186 patients (80.6%) could be linked to the SNDS. Non-linked patients were mainly children younger than 10 who had little or no spirometry data. The sensitivity analyses validated the linkage.Conclusions: We performed one of the first linkages between a French registry and the SNDS. These two data sets include robust and complementary data, leading to an exhaustive database and opening new prospective to study the management of CF patients.

PO-2612 Metoprolol Versus Carvedilol In Patients With Heart Failure, Chronic Obstructive Pulmonary Disease, Diabetes Mellitus, And Renal Failure. 1400681 Maurizio Sessa University of Copenhagen Metoprolol Versus Carvedilol In Patients With Heart Failure, Chronic Obstructive Pulmonary Disease, Diabetes Mellitus, And Renal Failure. Geriatric Pharmacoepidemiology Background: In heart failure patients, it is not uncommon to find patients with heart failure, chronic obstructive pulmonary disease, and diabetes mellitus, but currently, none of the most authoritative clinical practice guidelines provide recommendations on which β-blocker should be used. A recent study suggested that physicians prefer carvedilol to other β-blockers in this clinical scenario. However, evidence supporting this clinical practice is missing.Objectives: This study aims to compare the survival and the risk of heart failure, chronic obstructive pulmonary disease, diabetes mellitus, hypoglycemia, and renal failure hospitalizations in geriatric patients with heart failure, chronic obstructive pulmonary disease, and diabetes mellitus exposed to carvedilol or metoprolol. Additionally, this study aims to identify the most import predictors for these study outcomes by using an artificial intelligence technique and to evaluate subpopulations of interest such as patients with concurrent lipid disorder or chronic kidney disease.Methods: Data sources were Danish administrative registers. Patients aged ≥65 and having heart failure, chronic obstructive pulmonary disease, and diabetes mellitus were followed for 1 year from the first β-blocker prescription redemption. Patients' characteristics were used to 1:1 propensity score match carvedilol and metoprolol users. A Cox regression model was used to compute the hazard ratio (HR) of study outcomes. For statistically significant associations, a conditional inference tree was used to assess predictors most associated with the outcome.Results: In total, 1,424 patients were included. No statistically significant differences were observed for survival (HR 0.86; 95% confidence interval [CI] 0.67 to 1.11, p = 0.240) between carvedilol/metoprolol users. The same applied to chronic obstructive pulmonary disease (HR 0.88; 95% CI 0.75 to 1.05, p = 0.177), diabetes mellitus (HR 0.95; 95% CI 0.82 to 1.10, p = 0.485), hypoglycemia (HR 0.88; 95% CI 0.47 to 1.67, p = 0.707), and RF (HR 1.25; 95% CI 0.93 to 1.69, p = 0.142) hospitalizations. Carvedilol users had a 38% higher hazard then metoprolol users of heart failure hospitalization during the follow-up period (HR 1.38; 95% CI 1.19 to 1.60, p <0.001). Artificial intelligence identified carvedilol exposure as the most important predictor for heart failure hospitalization.Conclusions: This study found an increased risk of heart failure hospitalization for carvedilol users compared with metoprolol users with this triad of diseases.

PO-2614 Application Of Quantitative Bias Analysis For Unmeasured Confounding In Pharmacoepidemiology 1400683 Jeremy Brown London School of Hygiene and Tropical Medicine Application Of Quantitative Bias Analysis For Unmeasured Confounding In Pharmacoepidemiology Methods in Pharmacoepidemiology -> Confounding/Bias - Methods to avoid bias and confounding, including confounder summary scores (e.g., propensity scores, disease risk scores) Background: Confounding is a major concern in pharmacoepidemiological studies. Measured confounders can be accounted for through stratification, regression, or other methods. However, it is often unclear whether all confounders have been measured, and therefore it is not known to what extent associations observed are causal or due to unmeasured confounding.Objectives: To apply and compare quantitative bias analysis methods for unmeasured confounding.Methods: We applied two quantitative bias analysis methods, probabilistic bias analysis and e-value calculation, to a study we previously conducted where unmeasured confounding was considered likely. These methods quantify the potential impact of unmeasured confounders. In the cohort study conducted in the UK Clinical Practice Research Datalink, comparing new users of proton pump inhibitors (n=733,885) to new users of H2 receptor antagonists (n=124,410), we identified an increased risk of all-cause mortality associated with proton pump inhibitor prescription (adjusted hazard ratio 1.38, 95% CI 1.33-1.44). For the probabilistic bias analysis we repeatedly sampled (Monte Carlo sampling - 10,000 iterations) the prevalence of the putative unmeasured binary confounder (uniformly between 0.1 and 0.5), the strength of association between exposure and confounder (uniformly on log scale to obtain odds ratios between 1.0 and 6.0), and the strength of association between the confounder and outcome (uniformly on log scale for hazard ratios between 1.0 and 3.0). On each iteration we calculated a bias-adjusted hazard ratio using sampled parameters. We ran additional simulations increasing the number of unmeasured confounders (assuming independent multiplicative effects).Results: Using the e-value formula, for an unmeasured confounder to fully account for the observed lower bound of the hazard ratio of 1.33, the confounder would need to be associated with both treatment and outcome by a risk ratio of at least 1.9. In probabilistic bias analysis, the median bias-adjusted hazard ratio (2.5th, 97.5th percentiles) was 1.27 (0.99, 1.38) with one unmeasured confounder, 1.22 (0.91, 1.36) with two unmeasured confounders, and 1.13 (0.80, 1.34) with three unmeasured confounders. Conclusions: A single unmeasured confounder would need to be strongly associated with the exposure and outcome to fully account for the observed association, whereas multiple unmeasured confounders would only need be associated more weakly. Quantitative bias analysis deserves wider application in pharmacoepidemiology given the susceptibility of studies in the field to unmeasured confounding.

PO-2615 Digital Tools For The Communication Of Risk Minimization Measures For Medicines 1400684 Michael Forstner PrimeVigilance Digital Tools For The Communication Of Risk Minimization Measures For Medicines Benefit-Risk Assessment, Communication, and Evaluation (BRACE) Background: The primary goal of a Risk Minimization Plan (RMiP) is to develop activities that help patients and healthcare providers understand important risks associated with a drug and to reduce the chance that a serious side effect might occur. The activities implemented by a drug company as part of this plan are called risk minimization measures (RMM). Implementing an RMiP consists of implementing the RMMs, assessing the usefulness of those activities, and using the assessment results to decide if the RMiP needs to be changed.Objectives: Despite the widespread use of mobile phone apps and other electronic means for the dissemination of information to patients, RMMs are still mostly implemented as paper-based tools, consisting of e.g., educational material packs or patient alert cards. Numerous evaluations by regulators and companies have shown the relative inefficiency of these tools. In addition, paper based tools require an offline means of effectiveness evaluation, such as a survey, and they do not lend themselves to two-way communication. Early attempts to design electronic tools for RMMs consistent mostly of web-versions of the paper material used for educational purposes but did not make use of the full capabilities of digital tools.Methods: We review different approaches and show that apps designed to deliver information about the safe use of medicines can also be used to allow patients to report experiences pertaining to the safety of the drug and to quality-of-life related issues they may experience in connection with their disease and/or the drug.Results: The two-way communication allows for a personalized delivery of safety relevant messages to patients and for a real-time evaluation of the benefit-risk profile by feeding into safety signal management tools directly. The evaluation of process and outcome related parameters to demonstrate the effectiveness of risk minimization is simplified as the collection of these parameters is linked to the use of the tool. Furthermore, the collection of meta-data related to the use of the app, e.g., geo-spatial data, behavioral aspects concerning the use of the app, or the frequency, time and duration of use may help to evaluate important real-world parameters associated with the safe and efficacious use of the medicine. The inherent biases associated with data obtained from such sources and ways of accounting for them are discussed as well.Conclusions: The use of truly digital tools for patient and HCP education has the potential to change the communication on benefits and risks of drugs substantially and integrate risk management of medicines better into clinical care.

PO-2616 Comparative Risk Of Parkinsonism Associated With Olanzapine, Risperidone And Quetiapine In Older Adults 1400685 Te-yuan Chyou University of Otago Comparative Risk Of Parkinsonism Associated With Olanzapine, Risperidone And Quetiapine In Older Adults Pharmacovigilance -> Neurological/Mental Health Background: In New Zealand (NZ), clinical trials provide limited information on the comparative risk of Parkinsonism associated with second-generation antipsychotics (SGAs) in older adults.Objectives: This study examines the incidence of Parkinsonism in new users of SGAs in older adults (≥ 65 years). As secondary analyses, we examined the risk of Parkinsonism by type and dose of SGA and also the age-sex interactions.Methods: We used the NZ national minimum data set and the pharmaceutical collections to identify older adults who had a diagnosis of Parkinsonism and who started taking olanzapine, risperidone or quetiapine between 01/01/2005 and 31/12/2016. Confounding was mitigated using propensity score methods. The Cox proportional hazard (COXPH) regression model with inverse treatment probability weighting (ITPW) was used to evaluate the risk of Parkinsonism associated with SGAs, using quetiapine as the reference. We used the generalized propensity score method, where daily doses as continuous variables were used as exposure variables, to evaluate the dose-response risk of Parkinsonism associated with SGAs.Results: After ITPW adjustment for confounders including comorbidity, the COXPH model showed that compared to quetiapine, the use of olanzapine and risperidone were associated with an increased risk of parkinsonism (ITPW-adjusted hazard ratio (ITPW-HR) 1.76, 95% confidence interval 1.57 to 1.97; and ITPW-HR 1.31 (1.16 to 1.49) respectively). The dose-response risk of Parkinsonism was highest for olanzapine (HR 1.69 (1.40 to 2.05)) and the least for quetiapine (HR 1.22 (1.14 to 1.31)). Age-sex interactions revealed risk of Parkinsonism in the 65-74yr group was higher for both sexes with risperidone compared to olanzapine, but the risk increased significantly with olanzapine for both sexes in the 85yr+ group.Conclusions: The study found that quetiapine use might be associated with a relatively lower risk of Parkinsonism than olanzapine or risperidone in older adults. Prescribers should exercise caution when using olanzapine in the oldest old.

PO-2617 Prescribing Trends Of Direct Acting Antivirals For The Treatment Of Hepatitis C In Ontario 1400686 Mina Tadrous Prescribing Trends Of Direct Acting Antivirals For The Treatment Of Hepatitis C In Ontario Drug Utilization Research -> Health policy and governance Background: The ease and effectiveness of direct acting oral antivirals (DAAs) treatments provide opportunity to expand treatment for chronic hepatitis C (HCV). Reimbursement mechanisms for DAAs have changed on two occasions since these drugs were added to Ontario’ publicly funded drug formulary. Whether these changes have appreciably modified prescribing patterns and increased access to DAAs is unknown.Objectives: The purpose of this study was to summarize the utilization of DAAs in Ontario between 2012 and 2018, and describe changes in prescribing physician specialties. A secondary objective was to describe the characteristics of people who received treatment through public reimbursement in 2018.Methods: We conducted a repeated cross-sectional study for DAAs reimbursed by the public drug program in Ontario from January 1, 2012 to December 31, 2018. We measured the quarterly number of users, overall and by prescriber specialty. Characteristics of those receiving DAA treatments in 2018 were examined overall and by prescriber specialty.Results: A total of 27,116 individuals received a publicly-funded DAA prescription between Q1-2012 and Q4-2018. Nearly two-thirds (n=17,813; 65.7%) of all DAAs were prescribed by gastroenterologists, hepatologists or infectious disease specialists. Only 9.6% of DAA recipients were treated by general practitioners. Utilization of DAAs had three major phases of increased uptake: (1) the introduction of DAAs to Ontario’s public drug formulary as a prior authorization benefit in Q1-2015; (2) expanded listing of DAAs as limited use products on the formulary in Q1-2017; and (3) the introduction of newer DAAs in Q2-2018. In 2018, 2,538 unique individuals received publically funded DAAs. The majority were over age 50 (59.8%), male (63.8%) and living in urban (88.8%) neighborhoods of lower socioeconomic status (41.7% in lowest income quintile neighborhoods). There were important differences among patients who received treatment from gastroenterologists/hepatologists (GH).For example, patients who received treatment from a general practitioner (GP) or infectious disease specialist (IDS) were more likely to have a mental health diagnoses (GP: 77.7%; IDS: 71.2% v GH: 41.6%) and be receiving opioid agonist therapy (GP: 54.5%; IDS: 51.6% v GH: 12.4%).Conclusions: HCV elimination guidelines call for an expansion of the range of who delivers HCV treatment. Changes in the listing criteria of DAAs in Ontario’s public drug program has led to increased uptake of these agents. However, there does not appear to be increased prescribing of DAAs among.

PO-2618 Assessment Of A Claims-based Osteonecrosis Of The Jaw Algorithm In The United States 1400687 Lauren Stevens Humana Florence Wang Optum Epidemiology Assessment Of A Claims-based Osteonecrosis Of The Jaw Algorithm In The United States Methods in Pharmacoepidemiology -> Measurement Methods - Validation methods to confirm data (including Patient Reported Outcomes development and validation); methods to deal with missing data Background: Osteonecrosis of the jaw (ONJ) is a rare condition associated with certain drug therapies. Pharmacoepidemiologic studies often rely on electronic healthcare data to evaluate potential occurrence of adverse events following drug exposure. To date, few studies have been conducted to develop and validate claims-based ONJ identification algorithms.Objectives: To assess performance of a claims-based ONJ algorithm by chart review of potential cases among postmenopausal (PM) women and women with PM osteoporosis (PMO).Methods: PM women and women with PMO were identified in a large US commercial health insurance database affiliated with Optum. Potential ONJ cases were identified by ICD-9 (526.4, 522.7, 526.5, 733.45) and ICD-10 (M27.2, K04.6, M27.3, M87.180, M87.08, M87.28, M87.38, M87.88, M87.9) diagnosis codes; 200 potential ONJ cases were selected for chart retrieval, with the goal of obtaining 100 charts in each coding era (ICD-9, ICD-10). Procured charts were redacted with respect to identifiable information and osteoporosis treatment received, then reviewed by a practicing oral surgeon who applied a case definition to determine final ONJ case status. Positive predictive values and 95% confidence intervals were calculated overall, and by cohorts and coding eras. Baseline characteristics of all potential cases were assessed. Two potential algorithm refinements were explored: using a restricted set of ICD diagnosis codes that included only a subset of codes that identified true cases of ONJ; requiring antibiotic use after ONJ diagnosis.Results: 1,273 potential cases were identified by ONJ diagnosis codes. Of the 200 potential cases selected, 104 (52%) charts were procured, and six cases were confirmed across the two cohorts (PPV 5.8%, 95% CI 2.2, 12.1). Baseline characteristics among potential cases were largely similar across all strata, and among those with procured and non-procured records. Applying potential algorithm refinements resulted in marginal PPV improvement.Conclusions: This study identified only a small number of confirmed cases, and the resulting PPVs were generally low, but consistent with previously reported studies. Potential refinements to the algorithm resulted in minimal improvements. To our knowledge, our study is the first to report on the identification of ONJ using ICD-10 codes in the US.

PO-2619 The Impact Of Changing Biosimilar Policies For Rheumatic Conditions And Inflammatory Bowel Disease In Ontario, Canada 1400688 Mina Tadrous Spotlight Poster Session Spotlight Poster Biologics & Biosimilars Background: Biologic drugs are important treatment options for several conditions, including rheumatic conditions, gastrointestinal disease, and diabetes. Yet, these medications are associated with high costs. Recently available biosimilars are similar, lower cost medications with no clinically meaningful differences compared to their innovator biologics. Two drugs currently have available biosimilars but are listed differently. We sought to determine the impact on biosimilar utilization when different listing strategies are used.Objectives: We sought to determine the impact on biosimilar utilization when different formulary listing strategies are used.Methods: We conducted a population-based time-series study between January 2010 and June 2019 to assess the quarterly utilization of biologics used for the treatment of rheumatic conditions and inflammatory bowel disease (IBD) with currently available biosimilars (etanercept and infliximab) through the public drug program in Ontario, Canada. Utilization was defined as the number of individuals who were dispensed a prescription for a publicly-funded biologic or biosimilar per quarter. Trends in utilization and expenditures for biologics were forecasted to June 30, 2022. We used Holt-Winters’ exponential smoothing models to forecast utilization and costs.Results: Between Q1 2010 and Q2 2019, the number of biologic users for rheumatic and IBD indications increased by 133% from 5,225 to 12,178 users. Among biologics indicated for rheumatic conditions and IBD with a currently available biosimilar, 16.7% of users (N=1,196 of 7,158) were treated with a biosimilar in Q2-2019. By Q2-2022, this was similar for both drugs. It is estimated that 2,717 users (95% CI: 1,899, 3,536) out of 8,142 biologic users (95% CI: 7,438, 8,847) with a currently available biosimilar will be treated with biosimilars, representing a proportion of 35.1% (95% CI: 24.5%, 45.6%).Conclusions: The utilization of biologic drugs for rheumatic and IBD indications are substantial and expected to rise over time. Biosimilars current uptake is low and did not different greatly between drugs. Traditional formulary-based policies may not be effective at significantly altering biosimilar uptake.

PO-2622 Impact Of Two Recent Major Drug Shortages On Drug Utilization 1400689 Mina Tadrous Impact Of Two Recent Major Drug Shortages On Drug Utilization Drug Utilization Research -> Health policy and governance Background: Over the last decade the number of drug shortages has dramatically increased in Canada and worldwide. Between 2015 to 2017, there were over 1,000 reported drug shortages and the annual number appears to be increasing. Two recent major shortages occurred impacting access to bupropion and valsartan- two very commonly used treatments.Objectives: To measure the impact of product shortages on drug utilization of two recent drug shortages in Ontario Canada.Methods: We conducted a repeated cross-sectional study of bupropion and valsartan between July 1, 2012 and March 31, 2019. Both drugs experienced a significant drug shortage in 2018 and 2019, respectively. Using the public claims data in Ontario. We analyzed the quarterly number of drug claims for each of these medications. We analyzed all drugs including those impacted and not impacted by the specific shortage. We reported the number of claims quarterly for each drug class before and after the drug shortage occurred.Results: We found that the number of claims for bupropion did not significantly decrease after the drug shortage was reported in Q1- 2018, with the number of claims only slightly plateauing at 236,017 claims and no significant reduction in the trend. In contrast, valsartan claims dropped dramatically from 191,732 claims after the shortage occurred to 56,548 claims in Q1-2019, a 71% reduction.Conclusions: Our results highlight the variable impact of shortages on utilization. Although both drug shortages were prolonged in their limitation of products the extent and number of products affected may greatly influence the impact of the drug shortage. This work highlights potential differences in drug shortages that should be used to inform policies to help curb the growing problem.

PO-2624 The Impact Of Immortal Time Bias And The Adjustment Using Modified Kaplan-meier Method And Time-dependent Cox Proportional Hazard Model: A Case Study 1400690 Tomomi Kimura Astellas The Impact Of Immortal Time Bias And The Adjustment Using Modified Kaplan-meier Method And Time-dependent Cox Proportional Hazard Model: A Case Study Methods in Pharmacoepidemiology -> Confounding/Bias - Methods to avoid bias and confounding, including confounder summary scores (e.g., propensity scores, disease risk scores) Background: Immortal time bias is well known, but often overlooked, in the analysis of time-varying exposure during the hospitalization. In our study, we tried to evaluate if the switching antibiotics from intravenous (IV) to oral can shorten the length of hospital stay (LoS) in hospitalized community-acquired pneumonia (CAP) patients, while maintaining clinical effectiveness.Objectives: To calculate adjusted Kaplan-Meier estimates of time to discharge and death, the difference in median time to discharge and adjusted hazard ratios for time to discharge and death in CAP patients with and without switching to oral antibiotics.Methods: Hospitalizations for adult CAP patients in 2010-2018 in a hospital-based administrative data were analyzed. CAP severity was determined by A-DROP score. Discontinuation of IV antibiotics followed by immediate start of oral antibiotics during the stay was defined as treatment switch, which was considered as time-varying exposure. Adjusted Kaplan-Meier estimates of time to discharge and death were estimated by modified Kaplan-Meier method in which the risk set of the time-dependent exposure is dynamically update over time. Also, adjusted hazard ratios for discharge and death were estimated with time-dependent Cox proportional hazard model. In addition, two sensitivity analyses were performed where death was treated as either a competing event or censor.Results: In total, 210,314 hospitalizations were analyzed and median LoS (days) was 8 (mild CAP) to 16 (extremely severe CAP). After stratifying by CAP severity, unadjusted median time to discharge was similar between the switch and non-switch groups, while adjusted median time to discharge (days) was 6 to 15 days shorter in the switch group compared to the non-switch group. Unadjusted difference of Kaplan-Meier estimates of time to death between the switch and non-switch groups became wider as CAP severity worsened, while the differences shrank after the adjustment. Adjusted hazard ratios (95% confidence intervals) for time to discharge and death were 2.4 (2.3-2.4) and 0.7 (0.5-1.2) in mild, 1.9 (1.8-1.9) and 0.6 (0.5-0.6) in moderate, 1.7 (1.7-1.8) and 0.6 (0.5-0.6) in severe and 1.6 (1.5-1.7) and 0.6 (0.5-0.7) in extremely severe group, respectively. Sensitivity analyses showed robustness of the findings.Conclusions: The impact of immortal time bias was huge and appropriate analysis methods for retrospective database study are critical to assess the attribution of time-varying exposures.

PO-2626 Knowledge, Attitude And Practice Of Healthcare Professionals Towards Pharmacovigilance In Alexandria, Egypt- A Cross Sectional Survey 1400691 Mai Salama Medical research institute, Alexandria University, Egypt. Knowledge, Attitude And Practice Of Healthcare Professionals Towards Pharmacovigilance In Alexandria, Egypt- A Cross Sectional Survey Pharmacovigilance -> Other Background: The concept of Pharmacovigilance is still new to healthcare professionals (HCPs) in Arabian countries.Objectives: To compare the knowledge, attitude and practice towards pharmacovigilance between health care professionals in Alexandria, Egypt.Methods: A cross-sectional survey was conducted among 547 pharmacists and physicians in three different health sectors in Alexandria, Egypt from August 2017 to March 2018. The questionnaire consisted of 20 questions. The data were analyzed using SPSS version 20. Bivariate analysis using Chi-square test and multivariate logistic regression were used. The main outcome was measuring knowledge, attitude and practice towards pharmacovigilance. Pareto chart was used to determine the most common barriers affecting the reporting of adverse drug reactions (ADRs).Results: The survey was completed by 232 (42.4 %) physicians and 315 (57.6 %) pharmacists. The odds of high knowledge and positive attitude were around 6 times more among pharmacists than that among physicians (OR= 6.60, 95% CI: 2.31-18.85), and (OR= 5.66, 95% CI: 2.26-14.15) respectively. The odds of high practice was more than twice among pharmacists than that among physicians (OR= 2.62, 95% CI: 1.35-5.05). Major barriers against reporting ADRs were lack of time and difficulty in deciding whether ADR has occurred or not.Conclusions: The limited knowledge about pharmacovigilance among physicians could have affected ADRs reporting incidence. Educational intervention and a practical training program need to be applied by the health authority to physicians in order to enhance the pharmacovigilance and drug safety culture in Egypt.

PO-2629 Validation Of A Us Health Insurance Claims-based Algorithm To Identify Acute Exacerbations Of Chronic Obstructive Pulmonary Disease 1400692 Kieran Rothnie GlaxoSmithKline Validation Of A Us Health Insurance Claims-based Algorithm To Identify Acute Exacerbations Of Chronic Obstructive Pulmonary Disease Methods in Pharmacoepidemiology -> Measurement Methods - Validation methods to confirm data (including Patient Reported Outcomes development and validation); methods to deal with missing data Background: Acute exacerbations of chronic obstructive pulmonary disease (AECOPD) are important events which may lead to adverse outcomes and are a common endpoint in observational research. However, the algorithms to identify AECOPD in US administrative claims data have not been validated.Objectives: To develop claims based algorithms identifying moderate and severe AECOPD, separately, in two US healthcare systems and to evaluate their positive predictive value (PPV) and negative predictive value (NPV).Methods: Computerized healthcare databases (inclusive of insurance claims and electronic health records [EHR] of medical services) from Kaiser-Permanente Mid-Atlantic States and Reliant Medical Group were used (Q1 2010-Q3 2015). Eligible patients were identified using ICD-9 diagnosis codes for COPD. The first prescription fill for a COPD maintenance medication was defined as the index date after which AECOPD were identified for each patient. The algorithm to identify potential moderate AECOPD consisted of outpatient or emergency department visits associated with AECOPD related diagnosis codes and antibiotic or systemic steroid dispensings. A second algorithm to identify potential severe AECOPD consisted of inpatient visits associated with AECOPD related diagnosis codes. Objective criteria for exacerbations were developed from Global Initiative for Chronic Obstructive Lung Disease guidelines. To estimate PPV, a random sample of 300 potential moderate AECOPD and 250 potential severe AECOPD identified by the claims algorithms were abstracted from the EHR and adjudicated by physician review. To estimate NPV, a random sample of 200 patients without any AECOPD identified by the claims algorithms (100 without moderate and 100 without severe AECOPD) during the two years following the index date were abstracted from the EHR and adjudicated by physician review to identify any AECOPD in the EHR missed by the algorithm.Results: The PPV was 98.3% (95% CI: 96.1-99.5) for the moderate algorithm and 96.0% (95% CI: 92.5-98.2) for the severe algorithm. The NPV for the moderate algorithm was 75.0% (95% CI: 65.3-83.1) and was 95.0% (95% CI: 88.7-98.4) for the severe algorithm. Results were consistent across both sites.Conclusions: The PPV for both algorithms and NPV for severe algorithm was very high, which validates the combinations of diagnoses and procedures used in the algorithms. The NPV for the moderate algorithm was modest. Further replication and validation of the algorithm to the post ICD-10 era would bolster the validity and relevance of the findings. Funding: GSK (210043).

PO-2632 Finnished Geneness: Linking Biobank With Electronic Health Record Data - Trk Fusion In A Predefined Finnish Cancer Cohort 1400693 Wei Zhang Bayer AG Spotlight Poster Session Spotlight Poster Genomics Background: Selective tropomyosin receptor kinase (TRK) inhibitors, indicated for patients with solid tumors harboring a gene fusion involving one of the three neurotrophic TRK genes (NTRK1-3), are now approved in the US and Europe. However, systematic estimates of NTRK gene fusion frequency and knowledge on its disease characteristics in each tumor entity are limited. This type of study, which requires linkage of genetic testing data with patients’ electronic health records (EHR), is currently lacking.Objectives: This study aims to evaluate the ability to characterize biobank samples with linkage to secondary care EHR to: 1. Describe the NTRK gene fusion frequency amongst papillary thyroid carcinoma (PTC) cohort. 2. Describe patient demographics, natural history of disease, and clinical journey.Methods: The Southwest Finland province is home to 470,000 residents of Caucasian descent with minimal migration and comprehensive cancer care within the hospital. A retrospective cohort of histologically-confirmed PTC patients with research-consented archival formalin-fixed paraffin-embedded tumour samples were obtained from Auria Biobank. All samples were first stained for pan-TRK protein expression by immunohistochemistry (antibody clone EPR17341) followed by next generation sequencing using the Illumina TruSight170 kit. Characterized genomic data were linked to the Turku University Hospital EHR. Observation period began 1 year prior to diagnosis till end of record/death to describe NTRK gene fusion, other gene alterations, patient demographics, and clinical journey.Results: EHR data is available from 1993 but most complete from 2005 onwards, therefore a cohort of 234 adult PTC patients, diagnosed between 2005 - 2017, were identified. Complete structured data providing information on dates, therapies, and corresponding results are linkable and available for patient demographics, tumour characteristics, comorbidities, operations, chemotherapy, radiotherapy, drugs, lab tests, hospitalizations, and overall survival datasets. More detailed clinical information such as lifestyle risk factors and treatment responses are in unstructured format.Conclusions: Linking biobank and EHR data in a Finnish cohort provides the non-interventional platform needed to study natural history of gene-driven tumour entities. Additional data linkages, such as Statistics Finland and Finnish Cancer registry, is possible. High-level information can be readily found in structured format while more detailed clinical information may require further manual data extraction.

PO-2635 Utilization Of A Booster Dose Of Influenza Vaccine Among Adults With End-stage Renal Disease In The United States 1400694 Leah McGrath NoviSci Utilization Of A Booster Dose Of Influenza Vaccine Among Adults With End-stage Renal Disease In The United States Vaccines Background: Evidence suggests that neither the standard-dose nor high-dose influenza vaccines are effective among patients with end-stage renal disease (ESRD). A second, booster dose of standard-dose vaccine may provide better protection than a single dose, but it is unknown whether a booster dose is used in clinical practice.Objectives: To compare baseline characteristics of patients with ESRD who receive one versus two doses of influenza vaccine in a given influenza season, overall and by influenza season.Methods: Patients with ESRD receiving hemodialysis were identified in the United States Renal Data System during each influenza season from 2010 to 2016. Patients were eligible for multiple seasons. Standard-dose vaccines of any type were identified between August 1 and the end of the influenza season, except for the season starting in 2016, where the data ended on December 31. Baseline demographics were assessed on the date of the first vaccination date. Healthcare utilization measures were assessed during the 30 days before the first vaccination date. We compared patient characteristics of one vs. two-dose recipients using a standardized mean difference (SMD).Results: Of 1,076,535 influenza vaccine recipients, 2.2% received at least two doses of standard-dose vaccine during a single influenza season. This proportion was similar for all influenza seasons. There was little difference between patients with one and two doses with regards to demographics. However, two-dose recipients were more likely to be dual-eligible for Medicaid than one-dose recipients (64% versus 51%; SMD = 0.26). Similarly, two-dose recipients had a higher prevalence of skilled nursing facility stays in the 30 days prior to the vaccination date (10% versus 4%, SMD = 0.26). The pattern was similar for hospitalization days. Vaccination was administered more often in the outpatient setting than in the dialysis clinic, regardless of dose. Among two-dose recipients who received the first dose in the dialysis clinic, 85% received the second dose in an alternate setting.Conclusions: A small proportion (~2%) of vaccinated patients received two doses in one influenza season. Some patients received the second dose in a different setting from the first, perhaps in part due to an inability to transfer medical history data across health care facilities resulting in inadvertent re-vaccination, or a true perceived need for a booster dose. Investigating the effectiveness of a two-dose vaccination strategy could be possible if differences in health care utilization can be addressed in the analysis.

PO-2637 Intentional Self-harm Classification In Medicare: An Interrupted Time Series Analysis Comparing Icd-9 Definitions To Corresponding Icd-10 Definitions 1400695 Richard Swain US Food and Drug Administration Intentional Self-harm Classification In Medicare: An Interrupted Time Series Analysis Comparing Icd-9 Definitions To Corresponding Icd-10 Definitions Methods in Pharmacoepidemiology -> Measurement Methods - Validation methods to confirm data (including Patient Reported Outcomes development and validation); methods to deal with missing data Background: Epidemiologic studies often rely on International Classification of Diseases, Clinical Modification (ICD-CM) codes to classify outcomes. The consistency of ICD definitions from ICD-9 to ICD-10 has not been fully described. Previous studies have estimated validity of ICD-9 codes to classify intentional self-harm (ISH) recorded in medical claims data. With several years since the transition to ICD-10 (Oct 1, 2015), there remains a need to adapt ICD-9 definitions of ISH and estimate their performance in ICD-10.Objectives: To explore temporal patterns in ISH claims, and to perform an Interrupted Time Series (ITS) analysis to examine estimated ISH rates across the ICD-9 to ICD-10 transition.Methods: We identified two ICD-9 ISH definitions with high positive predictive value and adapted them to ICD-10: ISH diagnosis codes (ISH-DC) and an algorithm validated by Patrick and colleagues (PA) requiring injury and mental disorder diagnoses on an inpatient claim. We estimated monthly rates of beneficiaries with ISH events using ISH-DC and PA based on monthly Medicare fee-for-service (FFS) enrollment from 2012 to 2018. We conducted an ITS analysis, adjusting for seasonality using a Fourier transform and autocorrelation using an autoregressive moving average model with order chosen to minimize Akaike information criterion.Results: We identified 2,456 beneficiaries with inpatient ISH events captured by ISH-DC, 4,333 captured by PA, and 4,989 captured by ISH-DC or PA, out of 32,194,551 FFS Medicare enrollees in January 2012, yielding rates per million enrollees (M) of 76.3, 134.6, and 155.0, respectively. By January 2018, rates decreased to 63.0, 106.0, and 126.0, respectively. Each ISH definition displayed seasonal variability, with rates approximately 20% higher during late summer (Jul-Aug) compared to winter. The ITS analysis showed a slight increase in level for event rates per M for ISH-DC [3.4; (95% C.I. 1.7, 5.1)] and PA [11.5 (8.4, 14.6)] across the ICD-9/10 transition. For the ITS analysis classifying ISH as events identified by ISH-DC or PA, the ICD-9 to ICD-10 transition produced a similar estimated effect on level [11.2 (8.5, 13.8)].Conclusions: The adapted ISH definitions resulted in consistent estimates across the ICD transition, with small differences in level indicating a slight increase in the number of ISH events classified in ICD-10. These findings suggest the validity of the adapted ISH definitions may be consistent with the original ICD-9 versions. Further validation efforts are needed to assess the performance of a combined ISH definition.

PO-2638 Cardiovascular Safety Of Sulfonylureas In Type 2 Diabetes: Effect Modification By Obesity? 1400696 Karine Suissa Brigham and Women's Hospital, Harvard Medical School Spotlight Poster Session Spotlight Poster CER Background: Pharmacokinetic studies suggest that adiposity may modify antidiabetic medication efficacy and safety via altered drug distribution and clearance. While sulfonylureas have been linked to a higher cardiovascular (CV) risk, with inconsistent results, the role of obesity remains unclear.Objectives: Our main objective was to evaluate the effect of obesity on CV risk with sulfonylureas as first-line therapies for type 2 diabetes in a clinical practice.Methods: Using the UK’s Clinical Practice Research Datalink, we conducted a cohort study of patients initiating antidiabetic treatment. Initiators of sulfonylureas monotherapy were matched with initiators of metformin monotherapy on a high-dimensional propensity score. Patients were censored at the time of discontinuation or addition of another antidiabetic drug. We used Cox proportional hazards models within BMI categories (2 (normal weight; 26%), 25-30 kg/m2 (overweight; 42%), ≥ 30 kg/m2 obese; 32%)) to estimate hazard ratios (HR) of major adverse cardiovascular events (MACE), individual components of MACE, and all-cause mortality, comparing sulfonylureas with metformin.Results: The initial cohort included 94,750 patients with type 2 diabetes initiating sulfonylureas or metformin. We matched 13,999 initiators of sulfonylureas to an equal number of metformin initiators. The overall HR of MACE comparing sulfonylureas with metformin was 1.23 (95% CI: 1.08-1.39); the HR was 1.33 (95% CI: 1.02-1.74) among obese patients, 1.27 (95% CI: 1.04-1.55) for overweight patients, and 0.99 (95% CI: 0.77-1.28) among normal weight patients. Similarly, the risk of all-cause mortality, CV death and ischemic stroke with sulfonylureas was highest in obese individuals, with evidence of interactions for all-cause mortality (p-interaction: <0.0001) and stroke (p-interaction: 0.05).Conclusions: Sulfonylureas as a first-line monotherapy for type 2 diabetes is associated with a higher risk of MACE compared with metformin. Our data suggest that this effect might be strongest in obese patients, with no increased risk in normal weight patients. To ensure the best response to treatment, BMI level may be considered when choosing metformin over sulfonylureas for first-line treatment.

PO-2639 Hypoglycemia-related Ed Visits And Hospitalizations Among Initiators Of Insulin Analogues Compared To Nph Insulin 1400697 Marie Bradley US Food and Drug Administration Hypoglycemia-related Ed Visits And Hospitalizations Among Initiators Of Insulin Analogues Compared To Nph Insulin Benefit-Risk Assessment, Communication, and Evaluation (BRACE) Background: In clinical trials the risk of severe hypoglycemia did not differ between insulin analogues and neutral protamine hagedorn insulin (NPH) for patients with type II diabetes mellitus (T2DM). A recent observational study examining emergency department (ED) visits and hospitalizations for hypoglycemia in T2DM patients found similar results.Objectives: To examine the risk of ED visits and hospitalizations for hypoglycemia among older T2DM patients who initiated long acting insulin or NPH in real world settings.Methods: A retrospective new-user cohort study of Medicare beneficiaries aged ≥65 years initiating insulin glargine (407,018), insulin detemir (141,588) or NPH (26,402) who did not use other insulins, from January 2007 to July 2019 was conducted. The primary outcome was time to first hypoglycemia-related ED visit or hospitalization, defined using a modified validated algorithm. Propensity score weighted Cox proportional hazards regression was used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). In the secondary analysis, recurring hypoglycemia events were modeled using the Anderson-Gill model.Results: There were 7,347 hypoglycemia-related ED visits or hospitalizations during the study period; 5,194 for glargine, 1,693 for detemir, and 460 for NPH with a median follow-up across the three cohorts of 0.37 years (interquartile range: 0.20 to 0.76). Crude incidence rates (with 95% CIs) were 1.74 (1.69, 1.78), 1.67 (1.59, 1.75), and 3.07 (2.79, 3.35), for glargine, detemir and NPH, respectively. Initiators of glargine and detemir had reduced risk of an outcome event compared to NPH initiators, HRs (95% CIs) 0.69 (0.61-0.77) and 0.69 (0.61-0.79), respectively. When adjusting for concomitant use of oral diabetes medication the HRs did not change substantially; 0.71 (0.63-0.80) and 0.72 (0.63-82). The HR remained unchanged for the recurrent event analysis and the risk of recurrent ED visits and hospitalizations increased with the number of prior events.Conclusions: Initiation of long acting insulin analogues was associated with a lower risk of hypoglycemia-related ED visits and hospitalizations compared with NPH in older T2DM patients.

PO-2644 Tailoring Type Ii Diabetes Treatment: Effect Of 5-htt Polymorphism On Hba1c Levels After Metformin Initiation 1400698 Taichi Ochi University Medical Center Groningen Spotlight Poster Session Spotlight Poster Genomics Background: The serotonin transporter (5-HTT) has been associated with metformin intolerance and may play a role in its efficacy. Tailoring metformin dosage depending on 5-HTT polymorphisms is a potential step towards personalizing the treatment for patients with type 2 diabetes.Objectives: To investigate the effect of 5-HTT polymorphisms on change in HbA1c levels six months after metformin initiation.Methods: Participants of PROVALID (PROspective cohort study in patients with type 2 diabetes mellitus for VALidation of biomarkers) within the GIANTT (Groningen Initiative to Analyse Type 2 Diabetes Treatment) cohort who initiated metformin were genotyped for combined SERT 5-HTTLPR/rs25531 (L*L*, L*S* and S*S*) and 5-HTT VNTR (STin 2.12, 12/- and 10/-) polymorphisms, respectively.Differences in distribution of patient characteristics by 5-HTTLPR genotype tested by ANOVA or Chi-square. Multiple linear regression was conducted to determine change in HbA1c level from baseline to six months, adjusted for differences at start in age, sex, 5-HTTLPR/VNTR genotype groups, triglyceride level, low density lipoprotein level, serum creatinine, comorbidities and use of medications that may affect the outcome.Results: Data on 320 participants, 184 men and 136 women, were analyzed. Mean age was 58.6 ± 8.8 years and index HbA1c was 58.72 ± 14.43 mmol/L. 5-HTTLPR was characterized as L*L*, L*S* and S*S* in 94, 150 and 76 patients. 5-HTT VNTR was categorized as STin2.12, STin2.12/- and STin2.10/- in 74, 157 and 89 patients.Compared to the reference group of L*L*, patients with S*S* had lesser change in HbA1c (B = -6.26, p = 0.044). However, significance was lost after adjusting the regression analysis for further confounders (proton pump inhibitor prescription, low density lipoprotein, serum creatinine), but an effect remains (B = -5.25, p = 0.07). 5-HTT VNTR STin2 polymorphisms were not found to be a significant predictor for difference in HbA1c in six months. Of note, sex and triglyceride level at index were found to be significant covariates in difference in HbA1c levels for the six-month period.Conclusions: Change in HbA1c after metformin initiation may be smaller in patients with S*S* genotype. Other 5-HTT genotyping for polymorphisms did not provide evidence towards its influence on HbA1c change over the course of six months.

PO-2646 Strategies To Approach The Judicialization Of Health Technology In Latin America And Caribbean 1400699 sueli yamauti Luciane Lopes University of Sorocaba, S�o Paulo, Brazil Strategies To Approach The Judicialization Of Health Technology In Latin America And Caribbean Drug Utilization Research -> Health policy and governance Background: The judicialization of health technology in Brazil is a phenomenon that concerns public institutions, as it is growing, cumulative and increases inequities. In the literature, there are several suggested strategies to approach it. On the other hand, several strategies were implemented in Brazil. However, it is not known what type of arrangements for the health system they used in their implementation and whether there are similarities with those suggested in the literature.Objectives: To analyze the feasibility of implementing the proposed strategies to approach the judicialization of health technology in Latin America and the Caribbean in the Brazilian context.Methods: Comparative cross-sectional analysis among the 45 strategies proposals to approach the judicialization of health technology in the public sector, extracted from the literature review by Pinzón-Flórez et al. with the 78 strategies already implemented in Brazil, identified and categorized by Yamauti et al. as to the correspondence between them. The strategies were classified according to six World Health Organization's health system arrangements.Results: Eight strategies proposed in the study by Pinzón-Flórez et al. are consistent with the Brazilian government's policies or programs; 27 are similar in terms of the activities developed by the strategies implemented in some Brazilian context and; 10 have no analogy with the strategies implemented in Brazil. The 78 strategies implemented work using at least two arrangements for health systems and 96.2% of them belong to the service delivery arrangement.Conclusions: The proposed strategies to approach the judicialization of health technology in Latin America and the Caribbean are feasible to be implemented both in the form of government policies or programs and as institutional strategies in the Brazilian context. However, the strategies implemented are complex, encompass various health system arrangements and require a legal apparatus along with a public health system that has an integrated service delivery. These characteristics must be evaluated before implementing them in other contexts.

PO-2650 Online Information Discrepancies In Patient Information Regarding Medicine Use During Pregnancy And Lactation: A Conception Study 1400700 Ulrika N�rby Health and Medical Care Administration Online Information Discrepancies In Patient Information Regarding Medicine Use During Pregnancy And Lactation: A Conception Study Pregnancy and Lactation Background: Consistent, adequate information on the safety of medicines during pregnancy and lactation is important for women’s therapeutic decision making. Discrepancies between information resources may cause confusion, resulting in non-adherence to therapy with risks for both the mother, fetus or breastfed child.Objectives: To analyze the frequency and nature of discrepancies between different online information sources for patients in four European languages.Methods: The study was performed on internet data sources, primarily via Google, in Swedish, Dutch, French and English. The medicines analysed were ibuprofen, ondansetron, olanzapine, fingolimod, methylphenidate and adalimumab, i.e. medicines commonly used during pregnancy and/or for treating chronic diseases in reproductive age. The top search hits with considerable information that could be classified into these data source categories, were collected: i) regulatory sources, ii) scientific sources, e.g. Teratology Information Services (TIS), iii) blogs/forums/social media, iv) commercial websites for patients. The recommendation was categorized as a) Can be used, b) Individual benefit-risk assessment, c) Should not be used, d) Trimester specific, e) Not classifiable. Descriptive analysis was used to identify the discrepancies.Results: Pregnancy recommendations for adalimumab and fingolimod were consistent in all data sources in 3 out of the 4 languages, those for olanzapine were consistent in 2 languages, and pregnancy recommendations were consistent in only 1 language for ibuprofen, methylphenidate and ondansetron. Lactation recommendations for ondansetron were consistent in all data sources in 2/4 languages, for fingolimod and methylphenidate in 1/4 languages. Recommendations for adalimumab, ibuprofen and olanzapine were inconsistent in all languages. TIS center recommendations between countries were discrepant in 7% (2/27) of the pregnancy and 32% (7/22) of the lactation statements.Conclusions: Information discrepancies are frequent for medicine use during pregnancy/lactation, especially for lactation. All major inconsistencies are however important. Compared to other sources, different TIS centers offer improved consistency in recommendations, indicating improved consensus on a scientific level. More work is needed to harmonize information both within and between countries, to avoid conflicting messages.

PO-2651 Correlation Between Antibiotic Consumption And Resistance At An Intensive Care Unit In Serbia 1444330 Ana Tomas University of Novi Sad, Faculty of Medicine Correlation Between Antibiotic Consumption And Resistance At An Intensive Care Unit In Serbia Drug Utilization Research -> Trends and comparisons Background: The paucity of information on quantity and quality of antimicrobial prescribing is one of the important barriers to the successful development and implementation of antimicrobial stewardship programs in Serbia.Objectives: To describe antimicrobial prescribing practices and antimicrobial resistance among gram negative pathogens and to relate exposure to antimicrobials to the development of antimicrobial resistance.Methods: A retrospective, observational study was performed in an intensive care unit in the largest tertiary health care center in Vojvodina, Serbia from 2014 to 2018. Antibiotic prescription data were analysed according to the WHO anatomical therapeutic chemical (ATC)/defined daily dose (DDD) methodology; antibiotic resistance was expressed as incidence density adjusted for total inpatient-days, Targeted organisms included Klebsiella pneumoniae, Pseudomonas aeruginosa and Acinetobacter spp. Individual trends were determined by linear regression, while possible associations between antibiotic prescription and resistance were evaluated using cross-correlation analysis.Results: Overall decrease in antibiotic utilization was observed. The prescription rates of piperacillin-tazobactam increased significantly, while consumption of 3rd and 4th generation cephalosporins and fluoroquinolones decreased. Results over 5 years indicated significantly rising incidence densities of doripenem resistant Acinetobacter spp., piperacillin/tazobactam resistant P. aeruginosa and doripenem, ertapenem, meropenem and colistin resistant Klebsiella pneumoniae. Cross-correlation analysis demonstrated positive association between the use of glycopeptides (R=0.76) and resistance to piperacillin/tazobactam in P. aeuruginosa. Strong positive correlation (R=0.97) has been found between the colistin use and resistance to doripenem in Acinetobaccter spp. For K. pneumoniae, resistance to colistin was strongly positively associated with use of colistin (R=0.89), carbapenems (R=0.73) and imidazoles (R=0.75). Positive association between the use of colistin and carbapenems and carbapenem resistance in K. pneumoniae was identified. There was an inverse relationship between the use of 1st and 2nd generation cephalosporins with the resistance to colistin and carbapenems in K. pneumoniae.Conclusions: In conclusion, our data demonstrated increase in resistance and a correlation between prescription of and Gram-negative bacterial resistance to several antimicrobial agents in Serbian hospital.

PO-2656 Evaluation Of Spontaneous Adverse Event Reports For Drug Abuse, Misuse, Dependence Or Off-label Use Attributed To Pregabalin Within The Fda Adverse Events Reporting System (faers) 1400701 Ahmed Elkersh High Institute of Public Health, Alexandria University Evaluation Of Spontaneous Adverse Event Reports For Drug Abuse, Misuse, Dependence Or Off-label Use Attributed To Pregabalin Within The Fda Adverse Events Reporting System (faers) Pharmacovigilance -> Neurological/Mental Health Background: Pregabalin abuse and dependence have become increasingly reported. However, no study has analyzed spontaneous reports of adverse events with this drug in the most recent years that led to worldwide regulatory actions.Objectives: To describe and evaluate spontaneous adverse event reports of drug abuse, misuse, dependence or off-label use attributed to pregabalin.Methods: We identified adverse events reports listing pregabalin in the FDA Adverse Events Reporting System (FAERS) from January 2014 to September 2019. We used the terms “pregabalin,” and “Lyrica,” to identify pregabalin. The “Intentional product misuse,” “drug abuse,” and “off-label use” MedDRA terms were used to identify the drug abuse, misuse, dependence and off-label use events. We calculated the proportional reporting ratio (PRR) and the reporting odds ratio (ROR) for this drug.Results: Among 3,351,770 FAERS reports between January 2014 and September 2019, around 28890 (1%) reports mentioned pregabalin. There were 168064 (5%) drug abuse, misuse, dependence and off-label use reports during the same period of which 3723 (2%) was attributed to pregabalin. The PRR was 2.605 (95% CI: 2.527, 2.685), and the ROR was 2.843 (95% CI: 2.746, 2.943).Conclusions: Although FAERS is subject to significant limitations, the results suggest that pregabalin has a statistically significant association with an increased incidence of drug abuse, misuse, dependence and off-label use. pregabalin prescribers should evaluate the patient’s history of drug abuse and monitor carefully all possible adverse effects.

PO-2657 Sensitivity Of Treatment Patterns To Grace Period Selection In Medications With Tolerability Issues: A Real-world Example Of Hedgehog Inhibitor Use In Basal Cell Carcinoma 1400702 Wenzhen Ge Regeneron Pharmaceuticals Jessica Jalbert Regeneron Pharmaceuticals Sensitivity Of Treatment Patterns To Grace Period Selection In Medications With Tolerability Issues: A Real-world Example Of Hedgehog Inhibitor Use In Basal Cell Carcinoma Drug Utilization Research -> Trends and comparisons Background: Hedgehog inhibitors (HHIs) are oral targeted therapies approved for the treatment of advanced basal cell carcinoma (BCC). As patients commonly experience adverse events that may be managed through treatment interruptions, we hypothesized that HHI treatment patterns may be sensitive to grace period (GP) duration.Objectives: To assess the impact of GP selection on patterns of HHI treatment interruptions (TI), treatment duration, and re-initiations (RI) among patients with BCC.Methods: This was a retrospective cohort study using MarketScan Commercial/Medicare databases (01/01/2013-09/30/2018). We identified new users of HHIs (index date = date of the first dispensation), ≥18 years of age who were continuously enrolled for ≥6 months prior to the index date (i.e. baseline) with ≥1 baseline BCC diagnosis. TIs were defined as a lack of dispensation following the exhaustion of days’ supply and allotted GP. RI was defined as ≥1 HHI dispensation after TI. The Kaplan-Meier method was used to estimate time to HHI TI and, among patients with a TI, incidence of RI. HHIs are generally dispensed in 30-days’ supply and indicated as long as a patient derives a clinical benefit, however, since TIs are commonly employed during HHI therapy, sensitivity analyses were conducted using GPs of 14, 30, 60, 90 and 120 days.Results: We identified 469 patients with a BCC diagnosis initiating HHIs. The mean (SD) age was 67.6 (15.8) years, 64.2% were men, 51.2% were covered by commercial insurance and 99.2% initiated vismodegib. Using a GP of 14, 30, 60, 90 and 120 days, the risk of HHI TI was 79.2%, 68.8%, 60.0%, 55.4% and 51.4% at 6 months and 94.8%, 91.3%, 88.2%, 83.2% and 80.2%, at 1 year, respectively. Median HHI treatment duration ranged from 94 days (95% CI: 90-109) using a GP of 14 days to 173 days (95% CI: 156-194) using a GP of 120 days. At 6 months following TI, incidence of RI was 35.8%, 19.8%, 11.3%, 6.9% and 4.9% using a GP of 14, 30, 60, 90 and 120 days, respectively. The incidence of HHI RI at 1-year following TI ranged from 40.8% using a 14-day GP to 13.4% using a 120-day GP.Conclusions: Characterizing real-world persistence on HHIs is challenging due to the use of drug holidays, as TIs, treatment duration, and RIs were sensitive to the duration of the GP selected. Real-world studies of medications with tolerability issues should consider the use of drug holidays in order to accurately characterize drug exposures.

PO-2659 Multi-level Propensity Scores In Comparative Effectiveness Research On Medical Devices: A Pragmatic Example Of Partial Vs Total Knee Replacement Using National Joint Registry Data Compared To Rct Findings 1400703 Daniel Prieto-Alhambra CSM, NDORMS, University of Oxford Spotlight Poster Session Spotlight Poster Medical Devices Background: Propensity scores (PS) are widely used in pharmacoepidemiology, but little is known about how surgeon-related confounders should be incorporated in PS in the study of implantable devices.Objectives: To test the performance of different methods to account for surgeon level confounders to emulate the TOPKAT surgical RCT comparing partial (PKR) vs total knee replacement (TKR) [Beard D et al. Lancet 2019].Methods: We identified trial eligible patients undergoing PKR or TKR in the UK National Joint Registry linked to hospital inpatients data and patient reported outcome measure. We calculated PS using four different strategies: a) Using 18 clinically defined patient confounders in logistic regression (LR), b) as A but adding surgeon volume in LR c) LR random effects model with surgeon pseudonymised id as a random intercept (REM), and d) like C but with both surgeon- and patient-level confounders included in the PS equation. We used the resulting PS to compute 4 different stabilized inverse probability weights. We used non-parametric marginal estimators to estimate post-operative Oxford Knee Scores (OKS) differences as treatment effects between PKR and TKR. We compared the resulting estimators to TOPKAT and specified four agreement criteria: I20.05, full coverage of the estimate and 95CI, and statistical significance agreement.Results: We identified 602 and 114,871 patients undergoing PKR and TKR, and 2,625 surgeons. Mean surgical volume in the previous year was 5 surgeries for PKR and 67 for TKR. PS based on patient variables (method a) resulted in a postoperative OKS difference of 0.87 95%CI (-0.20 to 1.94); PS using multilevel modelling (method c) resulted in 1.50 (-0.44 to 3.44). Adding surgeon volume in a LR-based PS (method b) resulted in an estimator of 1.41 (0.25 to 2.57), compared to 2.11 (0.16 to 4.06) for the REM-based PS including patient and surgeon variables (method d). Compared to TOPKAT (treatment effect 191 (0.20-3.62)), method b passed all four agreement criteria, followed by method d with 3/4.Conclusions: Methods including surgeon-level confounders in the PS either in single level or multi-level models (methods b and d) are the preferable methods, indicating the importance of incorporating surgeon features in PS models for the study of implantable devices.

PO-2663 Quantification Of Adverse Drug Reactions Related To Drug Switches In The Netherlands 1400704 Pieter Glerum Quantification Of Adverse Drug Reactions Related To Drug Switches In The Netherlands Pharmacovigilance -> Other Background: From a pharmacological perspective, it is unexpected that a different clinical profile is observed for a generic drug as compared to a brand-name drug. Nor that an adverse drug reaction (ADR) is experienced following a switch between bioequivalent drug products. However, ADRs related to drug switching are reported and there is a need to systematically study these, to elucidate points for improvement on the pharmacological and psychological assumptions of generic interchangeability.Objectives: The primary aim is to calculate the relative number of reported ADRs per number of drug switches, for active substances for which a relatively high number of ADRs was reported in the Netherlands. In addition, we examined whether different peaks of ADRs would be observed using the absolute number of ADRs or using switch-corrected number of ADRs.Methods: This is a retrospective cohort study in the Dutch patient population on active substances for which a relatively high number of ADRs related to drug switching was reported. For this, we analyzed drug switches from the National Health Care Institute in the Netherlands and reported ADRs related to switching from Lareb, the Netherlands Pharmacovigilance Centre, between June 1, 2009 and December 31, 2016 for a selection of 20 active substances. We also compared pharmacovigilance analyses based on the absolute, switch-corrected and user-corrected numbers of ADRs.Results: For our selection of 20 active substances, 1,348 reported ADRs and over 23.8 million drug switches were identified. There was no overall correlation between the number of ADRs and the number of switches, but on average we found 5.7 reported ADRs per 100,000 switches. The number was relatively high for rivastigmine, levothyroxine, methylphenidate and salbutamol, with 74.9, 50.9, 47.6 and 26.1 ADRs per 100,000 switches, respectively. When comparing analyses using the absolute number and the switch-corrected number of ADRs, we demonstrate that different active substances would be identified as having a relatively high number of ADRs, and different time periods of increased numbers of ADRs would be observed. We also demonstrate similar results when using the user-corrected number of ADRs instead of the switch-corrected number of ADRs, allowing for a more feasible approach in pharmacovigilance practice.Conclusions: The number of drug switch-related ADRs is relatively high for rivastigmine, levothyroxine, methylphenidate and salbutamol. Further, we demonstrate that analyses of drug switch-related ADRs leads to different results when the number of reported ADRs is corrected for the number of drug switches.

PO-2666 The Predictors Of Major Gastrointestinal Bleeding Among Oral Anticoagulant New Users With Atrial Fibrillation 1400705 Jakub Qazi Faculty of Pharmacy, Universit� de Montr�al The Predictors Of Major Gastrointestinal Bleeding Among Oral Anticoagulant New Users With Atrial Fibrillation Safety End Points -> Cardiovascular Background: The real-world characteristics of patients with major gastrointestinal bleeding (GIB) have been well-studied among warfarin users, but less so among direct oral anticoagulant (DOAC) users in patients with non-valvular atrial fibrillation (NVAF).Objectives: To assess the cumulative incidence of GIB for each DOAC (dabigatran, rivaroxaban and apixaban) and warfarin. To identify GIB predictors for each class of treatment.Methods: We built a cohort of hospitalized NVAF patients using Quebec hospitalization (Med-Echo) and drugs and medical services (RAMQ) administrative databases from 2011 to 2018. Each participant needed to be enrolled in the insurance plan for a year prior to and at cohort entry, which was defined as the first OAC claim. We determined age, sex, co-morbidities during the AF hospitalisation and 3 years prior to cohort entry using ICD-9 or ICD-10 codes as well as concomitant drug use 2 weeks prior to cohort entry. Patients were new users, defined as having no OAC claims in the year prior to cohort entry. The exposures were warfarin

PO-2668 Cost Of Illness Studies Using The Clinical Practice Research Datalink: A Review Of Costing Methodologies 1400706 Elaine Stamp PHMR Ltd Preveina Mahadevan PHMR Ltd Cost Of Illness Studies Using The Clinical Practice Research Datalink: A Review Of Costing Methodologies Health Economics/Outcomes Research Background: The linked Clinical Practice Research Datalink (CPRD)-Hospital Episode Statistics (HES) dataset is increasingly being used to investigate resource utilisation and costs associated with specific patient groups. In order to ensure this work produces reliable and comparable findings it is important that they use appropriate and consistent methods.Objectives: To review methodologies used to estimate costs of resource utilisation from the CPRD.Methods: The Ovid MEDLINE(R) database was searched for keywords relating to studies that investigated cost of illness (COI) using CPRD data. All COI studies published between 2017 to 2019 were identified and screened. Information on the source of cost data and the methodology used to attribute costs to episodes of healthcare resource use were extracted and reviewed. All data extraction was performed by two independent reviewers and all decisions were reached by consensus with an additional third reviewer.Results: A total of 30 COI studies were identified. Of the 27 studies that costed GP visits, 26 (96%) used Personalised Social Services Resources Unit (PSSRU). Of 25 studies that included GP prescription costs, 10 studies (40%) used the Prescription Cost Analysis (PCA), 11 (44%) used the British National Formulary (BNF), 3 (12%) used NHS Drug Tariff, one study used the RESIP Gem- script Code Dictionary and one study used the Dictionary of Medicines and Devices browser. One study used PCA and BNF data and one used PCA and NHS Drug Tariff data. Studies which were non-specific regarding the source of cost data included one for GP data and two for prescriptions. Of 26 studies that costed inpatient admissions, 17 (65%) used NHS reference costs, 7 (27%) used the National Tariff and 2 (8%) used both these sources. Amongst these, 20 (77%) stated they additionally used the Healthcare Resource Grouper (HRG) software. Of only 18 studies that costed outpatient visits, 16 (89%) used NHS reference costs where 9 studies (56%) defined the cost by speciality, 2 (11%) used National tariff and one study used PSSRU.Conclusions: While there is some consistency in the methodologies used to attach costs to resource utilisation data from the CPRD, variation in the choice of sources of unit cost data and their application was observed. Incomplete reporting of methodologies also limited the extent to which the consistency of approaches could be determined. A more standardised and transparent approach to calculating total and relative costs using the CPRD-HES database is needed, this will allow for comparison of results and effective evaluation of methodologies.

PO-2669 Addressing Treatment Implementation Bias In The Construction Of High Dimensional Propensity Scores 1400707 Nicolas Thurin Universit� de bordeaux Addressing Treatment Implementation Bias In The Construction Of High Dimensional Propensity Scores Methods in Pharmacoepidemiology -> Confounding/Bias - Methods to avoid bias and confounding, including confounder summary scores (e.g., propensity scores, disease risk scores) Background: Formally, covariate for propensity score (PS) construction are assessed before treatment onset. However differences occurring in patient journeys after the decision to treat and before the treatment onset may bias the score (e.g. pre-chemotherapy assessment).Objectives: To address treatment implementation bias induced by differential patient journeys between the decision to treat and the treatment initiation.Methods: A comparative effectiveness study was set to compare cancer treatments using the French nationwide healthcare database (SNDS). Successive high-dimensional PS (hdPS) were constructed to estimate the probability to be treated by an oral versus intravenous agent, aiming to minimize standardized differences (SD) >10% and C-statistic value, after 1:1 matching. All models included 100 empirical covariates and differed by the covariate assessment period length and extra fixed variables.Results: Model 1 used a pre-exposure covariate assessment period of 1 year and showed 2 different patterns of hdPS distribution, leading to 273 matched patient-pairs out of 1213 potential (C-statistic =0.713). Model 2 excluded the month preceding treatment initiation from the covariate assessment period [-12; -1 months], leading to 830/1213 matched pairs (C-statistic =0.586; 12 independent variables with SD >10%). Model 3 extended the covariate assessment period [-24; -1 months], and included dispensing of previous anticancer agents as fixed variables. Cancer age and hdPS were used for matching, leading to 716/1213 patient-pairs (C-statistic =0.603; 5 independent variables with SD >10%).Conclusions: Pre-exposure window should be routinely assessed and potentially excluded when constructing PS, especially when patients have followed distinct care pathways. Ideally, covariate assessment period should stop at the time treatments are decided upon and not yet started.

PO-2671 Increased Risk Of Falls And Fractures In Patients With Psychosis And Parkinson Disease 1400708 Joan Forns RTI Health Solutions Increased Risk Of Falls And Fractures In Patients With Psychosis And Parkinson Disease Disease Epidemiology/Clinical Course -> Neurological/Mental Health Background: Psychosis is a common complication of Parkinson disease (PD), particularly at advanced stages of the disease, with prevalence up to 75%. Both PD and PD with psychosis (PDP) have been implicated as risk factors for falls and fractures due to cognitive and motor impairment. Yet it has not been established whether the risk of falls and fractures is higher in patients with PDP compared with patients with PD.Objectives: To evaluate whether the risk of falls and fractures differed in patients with PD or PDP at a similar disease stage.Methods: This study was conducted using the MarketScan (IBM Watson Health) commercial and Medicare claims databases from 2008 to 2018. Patients with a PD diagnosis without psychosis were followed from the first identified PD diagnosis date during the study time period. After a noted psychosis diagnosis, patients were included in the PDP cohort. Patients with PD and patients with PDP were matched at similar points in their disease trajectories using a sequential propensity score matching approach. Falls and fractures were captured both as composite and separate outcomes. The incidence rates (IR), expressed per 100 person-years, and 95% confidence intervals (CI) of falls and fractures were estimated. Incidence rate ratios (IRRs) compared patients with PDP with patients with PD without psychosis in the matched samples.Results: We identified 154,306 patients with PD without psychosis who did not have falls or fractures before the index date; the IR for falls and fractures was 11.41 events (95% CI, 11.29-11.53). Most of the events were falls. There were 12,127 (7.8%) patients with a subsequent diagnosis of PDP that met the inclusion/exclusion criteria. These patients with PDP had higher prevalences of most comorbidities and risk factors for falls and fractures than those without psychosis. The IR of falls and fractures among patients with PDP was 29.03 (95% CI, 28.27-29.81); falls were most commonly recorded. After matching, 24,144 (15.6%) patients with PD without psychosis and 12,077 (99.6%) with PDP were retained. Within the matched groups, patients with PDP had higher incidences of falls and fractures than those without psychosis (IRR = 1.44; 95% CI, 1.39-1.49). The increased rate was noted separately for falls (IRR = 1.48; 95% CI, 1.43-1.54) and any fractures (IRR = 1.17; 95% CI, 1.08-1.27) as well as within specific types of fracture, including pelvis and hip fractures.Conclusions: The present study suggests a modest but precise and consistent increased risk of falls and fractures in patients with PDP compared with patients with PD.

PO-2673 Differential Cancer Screening During Follow-up As A Source Of Bias In Pharmacoepidemiology 1400709 Sophie Mayer Differential Cancer Screening During Follow-up As A Source Of Bias In Pharmacoepidemiology Methods in Pharmacoepidemiology -> Analytical Methods - e.g., marginal structural models, instrumental variables, sensitivity analyses Background: Certain medications are associated with receipt of cancer screening through health-seeking behavior, comorbidity, and healthcare access and quality. In these settings, screening after the start of follow-up acts as an intermediate to cancer detection and may result in the appearance of non-causal relationships between medication use and cancer risk.Objectives: Evaluate the impact of screening during follow-up in pharmacoepidemiologic studies of cancer risk using simulation.Methods: We simulated 200 cohorts of 100,000 women in which “high healthcare access” impacted both treatment initiation and screening; true disease risk was independent of treatment and healthcare access (odds ratios [ORs]: 1.0). Disease detection within 5 years of drug initiation was affected by true disease status and ever-receipt of screening, with screening conferring a higher probability of cancer detection than through clinical surfacing (85 vs. 40%). For the purposes of this study, we assumed 100% screening specificity, and varied the associations between healthcare access and treatment or screening (ORs: 1.0, 2.5, 10.0). We also varied the marginal screening prevalence in the population between 10, 25, 50, 75, and 90%. We calculated observed 5-year risks and risk ratios (RRs) based on detected cases only, and computed mean bias.Results: In total, 45 scenarios were simulated. Because both screening and clinical surfacing had <100% sensitivity, marginal cancer risk estimates were consistently lower than true cancer risks, with greater bias observed at lower screening prevalence. There were no notable differences in the proportion screened and therefore no bias in the observed RR when healthcare access was not associated with either treatment or screening. In scenarios where healthcare access was associated with both treatment and screening, the differences in the proportion screened between treatment groups ranged from 1.7-25.5 percentage points, in line with literature supporting the existence of differential screening in some clinical settings (e.g., glucose-lowering drugs, statin adherence). The resulting mean bias in the ln(RR) ranged from 0.012 to 0.183; the greatest bias was consistently observed in scenarios with a marginal screening frequency of 50%.Conclusions: In the setting of a null treatment effect, strong associations between healthcare access, treatment, and post-initiation screening generated notable bias in the observed RR. The presence of shared predictors of treatment and outcome detection in observational research should be assessed and considered in the interpretation of study findings.

PO-2674 Characterization Of Women With Adenomyosis Identified In A Large U.s. Electronic Health Record Database With And Without Hysterectomy 1400710 Stephanie Chiuve AbbVie Characterization Of Women With Adenomyosis Identified In A Large U.s. Electronic Health Record Database With And Without Hysterectomy Disease Epidemiology/Clinical Course -> Other Background: Adenomyosis is the growth of endometrial tissue into the myometrium. Data on symptoms and comorbidities of adenomyosis have been limited primarily to women diagnosed via uterine histology after a hysterectomy. Although “endometriosis of the uterus” may be used as a surrogate for documentation, emerging hallmarks of adenomyosis do not overlap with endometriosis, and the lack of an ICD code for adenomyosis has limited studies of women living with adenomyosis.Objectives: To characterize symptoms and comorbidities of women with affirmed adenomyosis, stratified by timing of hysterectomyMethods: Women aged 18‐55 years for whom adenomyosis was mentioned within clinical notes from 2014-2018 were identified using natural language processing of the Optum Electronic Health Record (EHR) database. The anchor date was the date of first affirmed mention of adenomyosis. Covariates were assessed in the 12 months prior to anchor date (baseline period).Results: Among the 19,503 women with an affirmed mention of adenomyosis, 27% had an ICD code for “endometriosis of the uterus”. Among these women, 84% were ≥36 years old, 68% were white, and 45% were obese. In the baseline period, 50% of women had a transvaginal ultrasound, 6% had a pelvic MRI, and 52% had a hysterectomy. Of women who had a hysterectomy, 31% occurred prior to, 56% occurred on and 23% occurred after the anchor date. Gynecologic comorbidities among women without hysterectomy, or women with hysterectomy timed prior to, at, or after anchor date, respectively, were endometriosis (19%, 72%, 32% and 28%), uterine fibroids (18%, 59%, 42% and 26%) and gynecologic cancer (7%, 16%, 10% and 7%). Gynecologic symptoms among women without hysterectomy, or women with hysterectomy timed prior to, at, or after anchor date, respectively, were heavy menstrual bleeding (47%, 78%, 70% and 66%), anemia (19%, 31%, 24% and 21%), pelvic pain (31% 50%, 35% and 42%), dysmenorrhea (18%, 38%, 33% and 30%), and dyspareunia (7%, 14%, 10% and 11%).Conclusions: While symptoms and comorbidities of adenomyosis may be documented more consistently when diagnosis was made at or after hysterectomy, the ICD code for “endometriosis of the uterus” was documented for <1/3 of women with affirmed adenomyosis, missing 2/3 of diagnosed adenomyosis cases. This suggests that ICD-driven discovery is subject to misclassification and potential selection bias. Natural language processed clinical notes enabled the characterization of women living with adenomyosis that had not been studied previously.

PO-2677 Prevalent New Users: When Is Time Not Enough? 1400711 Michael Webster-Clark UNC Gillings School of Global Public Health Prevalent New Users: When Is Time Not Enough? Methods in Pharmacoepidemiology -> Study Design - Studies assessing study design (e.g., choice of comparator, self-controlled methods, avoiding immortal time bias) Background: The prevalent new user design estimates causal effects of initiating a new drug (Drug A) by matching initiators of A to non-initiators on time since initiating an older drug (Drug B). Initiators of A may switch directly from B (direct switchers) or start A after a period off B (delayed switchers).Objectives: We aimed to describe when matching only on time since initiating B may be biased.Methods: We simulated 40,000 individuals initiating B at time 0 and followed for 4 units of time or until “death”. After time 0, individuals could switch to A (direct switchers), stay on B, or discontinue B. Discontinuers of B could then restart B or start A (delayed switchers). Stopping, switching, or restarting were a function of three covariates (two binary and one continuous) such that approximately 50% of A initiators were direct switchers and 50% were delayed switchers. We present the results of two illustrative scenarios here. In scenario 1, outcomes depended on time since starting B (RR: 1.1 per unit time). In scenario 2, outcomes depended on total exposure to B (RR: 0.70 per total exposure). Outcomes were also a function of current treatment and the three covariates. In each scenario, we estimated a hazard ratio (HR) of death had the A initiators instead continued on or restarted B. We implemented two analytic approaches to estimate standardized morbidity ratio weights to adjust for confounding. In the first, we fit logistic regression models stratified by time since starting B (analogous to matching on time since starting B). In the second, the models were stratified by full treatment history (analogous to matching on B treatment history, accounting for stopping and restarting). We estimated mean values and 95% confidence intervals (CIs) from the 500 replicates.Results: When time since B initiation was a cause of the outcome (scenario 1, true HR: 0.60, crude HR: 0.76, 95% CI 0.72-0.79), both analytic approaches were unbiased (time-stratified: 0.60, 95% CI 0.57-0.63; treatment history-stratified: 0.60, 95% CI 0.57- 0.63). When total exposure to B was a cause of the outcome (scenario 2, true HR: 0.93, crude HR: 1.26, 95% CI 1.19-1.34), however, time-stratified analysis was biased (HR: 1.03, 95% CI 0.97-1.10) and analysis that stratified on treatment history was not (HR: 0.93, 95% CI 0.87, 1.00).Conclusions: If outcomes depend only on measured covariates and time since starting B, conditioning on time since B initiation and adjusting for confounders estimates an unbiased treatment effect in initiators of A. If outcomes depend on past treatment in other ways (e.g. total exposure to B, recently restarting B, or time spent off treatment), matching only on time since initiating B is insufficient.

PO-2678 Individualized Glycemic Control In Type 2 Diabetes In The Us And Europe 1400712 Tongtong Wang Merck & Co., Inc. Individualized Glycemic Control In Type 2 Diabetes In The Us And Europe Disease Epidemiology/Clinical Course -> Diabetes Background: Clinical guidelines in type 2 diabetes recommend individualization of glycemic goals, however clinical databases and real-world studies typically do not capture this measure.Objectives: We used the 2018 Adelphi T2DM Disease Specific Programme (where individualized HbA1c goals are available) to understand the relationship between individual glycaemic goals, factors associated with goal attainment and treatment modification.Methods: The 2018 Adelphi T2DM Disease Specific Program included a survey involving 730 physicians and 8794 patients in US and 4 European countries (Germany, Italy, Spain, and United Kingdom). Data on age, sex, body mass index (BMI), HbA1c, comorbid medical conditions and prescriptions were collected. Antihyperglycemic treatment regimens included were Metformin, Sulfonylureas (SUs), Thiazolidinediones (TZDs), dipeptidyl peptidase-4 inhibitors (DDP4i), glucagon-like peptide-1 receptor agonist (GLP-1RA), sodium-glucose cotransporter-2 inhibitors (SGLT2i), insulin plus any combination of the above. Patients had been prescribed current antihyperglycemic therapy >3 months were included in the analysis. Physician-reported target HbA1c values for individual patients were used to assess mean individual HbA1c goal.Results: Our results suggested that the most recent mean HbA1c was 7.3%, mean individual HbA1c goal was 6.8% [Standard Deviation (SD)=0.69] with 38% patients having achieved their individualized goal. Patients achieving their individual HbA1c goal were older (62 vs 59 years), had lower BMI (29.5 vs 30.3), slightly lower diabetes duration (6.4 vs 6.8 years) or more likely to have chronic kidney disease (8.3% vs 7.5%). Mean HbA1c at diagnosis of type 2 diabetes was 8.4% (SD=1.8). Switch from one line of therapy typically occurred at higher mean HbA1c levels >8%. More specifically, the mean HbA1c in 2,348 patients who switched from 1st line of therapy was 8.3% (SD=1.4%). The mean HbA1c in 624 patients who switched from 2nd line of therapy was 8.4% (SD=1.3%). The mean HbA1c in 126 patients who switched from 3rd line of therapy was 8.5% (SD=1.6%). The mean HbA1c in 17 patients who switched from 4th line of therapy was 9.3% (SD=1.8%).Conclusions: The mean individualized HbA1c goal was 6.8% and was attained by 38% of the patients. Switch from one line of therapy typically occurred at higher mean HbA1c levels >8%.

PO-2681 Asthma Maintenance Medication Users In The Sentinel Distributed Database 1400713 Marie Bradley US Food and Drug Administration Asthma Maintenance Medication Users In The Sentinel Distributed Database Drug Utilization Research -> Trends and comparisons Background: Recent legislation mandates that the US Food and Drug Administration evaluate the potential use of real-world evidence (RWE) to support regulatory decision making. The feasibility of using the Sentinel Distributed Database (Sentinel) to evaluate comparative effectiveness and safety of asthma maintenance medications (AMM) in real world settings is unclear.Objectives: To examine demographic and clinical characteristics of asthma patients dispensed AMM in Sentinel.Methods: We conducted a retrospective cohort study in Sentinel from 2008 to 2018. We defined asthma patients as enrollees aged 18 years or greater with an index event of an ICD-9 or -10 code for asthma and with 1 or more dispensings of AMM during the 12 months prior to the index event (baseline period). AMM included inhaled corticosteroids (ICS), long-acting beta-agonists (LABA), long-acting muscarinic antagonists (LAMA), inhaled combination products, leukotriene receptor antagonists (LTRA), and biological products for asthma. Asthma patients with ICD codes for common respiratory comorbidities in the baseline period were excluded. We calculated summary statistics for selected demographic and clinical characteristics during the 12-month baseline period and 12-month follow-up periods.Results: A total of 3.32 million asthma patients met our eligibility criteria; their mean age was 57 years (standard deviation [SD] = 15.2 years) and 69.3% were female. During the baseline period, the mean Charlson/Elixhauser combined comorbidity score was 1.8 (SD = 1.8), and the mean number of ambulatory encounters and filled prescriptions were 19 (SD = 16.4) and 42.3 (SD = 35.5), respectively. In the combined baseline and follow-up periods, the proportion of patients that filled ≥2 prescriptions for ICS was 34.1%, for ICS/LABA combination products was 43.8%, and for LTRA 37.2%. During the follow-up period, 14% of patients had ≥1 respiratory event coded as asthma exacerbation (AEx), 37.9% of patients filled ≥1 oral corticosteroid (OCS) prescriptions, and 1.9% had an AEx that required hospitalization. Conclusions: To our knowledge, these data represent the first comprehensive characterization of asthma patients in Sentinel. The burden of healthcare encounters, prescriptions, and comorbidities provides insight on the health status of these asthma patients. The inconsistency between the proportion of patients with ICD codes for AEx and the proportion filling OCS prescriptions warrants further examination. These exploratory data inform the feasibility of using Sentinel for future analyses of the potential benefits and risks of asthma medications.

PO-2685 Applying Machine Learning To Define Phenotypes Of Women With Endometriosis Undergoing Laparoscopic Surgery 1400714 Andrea Chomistek Optum Epidemiology Applying Machine Learning To Define Phenotypes Of Women With Endometriosis Undergoing Laparoscopic Surgery Disease Epidemiology/Clinical Course -> Other Background: Endometriosis is an often chronic, inflammatory gynecologic condition; endometriosis-related pain, dysmenorrhea (DYSM), and dyspareunia (DYSP) can impair quality of life. Laparoscopic surgery is used as first-line therapy to remove or ablate endometriosis to relieve pain or performed after failure of medical therapy.Objectives: To identify phenotypes of women who have had laparoscopic surgery to treat endometriosis defined by patterns of clinical covariates.Methods: This study was conducted in the Optum Research Database, a claims database from a large U.S. health insurer. Women aged 18-55 years who had a laparoscopic surgery (LS) within the abdomen or pelvic cavity (excluding hysterectomy and oophorectomy), in conjunction with an ICD-10 code for endometriosis within 30 days of surgery, between July 2016 and June 2018 were identified. Machine learning using an agglomerative hierarchical clustering analysis identified subgroups based on comorbidities, procedures, and medications within the 6 months prior to LS and a phenotype was defined for each subgroup based on clinical characteristics.Results: The cohort included 4,878 women with LS with a mean age of 34 (SD: 8) years. Pain (pelvic, abdominal, lower back) was the most prevalent symptom (81%) and opioid analgesics (52%) and prescription nonsteroidal anti-inflammatory drugs (46%) were the most prevalent medications.An 8-cluster solution was identified based on cluster statistics and covariate distributions. Among the 8 clusters, one cluster predominantly had endometriosis of the ovary, along with low frequency of pain medication, pain and mental health conditions (6%). Four clusters predominantly had endometriosis of the pelvic peritoneum and the phenotypes identified were: 1) low medication use, DYSM, DYSP, pain comorbidities and mental health conditions (11%); 2) high DYSM and DYSP but low pain medication (14%); 3) infertility and DYSM (5%); and 4) high DYSM, DYSP, pain comorbidities, pain medication, and mental health conditions (11%). Finally, 3 clusters had endometriosis of multiple sites and the phenotypes identified were: 1) high cardiometabolic comorbidities and pain medication use, but low DYSM and DYSP (11%); 2) high DYSM, DYSP, pain comorbidities, pain medication and poor mental health (21%); and 3) heavy menstrual bleeding and leiomyoma (21%).Conclusions: Among patients with endometriosis who underwent laparoscopic surgery, we identified several phenotypes based on varying patterns of clinical covariates. Future studies might consider whether unmet needs for surgical avoidance differ by patient phenotype.

PO-2687 Incidence And Risk Factors Of 30-day Readmission After Hospitalization For Chemotherapy Among Acute Lymphocytic Leukemia Patients 1400715 Phuong Tran University of Florida College of Pharmacy Incidence And Risk Factors Of 30-day Readmission After Hospitalization For Chemotherapy Among Acute Lymphocytic Leukemia Patients Rare Disease Background: Due to immunodeficiency and intensive treatment, patients with acute lymphocytic leukemia often require hospitalization for chemotherapy, placing them at high risk for unintentional readmissions.Objectives: To identify the incidence, top causes, and risk factors of 30-day readmission after chemotherapy admission among children and adults with acute lymphocytic leukemia.Methods: This was a retrospective cohort of acute lymphocytic leukemia patients who received at least one inpatient chemotherapy treatment in 2016 using National Readmission Database. We excluded patients without risk for readmission such as those that died or transferred during the index admission, not a resident in the state, or index admission was in December. Causes of readmission were classified by Clinical Classifications Software Refined using primary and secondary diagnoses codes. For risk factor analysis, we applied 3 different outcome definitions: (1) nonelective readmission which was based on readmission type reported by hospitals, (2) unplanned readmission was defined by the algorithm developed by the Center for Medicare and Medicaid, and (3) unintended readmission was the combination of the two previous outcome measures. Patients were followed 30 days after discharge to assess the outcomes. We used a weighted survey procedure to calculate incidence and unweighted, multivariable Poisson regression with robust estimates for risk factors analysis, including age, sex, disease states, disease severity, length of stay, daily charge, expected payers, hospital characteristics, and type of admission. Sensitivity analyses include stratifying by age group, disease states and admission type.Results: Weighted readmission rates for nonelective readmission, unplanned readmission, and unintended readmission were 33.3%, 22,4%, and 18.5%, respectively. Top 3 causes for readmission were neutropenia/agranulocytosis (27.8%), septicemia (15.3%) and aplastic anemia (11.5%). Risk ratios for unintended readmission were 1.21 (1.08-1.36) for nonelective vs elective admission, 1.19 (1.06-1.33) for public vs private insurance enrollees, 0.96 (0.95-0.98) for length of stay (days), 0.77 (0.62-0.95) for large, teaching, 0.87 (0.70-1.08) for small teaching vs nonteaching hospitals.Conclusions: Risk factors for readmission vary based on outcome definitions. Using the most stringent definition, opportunity exits to reduce readmission among patients with acute lymphocytic leukemia.

PO-3691 Revision Risk Of Lower Joint Replacement For Non-steroidal Anti-inflammatory Drug Users: A Retrospective Cohort Study 1400716 Mohammad Bakhriansyah School of Medicine Lambung Mangkurat University Revision Risk Of Lower Joint Replacement For Non-steroidal Anti-inflammatory Drug Users: A Retrospective Cohort Study Safety End Points -> Other Background: Only a few studies asses the impact of non-steroidal anti-inflammatory drugs (NSAIDs) on the revision risk of lower joint replacement (LJR). Moreover, the results are conflicting. Objectives: To assess the risk of revision LJR (i.e., hip and knee replacements) associated with NSAID use compared to non-use.Methods: A retrospective cohort study was performed in the UK Clinical Practice Research Datalink among patients aged ≥40 years with a primary LJR between January 2000-December 2018. Patients were followed until either date of revision surgery, lost to follow-up, or end of study. Use of NSAIDs (either conventional NSAIDs or selective COX-2 inhibitors), as well as potential confounders was assessed time-dependently. Cox-proportional hazard analysis was used to estimate hazard ratios (HRs) and 95% confidence intervals (95% CIs). Results: At cohort entry, 155,490 patients with a primary LJR were identified. 26,946 (17.3%) and 4,646 (3.0%) patients were prescribed conventional NSAIDs and selective COX-2 inhibitors, respectively and 123,898 (79.7%) were non-users. Compared to non-use, the adjusted HRs of revision surgery for different durations of current use of conventional NSAIDs were 12 months: 2.08 (1.69-2.56). For current use of selective COX-2 inhibitors, the HRs were 12 months: 2.07 (1.29-3.30). Recent and past use of both conventional NSAIDs and selective COX-2 inhibitors were also associated with a higher risk compared to non-use.Conclusions: Compared to non-use, current use of conventional NSAIDs and selective COX-2 inhibitors for most durations of use were associated with a higher risk of revision of LJR. A higher risk lasted for >3 months after discontinuation. We can not exclude that confounding by indication (severity of pain) might explain our results.

PO-3693 Long-term Survival And Healthcare Utilization Following Deep Brain Stimulation For Parkinson's Disease In Ontario, Canada 1400717 James Crispo University of British Columbia Long-term Survival And Healthcare Utilization Following Deep Brain Stimulation For Parkinson's Disease In Ontario, Canada Medical Devices Background: Regional variations in the use of deep brain stimulation (DBS) surgery for Parkinson’s disease (PD) may be attributed to long-term safety concerns. Presently, there are limited data on long-term outcomes following DBS surgery for PD in Ontario, Canada.Objectives: The objectives of our study were (1) to compare the long-term survival of PD patients who received DBS surgery to that of matched controls and (2) to examine whether undergoing DBS surgery for PD was associated with differences in the use of publicly funded healthcare services.Methods: Using administrative health data (Ontario, Canada), we retrospectively examined outcomes following DBS surgery within a cohort of individuals diagnosed with PD between 1997-2012 at age 40 years or older. Patients receiving DBS surgery were matched with up to 4 non-DBS controls by age, sex, PD diagnosis date, time with PD, and the logit of the propensity score (clinical and sociodemographic factors). Differences in survival between DBS and control groups were assessed using the log-rank test and marginal Cox proportional hazards regression. Cumulative incidence function curves and marginal subdistribution hazard models were used to assess the effects of DBS surgery on the probability of long-term care placement and the use of home care services. For all analyses, death was classified a competing risk.Results: A total of 23,290 individuals with PD were included in our unmatched cohort. There were 811 patients (260 DBS recipients, 551 controls) included in our survival and long-term care placement analyses, while 798 patients (258 DBS recipients, 540 controls) were included in our home care analyses. Relative to controls, patients who underwent DBS surgery for PD did not experience a significant survival advantage (log-rank test p = 0.50; HR: 0.89, 95% CI: 0.65-1.22). However, patients who underwent DBS surgery prior to 65 years of age experienced a significant survival advantage compared to patients within the same age group who did not receive DBS (HR: 0.49, 95% CI: 0.28-0.84). Long-term care placement was similar between DBS and non-DBS groups (Gray’s test p = 0.82; HR: 1.11, 95% CI: 0.75-1.65); however, patients who received DBS surgery were more likely to utilize home care services (Gray’s test p < 0.01; HR: 1.59, 95% CI: 1.32-1.90).Conclusions: Undergoing DBS surgery for PD may offer a survival advantage for patients who receive DBS at an earlier age. Future studies should examine the effects of DBS on survival in other populations with PD, and whether survival benefits may be attributed to effects on PD or comorbidities that may influence mortality.

PO-3695 Occurrence Of Metabolic Risk Factors Stratified By Psoriasis Severity A Swedish Population-based Matched Cohort Study 1400718 David Hagg Karolinska Institutet Spotlight Poster Session Spotlight Poster Biologics & Biosimilars Background: Psoriasis is a chronic immune-mediated inflammatory disease which can affect up to 8% of the population in Nordic countries. Patients with psoriasis experience an increased risk of developing metabolic risk factors such as diabetes mellitus (DM), hypertension (HTN) or hypercholesterolemia. Understanding whether patients with psoriasis are being actively managed regarding metabolic risk factors is a research gap in the literature. Based on prior work that demonstrated an increased risk of HTN and DM among Swedish patients with severe psoriasis during a one-year period before anti-psoriasis drug prescription, the current study aims to understand the time to occurrence of metabolic risk factors such as DM, HTN and hypercholesterolemia.Objectives: To assess relative risk of metabolic risk factors in patients stratified by psoriasis severity compared with population controls.Methods: Retrospective cohort study conducted using national Swedish registers. Adult patients with psoriasis were selected if they had a dispensing of anti-psoriasis prescription (2007-2013) and at least one diagnosis within 5 years before the dispensing date. The patients with psoriasis were matched 1:10 to controls from the general population on birthyear, sex and county. The cohort was further divided into three disease severity groups (mild, moderate or severe) based on their dispensed anti-psoriasis medication. We applied flexible parametric modelling to understand the risks of the incident comorbidities HTN, hypercholesterolemia and DM among patients with psoriasis from 6 months through 10 years. Hazard ratios and predicted risk of comorbidities are reported.Results: The hazard of HTN, hypercholesterolemia and DM is higher among psoriasis patients compared with population controls, and the hazard ratio increased following the psoriasis severity. For example, HRs of HTN for patients with mild, moderate and severe psoriasis are 1.29 (95% CI: 1.27-1.32), 1.35 (95% CI: 1.32-1.38) and 1.73 (95% CI: 1.64-1.82), respectively. The predicted risk of HTN, hypercholesterolemia, and DM among patients with severe psoriasis at year ten was 0.58 (95% CI: 0.56, 0.59), and 0.22 (95% CI: 0.21, 0.24) respectively while it was 0.42 (95% CI: 0.41, 0.43), 0.30 (95% CI: 0.30, 0.30), 0.11 (95% CI: 0.10, 0.11) among controls, respectively. The predicted risk at year ten was similar among patients with mild or moderate psoriasis.Conclusions: The HRs and the predicted risks of metabolic risk factors are higher among patient with psoriasis compared with matched controls, and more prominent among the severe psoriasis group.

PO-3699 Real-world Data Collection In Early Access Programs: An Additional Source Of Data To Inform Benefit-risk Assessment In The Pre-approval Phase 1400719 Mira Soni Evidera Real-world Data Collection In Early Access Programs: An Additional Source Of Data To Inform Benefit-risk Assessment In The Pre-approval Phase Benefit-Risk Assessment, Communication, and Evaluation (BRACE) Background: Early access programs (EAPs) provide patients with life-threatening diseases with access to investigational products prior to market approval. Across EAPs, safety data collection is usually required, but guidance on effectiveness data is limited and can vary by local regulations. The ‘real-world data’ (RWD) collected in EAPs could supplement trial data in the pre-approval phase, as the investigational drug is provided to a wider range of populations than would be eligible for a trial and patients are monitored as per expected usual practice. However, careful consideration of RWD collection ahead of EAP implementation is needed to ensure data quality to support pre-approval benefit-risk assessment.Objectives: We developed an RWD framework to collect long-term effectiveness and quality of life (QoL) outcomes as part of an oncology EAP in the United Kingdom (UK). To our knowledge this was the first UK EAP with a voluntary observational study component. Our aim is to share knowledge on optimization of RWD collection in EAPs for future use in benefit-risk assessment.Methods: During RWD framework development, input/approval on the observational study were sought from the ethics committee and regulatory body stakeholders. As drug supply requests in the UK EAP were physician-led, it was not possible to know in advance which sites would participate, so special circumstances were required to set-up and train sites for data collection within a couple of days of them submitting a drug supply request. For physicians, data collection was limited to key outcomes [disease progression, survival] collected from patient medical records. QoL questionnaires were completed by patients at home.Results: To ensure data quality to support pre-marketing benefit-risk assessment, the following aspects should be considered ahead of EAP implementation: (1) primary research objectives should be planned and vetted with key stakeholders; (2) expected sample size should be accounted for and ways to improve the reliability of data (e.g. data pooling in multi-country EAPs); (3) a robust data collection platform should be utilized that is easy and convenient for physicians; (4) electronic questionnaires are recommended rather than paper questionnaires; (5) a rigorous safety reporting plan should be in place.Conclusions: Regulatory bodies are increasingly looking to incorporate RWD in pre-approval benefit-risk assessment. EAPs could contribute RWD, however the related setting and challenges need to be taken into account to ensure quality data that can be considered for drug approval.

PO-3700 Analysing Patient Level Data From Norwegian Registries: Challenges And Considerations. 1400720 Sophie Graham Evidera Analysing Patient Level Data From Norwegian Registries: Challenges And Considerations. Methods in Pharmacoepidemiology -> Analytical Methods - e.g., marginal structural models, instrumental variables, sensitivity analyses Background: The Norwegian patient and prescription registries provide linked data without dates of event occurrences. Instead, the dataset contains so-called “difference days”, which represent the difference between an event date and a random date. Although the difference days from event occurrences are provided, difference days from known event dates such as the end of study follow-up are not included.Objectives: To investigate the impact of absent information on patients’ end of follow up on a treatment pattern analysis in the Norwegian registries.Methods: Patients were identified with a venous thromboembolism (VTE) diagnosis that occurred between 2013 and 2017, and a dispensation of an oral anticoagulant within 30 days after VTE diagnosis. Treatment discontinuation was defined as no new dispensation of oral anticoagulation during the days’ supply of the previous prescription + 30-day grace period. Kaplan Meier (KM) methods were used to assess time to treatment discontinuation. Patients were censored at the earliest of death or at their last ‘known’ date in the data. The last ‘known’ date was either the last dispensation date or last encounter (outpatient or inpatient) that was identified in the patient’s record. In the absence of date of end of follow up, it was not possible to determine whether patients initiating treatment in 2017 had discontinued treatment or reached the end of follow up. Therefore, a sensitivity analysis was conducted to assess the discontinuation rates in all treatment groups excluding patients who initiated treatment in 2017.Results: In the main analysis, 3,557 warfarin, 3,088 apixaban and 5,828 rivaroxaban patients were identified. Warfarin patients were younger than apixaban (median [IQR]: 64 [26] vs. 66 [24] years; p=0.0045), but older than rivaroxaban (61 [24]; p<0.0001). The median (IQR) months to treatment discontinuation was 6.4 (3.7-15.9), 6.3 (3.4-10.6) and 6.0 (3.9-9.2) for apixaban, warfarin and rivaroxaban patients, respectively. In the sensitivity analysis, the median (IQR) months to treatment discontinuation was 6.3 (3.7-13.8), 6.3 (3.5-10.6) and 5.9 (3.9-8.9) for apixaban, warfarin and rivaroxaban patients, respectively.Conclusions: A treatment pattern analysis in the Norwegian registries can be conducted by censoring patients at their last ‘known’ encounter in the data. Excluding the final calendar year (sensitivity analysis) had little impact on the results, however, further analysis using different methodologies is recommended to more accurately address this issue.

PO-3702 To What Extent Are Topical Tacrolimus Or Pimecrolimus Associated With Increased Risk Of Skin Cancer And Lymphoma? Long-term Results From Joelle Study 1400721 Alejandro Arana RTI Health Solutions Spotlight Poster Session Spotlight Poster Databases Background: Observations of transplant recipients receiving systemic calcineurin inhibitors have raised concern of a potentially increased risk of lymphoma and skin cancer with topical use of tacrolimus (TAC) or pimecrolimus (PIM) for atopic dermatitis.Objectives: To estimate the long-term risk of skin cancer and lymphoma associated with use of topical TAC or PIM in adults and childrenMethods: A multinational cohort study was conducted using available health data from 2002 through 2017 in the United Kingdom’s Clinical Practice Research Datalink, the PHARMO Database Network in the Netherlands, and the Danish and Swedish nationwide health registers. Mantel-Haenszel methods were used to estimate incidence rate ratios (IRRs) and differences in children and adults of study malignancies, comparing new users of topical TAC and topical PIM with two propensity score-matched cohorts of users of moderate- to high-potency topical corticosteroids. Malignancies included Hodgkin’s lymphoma, non-Hodgkin’s lymphoma other than cutaneous T-cell lymphoma (CTCL), CTCL, and skin cancer (melanoma and nonmelanoma).Results: The study included 126,908 adults and 32,605 children initiating treatment with topical TAC and 61,841 adults and 27,961 children initiating treatment with topical PIM. Follow-up was 10 years or more for 19.4% of adults and for 31.9% of children. Among adults and children, the median number of topical TAC or PIM prescriptions was one, and for the 80th percentile, three. In adults, IRRs were not increased other than the IRR of CTCL for topical TAC: 1.80 (95% confidence interval [CI], 1.25-2.58) corresponding to an excess rate of 3 cases per 100,000 person-years (95% CI, 1-6). The IRR of melanoma for topical PIM was 1.21 (95% CI, 1.03-1.41) for an excess rate of 10 cases per 100,000 person-years (95% CI, 1-18). The IRR of nonmelanoma skin cancer for topical PIM was 1.28 (95% CI, 1.20-1.35) for an excess rate of 91 cases per 100,000 person-years (95% CI, 68-114). In children, results were inconclusive due to few events. For adult users of topical TAC in whom the time since first exposure to treatment was ≥ 5 years, the IRR for CTCL was 0.25 (95% CI, 0.03-1.87).Conclusions: Overall, we found little evidence associating use of topical TAC and topical PIM with skin cancer and lymphoma; confounding by indication, surveillance bias, and reverse causation may have influenced these results. If there were a causal association, our data indicate that any public-health impact of CTCL associated with the use of topical TAC and of skin cancer associated with the use of topical PIM would be low.

PO-3704 Exploratory Analyses Of Modafinil Utilization During Pregnancy And Diagnoses Of Malformations In The Offspring 1400722 Ulrike Haug Leibniz Institute for Prevention Research and Epidemiology - BIPS Anna Julia Witzleb Exploratory Analyses Of Modafinil Utilization During Pregnancy And Diagnoses Of Malformations In The Offspring Pregnancy and Lactation Background: Modafinil is indicated for excessive sleepiness in patients with narcolepsy. Based on data from pregnancy registries and case reports, modafinil is suspected to cause severe congenital malformations in the child when used during pregnancy. It is therefore contraindicated in pregnant women which has been stressed in 2019 by a warning that was issued by the manufacturer.Objectives: To do exploratory analyses on the utilization of modafinil during pregnancy and diagnoses of malformations in the offspring based on German claims data.Methods: Using the German Pharmacoepidemiological Research Database (GePaRD; health claims data covering 17% of the German population), we selected pregnancies ending between 2006 and 2016 (overall ~1.5 million pregnancies) and with at least one modafinil dispensation during pregnancy. For live births, the mother’s and baby’s data were linked and in- and outpatient diagnosis codes indicating a malformation in the child were explored.Results: We identified 28 pregnancies in 24 women with at least one modafinil dispensation during pregnancy. Of these women, 71% had at least one diagnosis code of narcolepsy. In 26 pregnancies there was at least one modafinil dispensation during the first trimester, in seven during the second and in four during the third trimester. Overall, 71% of the 28 pregnancies ended in a live birth and 25% in induced abortions. Out of 20 live born children, 18 could be linked to their mothers’ data. Four of these 18 children had diagnosis codes indicating a severe malformation (atrial septum defect: n=1; plagiocephaly: n=2, asymmetric face: n=1).Conclusions: In our study period, we observed 28 pregnancies with at least one modafinil dispensation. In the majority, this dispensation was in the first trimester. Although the sample size is low, the high proportion of induced abortions is striking and also the proportion of children with diagnosis codes for malformations support the safety concerns regarding modafinil exposure during pregnancy. An important next step would be the pooling of data from several databases to allow risk quantification.

PO-3706 Polyneuropathy In Early Type 2 Diabetes Is Associated With Higher Risk Of Subsequent Cardiovascular Disease: Evidence From The Danish Centre For Strategic Research In Type 2 Diabetes (dd2) Cohort 1400723 Sia Nicolaisen Aarhus University Polyneuropathy In Early Type 2 Diabetes Is Associated With Higher Risk Of Subsequent Cardiovascular Disease: Evidence From The Danish Centre For Strategic Research In Type 2 Diabetes (dd2) Cohort Disease Epidemiology/Clinical Course -> Diabetes Background: Diabetic polyneuropathy (DPN) may be associated with subsequent risk of cardiovascular disease (CVD) and mortality in patients with early type 2 diabetes (T2D), but evidence is sparse.Objectives: We used follow-up data in the DD2 research database to compare the risks of CVD and mortality among T2D patients with and without DPN.Methods: Patients were assessed for DPN by a neuropathy questionnaire survey conducted in 2016 in the DD2 cohort of recently diagnosed T2D patients and linked to Danish prospective national healthcare registries. A score ≥4 on the Michigan Neuropathy Screening Instrument questionnaire (MNSIq) indicated DPN. We defined CVD as hospitalization with ischemic heart disease, stroke, peripheral artery disease, or heart failure. We followed the patients from the time of the neuropathy questionnaire survey and calculated adjusted incidence rate ratios (IRRs) of CVD and mortality using Poisson regression with adjustment for age, sex, HbA1c, BMI, smoking, alcohol consumption, LDL-cholesterol, systolic blood pressure, antihypertensive and lipid lowering medication, diabetes duration, albumin-creatinine-ratio, and prior cardiovascular event.Results: A total of 5,028 T2D patients were included in the study. Median diabetes duration at the time of the questionnaire was 4.6 (IQR 3.5-5.7) years. In total, 818 (16.3%) patients had DPN. Patients with DPN were slightly younger than those without DPN (median 62.3 [IQR 54.8-70.1] vs. 65.7 [IQR 57.2-71.2] years) and a larger proportion were women (47.2 vs. 41.1%). Among patients with DPN, 30.3% had a prior cardiovascular event versus 19.7% of patient without DPN. Use of cardiovascular drugs was higher for antihypertensive drugs (75.8 vs. 72.7%) and aspirin (26.0 vs 21.9%) among patients with DPN, but lower for lipid-lowering drugs (70.8 vs. 72.3%). During a median of 2.2 (IQR 2.2-2.2) years of follow-up, 10.8% of the patients were diagnosed with CVD and 5.0% died. After adjustment, the risk of CVD was 57% (IRR 1.57 [95% CI 1.27-1.94]) and mortality was 20% (IRR 1.20 [95% CI 0.76-1.89]) higher in individuals with DPN compared to those without. Sensitivity analyses among patients without a history of CVD and complete case analyses showed similar results.Conclusions: We found that DPN defined by MNSIq ≥4 associates with significantly higher risk of CVD after adjustment for well-known CVD risk factors. Notably, these associations were found in patients with early T2D emphasizing the importance of screening for this neglected diabetic complication.

PO-3708 Three Strategies To Measure Frailty And/or Complex Health Needs In Real World Data: An Analysis Of Uk Primary Care And Hospital Linked Data 1400724 Daniel Prieto-Alhambra CSM, NDORMS, University of Oxford Three Strategies To Measure Frailty And/or Complex Health Needs In Real World Data: An Analysis Of Uk Primary Care And Hospital Linked Data Geriatric Pharmacoepidemiology Background: Complex health needs and frailty are key confounders in geriatric pharmaco-epidemiology. Although a number of definitions exist, none are formally accepted or adopted in geriatric pharmacoepidemiology. Recent NICE guidelines recommend using primary care electronic health records to identify markers of frailty, including the electronic Frailty Index (eFI). Others have advocated for the measurement of polypharmacy and/or healthcare resource use as alternative methods for the identification of frail patients in routine data.Objectives: We tested the use of three definitions of frailty, and assessed the overlap between the resulting three cohorts. Socio-demographics, co-morbidity, and common drug/s use were then characterised in the three cohorts.Methods: All participants aged 65 or older on 1 Jan 2009, registered for 1 year or above were identified from CPRD GOLD, a primary care database including >10 million UK people. Those registered in practices with no linkage to Hospital Episode Statistics (HES) data were excluded. Three cohorts were identified based on the top quintile of three indices measured in the previous year: 1.healthcare resource use defined as the number of unplanned hospital admissions [cohort1], 2.score in electronic frailty index (eFI) [cohort2], and 3.polypharmacy estimated as the number of different drugs prescribed in GP records [cohort3]. Baseline characteristics and overlap between the cohorts were described.Results: 464,886 participants were eligible for the study, of whom 73,724, 104,885, and 100,965 constituted Cohorts 1, 2 and 3 respectively. 20,741 (4.5%) patients were included in all three cohorts. Mean age (SD) was very similar for all three cohorts (77.0 (7.7), 77.6 (7.4) and 77.2 (7.3) respectively). Cohort 3 had the lowest proportion of males: 38.8% vs. 43.2% in cohort 1 and 41.0% in cohort 2. Preventative medicines, such as, bisphosphonates, statins and diuretics were prescribed to a higher proportion of patients in the three cohorts compared with the background (non-frail) population in the year prior to start (e.g. statins were prescribed to 47.8% in cohort 1 vs. 38.9% in the background population). Conclusions: All three strategies are feasible for the identification of elderly frail patients in UK data, but the overlap between all three is small. Further research is needed on the best strategy/ies to measure frailty in real world data.

PO-3709 Different Safety Reporting Profiles Of Infliximab Biosimilar From Originator : Comparison Among Countries 1400725 Kimie Sai National Institute of Health Sciences Different Safety Reporting Profiles Of Infliximab Biosimilar From Originator : Comparison Among Countries Biologics/Biosimilars Background: Usage of biosimilar (BS) of infliximab (IFX) has been encouraged worldwide. One important issue of BS usage is a possible interpopulation difference in its safety profiles; e.g., similar to the originator, immunogenicity of BS might be different among individual/populations and could lead to different efficacy/safety profiles among regions. However, information of safety profiles of BS in real world and possible regional differences is still limited.Objectives: To examine possible differences in adverse drug reaction (ADR) reporting profiles between IFX originator and BS, and to explore those regional differences.Methods: ADRs reports of IFX originator (Remicade) and its CT-P13 BS (Remsima/Inflectra/BS 1) were examined. JADER was used in analysis for Japan (Apr 2015 - Dec 2019) and VigiBase (serious report type) was used for global analysis, i.e., Korea, UK, France and USA (Jan 2014 - Dec 2019). Number/percent for ADR reports of either IFX (originator or BS) and reporting odds ratios (RORs) (target vs. non-target drugs) based on MedDRA preferred terms and 27 categories for System Organ Class (SOC) were compared between the originator and the BS. For evaluation of differences in ADR reporting profiles, lower limit of 95% confidence intervals (> 1) for ROR (BS vs. originator) was used.Results: Major ADRs (either the top or top two SOC) for both originator and BS was “Infections and infestations” in Japan, Korea and UK, “Skin and subcutaneous tissue disorders” in France, and “General disorders and administration site conditions” in USA. Significantly higher ROR (BS vs. originator) was observed in 5 SOC categories in Japan, 2 for Korea, 3 for France, 1 for UK and 7 for USA, of which profile was different among regions. “Injury, poisoning and procedural complications“, attributed to “infusion related reaction” was raised to the second BS ADRs in Japan [ROR (95%CI):2.98 (1.70-5.23)], and to the third, but not due to infusion related reaction, in Korea [2.10 (1.08-4.10)]. “General disorders and administration site conditions” was raised to the top BS ADRs, mainly “drug ineffective”, in France [2.33 (1.75-3.12)] and, mainly “pain” and “death”, in USA [1.36 (1.05-1.76)].Conclusions: This study suggests that the major ADRs (the top or top two) by IFX would be mostly consistent between originator and BS in each country, but different ADRs might be also raised as major BS ADRs in a region-specific manner. The reason of regional difference might be due to differences in medical practice and population differences among regions. However, further confirmation study is still needed in a larger number of case reports.

PO-3711 Glucocorticoid Use And Breast Cancer Risk In A Prospective Cohort Study Of Postmenopausal Women. 1444136 Manon cairat Glucocorticoid Use And Breast Cancer Risk In A Prospective Cohort Study Of Postmenopausal Women. Disease Epidemiology/Clinical Course -> Cancer Background: Trough their anti-inflammatory properties, glucocorticoids are hypothesized to prevent breast cancer. However, glucocorticoids are known to decrease glucose tolerance and to interact with the oestrogen pathways, two mechanisms involved in breast cancer development. Experimental data suggested that glucocorticoids could play a complex/dual role in breast cancer initiation and progression but epidemiological evidence are scare regarding the associations between glucocorticoid use and breast cancer risk.Objectives: We investigated the association between glucocorticoid use and breast cancer incidence, overall and by subtypes, in the French E3N prospective cohort.Methods: E3N includes 98,995 French women born between 1925 and 1950 and insured by a health insurance plan that covers mostly teachers. Information on lifestyle, anthropometric and reproductive factors, and medical history were collected biennially by self-questionnaires and matched with data from a drug reimbursement database pertaining to years 2004-2014. Women with at least two reimbursements in any 3-month period since 2004 were defined as ever exposed to glucocorticoids. Considering exposure as time-varying, multivariable Cox regression models were used to estimate hazard ratios (HRs) for the association of glucocorticoid use with breast cancer risk, overall and by breast cancer subtypes.Results: In the current analysis, 62,390 postmenopausal women were followed between 2004 and 2014 (9 years on average, starting at a mean age of 63 years; 2,887 breast cancers). Of them, 28% had been exposed to glucocorticoids during follow-up. In multivariable models, there was no statistically significant association between ever use of glucocorticoids and breast cancer risk overall [HR= 0.96 (0.90 - 1.06)]. However, we found a statistically significant heterogeneity according to breast cancer subtypes suggesting an increased risk of in situ breast cancer and a decreased risk of invasive breast cancer [HRin situ= 1.32 (1.00 - 1.75), HRinvasive=0.87 (0.77 - 0.98), Pheterogeneity=0.01] associated with glucocorticoid ever use.Conclusions: Glucocorticoid use was associated with an increased in situ breast cancer risk and a decreased invasive breast cancer risk in this large, prospective cohort of postmenopausal women with 10 years of exposure information. Associations according glucocorticoid types, main breast cancer risk factors as well as potential confounding effect by other drugs/comorbidities will be explored and presented at the conference.

PO-3712 The Risk Of Myocardial Infarction And Venous Thromboembolism After Partial Or Total Knee Arthroplasty: A Self Controlled Case Series 1400726 Daniel Prieto-Alhambra CSM, NDORMS, University of Oxford Spotlight Poster Session Spotlight Poster Medical Devices Background: Prior studies comparing patients undergoing knee replacement (KR) with non-KR controls have described a short-term increased risk of post-operative myocardial infarction (MI) and venous thromboembolism (VTE). Residual confounding could bias these findings. Despite seldom use in surgical epidemiology, self-controlled case series (SCCS) controls for time-fixed within-person confounding by design and can therefore be used to quantify any excess risk associated with knee replacement.Objectives: To estimate the increase in risk of MI and VTE in the 3 months (91 days) after two types of KR: partial KR (PKR) and total KR (TKR).Methods: Data from two UK databases (National Joint Registry and Hospital Episode Statistics) were linked to identify KR and outcomes. Adults (18+) undergoing KR who experienced a) MI and b) VTE within 1 year either side of first KR were identified for the SCCS cohorts. Six months wash-outs were required for a new outcome of the same type to be considered a new event. Incidence rate ratios (IRR) and 95% confidence intervals (CI) were calculated for up to 3 months post-surgery compared to the 6-12 months prior to and 3-12 month post surgery together (reference period). A 6-month pre-exposure period was included prior to surgery date. Interactions with gender, ASA stage and TKR v PKR were assessed using a cut-off p-value <0.1. The assumptions of SCCS were assessed in sensitivity analyses: 1) assessing the risk for the first adverse event only, and 2) excluding patients who died within the 3 months post-KR.Results: Of 354,494 patient undergoing KR, 6,947 had 7,207 MIs and 4,227 had 4,284 VTEs. Patients who had an MI were on average 73 years old, with 61% male. 64% were ASA 1 or 2 and 5% had a partial KR. Patients who had a VTE were on average 72 years old, with 43% male. 78% were ASA 1 or 2 and 3% had a partial KR. In the 3 months post KR, IRR were 1.22 (95% CI 1.15, 1.30) and 9.26 (8.68, 9.89) for MI and VTE respectively. Interactions were identified for gender and ASA stage with the pre-exposure period only. Patients undergoing PKR were not at increased risk of MI in the 3 months after surgery, IRR 0.74 (0.53, 1.02), and at a lower increased risk of VTE than patients undergoing TKR (5.51 [3.87, 7.84] vs 9.43 [8.82, 10.08]). The assumptions of SCCS held with little change in the no death and first event models.Conclusions: Patients are at a 20% and 9-fold increased risk of MI and VTE respectively in the 3 months post KR surgery compared to the reference period. No strong interaction was seen with gender or ASA stage, but patients undergoing PKR were not at excess risk of MI and at lower risk of VTE.

PO-3713 Polypharmacy Among Hospitalized Children In The Capital Region Of Denmark 1444217 Thomas Jensen Polypharmacy Among Hospitalized Children In The Capital Region Of Denmark Pediatric Pharmacoepidemiology Background: The prevalence of polypharmacy is increasing in both children and adults. The implications are not fully known, but it has amongst others been associated with drug-drug interactions, increased risk of adverse drug reactions, hospital admissions and increased costs. Previous studies have found the prevalence highly variable in terms of study design, patient population and time period.Objectives: We therefore aimed to investigate the prevalence of polypharmacy in a Danish pediatric inpatient population.Methods: Using the Danish National Patient Register we identified all children (between 1 and 17 years of age) hospitalized in the Capital Region of Denmark between January 1st, 2012 to May 15th, 2016. All in-hospital administrated drugs in the study period were identified from the administrative in-hospital drug use database in the Capital Region of Denmark. The register contains information on all in-hospital drug administrations on a personal level between January 1st, 2012 to May 15th, 2016 in the Capital Region of Denmark. Polypharmacy was defined as treatment with 5 or more drugs with different ATC-codes within the admission period.Results: We identified 66,897 hospital admissions in the study period. The mean age of the hospitalized children was 8.3 years and 52.5 % were boys. The median number of admission days per admission was 2 (interquartile range [IQR] 2-4) and the median number of different drugs administrated per admission was 2 (IQR 0-5). 26.3 % of the children were treated with five or more different drugs during the admission. The prevalence of polypharmacy increased with age: 21.0 % (1-4 years), 23.5 % (5-8 years), 26.9 % (9-13 years) and 33.9 % (14-17 years). The most commonly administrated drug was acetaminophen (15.9 %).Conclusions: Polypharmacy among hospitalized children within the Capital Region of Denmark is common and increases with age. Further investigation is warranted.

PO-3715 Changes In Utilization Of Antihypertensive Drugs Following Angiotensin Receptor Blocker Product Recalls In Germany 1418442 Salka Enners German Institue for Drug Use Evaluation Changes In Utilization Of Antihypertensive Drugs Following Angiotensin Receptor Blocker Product Recalls In Germany Drug Utilization Research -> Changing drug utilization (interventions and implementation research) Background Angiotensin receptor blockers (ARBs) are one of the most common drug classes, used by approximately 8 million patients in Germany. Discovery of unacceptable amounts of potentially carcinogenic nitrosamines resulted in more than 30 recalls of generic ARBs, primarily valsartan, between July 4 and December 20, 2018, and in subsequent drug shortages. Objectives We explored the potential impact of the recalls on the dispensings of valsartan, all ARBs, and other antihypertensives (AHTs). Methods We performed a descriptive drug utilization study using the DAPI database of anonymized dispensing data from >80% of community pharmacies at the expense of the statutory health insurance (SHI) funds. Nearly 88% of Germany’s population is insured by the SHI system. We included all oral mono and fixed-dose combination products of ARBs, ACE inhibitors (ACEi), calcium channel blockers, and beta-blockers. For the analysis of ARBs individually, we only included mono preparations and combinations with hydrochlorothiazide (HCTZ) because only those were affected by the recalls. We calculated utilization as defined daily doses per 1,000 inhabitants per day (DID). Time trends were analyzed on a quarterly level. Results The market shares for valsartan and valsartan/HCTZ drug products recalled in July and November 2018 were both 45% of all valsartan and valsartan/HCTZ-containing drugs. Valsartan dispensings decreased substantially and continuously after first recalls in July 2018 (2Q18: 41.1 DID, 4Q19: 16.9 DID). A simultaneous increase of other ARBs, primarily candesartan was observed. Other ARBs: 2Q18: 77.7 DID, 4Q19: 121.9 DID, of which candesartan was: 2Q18: 57.7 DID, 4Q19: 99.8 DID. Effects on utilization of ACEi and AHTs overall were not observed. ACEi: 2Q18: 212.3 DID, 4Q19: 212.5 DID; AHTs: 2Q18: 520.9 DID, 4Q19: 544.2 DID. Similarly, there was no impact on the moderate, but continuous increase of the total volume of ARB dispensings (2Q18: 129.4 DID, 4Q19: 150.2 DID). Conclusions We noted a substantial decrease in utilization of valsartan but not losartan, irbesartan, olmesartan, or telmisartan after generic ARB product recalls. The total volume of ARB further increased and AHT utilization overall remained nearly unchanged suggesting shifts within the class of ARBs, mostly to candesartan.

PO-3717 Utilization Of Spironolactone In Patients With Reduced Or Preserved Ejection Fraction Heart Failure 1400727 Sruthi Adimadhyam Harvard Pilgrim Health Care Institute Utilization Of Spironolactone In Patients With Reduced Or Preserved Ejection Fraction Heart Failure Drug Utilization Research -> Trends and comparisons Background: Spironolactone is an aldosterone antagonist indicated for the management of reduced ejection fraction (rEF) heart failure (HF). In the TOPCAT trial, spironolactone significantly lowered the incidence of HF hospitalizations among patients with preserved ejection fraction (pEF) HF. Real world utilization of spironolactone in pEF is unknown.Objectives: To evaluate the use of spironolactone following a diagnosis of rEF or pEF and characterize the first treatment episode.Methods: We conducted a retrospective cohort study using data from the MarketScan® Research Databases (7/2010 - 9/2018). We identified patients with rEF or pEF using diagnosis and procedure codes from a previously validated algorithm. Patients with a diagnosis of left or systolic HF with a history of cardiomyopathy, implantable cardioverter defibrillator, or myocardial infarction in 183 days prior HF diagnosis were classified as rEF, while those with diastolic or unspecified HF with no history of cardiomyopathy or implantable cardioverter defibrillator were classified as pEF. We required patients to be continuously enrolled in the 183 days prior to cohort entry and have no history of spironolactone use. Follow-up started on the day of diagnosis and ended at the earliest occurrence of a spironolactone dispensing, disenrollment, death, or end of data. We calculated the proportion of HF patients initiating spironolactone. For those initiating treatment, we estimate the dose and duration of the first treatment episode. Results were stratified by age group, sex, and year of HF.Results: Among 178,334 patients with rEF, 65.3% were male, and mean age was 68.5 ± 14.6 years. Among 1,361,449 patients with pEF, 49.8% were male, and mean age was 64.4 ± 15 years. Proportion of patients that initiated spironolactone following rEF was 17.2% versus 4.2% after pEF. The median time to initiation of spironolactone after rEF was 79 (12 - 319) days versus 226 (40 - 668) days after pEF. The median duration of first treatment episode in rEF was 120 (34 - 306) days with a median dose of 25 (25 - 25.3) milligrams/day. The median duration of first treatment episode in pEF was 100 (30 - 271) days with a median dose of 25 (25 - 50) mg/day.Conclusions: Our study indicates low utilization of spironolactone in patients with pEF compared to rEF. Future utilization studies will be needed to evaluate the influence of the TOPCAT trial findings in the management of patients with pEF HF. We intend to replicate this study in the Sentinel Distributed Database.

PO-3718 Utilisation And Safety Of Rivaroxaban For Acute Coronary Syndrome (acs) In Secondary Care In The Uk: Results From The Rivaroxaban Observational Safety Evaluation Post Acs (rose-acs) Study 1400728 Alison Evans Drug Safety Research Unit Utilisation And Safety Of Rivaroxaban For Acute Coronary Syndrome (acs) In Secondary Care In The Uk: Results From The Rivaroxaban Observational Safety Evaluation Post Acs (rose-acs) Study Pharmacovigilance -> Cardiovascular Background: The ROSE-ACS Specialist Cohort Event Monitoring (SCEM) study was conducted in secondary care to monitor the safety and use of rivaroxaban for the prevention of atherothrombotic events in adults after an ACS. The targeted safety outcome was bleeding risk.Objectives: To estimate the short-term (12 week) cumulative incidence and incidence rate of major bleeding in patients (pts) prescribed rivaroxaban post ACS and describe drug utilisation in the secondary care setting in England and Wales.Methods: Pts identified through clinical speciality groups supported by UK Clinical Research Networks Sept 2015 - Oct 2018. Pt characteristics, drug utilisation and bleeding outcomes collected via questionnaires completed by specialists sent at baseline and after 12 weeks of observation. A contextual cohort of pts receiving standard treatment also included in order to compare reasons for choice of anticoagulation type. However the study did not allow any direct comparisons between the two cohorts and therefore bleed incidence results are not presented.Results: The final cohort consisted of 124 pts; median age was 59.5 years [IQR 51-68] and the majority were male (83.1%). Most pts were initiated treatment with rivaroxaban for ST-elevation myocardial infarction (STEMI) (n=64, 51.6%) with 53 (42.7%) treated for Non-ST-elevation myocardial infarction (NSTEMI). Rivaroxaban was largely prescribed in accordance with the product label; 116 pts (93.5%) initiated on a daily dose of 5mg. The majority of pts were started rivaroxaban within 3 days of being admitted (n=102, 82.3%), with the highest frequency on day 2 (n=39, 31.5%). One major bleed was reported which was within the gastrointestinal site (cumulative incidence 0.8%, 95% CI [0.0, 4.4]; incidence rate 4.2 per 100 person years, 95% CI [0.6, 29.6]). A further 13 bleeds were reported as requiring medical attention (10.5%, 95% CI [5.7, 17.3]; 54.3 per 100 person years, 95% CI [31.5, 93.5]) and 2 were classified as minimal bleeds (1.6%, 95% CI [0.2, 5.7]; 8.3 per 100 person years, 95% CI [2.1,33.4]). There were 3 (2.4%) deaths (all as a result of myocardial infarction) and no reported pregnancies.Conclusions: The ROSE-ACS study shows that although use of rivaroxaban for ACS in the UK is low, it is largely being prescribed to populations in accordance with prescribing recommendations. Frequency of reported bleeding events were low and no new safety signals were identified. The SCEM design provides a framework suitable to evaluate the safety of newly marketed medicines in the secondary care setting.

PO-3719 St. John's Wort: Utilization During Pregnancy And Pregnancy Outcomes 1400729 Wiebke Schaefer Leibniz Institute for Prevention Research and Epidemiology - BIPS St. John's Wort: Utilization During Pregnancy And Pregnancy Outcomes Drug Utilization Research -> Other Background: Pregnant women may be prone to using complementary and alternative medicine, such as St. John’s Wort (St. John’s) for the treatment of mild depressive disorders. However, little is known about the utilization of St. John’s during pregnancy. In Germany, high-dose preparations of St. John’s can be reimbursed by statutory health insurances, enabling to study exposure to St. John’s based on German claims data.Objectives: To characterize pregnancies exposed to high-dose preparations of St. John’s and to describe the outcomes of these pregnancies.Methods: Using the German Pharmacoepidemiological Research Database (GePaRD) with data on ~1.5 million pregnancies between 2006 and 2016, we identified pregnancies with exposure to high-dose St. John’s via outpatient dispensations. We characterized these pregnancies regarding the timing of dispensation of St. John’s preparations, the use of other antidepressants in the year before or during pregnancy, concomitant use of potentially interacting drugs, as well as the relative frequency of pregnancy outcomes compared to all other pregnancies.Results: We identified 496 pregnancies with dispensation of a high-dose St. John’s preparation in at least one trimester. Of these, 386 (78%) had a dispensation during the first trimester only, and 16 pregnancies (3%) had a dispensation in all trimesters. About one third already had a dispensation of St. John’s in the year before pregnancy. In 21% of all pregnancies exposed to St. John’s, the mothers had a dispensation of other antidepressants in the year before pregnancy and 12% during pregnancy (9% in the first trimester). Concomitant use of potentially interacting drugs during pregnancy was infrequent, with less than 1% of all pregnancies being exposed to such drugs in one of the trimesters. Compared to the rates previously published for all pregnancies in GePaRD, pregnancies exposed to St. John’s less frequently resulted in live birth (89% vs. 94%), while the rates for induced abortions/stillbirths and ectopic pregnancies were about two times higher (7.9% vs. 3.8% and 2.0% vs. 1.4%, respectively). All non-live births except for one stillbirth occurred in pregnancies exposed during the first trimester only.Conclusions: Our results suggest that St. John’s is used as an alternative to other antidepressants in some pregnancies, but is also newly dispensed for treatment during pregnancy. Further investigations are needed to explore potential reasons why the rates for non-live births were higher among pregnancies exposed to St. John’s compared to other pregnancies.

PO-3723 Risk Of Stroke And Bleeding In Atrial Fibrillation Treated With Apixaban Compared To Warfarin 1400730 Marie Bradley US Food and Drug Administration Spotlight Poster Session Spotlight Poster BRACE Background: Apixaban use has been increasing in recent years in the US. A previous FDA study reported a favorable benefit risk for apixaban compared to warfarin for stroke prevention in older nonvalvular atrial fibrillation (NVAF) patients. However, that study was restricted to those aged 65 years or more and it remains unclear whether this favorable benefit risk profile persists in other populations including younger users. We examined if a similar benefit risk profile was observed in the Sentinel Distributed Database (SDD), and if it varied by age group.Objectives: To examine if a similar benefit risk profile was observed in the Sentinel Distributed Database (SDD), and if it varied by age group.Methods: Patients, 21 years and older initiating apixaban or warfarin for NVAF, between December 28, 2012 and June 30, 2018, were identified in the SDD. Cox proportional hazards regression was used to estimate the hazard ratios (HR) and 95 % confidence intervals (95% CI) for each outcome [ischemic stroke, gastrointestinal (GI) bleeding and intracranial hemorrhage (ICH)] in propensity score matched apixaban users compared to the warfarin users. A subgroup analysis by age (21-64, 65-74 and 75+ years) was conducted.Results: After matching, 55.3% and 58.4% (n=55,038) of the apixaban and warfarin users were included in the main analysis. GI bleeding was the most common outcome. The HR (95%CI) for GI bleeding, ICH and ischemic stroke in apixaban users compared to warfarin users was 0.57 (0.50-0.66), 0.53 (0.40-0.70) and 0.56 (0.45-0.71) respectively. The reduced risk of these outcomes in apixaban compared to warfarin users persisted across age groups and was lowest in those aged 21-64 years.Conclusions: In NVAF patients initiating either apixaban or warfarin for stroke prevention, apixaban was associated with a decreased risk of GI bleeding, ICH and ischemic stroke compared to warfarin, and this reduced risk was seen in all age groups.

PO-3724 Preconception Antibiotic Use And Spontaneous Abortion: A Prospective Study 1400731 Holly Crowe Boston University Preconception Antibiotic Use And Spontaneous Abortion: A Prospective Study Pregnancy and Lactation Background: Antibiotic use may increase the risk of spontaneous abortion (SAB) through increased oxidative stress or disruption of the reproductive tract microbiome.Objectives: To examine the association between antibiotic use and SAB risk in a North American preconception cohort study.Methods: From 2013 through 2019, we enrolled 10,921 female pregnancy planners; 5,793 conceived during 12 months of follow-up and 1,084 (19%) experienced SAB. At baseline, we collected data on antibiotic use in the past 4 weeks, including type and indication for use. Pregnancy losses at <20 gestational weeks were reported on follow-up questionnaires completed preconceptionally and in early and late pregnancy. We used Cox regression models, with gestational weeks as the time scale, to estimate hazard ratios (HR) and 95% CIs, adjusting for demographics, lifestyle, and reproductive history.Results: Overall, 458 women (8%) reported using antibiotics in the four weeks before baseline, most commonly penicillins (24%) and macrolides (9%). Respiratory infections (31%) and urinary tract infections (UTI) (27%) were the most common indications for antibiotic use. Overall, preconception antibiotic use showed little association with SAB (HR=1.05, CI: 0.85-1.30). Use of penicillins (HR=1.01, CI: 0.65-1.55), macrolides (HR=0.87, CI: 0.41-1.82), or nitrofurantoin (HR=1.11, CI: 0.53-2.34) was not appreciably associated with SAB, while use of cephalosporins was associated with increased risk (HR=1.63, CI: 0.87-3.06), although estimates were imprecise. Antibiotic use was associated with a slightly increased risk of SAB when used for UTI (HR=1.27, CI: 0.87-1.84) or pelvic/vaginal infections (HR=1.58, CI: 0.96-2.61), but not for respiratory infections (HR=0.94, CI: 0.64-1.37), indicating potential confounding by indication.Conclusions: Overall, our findings provide little support for the hypothesis that preconception antibiotic use is associated with SAB.

PO-3725 Sex Differences In The Risk Of Diabetes Mellitus Among Individuals With Psoriasis 1400732 Raymond Milan Research Institute of the McGill University Health Centre Sex Differences In The Risk Of Diabetes Mellitus Among Individuals With Psoriasis Disease Epidemiology/Clinical Course -> Diabetes Background: The choice of therapy for managing psoriasis varies according to its severity. It is also well established that psoriasis severity is associated with diabetes mellitus (DM) and severe psoriasis is more common in male than female individuals. However, sex differences in the risk of DM among individuals with psoriasis has not been examined.Objectives: We assessed the risk of DM by psoriasis severity and identified predictors in male and in female individuals with psoriasis.Methods: We conducted a retrospective cohort study using Quebec health administrative databases (1997 - 2015). Individuals ages ≥20 years were included if they received a first psoriasis diagnosis in 2000 - 2012 (Index date, ID), and either a second diagnosis or ≥1 claim for psoriasis treatment anytime afterward (N=27,582). Those who received a psoriasis treatment in the prior year, and those with a DM diagnosis in the prior or following year were excluded (to consider delays in DM diagnosis). Psoriasis treatments served as surrogates for disease severity: untreated, mild (topical agents), moderate (phototherapy), severe [conventional systemic agents (CSA) and biologic agents]. This definition was used as a time-varying exposure. DM was defined as 2 outpatient visits within 2 years or 1 hospitalization with a principal or secondary diagnosis for DM. Time-dependent multivariate Cox regression models (with 12-month lag) were used. Analyses were stratified by sex. Sensitivity analyses were conducted using other definitions for DM, different lag periods, and censoring individuals receiving methotrexate and cyclosporine because of their possible association with DM.Results: This study included 11,236 individuals with psoriasis (52.4% females). DM incidence was higher in males (20/1000 person-years vs 17/1000 person-years for females). DM risk increased with psoriasis severity in both sexes (severe vs untreated individuals: hazard ratio; 95% CI: males 2.16, 1.44-3.25; and females 2.27, 1.42-3.63). The effect of age on DM differed between sexes and psoriatic arthritis was associated with DM only in male individuals (1.57, 1.10-2.24). Results were robust in most sensitivity analyses, except for censoring both CSA in which severe psoriasis was not associated with DM in both sexes.Conclusions: Psoriasis severity, defined by the treatment received, is associated with DM in both males and females. Screening for DM should be considered for both sexes with high psoriasis severity and for male individuals with psoriatic arthritis.

PO-3727 A Us Multicenter Chart Review Study Of Patients With Metastatic Pancreatic Ductal Adenocarcinoma Receiving Liposomal Irinotecan After Gemcitabine-based Therapy 1400733 Rachel Bhak A Us Multicenter Chart Review Study Of Patients With Metastatic Pancreatic Ductal Adenocarcinoma Receiving Liposomal Irinotecan After Gemcitabine-based Therapy Disease Epidemiology/Clinical Course -> Cancer Background: Real-world data allows healthcare decision-makers to assess and manage therapeutic and economic options for patients, including those who would and would not have met eligibility criteria for randomized control trials (RCT) and are instead managed under usual care.Objectives: This retrospective multi-academic center chart review study describes real-world characteristics and outcomes of US patients receiving liposomal irinotecan for the management of metastatic pancreatic ductal adenocarcinoma (mPDAC).Methods: Patients with mPDAC treated with liposomal irinotecan were eligible. Initiation of liposomal irinotecan defined index date; covariates assessed included clinical characteristics and treatment patterns; real-world overall survival (rwOS) was assessed via Kaplan-Meier methodology. The target enrollment is 300 patients. The study centers included were Memorial Sloan Kettering Cancer Center, Cedars-Sinai Medical Center, Emory Winship Cancer Institute, Houston Methodist Cancer Center, Henry Ford Cancer Institute, and University of Pittsburgh Medical Center.Results: Data on 26 patients were available for initial analyses. Mean age was 68 years; 58% were female and 65% Caucasian. 54% of patients had stage IV disease at first diagnosis, and 17%, 65%, and 17% had index ECOG score of 0, 1, and 2, respectively. Common genetic mutations include KRAS (40%) and TP53 (40%). Prior to liposomal irinotecan, treatments received for metastatic disease include gemcitabine+nab-paclitaxel (77%) and fluorouracil (5-FU)/leucovorin (LV)+irinotecan+oxaliplatin (19%). Patients had received 0 (12%), 1 (23%), and ≥2 (65%) lines of therapy in the metastatic setting prior to liposomal irinotecan. Mean duration of liposomal irinotecan use was 3.0 months; liposomal irinotecan was mostly received with 5-FU (23%) or 5-FU/LV (69%). Median rwOS was 4.9 months (95% CI: 3.0, 6.3).Conclusions: Real-world data of the first 26 patients in this study show patients treated with liposomal irinotecan are older, sicker, and have had more lines of therapy than previously reported in RCT data.

PO-3729 Using Rwd To Assess Changes In Gfr Among Dkd Patients And The Rate Of Heart Failure 1400734 Seth Kuranz TriNetX Using Rwd To Assess Changes In Gfr Among Dkd Patients And The Rate Of Heart Failure Disease Epidemiology/Clinical Course -> Diabetes Background: Electronic medical records (EMRs) are a rich source of clinical data for the examination of disease progression in diabetic kidney disease (DKD) patients. EMR networks do not always have a clear source population or definition of time at risk.Objectives: The objectives of this analysis were to examine glomerular filtration rate (GFR) as a predictor of heart failure (HF) among DKD patients, and to assess the impact of varying person time (PT) definitions.Methods: The study was conducted using a US-based federated EMR network of 38M patients. DKD was defined as type 2 diabetes patients aged 20+ with an incident CKD diagnosis between 2015-2019 (index event (IE)). Patients were excluded if they had a kidney transplant, end stage renal disease, heart failure, or were deceased prior to or in the first year after the IE. Of those patients at risk for the outcome (n=76,240), 30,704 patients had at least one MDRD defined GFR measure in the year before and after the IE. Patients with increasing or decreasing GFR was further stratified by healthy (GFR ≥ 60 units) and unhealthy (15 ≤ GFR < 60 units) kidney function and categorized as follows: no change healthy (reference); no change unhealthy; increase from unhealthy to healthy; increase and remain unhealthy; decrease from healthy to unhealthy; decrease and remain unhealthy. The outcome was defined as the first HF diagnosis at least 1 year following the IE. Crude rate ratios (RRs) and differences (RDs) and 95% CIs were calculated. A Poisson regression model was used to calculate RRs adjusting for baseline confounders. The analyses were repeated with three varied definitions of PT, which involved adding 30-, 183-, and 365-day windows before and after each visit in the EMR during follow up. All patient characteristics were defined by ICD, LOINC, CPT, or RxNorm terminology.Results: Patients were 48% female with a mean age between 65 (12) and 71 (10) at the IE. In the unadjusted analyses, assuming a 30-day PT window, patients whose GFR decreased from healthy to unhealthy or decreased and remained unhealthy had 1.6 (1.2-2.2) and 1.5 (1.1-2.0) times the rate of HF. Accompanying RDs found 4.7 and 3.7 more cases of HF per 100,000 person days. Adjusted models [RR:1.4 (1.0-1.8); RR:1.1 (0.8-1.5), respectively] and increases in the PT window consistently attenuated results.Conclusions: Patients whose GFR decreased from healthy to unhealthy levels experienced the highest rate of HF with PT definitions strongly influencing the results. Researchers should clearly communicate assumptions around PT definitions and further analyses should continue to explore necessary assumptions when using EMR networks.

PO-3731 Falls And Fractures In Patients Treated For Parkinson’s Disease-related Psychosis: Comparing An Established Class To A More Recently Introduced Treatment 1400735 J Bradley Layton RTI Health Solutions Falls And Fractures In Patients Treated For Parkinson’s Disease-related Psychosis: Comparing An Established Class To A More Recently Introduced Treatment Safety End Points -> Neurological/Mental Health Background: Patients with Parkinson’s disease psychosis (PDP) are at increased risk for falls. PDP is commonly treated with atypical antipsychotics (AAP), although United States (US) labeling for AAP includes a warning of falls. Pimavanserin is an AAP approved by the US FDA for treatment of hallucinations and delusions associated with PDP.Objectives: To compare the risk of falls/fractures among patients diagnosed with PDP treated with pimavanserin as compared to AAP.Methods: We identified a cohort of patients aged ≥ 40 years initiating pimavanserin or another AAP (clozapine, quetiapine, risperidone, olanzapine, aripiprazole, brexpiprazole) following diagnosis of both PD and psychosis in US commercial and supplementary Medicare claims (2015-2018). Falls and fractures were identified through diagnosis and procedure coding. Patients initiating AAP were 2:1 matched with patients initiating pimavanserin using propensity scores, and the incidence rates (IR) of falls/fractures were compared with incidence rate ratios (IRR) and 95% confidence intervals (CI).Results: We identified 56 eligible patients initiating pimavanserin and 607 initiating another AAP. Pimavanserin initiators were younger (mean, 75 vs. 79 years, respectively) and had lower prevalences of many comorbidities and falls/fracture risk factors than initiators of another AAP, including use of ambulance/life support (55% vs 77%), heart failure (32% vs. 46%), stroke/brain injury (48% vs. 58%), paralysis (5% vs. 15%), and delirium (77% vs. 84%). However, patients taking pimavanserin had more indicators of PD- and PDP-related symptoms and treatment than patients taking another AAP, including dementia (96% vs. 90%), anticholinesterase inhibitors use (54% vs. 42%), and PD medications (96% vs. 75%). The crude IRs (cases/100 person-years) (95% CI) for falls/fractures were: pimavanserin, 20.1 (5.5-51.4); AAP, 45.8 (37.8-54.9). Matching retained 42 patients taking pimavanserin and 84 patients taking another AAP, although many covariates remained imbalanced. The resulting matched IRR (95% CI) was 0.43 (0.08-1.49).Conclusions: Sample sizes were too small to draw conclusions about the comparative risk of falls/fractures in users of pimavanserin versus another AAP. There were systematic differences between treatment groups, and pimavanserin initiators had greater PDP-specific care and treatment. This is an example of treatment dynamics where differences between the newly introduced drug and the established class are not adequately addressed through confounding adjustment.

PO-3732 Antidepressant Use Among Adults Patients In Primary Care In Brazil: Results Of A National Survey 1400736 Natalia Torres Prefeitura de Belo Horizonte Antidepressant Use Among Adults Patients In Primary Care In Brazil: Results Of A National Survey Drug Utilization Research -> Other Background: Antidepressant medicines are widely prescribed. In the last years, international population-based studies have identified growth in their use. Considering the limited number of studies on this topic in Brazil, it is important to investigate and describe characteristics of their use, providing information about a middle-income country.Objectives: To estimate the prevalence of antidepressant use and identify its associated factors among adults in primary care of the Brazilian National Health System (NHS).Methods: This study is part of the National Survey on Access, Use and Promotion of the Rational Use of Medicines in Brazil - Services, a cross-sectional and evaluative study, consisting of information collection from a representative sample of cities, primary health care services and patients in the five regions of Brazil. Patients were interviewed in primary health care services, with a specific questionnaire. An estimate of 1,800 patients would be interviewed by region of the country. At the end, 8,803 interviews were conducted, from July to December 2014. The outcome considered was use of any antidepressant in the past 30 days. We performed a descriptive analysis of the population and the antidepressants used. The association between antidepressant use and demographic, socioeconomic, clinical and lifestyle-related characteristics was evaluated using logistic regression model.Results: The prevalence of antidepressant use was 6.8% (95% CI 6.2 - 7.3). Most users who self-reported previous diagnosis of depression (70.6%) and who were extremely depressed/anxious (75.6%) did not report antidepressant use, suggesting underuse. The most commonly drugs used were fluoxetine, amitriptyline and venlafaxine. After all adjustments, the multivariate analysis found that the use of antidepressants is more likely in females, who self-reported depression and anxiety/depressive status, that practices polypharmacy and use psychotropics, have alcohol abstinence, a higher socioeconomic status and being of the white ethnicity/race.Conclusions: The prevalence of antidepressant use observed in Brazil is lower than in USA and some European countries. The factors associated with its use are consistent with the literature. The findings suggest that there may be inequity in access to drug treatment for depression in the Brazilian NHS. Although patients have a universal right to health in Brazil, the inequities found, especially for non-white individuals and those from lower social classes, show that it is necessary to improve actions to ensure access to treatment of depressive disorders by primary care users.

PO-3738 How Marketing Approval Changed The Use Of Tranylcypromine In The Netherlands 1400738 Angela van der Salm DADA Consultancy How Marketing Approval Changed The Use Of Tranylcypromine In The Netherlands Drug Utilization Research -> Trends and comparisons Background: Tranylcypromine (TCP) is an old MAO-inhibitor used in patients with treatment-resistant depression. It was taken off the market due to its safety profile in the mid-1960’s and later reintroduced but only available in the Netherlands on a named patient basis. Since June 2016 it is available again as authorized medicine.Objectives: Our aim was to investigate differences in patient characteristics of TCP users on a named patient basis versus those taking the authorized medicine.Methods: From the Out‐patient Pharmacy Database of the PHARMO Database Network, all patients were identified to whom TCP had been dispensed between 1 January 2009 and 31 December 2018. Patients were divided into 3 groups: new users up to July 2016 (new user-pre; using TCP on a named patient basis), new users after June 2016 (new user-post; using authorized TCP), and patients that switched from named patient TCP to authorized TCP after June 2016 (switchers).Patient characteristics of interest included age, gender, prescriber specialty, concomitant use of cardiovascular (CV) medication, and history of antidepressant (CNS) drug use.Results: There were 529 patients in the new user-pre group, 216 patients in the new user-post group, and 186 patients switched. The three groups showed a similar mean age (54 years; with switching occurring at the age of 59 years), and gender distribution (33% men vs 67% women). The vast majority of prescriptions was done by psychiatrists/specialists, but in switchers, TCP was prescribed three times more frequently by general practitioners (36% vs 12-13%).In the new user-post group, 36% of patients used 1 or more CV medications, vs 30% in the new user-pre group, and only 6% in switchers. TCP was prescribed to patients that did not meet the label requirement (i.e. previous use of two other antidepressants and lithium; ≥ 3 other CNS drugs) 1.5 times (95% CI of 1.02 to 2.17) more often in the new user-pre than in the new user post-group (22% vs 14%).Conclusions: This study has shown that characteristics of patients using TCP changed after the availability changed from a restricted named patient basis to an authorized medicine. Our results indicate a lack of specialist monitoring during switching, and improved but residual lack of adherence to the label requirement for initiating treatment.

PO-3741 The Spectrum Of Adenomyosis Identified In A Large Us Electronic Health Record Database 1400740 Anita Loughlin Optum The Spectrum Of Adenomyosis Identified In A Large Us Electronic Health Record Database Disease Epidemiology/Clinical Course -> Other Background: Adenomyosis, the presence of endometrial tissue within the myometrium, can co-exist conditions such as endometriosis and leiomyomas that have symptoms in common; thus it is difficult to delineate of symptom profiles specific to adenomyosis. Until recently, a definitive diagnosis was made by hysterectomy, recently use of MRI and transvaginal ultrasound allow for non-surgical diagnosis.Objectives: To examine the spectrum of the clinical presentation of adenomyosis.Methods: Chacteristics of women aged 18-55 years, with an affirmed mention of adenomyosis, identified in the natural language processed clinical notes of the Optum EHR database (2014-2018) were assessed. An anchor date was set as date of first qualifying adenomyosis mention. Baseline characteristics were identified in the 12 months prior to anchor date. A predictive probability (PP) model compared adenomyosis cases to a sample of symptomatic women aged 18-55 years without an affirmed adenomyosis mention. Adenomyosis cases were stratified by PP, and baseline characteristics were assessed across strata.Results: 19,503 women with adenomyosis were stratified from highest to lowest PP, specifically, 3,336 (PP, > 0.9); 3,651 (PP, 0.6-0.8); 3,523 (PP, 0.3-0.5); 3,857 (PP, 0.1-0.2); and 5,136 (PP, <0.1). Women in the highest PP strata were proportionally more likely to be 35+ years of age (93%, 90%, 88%, 83%, and 71%) and to have been pregnant (74%, 65%, 64%, 56%, and 32%), relative to women in the lowest PP strata. Likewise, women in the highest PP strata were more symptomatic ̶ heavy menstrual bleeding (HMB) (87%, 78%, 74%, 56%, and 23%), pelvic pain (52%, 44%, 39%, 33%, and 22%), dysmenorrhea (46%, 36%, 29%, 19%, and 10%), and gastrointestinal (GI) symptoms (51%, 47%, 44%, 41%, and 38%). General symptoms were predominant symptoms among women in lowest PP stratum, i.e., abdomen/ lower back pain (41%) and GI (38%), while the prevailing symptoms in women in higher PP strata were HMB (87%) and pelvic pain (52%). Women in the highest PP strata had more gynecologic comorbidities ̶ endometriosis (93%, 47%, 18%, 10% and 5%), leiomyomas (57%, 46%, 34%, 22%, and 8%), uterine polyps (13%, 8%, 5%, 4%, and 1%) and hysterectomy (51%, 23%, 11%, 5%, and 2%), relative to those in the lower PP strata.Conclusions: Using PP stratification of the cases, we show that the adenomyosis clinical presentation can ranged from mildly symptomatic to severely symptomatic, making recognition of disease and management of adenomyosis often difficult. Understanding symptom profiles may improve diagnostic suspicion, early detection, and lead to more treatment options for women with adenomyosis.

PO-3742 Dopamine Agonists And Risk Of Lung Cancer 1400741 Maria Hernandez-Con UPenn Center for Pharmacoepidemiology Research & Training Dopamine Agonists And Risk Of Lung Cancer Disease Epidemiology/Clinical Course -> Other Background: There is a clinical need to identify novel chemopreventive and therapeutic targets for lung cancer because it remains the leading cause of cancer deaths worldwide despite newer treatment options. Preclinical evidence suggests that dopamine receptor activation induces apoptosis and inhibits cancer cell proliferation in lung cancer. Dopamine agonists (DAs) are commonly used in Parkinson’s disease (PD) and restless leg syndrome (RLS).Objectives: To determine the association between long-term DA use and the risk of lung cancer.Methods: We conducted a population-based retrospective cohort study using private insurance administrative claims data from Optum®ClinformaticsTMData Mart (1/2006 - 12/2016). The study population included all adults ≥ 40 years who had ≥ 1 outpatient or inpatient ICD-9 or ICD-10 diagnosis for RLS recorded during the study period. We excluded PD patients because PD itself is associated with a lower overall risk of cancer. Cohort follow-up started on the date of first RLS diagnosis and ended on the earliest of the following: incident diagnosis of lung cancer, end of enrollment in the database or end of study period. The exposure of interest was cumulative duration (i.e., no use or ≤1 year [reference group], 1- 3 years and > 3 years) of DA (i.e., pramipexole, ropinirole or rotigotine) use, measured in a time-varying manner. The outcome was time to incident lung cancer. We constructed a multivariable Cox regression model to estimate the HRs and 95% CIs for lung cancer associated with cumulative durations of dopamine agonist use. The model was adjusted for the following covariates: age, gender, race/ethnicity, area of residence, chronic obstructive pulmonary disease (COPD), tobacco use and interstitial lung disease (ILD).Results: Over the 11-year study period, we identified 303,079 eligible patients with a diagnosis of RLS. The mean age of the study population was 62.9; 66.3% were women and 82.3% were white. The prevalence of any DA exposure in the study population was 48.2%. Compared to the reference exposure group, the crude HRs for lung cancer were 1.16 (95%CI 1.03 -1.3) and1.33 (95% CI 1.09 -1.62) for 1-3 years and > 3 years of cumulative DA use, respectively. In multivariable Cox regression analysis, the adjusted HR for lung cancer was 1.00 (95% CI 0.89-1.13) for 1-3 years of cumulative DA use and 1.09 (95% CI 0.89-1.34) for > 3 years of cumulative DA use.Conclusions: At doses typical for the clinical management of RLS, long-term DA use was not associated with the risk of lung cancer. Future studies are needed to determine whether there is a clinically apparent anti-cancer effect of DA exposure at higher doses as those used in animal experiments.

PO-3744 Heterogeneity Of Patient Preferences For Benefits And Risks Of Antiplatelet Therapies. 1400742 <