Introduction: The occurrence of time related biases, such as immortal time and time-window bias, have been demonstrated in several 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. Our objective was to systematically review studies on common classes of medications used during pregnancy, 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. Pregnancy outcomes of interest were pre-eclampsia, gestational hypertension, gestational diabetes, spontaneous abortion, stillbirth, preterm delivery, and small for gestational age. Articles were reviewed independently for eligibility by 2 reviewers, with discrepancies resolved by a third reviewer.
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. Spontaneous abortion (10 studies), and stillbirth and preterm birth (7 studies each) were the most commonly reported outcomes. 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.