PD13-04: Long-Term Competing Risks of Mortality Among Men with Biochemical Recurrence after Radical Prostatectomy

Timothy Daskivich, Lauren Howard, Christopher Amling, William Aronson, Matthew Cooperberg, Christopher Kane, Zachary Klaassen, Martha Terris, Stephen J. Freedland

Introduction: Men with biochemical recurrence (BCR) after radical prostatectomy (RP) have little information on long-term competing risks of mortality to inform prognosis and guide treatment decisions.  We sought to identify predictors of prostate cancer metastasis and mortality (PCSM) and other-cause mortality (OCM) and to quantify long-term mortality outcomes across these key clinical predictors.

Methods: We analyzed 1,225 men with BCR after RP from 2001–2017 in the VA Shared Equal Access Regional Cancer Hospital (SEARCH) database. Multivariable competing risks regression analysis was used to identify predictors of metastasis, PCSM, and OCM and to quantify cumulative incidence of these outcomes over time. Recursive partitioning analysis was used to identify optimum variable cutpoints for prediction of PCSM and OCM.

Results: Of the 1,225 men in our sample, 243 (20%) died of other causes and 68 (6%) died of prostate cancer over a median follow up of 5.6 years (IQR 2.7, 9.1) after BCR. Multivariable competing risks regression showed that higher preoperative D’Amico tumor risk and PSA-DT at BCR <9 months were significantly associated with both metastasis and PCSM (all p=0.001); for example, men with PSA-DT at BCR < 9 months had 16% and 9% probability of metastasis and PCSM at 10 years, respectively. Advanced age and worse comorbidity burden were significantly associated with OCM (both p=0.001); for example, cumulative incidence of 10-year OCM was substantially higher among men 70 years or older with CCI 1 (45%), 2 (40%), or 3+ (49%) compared with those with CCI 0 (20%).    RPA identified optimal variable cutpoints for prediction of PCSM—PSA at recurrence (<10 vs. =10 ng/ml),  D’Amico tumor risk (low/intermediate vs. high), and PSA doubling time (<9 vs. =9 months)—with the highest risk subgroup having 10-year PCSM of 21% (Figure).  RPA identified optimal variable cutpoints for prediction of OCM—age at BCR (<70 vs. =70 years), and CCI at BCR (for those <70,  CCI<2 vs. =2; for those =70,  CCI0 vs. =1)—with the highest risk subgroup having 10-year OCM of 50% (Figure).

Conclusions: Readily available clinical predictors such as PSA and PSA-DT at BCR, preoperative D’Amico tumor risk, age, and CCI can robustly stratify long-term risk of metastasis, PCSM, and OCM in men with BCR after RP. Source of

Funding: Department of Defense grant PC160720 (to TJD)