Abstract

Multiple investigators have argued that PSA may no longer be an accurate marker of prostate cancer biology. We determined whether the impact of PSA in predicting biochemical failure after radical prostatectomy has changed since the beginning of the PSA era.A total of 1,246 patients were identified from the Columbia University Comprehensive Urological Oncology Database who underwent radical prostatectomy by 1 of 3 surgeons between 1988 and 2003. Cox proportional hazards models were fit to the data to estimate the impact of PSA (logPSA) in predicting BCF (PSA 0.2 ng/ml or greater). To determine if the predictive impact of PSA changed over time, patients were classified based on year of surgery, and an interaction term between PSA and time was included. Finally concordance indexes were estimated to determine if the predictive ability of PSA has changed over time.In a Cox model including PSA, year of surgery and a year/PSA interaction term, the impact of PSA appears to change over time (p = 0.002). However, when correcting for the effects of stage and grade there was no significant change in the impact of PSA. In addition, concordance analysis indicated that the predictive ability of PSA has remained constant throughout the PSA era (0.65, 0.66 and 0.64 for each period, respectively).This study demonstrates that the predictive ability of PSA as a cancer outcomes biomarker has not changed significantly since the beginning of the PSA era. Despite suggestions to the contrary, PSA remains an important variable in predicting risk of BCF after RP.

View details for DOI 10.1016/S0022-5347(05)01022-0

View details for Web of Science ID 000236928400016

View details for PubMedID 16600724