Dynamic prediction of metastases after radical prostatectomy for prostate cancer

OBJECTIVE To develop a dynamic algorithm that predicts the risk of metastases from any time point after radical prostatectomy (RP). PATIENTS AND METHODS The study cohort consisted of 5741 RP patients who were treated from 1990-99. Patients were grouped into one of four clinical states at follow-up: State 1, prostate-specific antigen (PSA) undetectable; State 2, PSA 0.15-0.39 ng/mL; State 3, PSA = 0.4 ng/mL; and State 4, previous androgen deprivation or radiation therapy. Follow-up epochs (alive and at risk of systemic progression) at 0, 2, 4 and 6 years post-RP, cumulative incidence curves and multistate Cox models were used to assess the risk of metastases over the ensuing 5-year interval. Gleason score, seminal vesicle and surgical margin involvement, and PSA variables were evaluated as predictors. RESULTS Median follow-up was 11.7 years, with 4411, 4256 and 3983 patients followed with PSA at 2, 4 and 6 years, respectively. In total, 287 metastatic events occurred and the 5-year risk of metastasis was 0.4%, 2.1%, 8.7% and 12.6% for men in States 1, 2, 3 and 4, respectively. Independent predictors of metastasis by group included seminal vesicle involvement (all groups), Gleason score (groups 1, 3 and 4), current PSA (groups 3 and 4) and maximum past PSA (group 4). CONCLUSIONS We present a web-based prognostic tool for patients undergoing RP that is valid at many time points after surgery. Our tool predicts the development of metastases.