Predictive postoperative model for biochemical recurrence in patients with localized prostate cancer treated with radical prostatectomy as monotherapy.

Objectives: To identify the post-prostatectomy prognostic factors of biochemical recurrence (BCR) and develop a predictive model for BCR based on predictive pathological variables after radical prostatectomy (RP).
Methods: We retrospectively analysed patients with clinically localised prostate cancer treated with RP as monotherapy with a minimum follow up period of 12 months. We considered BCR to be the persistence or elevation of PSA levels after RP of> 0,4 ng/ml, and rising in the following determination. We performed uni-and multivariate analysis, using the logistic regression test to determine the variables associated with BCR. We developed a mathematical model to estimate BCR, based on the variables identified, with a logistic function equation and then designed an Excel spreadsheet to apply it. Calibration and discrimination were performed by way of a Hosmer-Lemeshow test and an ROC curve.
Results: 693 patients were included. Average age was 63.5 years and average follow up was 88.5 months. BCR was observed in 218 patients. The average time to BCR was 35.5 months, and 90% of the cases occurred in the first 7 years. In the multivariate analysis, the PSA, Gleason Score (GS) = 7(4+3), pathological stage pT3b and affectation of the surgical margin (SM) were identified as independent prognostic pathological variables related to BCR (p〈0,001). The above four variables were included into the equation of the model. Specificity and sensitivity were 90.6% and 50.2%. Its predictive capacity was 80.5% (CI 95% 76,80 -84.3).
Conclusions: PSA, GS = 7(4+3), pathological stage pT3b and PSM were found to be the independent prognostic pathological variables related to BCR-free survival. The predictive model developed permits BCR risk estimation with a reliability of 80.5%