Use of Gleason score, prostate specific antigen, seminal vesicle and margin status to predict biochemical failure...

We determine the importance of clinical and pathological variables for predicting biochemical progression in patients after surgery for specimen confined prostate cancer. We developed a simple scoring algorithm for biochemical progression in node negative cases and tested the algorithm performance on an independent group.

Our study included 2,518 patients with pT2N0 or pT3N0 disease treated between 1990 and 1993. Gleason score, preoperative prostate specific antigen (PSA), margin status, extraprostatic extension, seminal vesicle involvement, DNA ploidy and adjuvant treatment were primary variables analyzed univariately. The Cox proportional hazards model was used on 2,000 randomly selected patients to develop a multivariate scoring algorithm for the aforementioned factors to predict biochemical progression-free survival. The final model included Gleason score, preoperative PSA, margin status, seminal vesicle involvement and adjuvant treatment. The prognostic score derived from this model was validated by applying it to the remaining 518 patients. Harrell's measure of concordance (C) was used to compare competing models.

For patients who did not receive adjuvant therapy the derived score based on the Cox model coefficient was Gleason +1 (PSA 4 to 10), +2 (PSA 10.1 to 20), +3 (PSA greater than 20), +2 (positive seminal vesicle) and +2 (positive margin). The score was reduced by 4 if adjuvant hormonal therapy was given and by 2 for only adjuvant radiotherapy. The 5-year progression-free survival was 94% for scores less than 5, 60% for 10 and 32% for greater than 12 (C = 0. 718). Applying the score to the independent validation data set (518) resulted in 5-year progression-free survival of 96% for scores less than 5, 53% for 10 and 30% for greater than 12 (C = 0.759).

Progression-free survival determined by the model score group identified a wide range of risk levels for patients with specimen confined prostate cancer. This simple predictive model allows identification of patients at high risk for cancer progression with specimen confined disease who may be targeted for closer surveillance and adjuvant therapy, while those at lower risk may be simply observed.