Role of prostate-specific antigen as a predictor of outcome in prostate cancer.

Prostate-specific antigen (PSA) has been shown over the past few years to be a useful and sensitive marker for prostate cancer. During Phase III studies of the nonsteroidal antiandrogen, Casodex, in which different doses were compared with castration (either surgical or medical), serum PSA was measured on a regular basis. Attention was focused on the change in serum PSA from baseline after 3 months Casodex treatment and also on the number of patients receiving Casodex whose PSA returned to the normal range. Data from trials comparing Casodex, 50 mg/day, with castration showed a clear shortfall for Casodex compared with castration, in terms of percentage fall in PSA at 3 months, and also in the number of patients whose PSA fell into the normal range after 3 months. Subsequent analysis showed, however, that the PSA level was related to outcome in terms of time to progression. These data allowed the use of PSA to determine dose selection in a subsequent trial comparing Casodex, 100 mg/day or 150 mg/day, with castration. At the time of dose selection, changes in PSA showed a statistically significant difference between Casodex, 100 mg/day, and castration, but no significant difference between Casodex, 150 mg/day, and castration, either for the change in PSA at 3 months or for the proportion of patients whose PSA had fallen into the normal range. The idea that serum PSA levels can predict outcome in prostate cancer and that they are correlated with other measures of outcome, such as time to progression, is supported by these data. A decrease in PSA is not a true surrogate endpoint in that it cannot predict the outcome for an individual patient with complete accuracy, but it does correlate well with other measures of outcome, such as time to progression, for patient populations.