Prostate Cancer: Prognostic factors, markers of outcome and design of clinical trials

Phase III clinical trials to assess the clinical benefit of new treatment options often require large patient numbers and long follow-up, in particular in diseases with a long natural history, such as prostate cancer. In this thesis, we argue that in order to improve the efficiency of phase III prostate cancer clinical trials, a thorough understanding of prognostic factors of outcome is needed, as well as an exploration of potential predictive factors that might affect treatment benefit and cause heterogeneity of results. In addition, the development of surrogate endpoints of the long term clinical outcome is needed in order to speed up the drug development process by reducing the observation time required to assess the final endpoint. We assess the prostate specific antigen (PSA) as a potential surrogate endpoint. In assessing predictive factors of treatment benefit by retrospective analysis of existing clinical data, we recommend the greatest care for the confounding effect between prognostic factors and predictive factors, because factors that are predictive of the treatment effect will influence the prognostic analysis in pooled treatment arms analysis We also recommend that treatment effects in subgroups of patients be considered only if there is evidence of heterogeneity of results from a single statistical test for interaction, so as to avoid multiple testing problems that increase the chance of finding spurious statistical findings. The evidence that PSA progression is a surrogate endpoint for overall survival in endocrinally treated metastatic prostate cancer is moderate: we show that non null treatment effects on survival would potentially be identifiable only in very large new trials showing a very large effect on the PSA endpoint. The size of a study based on PSA should target a very precise estimation of the treatment effect onto the PSA endpoint rather than be based on statistical power. A study should not be stopped whenever statistically significant PSA results become available, we rather recommend that the follow-up should be continued to later document long term safety of the treatments and their impact on survival. With the proviso that longer term analysis of the trial be planned, effects on PSA endpoints might serve as a basis for accelerated drug approval, together with other trial data documenting safety and other patient benefits. We also suggest that a PSA endpoint could be used to sto! p the trial early on for futility in case no apparent effect on the PSA endpoint is shown and if there is strong evidence that the treatment effect is at least in part expressed through PSA changes. We recommend a thorough understanding of the treatment biological mechanisms of action and of the marker before defining a marker-based endpoint and using it in clinical trials.