Picking the winners in a sea of plenty

Selecting an experimental arm for a Phase III trial is based on the results of Phase II investigations. Historical results show that this paradigm leads to the failure of many experimental therapies in the Phase III setting. This is the result of failures in the Phase II design that include differences in the patient populations and basing sample size determinations on levels of benefit derived from surrogate end points that do not accurately reflect the end point of interest in the Phase III study. An additional factor is how to ensure that the experimental therapy chosen was the best available at the time.

We consider castrate metastatic prostate cancer, for which multiple regimens appear to have similar activity at this time. To assess superiority, we use a randomized Phase II/III design developed by Schaid et al. (D. J. Schaid et al., Biometrika, 77: 507-513, 1990) that allows multiple treatments to be tested at the same time and bases the determination to proceed from the Phase II study on the same clinical end point evaluated in the same population as the Phase III trial. A concurrent control group is also treated.

We demonstrate the integrative Phase II/III clinical trial design to evaluate two, three, or four experimental treatments with a survival-based end point in the same patient population. It includes a concurrent control in both the Phase II and Phase III portions of the study. The sample sizes in the Phase II component of the trial are comparable with those found in conventional single-arm Phase II trials.

The proposed design is valuable in situations where multiple regimens are available that appear worthy of evaluation in the Phase III setting, and where there is no adequate short-term surrogate end point for survival. The design is also useful in the evaluation of cytostatic agents where traditional response parameters may not identify potentially active drugs or, as is the case in advanced prostate cancer, in the evaluation of therapies that have a direct effect on prostate-specific antigen with an uncertain effect on survival.