Abstract
BACKGROUND: Because of the osseous distribution of prostate cancer metastases, progression is more readily identified than response in prostate cancer clinical trials. As a result, there is an increased focus on progression-free survival (PFS) as a phase 2 endpoint. PFS, however, is vulnerable to inter-study design variability. The authors sought to identify and quantify this variability and the resultant error in PFS across prostate cancer clinical trials. METHODS: The authors reviewed phase 2 clinical trials of cytotoxic agents in castration-resistant metastatic prostate cancer over 5 years to evaluate the policies determining extent of disease and the definitions of disease progression. A simulation model was created to define the degree of error in estimating PFS in 3 hypothetical cohorts (median PFS of 12, 24, and 36 weeks) when the frequency of outcome assessments varies. RESULTS: Imaging policies for trial entry were heterogeneous, as were the type, timing, and indications for outcome assessments. In the simulation, error in the reported PFS varied according to the interval between assessments. The difference between the detected and the true PFS could vary as much as 6.4 weeks per cycle, strictly resulting from the variability of assessment schedules tested. CONCLUSIONS: Outcome assessment policies are highly variable in phase 2 studies of castration-resistant prostate cancer patients, despite published guidelines designed to standardize authentication of disease progression. The estimated error in PFS can exceed 6 weeks per cycle, exclusively because of variations in the assessment schedules. Comparisons of PFS times from different studies must be made with circumspection.