Generation of virtual control groups for single-arm prostate cancer (PCa) adjuvant trials.

Abstract

239Background: It is difficult to construct a control group for trials of adjuvant Rx of PCa after radical prostatectomy (RP) due to ethical issues, patient acceptance. We present an algorithm involving 8 regression models to generate virtual control groups based on the Kattan post-prostatectomy nomogram. Methods: The online Kattan nomogram gives the probability of being progression-free at 2, 5, or 7yr after RP without adjuvant Rx. For each subject we estimate the progression-free survival (PFS) from these data by applying 8 models that use curve fitting and estimation of time to relapse. The models estimated time to 60%, 65%, . . . 95% chance of relapse for each subject. To prepare a virtual control group for a single-arm adjuvant Rx trial, the predicted PFSs generated from the optimum model are compared with the observed PFSs by the logrank test. Results: The 8 models were systematically applied to a training set of 153 post-RP cases without adjuvant Rx. The optimum model depended on the clinical features of the test population. Higher relapse risk required a higher stringency model. The algorithm was validated with an independent dataset of 155 post-RP patients. When analyzed by the optimum model there was high correlation between the group observed, predicted PFSs (p = 0.792 for no difference). We created a virtual control group for pts on a Ph II trial of adjuvant chemohormonal Rx post RP. The observed PFSs for these subjects were significantly longer than the PFSs predicted by the new algorithm, indicating that the adjuvant Rx is highly effective in prolonging PFS after RP in pts at high risk for PCa recurrence. This was confirmed by comparing observed PFSs with PFSs of a matched historical group. Conclusions: The algorithm can accurately generate control groups for single arm adjuvant Rx trials for PCa, facilitating development of new therapeutic strategies.