New statistical strategy for monitoring safety and efficacy in single-arm clinical trials.

Abstract

PURPOSE Efficacy and toxicity are both important outcomes in cancer clinical trials. Nonetheless, most statistical designs for phase II trials only provide rules for evaluating treatment efficacy, and moreover only allow early stopping after fixed cohorts of patients have been treated. We illustrate a new statistical design strategy for monitoring both adverse and efficacy outcomes on a patient-by-patient basis in phase II and other single-arm clinical trials. DESIGN The new strategy is used to design a phase II trial of the experimental regimen idarubicin plus cytarabine (ara-C) plus cyclosporine for treatment of patients with intermediate-prognosis acute myelogenous leukemia (AML). The design requires a maximum of 56 patients and provides continuous monitoring boundaries to terminate the trial if the toxicity rate is unacceptably high or the complete remission (CR) rate is unacceptably low compared with the rates of these events with the standard regimen of anthracycline plus ara-C. RESULTS The design has an 88% to 91% probability of stopping the trial early with a median of 15 to 18 patients if the toxicity rate of the experimental regimen is .05 to .10 above that of the standard and there is no improvement in the CR rate. If there is a .15 improvement in the CR rate and the toxicity rate is no more than .05 above that of the standard, then there is at least an 83% probability that the trial will run to completion. CONCLUSION The proposed monitoring strategy provides a flexible, practical means to continuously monitor both safety and efficacy in single-arm cancer clinical trials. The design strategy can be implemented easily using a freely available menu-driven computer program, and provides a scientifically sound alternative to the use of ad hoc safety monitoring rules.