Randomized phase II designs in cancer clinical trials: current status and future directions.

PURPOSE Randomized phase II (RPh2) designs are popular in cancer clinical trials because of the smaller sample size requirements when multiple treatments are being evaluated. We reviewed the use of RPh2 designs and give comments on future directions. DESIGN The trial design, statistical properties, conduct, data analysis, results, and reporting were examined in RPh2 trials reported from 1986 to 2002. RESULTS A statistical design was reported in only 46% of the 266 cancer trials, and approximately half of those provided inadequate information. Most studies applied randomization to achieve patient comparability, while embedding a one-sample phase II design within each treatment arm. Seventy-five percent of the trials' accruals were within +/- 10% of their targets. The average accrual rate was 3.3 patients per month. Planned interim analyses were reported in 27% of the trials, and 56% of the trials were stopped early; 69%, 13%, 13%, and 4% of the trial discontinuations were because of lack of efficacy, efficacy, toxicity, and slow accrual, respectively. Thirty-nine trials (14%) recommended or started phase III evaluations, with four positive reports in six phase III studies identified. CONCLUSION There is a trend of increasing use of RPh2 designs in cancer research. Continued improvement in study design, conduct, analysis, and reporting is required to enhance the quality of RPh2 designs. The accrual rate and success rate of the trials remain low, and therefore, futility stopping rules to terminate ineffective treatment arm(s) should be implemented more frequently. More innovative, flexible RPh2 designs are needed to facilitate the development of effective cancer treatments.