Our survey of 423 negative clinical trials indicates that 55% of trials had too few participants to detect a medium effect size in favor of the experimental over the standard treatment arm for their primary end point with at least 80% statistical power. Although underpowered negative clinical trials have been widely reported in the general medical and subspecialty literature, there are few reports relating to trials evaluating treatment of cancer. A review of 22 negative randomized oncology trials published in major general medical or oncology journals during a 1-year period found that 16 trials (73%) lacked adequate statistical power to detect a 50% improvement in median survival in favor of the experimental arm.