Selection bias in clinical trials of anaplastic glioma.

We report a retrospective analysis of survival, and attrition to a study, of patients with anaplastic glioma seen at a regional cancer center during its participation in a prospective randomized clinical trial. Median length of survival for all patients was 35 weeks. Median length of survival for study patients was 60 weeks. Study patients were not representative of all patients; they were younger and less disabled than the nonstudy patients. Survival predictions for patients with anaplastic glioma based on the results of clinical trials may be overly optimistic. Many trials study a nonrepresentative subset of patients with anaplastic glioma, excluding those who do poorly at the outset and those who have poor function.