Selection bias in clinical trials.

All research is subject to bias that may systematically distort results. Though over 50 types of bias in analytical research have been identified and many classification schemes proposed, this review focuses on the special problem of selection bias in clinical trials. We demonstrate the systematic nature of selection bias in clinical research. We describe the common sources of selection bias in clinical trials including the (inappropriate) use of historical controls, stage migration, inclusion/exclusion criteria, the use of multiple subset analyses, and investigator bias. We then move from the general to the specific, using the recent experience of high dose chemotherapy for breast cancer as an illustrative example. Finally, we suggest means to avoid falling into the many selection bias traps that often confront clinical researchers.