Importance of prognostic factors in cancer clinical trials.

The importance of prognostic factors in the design, analysis, and reporting of clinical trials is discussed. For many kinds of cancer, the variability in prognosis among patients is greater than the size of treatment differences usually seen. Consequently, failure to understand and adequately account for patient heterogeneity easily leads to unreliable claims and inefficient trials. Identified prognostic factors are generally of sufficient importance to demand attention in design and analysis, but rarely are sufficiently explanatory to render randomization unnecessary. Improvement in knowledge of prognostic factors is important for sharpening the focus and for improving the reliability, efficiency, and interpretability of clinical trials. Problems in the conduct of prognostic factor studies are also discussed, and the calculation of the proportion of variability explained by logistic regression models is illustrated for two examples.