Confidence intervals versus p-values for interpretation of clinical trial results: introduction.

The following three papers summarize the presentations at a Society for Clinical Trials annual meeting session on the relative merits of estimation versus testing for analysis of randomized clinical trials. By design, randomized clinical trials have internal validity. Whether they also possess quantitative external validity--generalizability of effect size to some population represented by the trial subjects--is one of the main points of disagreement among the three authors. It may be unrealistic to expect a resolution that applies across the wide variety of therapeutic areas and clinical trial goals. Extrapolation from clinical trial to clinical practice is often endorsed in connection with large trials having loose entry criteria and focusing on an objective, clearly meaningful clinical endpoint. By contrast, the relevance of estimates of effect size is less clear in the case of many clinical trials conducted in the course of drug development.