Post hoc analyses: after the facts

Prospective clinical trials are constructed with high levels of internal validity. Sample size and power considerations usually address primary endpoints. Primary endpoints have traditionally included events that are becoming increasingly less common and thus have led to growing use of composite endpoints and noninferiority trial designs in transplantation. This approach may mask real clinical benefit in one or the other domain with regard to either clinically relevant secondary endpoints or other unexpected findings. In addition, endpoints solely chosen based on power considerations are prone to misjudgment of actual treatment effect size as well as consistency of that effect. In the instances where treatment effects may have been underestimated, valuable information may be lost if buried within a composite endpoint. In all these cases, analyses and post hoc analyses of data become relevant in informing practitioners about clinical benefits or safety signals that may not be captured by the primary endpoint. On the other hand, there are many pitfalls in using post hoc determined endpoints. This short review is meant to allow readers to appreciate post hoc analysis not as an entity with a single approach, but rather as an analysis with unique limitations and strengths that often raise new questions to be addressed in further inquiries.