Composite outcomes in clinical trials: uses and problems.

Randomized clinical trials are the gold standard for proof of efficacy of a therapeutic or preventive intervention, because their design is the only one capable of establishing causation in medicine. Today drug licensing is based on RCTs that have shown improvement in patients' symptoms, prognosis or both. One of the most crucial steps in RCT planning is selection of the main outcome measures. These are normally classified as primary and secondary outcomes. The primary outcome is very important, since it constitutes the main basis for claiming efficacy in an RCT, so its choice and use must be particularly rigorous. Composite outcomes have become increasingly common in the medical literature. Even trials with a single primary outcome usually have several secondary ones (composite). Composite outcomes present several potential problems: clinical endpoints must be relevant and significant in terms of patient prognosis, and so non-validated surrogate endpoints should not be used; they must be carefully defined, should not be excessive in number, and, when they include all-cause mortality, should be carefully constructed. In this article we present and discuss the uses and problems of composite outcomes in RCTs.