Interpreting potential benefit on secondary outcomes when a trial fails to meet its primary endpoint

Randomized clinical trials are designed to test the effect of a therapy on the chosen primary endpoint. Beyond the primary endpoint, other clinically important endpoints are also ascertained. To avoid false discovery from testing an unlimited number of endpoints, a hierarchical-testing approach is commonly used where the primary endpoint and order of testing of secondary endpoints are pre-specified. When a trial fails to meet its pre-selected primary endpoint, a frequentist interpretation would consider all remaining endpoints to be inconclusive. Although this approach is commonly used for regulatory purposes, clinicians are not bound by such an approach that downgrades a large body of trial evidence, not based on biological ground-truth, but rather simply based on the subjective pre-trial selection of the primary endpoint and chosen sequence of hierarchical hypothesis testing. Secondary endpoint analyses in a trial that fails to meet its primary endpoint requires caution in interpretation as the risk of false-positive findings cannot be ignored. However, as detailed below, it may be prudent to carefully draw clinical conclusions in select circumstances, when the number of events is considerable, particularly when there is meaningful supportive evidence from other trials.