Re-Evaluating Recurrent Events in Heart Failure Trials: Patterns, Prognostic Implications, and Analytical...

Analyses using repeat hospitalizations (HFHs) are common in heart failure trials and typically assume that such repeat events occur randomly over time. Also, many think that using repeat events enhances statistical power. This article challenges those assumptions, using data from 4 heart failure trials of sodium-glucose cotransporter 2 inhibitors. We found marked within-patient time clustering of repeat events: risks of subsequent HFH and cardiovascular death are markedly elevated following a hospitalization, especially early on. The Lin-Wei-Yang-Ying and negative binomial models do not account for this. Alternative approaches using area under the curve and win ratio methods for the composite of cardiovascular death and all HFHs strengthened the treatment effect. But still, time-to-first event analyses tended to give the strongest evidence. Overall, some commonly used repeat event analyses appear not to be the best. It is time to rethink how best to use repeat events data in heart failure trials.