Composite End Points: Implications of Changing Compositions With Longer Follow-Up.

The use of composite end points (CEP) in clinical trials and observational studies is widespread, despite continuing controversies on methodological and interpretational aspects.1 It is commonly recommended to construct CEP only from component outcomes that represent different manifestations of similar pathophysiologic processes, are of similar patient importance, similar frequency, and with similar associations with the predictor/intervention.2In the cardiovascular field, where more than one third of trials use some kind of CEP, it has been shown that the effect estimates based on CEP are mostly driven by more frequent but less severe component outcomes.3 Fatal component outcomes in clinical trials, on the other hand, tend to occur with the lowest frequency and often feature the weakest treatment effects.4The possibly complicating role of study duration and follow-up time has attracted hardly any attention in this context. It has been suggested, for example, that long-term studies on valve implantation need to consider time-related component outcomes, such as valve failure.5 However, more generally, the contribution of fatal component outcomes may be expected to increase with substantial long-term follow-up, especially in chronic disease studies in the elderly. Fundamental to both examples—valve failure and higher age mortality—is the understanding that the relative composition of the CEP may change with increasing study duration. This could lead to study duration–dependent changes of the association of a risk factor with the CEP, not because of time-varying effects of the risk factor on the component outcomes, but because of a time-varying composition of the CEP.Because the relevance of such study duration–dependent CEP compositions has not been investigated to date, the aim of the present work is to empirically address this issue in a cardiovascular patient cohort followed during 13 years.The present work is based on the observational KAROLA prospective cohort study …