Why to avoid dual primary endpoints in cardiovascular trials

Background/Introduction The EMA guideline on the treatment of chronic heart failure (CHF) requires at least one adequately designed randomised controlled trial for confirming efficacy of a new drug [1]. Usually, mortality (all-cause or cardiovascular) should be the primary endpoint, either alone or as part of a composite endpoint with outcomes related to worsening heart failure (WHF). Functional capacity and clinical symptoms are only acceptable as primary endpoint in specific cases. If more than one primary endpoint is defined, a multiplicity adjustment strategy is required to control the study-wise type-1-error (T1E). Purpose From a statistical perspective, different multiplicity adjustment strategies can be applied. Traditionally, if, e.g., two co-primary endpoints were considered as primary, it was assumed that the null hypotheses for both should be rejected to conclude study success. Recently, the so-called dual primary endpoint concept has been proposed and applied in CHF trials. We aim to clarify differences between the multiplicity concepts, and illustrate implications for the interpretation of trial success. Methods With co-primary endpoints, no adjustment of the significance level for testing each null hypothesis is needed, but the sample size usually needs to be increased to maintain the pre-defined power of the study. In contrast, in the dual primary endpoint concept, an (uneven) Bonferroni split of the T1E between the primary endpoints is foreseen. While the T1E is formally controlled, study success can be claimed as soon as the null hypothesis for one of the primary endpoints is rejected even if there is no benefit or a detriment for the other primary outcome. This is obviously not in line with clinical and regulatory interpretation of the overall trial outcome [2]. The recently published SUMMIT trial defined dual primary endpoints: a composite of cardiovascular death or a WHF event and the change from baseline to 52 weeks in the Kansas City Cardiomyopathy Questionnaire clinical summary score (KCCQ-CSS) [3]. Consequently, one could claim study success based on a benefit in KCCQ-CSS alone despite a detriment in the composite primary endpoint, which is not in line with both the CHF and the multiplicity EMA guideline [1, 4]. To resolve this conflict, we propose to define them as co-primary by requiring superiority for KCCQ-CSS and non-inferiority to a pre-specified margin for the composite primary endpoint. Results The hypotheses of the considered multiplicity adjustment strategies are depicted in Table 1. The proposed co-primary endpoint concept combining superiority and non-inferiority can be implemented in two ways and allows flexible modelling of the minimum requirements for all endpoints. Conclusion The proposed approach only leads to a small increase in sample size while it aligns statistical with clinical interpretation of study success and follows the relevant EMA guidelines.