gms | German Medical Science

In heart failure trials efficacy is usually proven by a composite endpoint including cardiovascular death (CVD) and recurrent heart failure hospitalisations (HFH), evaluated with time-to-first-event analysis based on a Cox model. As a considerable fraction of events is ignored that way, recurrent event analaysis approaches were suggested, amongst others the semiparametric proportional rates models by Lin, Wei, Yang & Ying (LWYY model) [1] and Mao & Lin (Mao-Lin model) [2]. Both models focus on a certain composite event rate and target different estimands. In our work, we will characterize the behavior of the composite treatment effect estimates resulting from the Cox model, the LWYY model and the Mao-Lin model in clinically relevant scenarios. The latter are mimicked by joint frailty models [3], [4], that account both for different treatment effects on the two outcomes and for the positive correlation between HFH and CVD risk rates. The proportional rates assumption of the aforementioned composite endpoint approaches is in general violated in such situations. We apply simulations for finite-sample properties and least false parameter theory for asymptotic statements. Thereby the term “least false parameter” refers to the asymptotic estimate in a misspecified model, which can be derived numerically if the true data-generating process is known. In particular we show that, for all approaches, the positive correlation results in an attenuation of the estimate in case of beneficial effects on the two outcomes. Moreover, the treatment effect estimate decreases with increasing trial duration. The LWYY model and the Mao-Lin model yield very similar results in clinically relevant settings and are less affected by the outcomes' correlation than the time-to-first-event analysis. These results contribute to a better understanding on the behavior of recurrent event approaches, that recently has been claimed by regulators. Our theoretical investigations are motivated and compared with empirical results of the PARADIGM-HF trial (CT.gov identifier: NCT01035255), a large multicenter trial with 8399 chronic heart failure patients.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.