Another view on the analysis of cardiovascular morbidity/mortality trials†

摘要: In many morbidity/mortality studies, composite endpoints are considered. Although the primary interest is to demonstrate that an invention delays death, expected death rate often low studies focussing on survival exclusively not feasible. Components of endpoint chosen such their occurrence predictive for time death. Therefore, if non-fatal events censored by censoring no longer independent. As a consequence, analysis components cannot be reasonably performed using classical methods times like Kaplan–Meier estimates or log-rank tests. this paper we visualize impact disregarding dependent during and discuss practicable alternatives studies. context simulations provide evidence copula-based have potential deliver practically unbiased hazards endpoint. At same time, they require minimal assumptions, which important since all assumptions generally verifiable because censoring. there alternative ways analyze more appropriately accounting dependencies among endpoints. Despite limitations mentioned, these can at minimum serve as sensitivity analyses check robustness currently used methods. Copyright © 2010 John Wiley & Sons, Ltd.