BACKGROUND: Clinical trials in cardiology commonly consider time-to-event endpoints that are often influenced by competing risks. In the presence of competing risks, standard survival analysis techniques, such as the Kaplan-Meier estimator, can yield seriously biased results. Although methods to account for competing risks are well known in the statistical literature, they are rarely applied in clinical trials. DESIGN: Simulation study, to demonstrate the appropriate application and interpretation of the competing risks methodology with respect to time-to-event endpoints. METHODS: In this paper, different statistical approaches to account for competing risks are systematically compared, based on a simulation study and using the original data from a cardiology trial. RESULTS: Group comparisons in clinical trials that have competing time-to-event endpoints should be based on the cause-specific hazard functions. In contrast, group comparisons based on event rates should be carried out with care, as event rates are directly influenced by competing events. CONCLUSION: Ignoring or not fully accounting for competing risks may yield misleading or even erroneous results, which could hinder understanding of survival trends; therefore, it is important that competing risks methodology be routinely incorporated into clinical trial standards.
BACKGROUND: Clinical trials in cardiology commonly consider time-to-event endpoints that are often influenced by competing risks. In the presence of competing risks, standard survival analysis techniques, such as the Kaplan-Meier estimator, can yield seriously biased results. Although methods to account for competing risks are well known in the statistical literature, they are rarely applied in clinical trials. DESIGN: Simulation study, to demonstrate the appropriate application and interpretation of the competing risks methodology with respect to time-to-event endpoints. METHODS: In this paper, different statistical approaches to account for competing risks are systematically compared, based on a simulation study and using the original data from a cardiology trial. RESULTS: Group comparisons in clinical trials that have competing time-to-event endpoints should be based on the cause-specific hazard functions. In contrast, group comparisons based on event rates should be carried out with care, as event rates are directly influenced by competing events. CONCLUSION: Ignoring or not fully accounting for competing risks may yield misleading or even erroneous results, which could hinder understanding of survival trends; therefore, it is important that competing risks methodology be routinely incorporated into clinical trial standards.
Authors: Biniyam G Demissei; Douwe Postmus; Mattia A Valente; Pim van der Harst; Wijk H van Gilst; Edwin R Van den Heuvel; Hans L Hillege Journal: Eur J Epidemiol Date: 2014-11-07 Impact factor: 8.082