Predictors of medical events in patients enrolled in the cardiac insufficiency bisoprolol study (CIBIS): A study of the interactions between β-blocker therapy and occurrence of critical events using analysis of competitive risks

Abstract

Background The risk of occurrence of medical events in a clinical trial is competitive in nature; that is, in a given patient the risk of having a critical event depends on the amount of time elapsed since random assignment and on the previous events that may have occurred in the patient. The purpose of this study was to examine the relations between baseline variables, the interactions between treatment, bisoprolol, or placebo, and the occurrence of critical events during the CIBIS trial, a mortality and morbidity trial of β-blockade in patients with heart failure. Methods and Results A Cox model for censored data was used to analyze the relations between baseline variables, total deaths, permanent treatment withdrawals, nonlethal cardiovascular events, and their interactions with bisoprolol or placebo. We examined the influence of treatment on the occurrence of deaths, permanent treatment withdrawals, and nonlethal cardiovascular events by using the technique of event history analysis, which takes into account competitive risks between events. Compared with placebo, bisoprolol reduced mortality rates in patients with a left ventricular ejection fraction ≤20% (relative risk [RR] 0.49; 95% confidence interval [Cl] 0.27 to 0.88; P = .02). In patients whose baseline heart rate was in the upper fertile of distribution, permanent treatment withdrawals were less frequent in patients randomly assigned to bisoprolol than in patients randomly assigned to placebo (RR 0.50; 95% Cl 0.28 to 0.88; P = .02). Bisoprolol reduced the incidence of nonlethal cardiac events in patients in whom heart failure was present for at least 4 years (RR 0.44; 95% Cl 0.27 to 0.71; P < .01). Event history analysis revealed that among patients who died under treatment after having at least 1 nonlethal cardiovascular event, 20 patients were treated with placebo but only 7 patients were treated with bisoprolol (RR 0.41; 95% Cl 0.17 to 0.98; P < .05). Conclusions Some patients with heart failure derive more benefit from β-blocker therapy than others. In the CIBIS trial, they are those patients with the lower left ventricular ejection fractions and those who have nonlethal cardiovascular events but in whom β-blocker therapy is not permanently discontinued.