Article type
Year
Abstract
Background: Although the use of composite endpoints (CEPs) in randomized clinical trials (RCTs) is commonplace, their interpretation is problematic when they include components that vary greatly in importance (such as death and occurrence of angina), and when there is a large gradient of treatment effect across components.
Methods: We reviewed the use of CEPs in cardiovascular RCTs published in the Lancet, Annals of Internal Medicine, Circulation, European Heart Journal, JAMA, and New England Journal of Medicine, from January 1, 2002 to June 30, 2003. We categorized component endpoints of CEPs according to patient-importance as fatal, critical, major, moderate, or minor.
Findings: Of 114 identified RCTs that included a patient-important CEP, 68% reported complete component data for the primary CEP; almost all (98%) primary CEPs included a fatal endpoint. Of 84 CEPs for which component data were available, 54% (45 of 84) showed large or moderate gradients in both patient-importance and magnitude of effect across components. When analyzed by categories of patient-importance, the most important components were associated with lower control group event rates (medians of 3.3% to 3.7% for fatal, critical and major outcomes, and 12.3% and 8.0% for moderate and minor outcomes respectively). Patient-important components of greater importance were associated with smaller treatment effects than less patient-important ones (relative risk reduction of 8% for death, and 33% for components of minor patient-importance).
Interpretation: Our review suggests that the use of composite endpoints in cardiovascular trials is frequently complicated by large gradients in patient-importance and in magnitude of the effect of treatment across component endpoints. Higher event rates and larger treatment effects associated with less important components may result in misleading impressions of treatment impact. Cochrane review authors should avoid using composite endpoints, and report separately on components. Unfortunately, reporting procedures in original studies may make optimal reporting in systematic reviews a challenge. The Collaboration should discourage the use of composite endpoints.
Methods: We reviewed the use of CEPs in cardiovascular RCTs published in the Lancet, Annals of Internal Medicine, Circulation, European Heart Journal, JAMA, and New England Journal of Medicine, from January 1, 2002 to June 30, 2003. We categorized component endpoints of CEPs according to patient-importance as fatal, critical, major, moderate, or minor.
Findings: Of 114 identified RCTs that included a patient-important CEP, 68% reported complete component data for the primary CEP; almost all (98%) primary CEPs included a fatal endpoint. Of 84 CEPs for which component data were available, 54% (45 of 84) showed large or moderate gradients in both patient-importance and magnitude of effect across components. When analyzed by categories of patient-importance, the most important components were associated with lower control group event rates (medians of 3.3% to 3.7% for fatal, critical and major outcomes, and 12.3% and 8.0% for moderate and minor outcomes respectively). Patient-important components of greater importance were associated with smaller treatment effects than less patient-important ones (relative risk reduction of 8% for death, and 33% for components of minor patient-importance).
Interpretation: Our review suggests that the use of composite endpoints in cardiovascular trials is frequently complicated by large gradients in patient-importance and in magnitude of the effect of treatment across component endpoints. Higher event rates and larger treatment effects associated with less important components may result in misleading impressions of treatment impact. Cochrane review authors should avoid using composite endpoints, and report separately on components. Unfortunately, reporting procedures in original studies may make optimal reporting in systematic reviews a challenge. The Collaboration should discourage the use of composite endpoints.