Reproducibility of clinical events adjudications in a trial of venous thromboembolism prevention.

ABSTRACT

Essentials The reproducibility of Clinical Events Committee (CEC) adjudications is almost unexplored. A random selection of events from a venous thromboembolism trial was blindly re-adjudicated. 'Unexplained sudden deaths' (possible fatal embolism) explained most discordant adjudications. A precise definition for CEC adjudication of this type of events is needed and proposed. SUMMARY: Background When clinical trials use clinical endpoints, establishing independent Clinical Events Committees (CECs) is recommended to homogenize the interpretation of investigators' data. However, the reproducibility of CEC adjudications is almost unexplored. Objectives To assess the reproducibility of CEC adjudications in a trial of venous thromboembolism (VTE) prevention. Methods The PREVENU trial, a multicenter trial of VTE prevention, included 15 351 hospitalized medical patients. The primary endpoint was the composite of symptomatic VTE, major bleeding or unexplained sudden death (interpreted as possible fatal pulmonary embolism [PE]) at 3 months. The CEC comprised a chairman and four pairs of adjudicators. Of 2970 adjudicated clinical events, a random selection of 179 events (121 deaths, 40 bleeding events, and 18 VTE events) was blindly resubmitted to the CEC. Kappa values and their 95% confidence intervals (CIs) were calculated to measure adjudication agreement. Results Overall, 18 of 179 (10.1%, 95% CI 6.5-15.3%) adjudications proved discordant. Agreement for the PREVENU composite primary endpoint was good (kappa = 0.73, 95% CI 0.61-0.85). When analyzed separately, agreements were very good for non-fatal VTE events (1, 95% CI not applicable), moderate for all (fatal and non-fatal) VTE events (0.58, 95% CI 0.34-0.82), good for fatal and non-fatal major bleeding events (0.71, 95% CI 0.55-0.88), and moderate for all fatal events (0.60, 95% CI 0.40-0.81). Unexplained sudden death interpreted as possible fatal PE was responsible for nine of 18 (50%) discordant adjudications. Conclusion The reproducibility of CEC adjudications was good or very good for non-fatal VTE and bleeding events, but insufficient for VTE-related deaths, for which more precise and widely accepted definitions are needed.