Time-dependent event accumulation in a cardiovascular outcome trial of patients with type 2 diabetes and established atherosclerotic cardiovascular disease

BackgroundEstimating cardiovascular (CV) event accrual is important for outcome trial planning. Limited data exist describing event accrual patterns in patients with type 2 diabetes (T2D). We compared apparent CV event accrual patterns with true event rates in the Trial Evaluating Cardiovascular Outcomes with Sitagliptin (TECOS).MethodsCentrally adjudicated event dates and accrual rates for a 4-point major adverse CV event composite (MACE-4; includes CV death, nonfatal myocardial infarction, nonfatal stroke, or unstable angina hospitalization), MACE-4 components, all-cause mortality (ACM), and heart failure hospitalization were compiled. We used three graphical methods (Weibull probability plot, plot of negative log of the Kaplan-Meier survival distribution estimate, and the Epanechnikov kernel-smoothed estimate of the hazard rate) to examine hazard rate morphology over time for the 7 outcomes.ResultsPlots for all outcomes showed real-time constant event hazard rates for the duration of the follow-up, confirmed by Weibull shape parameters. The Weibull shape parameters for ACM (1.14, 95% CI 1.08-1.21) and CV death (1.08, 95% CI 1.01-1.16) were not sufficiently > 1 as to require non-constant hazard rate models to accurately depict the data. The time lag between event occurrence and event adjudication being completed, the adjudication gap, improved over the course of the trial.ConclusionsIn TECOS, the nonfatal event hazard rates were constant over time. Small increases over time in the hazard rate for fatal events would not require complex modelling to predict event accrual, providing confidence in traditional modelling methods for predicting CV outcome trial event rates in this population. The adjudication gap provides a useful metric to monitor within-trial event accrual patterns.

Automated summary

This study found that the incidence rate of cardiovascular events in patients with type 2 diabetes remained constant over time. Therefore, traditional modeling methods can be used to predict the occurrence rate of cardiovascular events in this population.