FIGURE 4.

HRs and 95% CIs¹ for the primary and secondary CVD outcomes, according to randomized assignment, where follow-up of noncompliant participants was censored. ¹Summary statistics were from weighted Cox regression models that stratified baseline hazard functions by multivitamin trial randomization group, age, sex, and recruitment cohort, and used the robust sandwich estimator for variance. Time-dependent weights were calculated as the inverse probability of compliance, where probabilities were estimated from a Cox regression model with baseline hazard functions stratified by age, sex, recruitment source, multivitamin-trial arm and cocoa extract trial arm, and included baseline history of CVD and hypertension status (time-dependent) as covariates. CIs were not adjusted for multiple comparisons. ²A participant's follow-up was censored at the first time they reported having missed >8 d of cocoa extract study pills per month, took personal nonstudy cocoa extract, or did not respond to a semiannual questionnaire. ³Total cardiovascular event was a composite of myocardial infarction, stroke, CVD death, CABG/PCI, unstable angina including hospitalization, carotid artery surgery, and peripheral artery surgery. ⁴Major cardiovascular event was a composite of myocardial infarction, stroke, and CVD death. CABG/PCI, coronary artery bypass graft and percutaneous coronary intervention; CVD, cardiovascular disease.