Study Objectives: To determine the clinical variables that best predict long- term continuous positive airway pressure (CPAP) adherence among patients with cardiovascular disease who have obstructive sleep apnea (OSA). Design: 12-mo prospective within-trial observational study. Setting: Centers in China, Australia, and New Zealand participating in the Sleep Apnea cardioVascular Endpoints (SAVE) study. Patients: There were 275 patients age 45-70 y with cardiovascular disease (i.e., previously documented transient ischemic attack, stroke, or coronary artery disease) and OSA (4% oxygen desaturation index (ODI) > 12) who were randomized into the CPAP arm of the SAVE trial prior to July 1, 2010. Methods: Age, sex, country of residence, type of cardiovascular disease, baseline ODI, severity of sleepiness, and Hospital Anxiety and Depression Scale (HADS) scores plus CPAP side effects and adherence at 1 mo were entered in univariate analyses in an attempt to identify factors predictive of CPAP adherence at 12 mo. Variables with P < 0.2 were then included in a multivariate analysis using a linear mixed model with sites as a random effect and 12-mo CPAP use as the dependent outcome variable. Measurements and Results: CPAP adherence at 1, 6, and 12 mo was (mean ± standard deviation) 4.4 ± 2.0, 3.8 ± 2.3, and 3.3 ± 2.4 h/night, respectively. CPAP use at 1 mo (effect estimate ± standard error, 0.65 ± 0.07 per h increase, P < 0.001) and side effects at 1 mo (-0.24 ± 0.092 per additional side effect, P = 0.009) were the only independent predictors of 12- mo CPAP adherence. Conclusion: Continuous positive airway pressure use in patients with coexisting cardiovascular disease and moderate to severe obstructive sleep apnea decreases significantly over 12 months. This decline can be predicted by early patient experiences with continuous positive airway pressure (i.e., adherence and side effects at 1 month), raising the possibility that intensive early interventions could improve long-term continuous positive airway pressure compliance in this patient population.