The extent to which drug adherence may affect survival remains unclear, in part because mortality differences may be attributable to "healthy adherer" behavioral attributes more so than to pharmacological benefits. To explore the relationship between drug adherence and mortality in survivors of acute myocardial infarction (AMI). Population-based, observational, longitudinal study of 31 455 elderly AMI survivors between 1999 and 2003 in Ontario. All patients filled a prescription for statins, beta-blockers, or calcium channel blockers, with the latter drug considered a control given the absence of clinical trial-proven survival benefits. Patient adherence was subdivided a priori into 3 categories--high (proportion of days covered, > or =80%), intermediate (proportion of days covered, 40%-79%), and low (proportion of days covered, <40%)--and compared with long-term mortality (median of 2.4 years of follow-up) using multivariable survival models (and propensity analyses) adjusted for sociodemographic factors, illness severity, comorbidities, and concomitant use of evidence-based therapies. Among statin users, compared with their high-adherence counterparts, the risk of mortality was greatest for low adherers (deaths in 261/1071 (24%) vs 2310/14,345 (16%); adjusted hazard ratio, 1.25; 95% confidence interval, 1.09-1.42; P = .001) and was intermediary for intermediate adherers (deaths in 472/2407 (20%); adjusted hazard ratio, 1.12; 95% confidence interval, 1.01-1.25; P = .03). A similar but less pronounced dose-response-type adherence-mortality association was observed for beta-blockers. Mortality was not associated with adherence to calcium channel blockers. Moreover, sensitivity analyses demonstrated no relationships between drug adherence and cancer-related admissions, outcomes for which biological plausibility do not exist. The long-term survival advantages associated with improved drug adherence after AMI appear to be class-specific, suggesting that adherence outcome benefits are mediated by drug effects and do not merely reflect an epiphenomenon of "healthy adherer" behavioral attributes.