BACKGROUND: Studies of survival after myocardial infarction (MI) are often based on intention to treat analyses of controlled trials. OBJECTIVES: Describe long-term survival after MI in France. METHODS: Six-year cohort study of patients recruited within 3 months after MI. Primary outcome was all-cause death. Vital status was verified in the national death registry. Analysis used Cox models with time-dependent variables and propensity scores. RESULTS: Five thousand five hundred and twenty-seven (5527) subjects were included, 62.1+/-13 years old, 77.6% male, 9.6% smokers, 16.7% diabetic, 13.3% with previous MI. Up to 99% of patients were initially prescribed secondary prevention drugs (aspirin and/or other antiplatelet agents, beta-blockers, statins or other lipid-lowering agents, angiotensin converting enzyme inhibitors or angiotensin receptor blockers); 73% had all four classes. Overall 6-year mortality was 13.1% [95% confidence interval 12.3 to 14.0%], 2.34 per hundred patient-years (% PY); 49% returned all or all but one of the possible questionnaires (compliant [C]), 50.8% did not (non-compliant [NC]). The main predictors for death were non-compliance with study protocol (death rates NC 2.98% PY, C 1.69%PY, hazard ratio (HR) 3.13 [2.63-3.57]); increasing age at inclusion (HR up to 15.7 [10.7-23.2] for age >/=80); diabetes (1.39 [1.17-1.65]); smoking at inclusion (1.76 [1.27-2.44]), previous MI (1.46 [1.22-1.75]). Beta-blockers (0.79 [0.64-0.96]), statins (0.68 [0.51-0.90]), and enrolment in physical rehabilitation programs (0.74 [0.62-0.89]) were associated with a lower death rate. CONCLUSION: Association of mortality with non-compliance to study protocol probably indicates general non-compliance with prevention. Analyses of treatment effects were hindered by paucity of events and of unexposed patients.