AIMS: The development of risk associated with diabetes mellitus during long-term follow-up after a myocardial infarction has not been studied in detail. We have studied time-related changes of risk of death during 10 years of follow-up in a cohort of patients not treated with thrombolytic therapy (the Glostrup cohort) and during 6 years in a cohort receiving such treatment in 40% of cases (the TRACE cohort). METHODS: A subgroup analysis of two cohorts: the Glostrup cohort, which consisted of consecutive cases of acute myocardial infarction who were admitted to one hospital between 1979 and 1983; the TRACE cohort which was comprised of patients with an acute myocardial infarction screened for entry into the Trandolapril Cardiac Evaluation study between May 1990 and June 1992. The Glostrup cohort consisted of 1954 patients and follow-up was for 10 years, The TRACE cohort consisted of 6676 patients and follow-up was for 6 years. Outcome measure was total death. RESULTS: A diagnosis of diabetes mellitus was present in 12% of the two study populations. In multivariate analysis, diabetes mellitus had an independent adverse effect on mortality which increased with time. In the Glostrup cohort risk ratio between day 0 and day 30 was 1.17 and increased to 2.51 (P=0.0002) 7-9 years after discharge from hospital. A similar increase in the risk ratio of diabetes mellitus on mortality was observed in the TRACE cohort (risk ratio for days 0-30 was 1.03, and for years 4-6 was 1.74 (P=0.0001). CONCLUSION: Diabetes mellitus has no independent influence on mortality immediately following an acute myocardial infarction, but has an important influence on long-term mortality which increases with time. The implication is that the effect of intervention against diabetes in patients with acute myocardial infarction and diabetes mellitus must be evaluated over a long course of time.