NT-proBNP is an excellent predictor of adverse events in patients with diabetes mellitus. Due to an aging population it is of interest to determine whether NT-proBNP can predict cardiac events with equal precision in subgroups with different ages. 1395 outpatients with diabetes mellitus were recruited for this prospective observational study. NT-proBNP, renal function, lipid status and other demographic variables were measured at baseline. The cohort was divided into three groups: Group I (609 patients under 60 years of age), group II (634 patients ranging from 60-75) and group III (152 patients older than 75). Patients were followed during a mean observation period of 11 months, 75 patients reached the defined endpoint, which was unplanned hospitalization due to a cardiac event. Mean age was 60 ± 30 years, mean HbA(1c) was 7.6% and mean NT-proBNP was 242 ± 437 pg/ml. In a multiple Cox regression model, age (hazard ratio (HR) 11.18, p < 0.01) and the absence of a cardiac disease (HR 0.49, p < 0.01) were important variables for short-term prognosis. The addition of the logarithm of NT-proBNP provided independent prognostic information (HR 1.81 p < 0.01) and significantly increased the explained variance of the model (χ(2 )= 22.93; d.f. = 1; p < 0.01). More importantly, the predictive power of this model was similar in different age-groups. The prognostic information of NT-proBNP was not influenced by age and this biomarker remained a reliable predictor of short-term cardiac events in patients with diabetes mellitus aged 75 years or older.