BACKGROUND: Cardiovascular diseases are the main cause of death worldwide, making their prevention a major health care challenge. In 2006, a German statutory health insurance company presented a novel individualised prevention programme (KardioPro), which focused on coronary heart disease (CHD) screening, risk factor assessment, early detection and secondary prevention. This study evaluates KardioPro in CHD risk subgroups, and analyses the cost-effectiveness of different individualised prevention strategies. METHODS: The CHD risk subgroups were assembled based on routine data from the statutory health insurance company, making use of a quasi-beta regression model for risk prediction. The control group was selected via propensity score matching based on logistic regression and an approximate nearest neighbour approach. The main outcome was cost-effectiveness. Effectiveness was measured as event-free time, and events were defined as myocardial infarction, stroke and death. Incremental cost-effectiveness ratios comparing participants with non-participants were calculated for each subgroup. To assess the uncertainty of results, a bootstrapping approach was applied. RESULTS: The cost-effectiveness of KardioPro in the group at high risk of CHD was [euro sign]20,901 per event-free year; in the medium-risk group, [euro sign]52,323 per event-free year; in the low-risk group, [euro sign]186,074 per event-free year; and in the group with known CHD, [euro sign]26,456 per event-free year. KardioPro was associated with a significant health gain but also a significant cost increase. However, statistical significance could not be shown for all subgroups. CONCLUSION: The cost-effectiveness of KardioPro differs substantially according to the group being targeted. Depending on the willingness-to-pay, it may be reasonable to only offer KardioPro to patients at high risk of further cardiovascular events. This high-risk group could be identified from routine statutory health insurance data. However, the long-term consequences of KardioPro still need to be evaluated.