Systematic reviews and meta-analyses attempt to minimise the biases from individual trials but can, on some occasions, miss or conceal beneficial treatment effects when trials, with varying patient populations, are analysed together. OBJECTIVE: To investigate the effect of 'splitting' diverse patient populations using part of a published meta-analysis of randomised controlled trials for the prevention of vascular events in high-risk patients. METHODS: Our study re-evaluated a section of results of a well-known meta-analysis of RCTs for the prevention of vascular events. The original review (ATC) estimated the combined treatment effect of dipyridamole plus aspirin (DP+ASA) compared to aspirin (ASA) alone using data from 25 clinical trials for all patients, regardless of their vascular morbidity profile. Patients who had previously experienced an MI, TIA/stroke, CABG, coronary angioplasty, intermittent claudication, peripheral grafts, angioplasty, haemodialysis and diabetes were 'lumped' together for analysis. We re-evaluated the data by meta-analysing sub-groups of trials according to the risk factor of the patient population (where there was more than one trial for the risk factor). RESULTS: The Relative Risk Reductions with [95% CI] for DP+ASA compared to ASA alone for all types of patients was calculated to be 5% [-5 to 15%]. However the RRRs were diverse when trial patient populations were split for analysis. RRRs ranged from 15% [4% to 26%] for patients with previous Stroke/TIA to -19% [-62% to 12%] for patients who had had CABG previous to entering the study. CONCLUSIONS: The usefulness of the ATC's summary statistics in decision-making for specific patient populations is limited, due to heterogeneous patient populations being combined together. In particular, our sub-group analysis revealed a statistically significant RRR in vascular events for patients treated with DP+ASA compared to ASA amongst patients with previous TIA/stroke.