Objectives The association between obesity and prognosis in patients with coronary artery disease (CAD) remains uncertain. We undertook a meta-analysis for the effects of body mass index (BMI) on mortality and cardiovascular events in these patients. Methods We identified studies that provided risk estimates for mortality or cardiovascular events on the basis of BMI in patients with CAD. Summary estimates of relative risks were obtained for five BMI groups: underweight, normal-weight, overweight, obese and grade II/III obese. Mortality was analysed separately as short-term (<6 months) and long-term (>= 6 months). Results Data from 89 studies with 1 300 794 patients were included. Mean follow-up of long-term estimates was 3.2 years. Using normal-weight as the reference, underweight was associated with higher risk of short-term mortality (2.24 (1.85 to 2.72)) and long-term mortality (1.70 (1.56 to 1.86)), overweight and obesity were both associated with lower risk of short-term mortality (0.69 (0.64 to 0.75); 0.68 (0.61 to 0.75)) and long-term mortality (0.78 (0.74 to 0.82); 0.79 (0.73 to 0.85)), but the long-term benefit of obesity disappeared after 5 years of follow-up (0.99 (0.91 to 1.08)). Grade II/III obesity was associated with lower risk of mortality in the short term (0.76 (0.62 to 0.91)) but higher risk after 5 years of follow-up (1.25 (1.14 to 1.38)). The similar J-shaped pattern was also seen for cardiovascular mortality and across different treatment strategies. Meta-regression found an attenuation of the inverse association between BMI and risk of mortality over longer follow-up. Conclusions Our data support a J-shaped relationship between mortality and BMI in patients with CAD. The limitation of current literature warrants better design of future studies.