Background: When the total number of events is small, meta-analyses with apparently positive results may arrive at false positive conclusions. This may be particularly so if one or more trials with few events are stopped early for apparent large treatment effects. We have noted dangers in the use of composite endpoints and suggested that metaanalyses should present results separately for each patient-important endpoint. One highly publicized, randomized trial of beta-blockers in high risk patients undergoing vascular surgery stopped early because of an apparent large decrease in the combined endpoint of cardiovascular death and myocardial infarction (total events 20). Prestigious guideline panels subsequently recommended beta blockers for patients undergoing vascular surgery at higher-than-average risk for cardiovascular events. Objectives: Unconvinced by these findings, we undertook a randomized trial that enrolled over 8000 patients to establish the role of beta blockers in patients undergoing non-cardiac surgery. Results: We found a reduction in the pre-specified composite primary outcome of cardiovascular death, myocardial infarction and non-fatal cardiac arrest (443 events, HR 0.83, 95% CI 0.70 to 0.99, p = 0.04). That result was driven by myocardial infarction (366 events, HR 0.70, 95% CI 0.56 to 0.86, p = 0.0007). Cardiovascular deaths were increased in patients receiving beta blockers (133 events, hazard ratio 1.30, 95% CI 0.92 to 1.83, p = 0.14); total mortality was higher in treated patients (226 events, hazard ratio 1.33, 95% CI 1.02 to 1.74, p = 0.03). Total stroke rate was also higher in patients receiving beta blockers (60 events, relative risk 2.17, 95% CI 1.26 to 3.73, p = 0.005). A meta-analysis of all trials demonstrates reduction in myocardial infarctions (433 events, RR 0.69, 99% CI 0.54 to 0.87, p < 0.0001), and provides additional evidence that beta blockers increase stroke (75 events, RR 2.19, 99% CI 1.06 to 4.50, p = 0.005).