Background: Trials of β blockers in patients undergoing non-cardiac surgery have reported conflicting results. This randomised controlled trial, done in 190 hospitals in 23 countries, was designed to investigate the effects of perioperative β blockers. Methods: We randomly assigned 8351 patients with, or at risk of, atherosclerotic disease who were undergoing non-cardiac surgery to receive extended-release metoprolol succinate (n=4174) or placebo (n=4177), by a computerised randomisation phone service. Study treatment was started 2-4 h before surgery and continued for 30 days. Patients, health-care providers, data collectors, and outcome adjudicators were masked to treatment allocation. The primary endpoint was a composite of cardiovascular death, non-fatal myocardial infarction, and non-fatal cardiac arrest. Analyses were by intention to treat. This trial is registered with ClinicalTrials.gov, number NCT00182039. Findings: All 8351 patients were included in analyses; 8331 (99.8%) patients completed the 30-day follow-up. Fewer patients in the metoprolol group than in the placebo group reached the primary endpoint (94 [5.8%] patients in the metoprolol group vs 290 [6.9%] in the placebo group; hazard ratio 0.84, 95% CI 0.70-0.99; p=0.0399). Fewer patients in the metoprolol group than in the placebo group had a myocardial infarction (176 [4.2%] vs 239 [5.7%] patients; 0.73, 0.60-0.89; p=0.0017). However, there were more deaths in the metoprolol group than in the placebo group (129 [3.1%] vs 97 [2.3%] patients; 1.33, 1.03-1.74; p=0.0317). More patients in the metoprolol group than in the placebo group had a stroke (41 [1.0%] vs 19 [0.5%] patients; 2.17, 1.26-3.74; p=0.0053). Interpretation: Our results highlight the risk in assuming a perioperative β-blocker regimen has benefit without substantial harm, and the importance and need for large randomised trials in the perioperative setting. Patients are unlikely to accept the risks associated with perioperative extended-release metoprolol.