The cardiovascular actions of the magnesium ion at pharmacological concentrations include coronary and systemic vasodilatation, platelet inhibition, and antiarrhythmic effects. Magnesium has also been reported to protect myocardial tissue in experimental models of ischaemia and reperfusion. Several small clinical trials in suspected acute myocardial infarction have suggested that early mortality can be reduced by intravenous infusion of magnesium salts in the acute phase, but none has been of sufficient size to be conclusive. We therefore conducted a randomised, double blind, placebo controlled study in 2316 patients with suspected acute myocardial infarction who received either intravenous magnesium sulphate (8 mmol over 5 min followed by 65 mmol over 24 h) or physiological saline. The primary outcome measure was 28-day mortality, which was ascertained in 99.3% of patients. The groups were well balanced for prognostic factors. By intention-to-treat analysis mortality from all causes was 7.8% in the magnesium group and 10.3% in the placebo group (2p = 0.04), a relative reduction of 24% (95% confidence interval 1-43%). Within the coronary care unit the incidence of left ventricular failure was reduced by 25% (7-39%) in the magnesium group (2p = 0.009). There was no significant difference between the groups in the incidence of heart block or the use of antiarrhythmic drugs, direct-current cardioversion, or temporary pacing. Myocardial infarction was confirmed in 65% of each group, with closely similar rises in cardiac enzymes. The side-effects of magnesium treatment were transient flushing, related to speed of injection of the loading dose, and an increased incidence of sinus bradycardia (2p = 0.02). Exploratory subgroup analyses of 28-day mortality did not indicate any effect modification by thrombolysis or aspirin, or by previous treatment with beta blockers, calcium antagonists, or diuretics. Intravenous magnesium sulphate is a simple, safe, and widely applicable treatment. Its efficacy in reducing early mortality of myocardial infarction is comparable to, but independent of, that of thrombolytic or antiplatelet therapy.