Hypomagnesemia has been hypothesized to play a role in coronary heart disease (CHD), but few prospective epidemiologic studies have been conducted. We examined the relation of serum and dietary magnesium with CHD incidence in a sample of middle-aged adults (n=13,922 free of baseline CHD) from 4 US communities. Over 4 to 7 years of follow-up, 223 men and 96 women had CHD develop. After adjustment for sociodemographic characteristics, waist/hip ratio, smoking, alcohol consumption, sports participation, use of diuretics, fibrinogen, total and high-density lipoprotein cholesterol levels, triglyceride levels, and hormone replacement therapy, the relative risk of CHD across quartiles of serum magnesium was 1.00, 0.92, 0.48, and 0.44 (P for trend=0.009) among women and 1.00, 1.32, 0.95, and 0.73 (P for trend=0.07) among men. The adjusted relative risk of CHD for the highest versus the lowest quartile of dietary magnesium was 0.69 in men (95% confidence interval 0.45 to 1.05) and 1.32 in women (0.68 to 2.55). These findings suggest that low magnesium concentration may contribute to the pathogenesis of coronary atherosclerosis or acute thrombosis.