Some recent epidemiologic studies have reported a nonlinear dose-response in the relationship between coffee consumption and health risks, such that the risks increase disproportionately to the increase in dose. Assuming caffeine contributes to the adverse health effects of coffee, a possible explanation for the nonlinear dose-response relationship is dose-dependent metabolism of caffeine. We examined the hypothesis that under chronic dosing conditions the metabolism of caffeine is dose-dependent. Nine healthy subjects were given, in randomized 5-day treatment blocks, placebo, 4.2 (low) and 12 (high) mg/kg/day caffeine in decaffeinated coffee, in six divided doses spaced throughout the day. On the third day of each dosing period, 25 mg of stable-isotope labeled caffeine (2-13C, 1,3-15N2) was given intravenously. Clearance of labeled caffeine fell from 0.118 (placebo treatment) to 0.069 (low dose; p less than 0.005) and to 0.54 (high dose; p less than 0.001) L/hr/kg. The formation and metabolite clearances of paraxanthine, the major primary metabolite of caffeine, also decreased comparing the low and high doses (p less than 0.05). We conclude that caffeine metabolism is dose-dependent, resulting in nonlinear accumulation of methylxanthines in the body. Dose-dependent metabolism of caffeine may explain in part why people who drink large amounts of coffee are at greater risk for cardiovascular disease.