Midazolam, a commonly used sedative and amnestic medication, has recently been shown to be largely metabolized in the liver by a cytochrome P450, termed CYP3A4. There is at least a tenfold intersubject variability in the liver content and catalytic activity of CYP3A4, which may in part account for the known interpatient differences in the kinetics of midazolam. To test this hypothesis, we determined the intravenous midazolam kinetics of 20 medically stable, hospitalized patients, whose hepatic CYP3A4 activities were determined with use of the [14C-N-methyl]erythromycin breath test. During the kinetic study, we also performed psychometric testing designed to quantitate the level of sedation and amnesia. We found a significant positive correlation between the erythromycin breath test results and weight adjusted clearance (in milliliters per minute per kilogram) of both total midazolam (r = 0.52; p = 0.03) and unbound midazolam (r = 0.61; p < 0.01). The relatively low dose of midazolam used (0.0145 mg/kg) produced significant but transient sedation and memory impairment in some of the patients. We conclude that interpatient differences in liver CYP3A4 activity in part account for the variations in midazolam kinetics. Our observations account for reported drug interactions involving midazolam and suggest that patients with low CYP3A4 activity may be most susceptible to prolonged amnestic effects occasionally produced by this short-acting benzodiazepine.