Therapy with tacrine, a promising new treatment for Alzheimer's disease, must be discontinued in up to 15% of patients because of hepatocellular toxicity. Recent studies using human liver microsomes have suggested that a single liver enzyme, cytochrome P450 1A2 (CYP1A2), catalyzes the major route of metabolism and elimination of tacrine, and also catalyzes the pathway(s) involved in the generation of reactive metabolites capable of covalent protein binding and cytotoxicity. Because CYP1A2 activity has been shown to vary up to 60- fold among patients, we proposed that a convenient measure of CYP1A2 activity, the [13C 3-methyl] caffeine breath test (CBT), might be clinically useful in identifying patients most susceptible to tacrine liver toxicity. To test this hypothesis, we administered the CBT to 37 patients with Alzheimer's disease before they began treatment with tacrine. Twenty patients received 2 mg/kg of [(13)C 3-methyl] caffeine. The remaining 17 patients received the commercially available CBT kit, which employs a constant 200-mg dose. The activities of two other major drug-metabolizing enzymes (cytochrome P450 3A4 and 2D6 [CYP3A4 and CYP2D6]) were also measured in these 17 patients. We found that the results obtained from the CBT protocol did not predict the peak serum alanine transaminase (ALT) observed in the patients. The measured CYP3A4 and CYP2D6 activities also failed to predict the susceptible patients. However, the result of the standardized-dose CBT correlated well with the logarithm of the steady-state plasma tacrine level obtained in 10 patients (R2 = .69, P = .003). We conclude that the CBT will not be clinically useful in determining the subset of patients most susceptible to tacrine hepatotoxicity. However, the correlation we observed between CBT results and tacrine blood levels is the first evidence supporting a critical role for CYP1A2 activity in the disposition of the drug in vivo.