Caffeine clearance was determined in 13 healthy control subjects and in 13 patients with histologically proven cirrhosis. On separate occasions, 70 mg, 200 mg, and 300 mg single doses of anhydrous caffeine were administered orally with decaffeinated coffee to each subject. Subjects were analyzed individually, acting as their own controls, thus reducing interindividual variability. The present study showed that caffeine exhibited dose-dependent pharmacokinetics, particularly in subjects who showed high initial clearance with the low dose (70 mg) of caffeine. There was a significant decrease in caffeine clearance with increasing dose from 70 mg to 300 mg (n = 26, p less than 0.01, Dunnett's test), indicating saturable caffeine metabolism in the dose range tested. These findings imply that if caffeine is to be used as a guide to deteriorating liver function, serial caffeine clearance estimations should be performed in each individual subject, with use of the same dose of caffeine each time.