The optimal conditions for performing the caffeine CO2 breath test (CBT) were investigated in smokers and nonsmokers. Caffeine labeled with 13C or 14C in all three (1, 3, and 7) methyl groups or specifically in the 1-, 3-, or 7-methyl groups were orally administered to healthy adults and the expiration of labeled CO2 was measured for 8 or 24 hr. The absolute rate of labeled CO2 excretion from trilabeled caffeine was proportional to the dose up to 3 mg/kg in all subjects. In smokers, the rate of labeled CO2 excretion averaged twice that in nonsmokers at all doses. A correlation was observed between the 2-hr cumulative CO2 excretion from trilabeled caffeine and the apparent oral metabolic clearance rate (MCR) of caffeine (R = 0.90). Monolabeled CBTs in smokers and nonsmokers demonstrated that 80% +/- 4% of labeled CO2 expired in the breath during the first 2 hr of a trilabeled CBT was derived from the 3 position; at 6 to 8 hr equal amounts were derived from the 3 and 7 positions. Little N-demethylation was observed from the 1 position at any time during the 8-hr test. The results indicate that the 2-hr cumulative excretion of labeled CO2 could be used to accurately predict the metabolic clearance rate of caffeine is the best CBT parameter for detecting the effect of smoking on caffeine N-demethylation. The data suggest that the primary routes of caffeine metabolism are 3-N-demethylation and ring hydroxylation and confirm that caffeine metabolites are N-demethylated primarily in the 3 and 7 positions.