Caffeine is used to phenotype subjects in vivo for the cytochrome P450 isoforms CYP1A2 and CYP2E1, and for N-acetyltransferase type 2 (NAT2). However, how much of the variation in phenotyping parameters may be attributed to variations in CYP1A2 and CYP2E1 activities has not been determined. Therefore, this study intraindividually compared enzyme activities and/or content in liver samples with pharmacokinetic parameters of caffeine in vivo after administration of a test dose in 25 patients undergoing hepatectomy. Parameters measured in vitro were the high affinity components of caffeine 3-demethylation and phenacetin 0-deethylation, microsomal CYP1A2 and CYP2E1 immunoreactivity, and cytosolic sulfamethazine N-acetylation. Caffeine parameters in vivo included caffeine clearance from plasma and/or saliva, paraxanthine/caffeine ratios in plasma and saliva, plasma theophylline/caffeine ratio, and several metabolite ratios from spot urine sampled 6 h postdose. Correlations between parameters were determined using weighted linear regression analyses. Caffeine clearance and paraxanthine/caffeine ratios correlated most highly to intrinsic clearance of caffeine 3-demethylation and to CYP1A2 immunoreactivity (r= 0.584-0.82), whereas urinary CYP1A2 ratios correlated less strongly with CYP1A2 parameters in vitro. Assignment of acetylator phenotype by urinary NAT2 ratios was concordant with sulfamethazine-N-acetylation in vitro. In contrast to CYP1A2 parameters in vitro, CYP2E1 immunoreactivity was not related to the theophylline/caffeine plasma ratio. CYP1A2 activity, thus, is the major determinant of caffeine clearance and the paraxanthine/caffeine ratios in vivo, of which the saliva ratio 6 h postdose appears as the most advantageous parameter. The results confirm that phenotyping using caffeine provides valid estimates of CYP1A2 and NAT2 activity.