The properties of caffeine render it an ideal substrate for a quantitative test of liver function. The aim of this study was to determine whether the caffeine breath test (CBT) using orally administered 13C-caffeine correlates reliably with plasma caffeine clearance and reflects varying degrees of liver dysfunction. The CBT was performed in 25 healthy controls; 20 subjects with noncirrhotic, chronic hepatitis B or C; and 20 subjects with cirrhosis. Plasma caffeine clearance was assayed simultaneously with the CBT in a cohort of these subjects. Over a broad range of caffeine clearances, the CBT exhibited a highly significant correlation with plasma clearance (r = 0.85, P <.001). Cirrhotic patients were characterized by significantly reduced CBT values (1.15 +/- 0.75 delta per thousand mg(-1)) compared with controls (2.23 +/- 0.76; P =.001) and hepatitic patients (1.83 +/- 1.05; P =.04). There was a significant inverse relationship between the CBT and Child-Pugh score (r = -.74, P =.002). The intraclass correlation coefficient between repeated CBTs in 20 subjects with normal and cirrhotic livers was 0.89. Although smoking was associated with an 86% to 141% increase in CBT in all groups, the CBT was able to distinguish control, hepatitic, and cirrhotic smokers (5.36 +/- 0.82, 3.63 +/- 1.21, and 2.14 +/- 1.14, respectively, P =.001). Multivariate analysis revealed that only smoking (P <.001) and disease state (P =.001) were significant predictors of the CBT. In conclusion, the 13C-CBT represents a valid indicator of plasma caffeine clearance and correlates reproducibly with hepatic dysfunction.