The effects of a variable daily administration of caffeine on fasting levels of caffeine in plasma and in saliva were measured in 24 patients with liver disease and hepatocellular dysfunction of variable degree. For 2 consecutive days the patients received either 250 mg of caffeine (in 2 separate doses of 125 mg each at 8:00 a.m. and at 6:00 p.m.) or a single dose of 125 mg at 6:00 p.m. Caffeine clearance was also measured and the results were correlated with the galactose elimination capacity and with antipyrine clearance. At the beginning of the study, fasting caffeine concentrations were largely variable, without any relation to liver function. A strict negative correlation between fasting caffeine and caffeine clearance was only observed after 2 days of controlled caffeine administration (rs = -0.814). Under these conditions, fasting caffeine also correlated with antipyrine clearance (rs = -0.671, n = 20) and with galactose elimination (rs = -0.565). Our data prove that fasting caffeine concentrations after an evening dose may be used as an index of liver function only in subjects under a strictly controlled dietary caffeine intake. The large availability of caffeine in the diet makes the compound scarcely reliable for a correct measurement of liver function based on a single sample, and correct clearance determination is needed.