Background: The volume of distribution at steady state for ethanol (V ss ) is thought to be identical to the total body water (TBW). We compared a two-compartment pharmacokinetic model with parallel Michaelis-Menten and first-order renal elimination with the classical one-compartment zero-order elimination model. Ethanol concentration-time profiles were established for breath, venous blood, and urine. The values of V ss obtained for ethanol were compared with TBW determined by deuterium oxide dilution.Methods: Sixteen healthy volunteers each received a 30-min intravenous infusion of ethanol on two occasions. Ethanol was measured in breath by a quantitative infrared analyzer and in blood and urine by headspace gas chromatography. Deuterium oxide was given as an intravenous injection and measured in serum by isotope-ratio mass spectrometry. Components of variation were calculated by ANOVA to determine the precision of the estimates of V ss and TBW.Results: Mean TBW, determined by deuterium oxide dilution, was 44.1 Ϯ 3.9 liters (ϮSD) for men, corresponding to 0.61 liters/kg, and 37.4 Ϯ 3.2 liters for women, or 0.54 liters/kg. Estimates of V ss from blood-ethanol pharmacokinetics were 87.6% of TBW according to isotope dilution and 84.4% for breath analysis with the two-compartment model. This compares with 95.1% and 95.4% for blood and breath alcohol, respectively, when the classical zero-order kinetic analysis is used. The precision of the estimates of V ss and TBW was between Ϯ1.56 and Ϯ2.19 liters (95% confidence interval).Conclusions: Ethanol does not distribute uniformly into the TBW. The precision of measuring V ss by ethanol dilution was comparable to estimates of TBW by isotope dilution. Results of noninvasive breath ethanol analysis compared well with use of venous blood for estimating V ss . The sophisticated twocompartment model was much superior to the classical one-compartment model in explaining the total concentration-time course of intravenously given ethanol.