The binding of drugs to plasma proteins is important to consider when concentrations in whole blood (eg, in forensic toxicology) are compared with therapeutic and toxic concentrations based on the analysis of plasma or serum. The plasma to whole blood distribution of diazepam (D) and its major metabolite nordiazepam (ND) was investigated under in vitro and ex vivo conditions. Studies in vitro were done by spiking whole blood with D and ND to give concentrations ranging from 0.1 to 1.0 microg/g. Venous blood was also obtained from hospital blood donors (n = 66) after informed consent. The hematocrit, hemoglobin, and water content of blood specimens were determined by routine procedures before D and ND were added to produce target concentrations of approximately 0.5 microg/g for each substance. The ex vivo work was done with blood specimens from hospital outpatients who were being medicated with D. Concentrations of D and ND were determined in body fluids by capillary column gas chromatography after adding prazepam as internal standard and solvent extraction with butyl acetate. The method limit of quantitation was 0.03 microg/g for both D and ND. The concentrations of D and ND were highest in plasma and lowest in erythrocytes. The plasma/blood (P/B) distribution ratios did not depend on drug concentration between 0.1 and 1.0 microg/g. The mean P/B ratios were 1.79:1 for D and 1.69:1 for ND when hematocrit was 45%. Furthermore, the P/B ratio for D (y) was positively correlated with blood hematocrit (x) and the regression equation was y = 0.636 + 0.025x (r = 0.86, P < 0.001). A similar strong association was found between the P/B ratio and hematocrit for ND (r = 0.79). P/B ratios of D and ND, blood hematocrit, hemoglobin, and the water content differed between sexes (P < 0.001). The overall mean P/B ratios for D and ND were 1.69 +/- 0.097 (+/- SD) and 1.62 +/- 0.08 (P < 0.001, n = 66) respectively when the mean hematocrit was 42.9 +/- 3.4 (+/- SD). For forensic purposes, it would be better to forgo making any conversion of a drug concentration measured in whole blood to that expected in plasma or serum; instead, therapeutic and toxic concentrations should be established for the actual specimens received.