The accuracy of different blood sampling methods used to characterize rapidly changing blood drug concentrations was examined both in vitro and in vivo. It was shown in vitro that blood sampling methods based on the fraction collection principle failed to characterize a "square wave" change in drug concentration, and there was a 9-16-s delay before achieving 95% of the expected drug concentration. Varying the catheter size and length did not improve the response. This observation is consistent with laminar and/or turbulent flow producing dispersion and mixing of blood of different drug concentrations in the catheter. A sampling method (flush and withdrawal) was developed to minimize these effects. In vivo studies showed that peak blood drug concentrations obtained using this method after an iv bolus of a drug were approximately 25-28% higher than those simultaneously obtained by methods based on fraction collection principles. It is concluded that blood sampling methods based on fraction collection principles can produce significant errors in measured blood drug concentrations. The error is greater the greater the rate of change of the blood drug concentrations.