The measurement of plasma volume (Vp) in humans and animals is frequently performed by the Evans blue dye dilution method. However, after injection of Evans blue into the circulation, no steady state is observed because of delayed mixing and progressive leakage of dye out of vascular space. Various methods of calculation have been proposed, either with a single blood sampling 5-10 min after dye injection (Single point method), or with extrapolation at time zero of a logarithmic decay (Log linear method). We propose a method based on a two-compartment hypothesis taking into account the initial mixing and the leakage phase in the time course of dye concentration. Nineteen Sprague-Dawley rats were studied in various conditions and blood sampling was performed before and 2, 4 and 6 min after injection of 200 μg Evans blue. A mathematical model was designed to describe the two-compartment hypothesis and allowed the calculation of Vp and Kout (rate of disappearance of dye from vascular space). A Bland and Altman representation evidenced an overestimation of Vp with previous methods and the great dispersion of results with the single point method, especially when using the 6 min point. Calculation of Kout revealed more accurate with the model than the Log linear method, especially when the mixing rate is slow. We suggest using the two-compartment model to measure Vp with Evans blue technique in rats. This method also allows precise evaluation of the rate of dye leakage, which could be a good marker of vascular permeability to albumin.