Résumé : Current methods for cytostatic dosimetry in isolation perfusion of the limbs are based on either limb tissue volume (LTV) or body weight. None of them take into account the actual blood volume intra- and extracorporal, including even the blood leakage if any, in which the pharmacokinetics take place. The present study describes a method which allows the assessment of the actual exchangeable blood volume. The latter is calculated by a formula based on three hematocrit measurements. Thirty-one cases entered the study. Exchangeable limb blood volume representing the limb vascular bed was found to average 340 +/- 148 (SD) ml for upper limb perfusion and 768 +/- 279 and 621 +/- 454 ml for iliac and femoropopliteal perfusion, respectively. There was a good correlation between exchangeable limb blood volume and limb tissue volume (LTV, r = 0.7), a poor one with body weight (r = 0.3), and no correlation at all with body surface. Melphalan dosage was calculated per ml of blood and applied at 20 to 40 micrograms/ml. Comparison between calculated dose and concentration measured by high performance liquid chromatography showed a high correlation (r = 0.963). Since there was a correlation between exchangeable limb blood volume and LTV, it was possible to derive a conversion for melphalan dosage where 13 mg/liter corresponds to 20 micrograms/ml in upper limb perfusion and 10 mg/liter corresponds to 40 micrograms/ml in lower limb perfusion. Comparison between calculated melphalan dosage based on our method and the LTV method showed a large dispersion of values in the latter (12 to 18% coefficient of variation) while the dispersion given by the body weight-based method increased 2-fold (16 to 31% coefficient of variation). It is concluded that the present dosimetry method is the most suitable up to the present for accurate prediction of cytostatic concentration in isolation perfusion.