A hazard of regional perfusion for melanoma is incomplete isolation, resulting in leakage of the cytostatic drug into the systemic circulation. Data were analysed retrospectively on 438 melphalan perfusions performed for melanoma of the extremities during the period 1978-1990; continuous isotopic measurement of systemic leakage was carried out. The cumulative systemic leakage after 60 min perfusion was 0.9 per cent (95 per cent confidence interval 0.7-1.1 per cent). Systemic leakage of > or = 1 per cent was detected in 12.6 per cent of perfusions, > or = 5 per cent in 6.2 per cent and > or = 10 per cent in 1.4 per cent. In 2.3 per cent of patients, systemic side-effects in the form of mild transient bone marrow depression occurred. Six variables related to the perfusion technique were assessed by multivariate analysis for their influence on systemic leakage. The level of isolation and diameter of the venous cannula emerged as significant factors. In addition, ligation of the internal iliac vein provided optimal isolation during iliac perfusion.