The incidence of long-term (> or = 3 months) neuropathy in 350 melanoma patients treated with single normothermic or 'mild' hyperthermic perfusion with melphalan in the period 1978 to 1990 was studied. Long-term neuropathy was encountered in 14 patients; in 10/51 patients (20%) after perfusion at the axillary level and in 4/247 patients (2%) after perfusion at the iliac level. After brachial and femoro-popliteal perfusions no long-term neuropathy was observed. Neuropathy, mainly consisting of paresis/paralysis of the hand and/or fingers, anaesthesia, and/or paraesthesiae, improved over a mean period of 16 (3-43) months in eight patients, but three patients still had serious neuropathy one year after perfusion. In another six patients little improvement was seen and four died with permanent neuropathy. Acute regional toxicity after perfusion and the application of 'mild' hyperthermia did not seem to influence the incidence of long-term neuropathy. This complication is probably a result of the isolating Esmarch rubber bandage being applied too tightly during perfusion at a proximal level. At the axillary level, where the brachial plexus lacks the protection from enveloping tissues, nerve damage is especially prone to occur. We recommend applying this bandage no tighter than is necessary to maintain the isolation of the circuit. This implies meticulous surgical isolation of the vascular system and accurate monitoring of leakage.