The most promising developments in the field of isolated limb perfusion have centred around the use of the recombinant cytokine tumour necrosis factor-alpha (rTNF-alpha) in combination with melphalan. While the results of clinical trials are impressive, the exact antitumour mechanisms of rTNF-alpha and its role in combination with melphalan remain unclear. Our aim was to study the antitumour activity of human rTNF-alpha with or without the combination of melphalan in a nude mouse human melanoma xenograft system. In a first attempt to define the maximal tolerated single dose of rTNF-alpha in this setting, 15 animals were exposed to increasing doses of rTNF-alpha (60-2500 microg/kg intraperitoneally). All but one animal survived and tumour growth was not influenced by these single dose applications of rTNF-alpha even at the very high doses. Anti-tumour activity of repeated application of melphalan (three times 9 mg/kg in group 2 and three times 6 mg/kg in group 3), of rTNF-alpha alone (nine doses of 50 microg/kg in group 4), and of rTNF-alpha in combination with melphalan (nine doses of 50 microg/kg rTNF-alpha and three times 6 mg/kg melphalan in group 5) was further compared with non-treated animals (group 1). Tumour growth was significantly inhibited in all animals treated with melphalan (group 2, 3 and 5), but was not decreased in animals treated with rTNF-alpha alone (group 4). Mean final tumour volumes and mean tumour weight were not different in group 2 (789 +/- 836 mm3, 0.38 +/- 0.20 g), group 3 (1173 +/- 591 mm3, 0.55 +/- 0.29 g) and group 5 (230 +/- 632 mm3, 0.37 +/- 0.29 g), but significant lower than group 1 (3156 +/- 1512 mm3, 2.35 +/- 0.90 g) and group 4 (3228 +/- 1990 mm3, 2.00 +/- 1.16 g). There were no significant differences between high and low dose melphalan treatment and between melphalan treatment in combination with rTNF-alpha. Histological examination did not show differences between treated and non-treated animals besides slightly inhibited mitotic activities of tumour cells in melphalan-treated animals. While tumour growth of human xenotransplanted melanoma in nude mice could be inhibited by melphalan, we failed to demonstrate any antitumour effect of rTNF-alpha. The combination of melphalan and rTNF-alpha did not enhance the antiproliferative effect of melphalan alone. Human xenotransplanted tumours on nude mice might not be the ideal experimental setting for studies of potential direct antineoplastic activity of rTNF-alpha, and these results support the concept that TNF-alpha exerts its antitumour activity indirectly, possibly by impairing the tumour vasculature and by activating the immune system.