TUMOUR NECROSIS FACTOR (TNFa) was discovered as a serum factor in mice treated with BCG and endotoxin, producing haemorrhagic and coagulative necrosis of tumours in recipient mice [I]. In fact, TNFa is the first cytokine which is able to produce very fast and effective necrosis of tumours more efficiently than chemotherapy itself. Therefore, efforts were made to clone the gene. In 1985, the human TNFu gene [2] was cloned and expressed in E. coli, followed in the same year, by the murine TNFa gene [3, 41. It is commonly accepted that the human TNFo structure is a non-glycosylated trimer of 157 amino acids with several cysteine bounds [S, 61. The trimer has three receptor binding sites apparently situated between each part of the trimer. There are two membrane receptors for TNFo of different molecular weights currently named~ 55 and~ 75 [7]. The recombinant TNFn (rTNFn) was made available for clinical trials, but unfortunately it was found, at that time, to cause fatal septic shock in humans (reviewed in ref [S]). It is not surprising, therefore, that phase I and II studies in humans were hampered by high levels of toxicity and seldom showed antitumour effects [9-151.

In 1988, we considered using isolated limb perfusion for administering efficient high dose TNFo combined to interferony and melphalan. The results in melanoma-in-transit-metastases and soft tissue sarcoma of the limbs have been astonishingly high in terms of complete response rate [161. During the past 5 years, we have treated, with our Dutch colleagues, approximately 200 cases. Side-effects were always acceptable and have been drastically reduced recently by improving the technique. A comprehensive study of the cases, including immunohistology, pathophysiology, immunology and biochemistry, allows us to propose that isolation perfusion of the limbs with high dose TNFcl is a model for biochemotherapy of cancer.