Summary In this study we determined the level of tumour necrosis factor alpha (TNF-a) in liver and tumour tissue samples obtained from patients with colorectal metastases confined to the liver, who were treated with isolated liver perfusion with TNF-a and melphalan. We adapted a standard enzyme-linked immunosorbent assay kit for the quantification of TNF-a in serum to measure the amount of this cytokine in solid tissue. For this purpose, we developed a buffer that lysed the tissues without affecting the TNF-a present. The minimum detection level was about 2 pg of TNF-a per mg tissue. Using this technique, we found a significant increase in the TNF-a level after perfusion in the liver tissue of all evaluable patients, which may explain the transient liver toxicity we observed in all patients. In tumour tissue, a significant TNF-a increase was observed in one out of five patients. The level of TNF-a in all liver tissue samples and some of the tumours after treatment by isolated liver perfusion was much higher than the peak serum concentrations obtained after systemic administration of the maximum tolerated dose of TNF-a. Furthermore, we demonstrated that the level of TNF-a in the liver tissue samples was about seven to eight times higher than in tumour tissue. We concluded that regional liver treatment resulted in a relatively high local level of TNF-a, but also that this cytokine did not preferentially accumulate in tumour tissue.