An isolated limb perfusion (ILP) model using soft tissue sarcoma-bearing rats was used to study prerequisites for an effective ILP, such as oxygenation of the perfusate, temperature of the limb, duration of the perfusion and concentration of tumour necrosis factor (TNF). Combination of 50 microg TNF and 40 microg melphalan demonstrated synergistic activity leading to a partial and complete response rate of 71%. In comparison to oxygenated ILP, hypoxia was shown to enhance anti-tumour activity of melphalan alone and TNF alone but not of their combined use. Shorter perfusion times decreased anti-tumour responses. At a temperature of 24-26 degrees C, anti-tumour effects were lost, whereas temperatures of 38-39 degrees C or 42-43 degrees C resulted in higher response rates. However, at 42-43 degrees C, local toxicity impaired limb function dramatically. Synergy between TNF and melphalan was lost at a dose of TNF below 10 microg in 5 ml perfusate. We conclude that the combination of TNF and melphalan has strong synergistic anti-tumour effects in our model, just as in the clinical setting. Hypoxia enhanced activity of melphalan and TNF alone but not the efficacy of their combined use. For an optimal ILP, minimal perfusion time of 30 min and minimal temperature of 38 degrees C was mandatory. Moreover, the dose of TNF could be lowered to 10 microg per 5 ml perfusate, which might allow the use of TNF in less leakage-free or less inert perfusion settings.