Background: In a prior randomized phase II trial comparing hyperthermic isolated limb perfusion (HILP) with melphalan and 4 different doses of TNFa (0.5 mg; 1 mg; 2mg, 3 mg/4mg superior/ inferior limb), there was no dose effect detected for the tumour response but systemic toxicity was significantly lower with low doses TNF-a (Bonvalot S et al. 2005). The objective of the present study was to confirm these datas on a larger population of locally advanced or recurrent STS of the extremity using low dose TNFa. Methods: Prospective database including 100 HILP (38-40°) with melphalan (10 mg/L) and TNFa 1 mg. Resection of the remnant tumour was performed 2 months after ILP. The main endpoint was complete response on MRI. Secondary endpoints were histological response, rate of amputation and toxicity. Results: Sex ratio was and median age was 43 years. Location/ median size were 30/48 mm and 70/72 mm for upper and lower limbs respectively. There were 26 multifocal tumors, 52 recurrences of which 18 arising in a pre irradiated field. Grade (FNCLCC) were 1 (23 pts), 2 (34 pts), 3 (43 pts). Fifty one pts had pre ILP systemic chemotherapy. A complete radiological response was obtained in 30% of pts while 48%, 9% and 13% of pts exhibited a partial, no change and disease progression. There was no mortality and no systemic toxicity. Local toxicity were (Wieberdink) grade 2 (16 pts), 3 (5 pts) and 4 (1 pt). A limb salvage surgery was achieved in 88% pts. With a median follow-up of 27 months, 3 years overall survival and local recurrence rate were 89% and 15% respectively. Age, gender, tumor size, recurrence, uni/multifocality, grade, pre op chemotherapy, pre irradiated field, were not predictive of the responses rates. Conclusions: This study confirm that 1mg TNFa is as effective as the standard dose without its systemic toxicity. A wide population of pts may benefit from this approach.