Between October 1969 and December 1986, 136 patients with recurrent limb melanoma were treated with hyperthermic antiblastic perfusion (HAP). This retrospective analysis is aimed at identifying tumor-related and treatment-related variables likely to influence tumor response, locoregional control, disease-free survival, and overall survival. Independent factors predicting a complete response (CR) were the number of lesions (P less than 0.0001) and the minimum tumor temperature (minT) (P = 0.03). Only a positive trend was observed for the drug dose (P = 0.08). However, the proportion of CR was significantly higher (57.7%; P = 0.02) in patients who had a minT of 41.5 degrees C or greater and who were given a dose equal to or greater than the standard dose than in patients treated with lower temperatures and/or lower drug doses. The occurrence of a CR significantly increased the rates of locoregional control (77%; P = 0.007), disease-free survival (55.6%; P = 0.006), and overall survival (68.6%; P = 0.03). Treatment optimization may provide further therapeutic improvements by increasing the incidence of CR. However, the overall survival rates also were influenced by the number of lesions (P = 0.0014), sex (P = 0.04), and the number of previous relapses (P = 0.01). Therefore, tumor aggressiveness also is crucial in determining the outcome of the disease, and only early treatment with HAP can reduce the risk of distant metastases.