The aim of this study was to define prognostic factors for survival, and especially for long-term survival in a mature data-set of patients with stage IV melanoma treated within a randomised trial of cytokine-based protocols. Long-term follow-up data on patients enrolled into a European Organization for Research and Treatment of Cancer (EORTC) trial comparing interferon-alpha (IFNalpha) plus interleukin-2 (IL-2) with or without cisplatin were collected. Univariate and multivariate Cox regression analyses were performed to define prognostic factors for survival. The characteristics of patients alive at 2 and 5 years after randomisation were compared with the entire cohort using the chi(2) test. The minimum potential follow-up of the 131 evaluable patients was 5 years. 18 patients (14%) were alive 2 years after randomisation, and 11 (8%) 5 years after randomisation. Pretreatment performance status (PS), serum lactate dehydrogenase (LDH) and tumour mass were significant predictors for survival, whereas site of metastases and number of sites were non-significant. PS and LDH were the only independent prognostic factors. All except 1 patient alive at 2 and 5 years had a pretreatment PS of 100%, and only three long-term survivors had elevated pretreatment LDH. There was no association between the site of metastases and long-term survival. Response to treatment was a major predictor for long-term survival, whereas addition of cisplatin did not impact upon overall survival probability or on long-term survival. The probability of long-term survival in stage IV melanoma patients after IL-2-based treatments is governed by pretreatment PS, serum LDH and response to treatment. Site of metastases, the basis for the M-subcategories of the new AJCC staging system, was not informative in this study.