This study was performed to assess the prognostic factors and the predictors of long-term (3-year) survival in patients with small cell carcinoma of the lung, accrued in one randomized trial, and to define patient subgroups showing significantly different survivals using recursive partitioning and amalgamation analysis. A total of 300 patients with small cell carcinoma of the lung were entered into a randomized study comparing cyclophosphamide, adriamycin and vincristine (CAV), cisplatin and etoposide (PE) and alternating treatments of CAV and PE. Of these, 286 patients were analysed for the present study of prognostic factors. Multivariate analysis showed that poor performance status (2-3) (P = 0.0001), extensive disease (P = 0.0015) and abnormally elevated serum lactate dehydrogenase (P = 0.0001) and alkaline phosphatase (P = 0.0013) were independently adverse pretreatment prognostic factors. Of limited disease patients, performance status (P = 0.029) and white blood cell count (P = 0.044) had a significant influence on the probability of 3-year disease-free survival. Using recursive partitioning and amalgamation analysis, three classes of similar prognosis were identified: the most favorable class was defined by knowledge of lactate dehydrogenase (normal), performance status (0-1) and serum sodium levels (normal) with median survival time of 16.0 months, and the class with the poorest prognosis was defined by knowledge of lactate dehydrogenase (elevated) and performance status (2-3) (median survival time 6.6 months). The intermediate class had a median survival time of 9.4 months. In conclusion, this subclassification system will be used for the design, implementation and interpretation of clinical studies as well as decision-making in individual patients.