The aims of this study were to identify prognostic factors in patients (pts) with small cell lung cancer and to identify dominant prognostic factors independent of disease stage, to define prognostic subsets through recursive partitioning and amalgamation (RPA) and to analyze the clinical characteristics of long-term survivors. The prognostic significance of 27 pre-treatment variables was evaluated in 144 pts seen at a single institution. The current study confirmed the superior outcome for pts with limited disease (LD) in terms of response, response duration, time to treatment failure and survival when compared to those with extensive disease (ED). None of the variables independently predicted for response in patients with LD. Response correlated significantly with a good performance status (PS) for pts with ED and for the whole group. A good PS was the most significant predictor for prolonged survival in pts with LD. In ED a longer survival was associated with a normal pre-treatment albumin value, absence of weight loss and female gender. When the whole group was considered, PS and number of metastatic sites were identified as the most influential factors for survival independent of disease stage. RPA analysis defined 3 prognostic subsets based on stage of disease, PS and number of metastatic sites. The best survival rates were seen in pts with LD with a good PS and pts with ED, only one metastatic site and a good PS. 11% of pts survived > 2 years (18% LD, 6% ED). A complete response to chemotherapy was the most important predictor for long-term survival. Comparison of the data from this study with published results of protocol studies showed similar outcomes.