Based on an increased frequency of early death (death within the first treatment cycle) in our two latest randomized trials of combination chemotherapy in small-cell lung cancer (SCLC), we wanted to identify patients at risk of early non-toxic death (ENTD) and early toxic death (ETD). Data were stored in a database and logistic regression analyses were performed to identify predictive factors for early death. During the first cycle, 118 out of 937 patients (12.6%) died. In 38 patients (4%), the cause of death was sepsis. Significant risk factors were age, performance status (PS), lactate dehydrogenase (LDH) and treatment with epipodophyllotoxins and platinum in the first cycle (EP). Risk factors for ENTD were age, PS and LDH. Extensive stage had a hazard ratio of 1.9 (P = 0.07). Risk factors for ETD were EP, PS and LDH, whereas age and stage were not. For EP, the hazard ratio was as high as 6.7 (P = 0.0001). We introduced a simple prognostic algorithm including performance status, LDH and age. Using a prognostic algorithm to exclude poor-risk patients from trials, we could minimize early death, improve long-term survival and increase the survival differences between different regimens. We suggest that other groups evaluate our algorithm and exclude poor prognosis patients from trials of dose intensification.