Can we predict very short term survival in small cell lung cancer?

We retrospectively reviewed the charts of 151 consecutive patients diagnosed as small cell lung cancer in our department and who had at least one course of chemotherapy. Nineteen patients died during the first 2 months, of the probability were reported to construct a receiver operating characteristic curve i.e. 13% of the population. The probability of dying within 2 months was investigated through a stepwise logistic regression. A performance status < or = 70 (Karnofsky index), an age > 60, a platelet count < or = 150,000/mm3, elevated alkaline phosphatase and a sodium < or = 135 mmol/l were independent predictors of a very short term survival and contributed to the equation for the probability of dying within a 2-month period. Sensitivity and specificity for various cutoff points characteristic curve allowing one to determine for a given patient his risk of being a very short term survivor. Such an approach could prevent inclusion of patients with high risk of early death in clinical trials and help to choose appropriate treatments for such poor risk patients.