Predicting the survival of terminally ill cancer patients can help in informing patients and their families, in programming therapy and assistance models, and in utilising existing resources correctly. Clinical prediction of survival (CPS) and Karnofsky performance status (KPS) are two factors which have already been described in the literature. The aim of our study was to verify their respective predictive value with regard to actual survival. In our study of 100 consecutive patients, the CPS obtained a higher prediction accuracy than that reported previously (correlation coefficient with actual survival = 0.51) and than that obtained with KPS alone (correlation coefficient = 0.37). The median difference between predicted and expected survival was only 1 week. The resultant predictivity could be further improved by integrating other prognostic factors studied in larger prospective, multicentric studies.