Planning terminal care for patients with malignant neoplasms is difficult, in part, because accurate measures of prognosis have not been defined. Using data from the National Hospice Study, we examined the correlation of 14 easily assessable clinical symptoms with survival in patients with terminal cancer. Performance status was the most important clinical factor in estimating survival time, but five other symptoms had independent predictive value as well (shortness of breath, problems eating or anorexia, trouble swallowing, dry mouth, and weight loss). We generated four parametric accelerated time survival models to estimate survival in patients with combinations of these symptoms and validated the log-normal model on the entire data set. This model was unaffected by patient age, sex, primary tumor type, or site. Our findings illustrate the value of biologically "soft" clinical data in predicting survival in patients with terminal cancer. The prevalence of similar symptoms among patients with cancer of various primary and metastatic sites also supports the concept of a common final clinical pathway in patients with advanced malignant neoplasms.