Accuracy of clinical prediction of survival for terminally ill cancer patients

Accurate estimation of survival is vital for effective palliative care. To verify the value of clinical prediction of survival (CPS), a prospective study was performed on 150 terminally ill cancer patients. The CPS was highly correlated with actual survival (AS), but the accuracy was not significantly superior to the prediction by performance status alone. Serious pessimistic error, defined as AS was at least 28 days and twice as long as CPS, was recognized in 13%, while serious optimistic error, defined as AS was less than 28 days and half as long as CPS, in 15%. The frequency of serious error was not significantly different by physicians' experiences, patients' age, sex, primary disease, and metastatic locations, but was significantly higher in cases with better performance status. Also, unexpected changes resulting in death were experienced in 42% of another 186 cases. The main underlying causes were pneumonia, bleeding, heart failure, intestinal perforation, cerebrovascular disease, hepatic/renal failure, hypoglycemia, sepsis and electrolyte imbalance. Clinical prediction was not sufficiently reliable and must be further improved.