A prospective study was carried out in the only tertiary oncology department in Oman to analyse the pattern of various prognostic factors in terminally ill cancer patients and their impact on these patients. Between September 1999 and February 2001, terminally ill cancer patients with solid tumours who fulfilled the eligibility criteria were included. All of them were coded Do Not Resuscitate (DNR). Clinical estimation of survival as well as recording of different symptoms and signs was carried out for each patient. Survival was calculated from the date of the DNR coding to the date of death. A total of 162 patients were included. Mean survival time was 41.5 days (median 10 days). The gastrointestinal tract (GIT) was the most common site of malignant disease, followed by the breast. Univariate analysis of evaluable patients showed that performance status (PS), dry mouth, presence of delirium, anorexia, peripheral oedema, absence of bone metastasis, low lymphocyte count and low albumin level had significant effects on survival. Multiple regression analysis showed that PS and oedema were the only independent predictors of survival. Clinical prediction of survival was correlated with observed survival. Patient's PS, presence of peripheral oedema and clinical estimate of survival are good predictors of survival in terminally ill cancer patients.