Despite recent advances in the diagnosis and treatment of cancer, cancer rates and the number of deaths from cancer continue to increase. In approximately one half of persons diagnosed with cancer, active treatment becomes ineffective at a certain point. At this point physicians are expected to predict the patient's life expectancy based on the clinical estimation of survival. This step becomes significant because of the difficulties that follow late referral or absence of referral to palliative care programs. Viganò and colleagues studied the accuracy of the clinical estimate of survival in patients who were seen at the onset of their terminal phase.

The study consisted of a cohort of sequentially selected patients with advanced cancer. All had terminal cancer of the breast, lung, gastrointestinal system or prostate gland. Further treatment to arrest or control cancer progression in these patients was deemed unavailable. Each patient was given a clinical estimate of survival by the treating oncologist. Follow-up was accomplished by monthly telephone interviews until the time of death.

Of the 248 patients in the study, 225 (91 percent) died during the study, with a median survival of 15.3 weeks. Only 25 percent of the clinical estimates of survival were accurate within one month of prediction. In more than 50 percent, life expectancy was overestimated, and in 23 percent it was underestimated. The median difference between clinical estimate of survival and actual survival was 1.1 months in the optimistic direction. There were no differences between the oncologists, and no other factors that were analyzed had an impact on these predictions.

This study was unique in that it included all patients at the end of life and not just those referred for hospice care. The majority of physicians overestimated survival times. Overestimation of survival has a major impact because it may cause some patients to be denied prompt access to palliative care.

The authors conclude that clinical estimate of life expectancy by itself is not an accurate predictor of prognosis. Better estimates of survival may be achieved by adding other clinical indicators to the prognostic model. Clinical estimate of survival should be considered one criterion for prognosis, not the sole benchmark.

editor's note: The provision of quality of care at the end of life has received renewed interest over the past few years. The biggest problem facing physicians is the inability to predict life expectancy. This study examined clinically based life expectancy predictions in terminal cancer patients made by the specialists who were providing their care. The results indicated that the physicians were overly optimistic in their predictions. This situation occurred in the care of cancer patients, where the progression of disease model is relatively predictable. In other, noncancer diagnoses, predicting life expectancy is even more difficult. Further studies need to be performed to provide physicians and their patients with better predictive tools at end of life.—k.e.m.