BACKGROUND
The cancer-specific death rate is a commonly used indicator in the assessment of progress against cancer. However, since the cause of death is often not substantiated and complete medical information is lacking, the validity of cancer-specific mortality rates is being questioned.
PURPOSE
We investigated the validity of the cancer-specific death rate by examining noncancer deaths of cancer patients in a large patient population.
METHODS
Data were obtained from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program on cancer patients diagnosed between 1973 and 1987, with follow-up complete through December 1987. The SEER database consists of 1.2 million records from nine population-based registries covering nine geographic regions of the United States. Rates of noncancer deaths in the U.S. population were obtained from the National Center for Health Statistics. Cancer mortality rates were subtracted from overall mortality rates to obtain noncancer death rates by sex and the 5-year age group for each calendar year. Excluded from the study were patients of races other than White and those diagnosed at age 85 years or more due to absence of noncancer death rate comparisons. Also excluded were cancer cases discovered at autopsy and in persons less than 20 years of age. The statistical analysis employed a log-linear model.
RESULTS
The ratio of patient-to-general-population noncancer death rates, as calculated by dividing the number of patient noncancer deaths per year by the number found in the matched U.S. population data and referred to as the noncancer relative hazard, is considered significant with values greater than 1 for those with all cancers combined and for the common solid tumors examined. Of the 12 leading causes of death other than cancer in the patient population, the most common causes were circulatory and respiratory failures. The noncancer relative risk of death decreased rapidly after diagnosis and also decreased with the patient's age at diagnosis. It increased slightly with the calendar year of diagnosis.
CONCLUSIONS
Because more noncancer deaths occurred shortly after diagnosis, it appears that this excess was caused by treatment of the cancer. Generally, cancer-specific death rates underestimate the mortality associated with a diagnosis of cancer. Therefore, because the degree of underestimation changes with time, an examination solely of cancer-caused mortality in assessing progress against the disease is incomplete.