Survivorship differences in geographical comparisons of cancer mortality: an urban-rural analysis.

Epidemiologists have traditionally used geographical comparisons of cancer site-specific mortality rates to draw aetiological inferences about neoplasms. This approach is based on the unproven assumption that cancer mortality and incidence rates are highly correlated within geographical areas. Since mortality is a function of both incidence and survival rates, geographical differences in cancer survivorship may confound area comparison of cancer mortality rates. To test this possibility, the survival experience of white male cancer patients residing in rural and urban areas is examined using cancer registry data. A multi-variable proportional hazards model is specified to determine the unique effect of geographical residence on survival. Only for cancers of the gastro-intestinal (GI) tract are there statistically significant differences in survivorship by geographical residence. This suggests that for most cancers, survival differences are not likely to play a confounding role in geographical comparisons of cancer mortality rates. However for GI cancers, survival differences should probably be considered in geographical-oriented analyses and their interpretation.