The impact of state-specific life tables on relative survival

Abstract

Background: Relative survival is based on estimating excess cancer mortality in a study population compared to expected mortality of a comparable population without cancer. In the United States, expected mortality is estimated from national life tables matched by age, sex, race, and calendar year to each individual in the study population. We compared five-year relative survival using state life tables to five-year relative survival using US decennial life tables. We assessed variations by age, race, and cancer site for all cancers combined, lung, colorectal, prostate, and female breast cancers. Methods: We used data from 17 National Cancer Institute Surveillance, Epidemiology, and End Results Program registries, including diagnoses from January 1, 2000 to December 31, 2009 with follow-up through December 31, 2010. Five-year relative survival was calculated using US-based life tables (USLT) and state-specific life tables (SLT). Results: Differences in SLT- and USLT-based survival were generally small (SLT < 4 survival percentage points lower than USLT). Differences were higher for states with high SES and low mortality and for prostate cancer. Differences were largest for all cancers combined, colon and rectum, and prostate cancer among males aged 85+ ranging from -10 to -17 survival points for whites and +9 to +17 for blacks. Conclusion: Differences between relative survival based on USLT and SLT were small and state-based estimates were less reliable than US-based estimates for older populations aged 85+. Our findings underscore the need to develop more appropriate life tables that better represent the varying mortality patterns in different populations in order to obtain accurate estimates of relative survival.