How generalizable are the SEER registries to the cancer populations of the USA?

Background We determined whether the current SEER registries are representative of the nation's cancer cases or the socio-demographic characteristics. Methods We used breast cancer (BC) and colorectal cancer (CRC) cases diagnosed 2004-2009 from the US Cancer Statistics (USCS) database. Cases were classified into groups residing in SEER coverage areas and the other areas. We compared difference between SEER and non-SEER areas in: age-race-specific proportions of late-stage BC or CRC, area demographics and socioeconomic factors, and data quality. Results For late-stage CRC diagnosis, SEER areas contained lower proportions of people with other race and higher proportions of Asian and Hispanic females aged <40, than non-SEER areas. For late-stage BC diagnosis, SEER and non-SEER estimates were comparable. SEER areas had lower percentages of whites and higher percentages of young people, were more urban, and had higher percentage of poor, lower educational attainment, and higher unemployment. SEER areas also tended to have a higher percentage of case completeness than non-SEER areas. Conclusion Overall, SEER registries were not significantly different from non-SEER areas in terms of average agerace-specific proportions of late-stage BC or CRC, except for estimates of late-stage CRC for other race and young Asian and Hispanic women. Although case completeness was better in SEER areas than non-SEER areas, SEER areas had greater economic disadvantage and greater minority diversity among the population. This study demonstrated a need for caution in using SEER data and discussed advantages of using the more complete USCS database.