Abstract B128: Geographic and socioeconomic variations in colorectal cancer incidence in the prospective NIH-AARP Diet and Health study

Abstract

Abstract Background: Persons from lower socioeconomic backgrounds have higher exposure to risk factors for, and higher mortality rate from, colorectal cancer (CRC), but there is a paucity of published population-based studies on the relationship between neighborhood socioeconomic context and CRC incidence. In this study, we used national prospective data to determine: 1) whether CRC incidence varied according to area-level socioeconomic context; and 2) if area-level differences in incidence are explained by clustering of CRC risk factors. Methods: We used data from the NIH-AARP Diet and Health Study, a prospective cohort of 566,402 persons residing in 6 US states and 2 metropolitan areas who were 50–71 years of age at baseline (1995–1996). Detailed data on diet, lifestyle and medical history were collected using questionnaires. Cancer incidence was ascertained from tumor registries through December 31, 2003. Participants were linked to the 2000 US Census data at the tract level. We used principal component factor analyses to derive an index of neighborhood socioeconomic deprivation from 10 census variables about education, poverty, employment and occupation. We restricted the analyses to 319,057 men and 214,298 women who did not have history of CRC at baseline and had complete data including valid geocoded information. Results: In single predictor Cox models with frailty, compared to persons residing in the lowest quintile of the deprivation index, those in the most deprived census tracts had a higher incidence of CRC (HR=1.27; 95% confidence interval [CI]: 1.16–1.40 for men and HR=1.32, CI: 1.15–1.52 for women). This increased risk was robust to adjustments for age, race-ethnicity and family history of colon cancer, body mass index, physical activity and dietary variables including calcium intake, smoking and a history of diabetes (HR: 1.13, CI: 1.03–1.24) for men and HR: 1.17, CI: 1.01–1.34 for women) and to further adjustment for use of non-steroidal inflammatory agents (HR=1.18; CI 1.05–1.34) for men; and HR= 1.19, CI: 0.99–1.42 for women). Significant heterogeneity was observed for the incidence of CRC among census tracts for both men and women (likelihood ratio test p-value <0.001). Conclusion: Our study found significant variation in the incidence of CRC by neighborhood of residence and neighborhood socioeconomic conditions after adjusting for recognized individual risk factors. Further studies are needed to explore other factors that may be associated with area-level differences in CRC incidence such as the use of screening. Citation Information: Cancer Prev Res 2010;3(1 Suppl):B128.