Mediating effects of cancer risk factors on the association between race and cancer incidence: analysis of the NIH-AARP Diet and Health Study

Abstract

Purpose: Racial disparities exist in the prevalence of cancer-related risk factors and incidence of cancer. The objective of this study is to determine if cancer-related risk factors mediate the association between race and cancer incidence.

Methods: We performed prospective analysis of data from the National Institutes of Health-American Association of Retired Persons Diet and Health Study, years 1995 through 2011. We compared differences in baseline characteristics between black and white participants using χ² tests and Wilcoxon tests, as appropriate. We determined risk of any cancer and the most common cancer types (i.e., breast, prostate, and colorectal) using Cox Proportional hazards models, adjusted for age, sex, marital status, education, health status, region, and adherence to guidelines on cancer-related risk factors (i.e., body mass index [BMI], smoking status, physical activity, nutrition, and alcohol consumption). We examined the mediation effect of cancer-related risk factors on the association between race and cancer incidence.

Results: Among 425,152 participants, 16,110 (3.79%) were black, and 409,042 (96.21%) were white. The white participants were more likely to be aged 65 years and older (35.33% vs. 25.93%), male (60.88% vs. 42.67%), married (70.37% vs. 48.26%), reside in Western US (30.14% vs. 23.88%), be physically active (46.72% vs. 41.94%), and have higher adherence scores (3.14 vs. 3.04). Blacks had reduced risk of breast cancer (adjusted hazard ratio [HR]: 0.82, 95% confidence intervals [CI]: 0.74-0.90) but higher risk of prostate (adjusted HR: 1.86, 95% CI: 1.75-1.98) and colorectal cancer (adjusted HR: 1.17, 95% CI: 1.05-1.31) compared with whites. Nutrition mediated the association between race and breast cancer (6.35% mediated, P < .01), whereas BMI mediated the association between race and colorectal cancer (7.99% mediated, P < .01).

Conclusions: Blacks were at reduced risk of breast cancer but increased risks for prostate and colorectal cancer incidence. Nutrition and BMI exerted small but significant mediating effects on the racial disparity in risk of breast and colorectal cancers, respectively.

Keywords: Cancer incidence; Mediation; Racial disparities; Risk factors.

Figure 1

National Institutes of Health-American Association of Retired Persons (NIH-AARP) Diet and Health Study population flowchart and cancer incidence. AA = Black.

Figure 2

These models represent the component effects of significant mediation effects on the association between Black race and breast cancer (top - A) and colorectal cancer (bottom - B) incidence. The mediator on the association between Black race and breast cancer is nutrition per SD (SD = 2.50). The mediator on the association between Black race and colorectal cancer is BMI per SD (SD = 5.07). β_a = represents the effect of race on the mediator, calculated from linear regression (mediators are continuous). β_b = represents the effect of the mediator on outcome, calculated from Cox regression and on the log hazard scale. β_c = the controlled direct effect of the exposure (race) on the outcome, calculated from Cox regression and on the log hazard scale.

Figure 2

These models represent the component effects of significant mediation effects on the association between Black race and breast cancer (top - A) and colorectal cancer (bottom - B) incidence. The mediator on the association between Black race and breast cancer is nutrition per SD (SD = 2.50). The mediator on the association between Black race and colorectal cancer is BMI per SD (SD = 5.07). β_a = represents the effect of race on the mediator, calculated from linear regression (mediators are continuous). β_b = represents the effect of the mediator on outcome, calculated from Cox regression and on the log hazard scale. β_c = the controlled direct effect of the exposure (race) on the outcome, calculated from Cox regression and on the log hazard scale.