|Home | About | Journals | Submit | Contact Us | Français|
Incidence rates in the U.S. show clear racial/ethnic disparities for colorectal cancer. We examined the extent to which ethnic differences in risk factors could explain the age-adjusted variation in the risk of colorectal cancer, overall and by stage at diagnosis, among 165,711 African Americans, Japanese Americans, Latinos, Native Hawaiians, and whites participating in the Multiethnic Cohort Study. Over a median follow-up period of 10.7 years, 2,564 incident cases of colorectal cancer were identified through SEER tumor registry linkages in Hawaii and California. Multivariable-adjusted Cox proportional hazard models were used to estimate relative risks (RR) and 95% confidence intervals (CI) for each ethnic group compared to whites. After accounting for known/suspected risk factors, Japanese Americans (men, RR = 1.27, 95% CI = 1.09-1.48; women, RR = 1.49, 95% CI = 1.24-1.78) and African American women (RR = 1.48, 95% CI = 1.23-1.79) remained at increased risk of colorectal cancer relative to whites; African American and Japanese American women were also at increased risk of advanced disease compared to whites. In site-specific analyses, after multivariable adjustment, African Americans (both sexes) and Japanese American women remained at increased risk for colon cancer, and Japanese Americans (both sexes) and Native Hawaiian men for rectal cancer compared to whites. The results of this study suggest that differences in the distribution of known/suspected risk factors account for only a modest proportion of the ethnic variation in colorectal cancer risk and that other factors, possibly including genetic susceptibility, are important contributors to the observed disparities in incidence.
Colorectal cancer is the third most frequently diagnosed cancer in both men and women in the U.S. and accounts for 9% of all cancer-related deaths.1 Factors associated with an increased risk of colorectal cancer include older age, male sex, family history of colon or rectal cancer, history of colorectal polyp, inflammatory bowel disease, obesity, physical inactivity, type 2 diabetes mellitus, alcohol consumption, smoking, and consuming a diet high in red and processed meat.1-3 Conversely, nonsteroidal anti-inflammatory drug (NSAID) use, hormone replacement therapy (HRT), and greater consumption of calcium, fiber, folate, and vitamin D may reduce risk of colorectal cancer.1-3
Clear racial/ethnic disparities in the incidence of colorectal cancer, stage at diagnosis, and mortality have been seen in recent Surveillance, Epidemiology, and End Results (SEER) program data. Age-adjusted incidence and mortality are highest for African Americans4-7 and lowest for Hispanics and Asians/Pacific Islanders.4, 6 When Asians/Pacific Islanders were separated into specific ethnic groups, Japanese Americans residing in California, Hawaii, and Washington State were found to have a particularly high incidence of colorectal cancer and Native Hawaiians an especially high mortality from colon cancer.8, 9 Differences in the age-adjusted incidence rates between ethnic groups in the U.S. suggest that biologic or environmental exposures vary sufficiently between subpopulations to affect risk of disease. As such, differences in risk factor prevalence between ethnic groups may, in part, explain ethnic variation in colorectal cancer risk.
For this report, we investigated this hypothesis by examining the extent to which differences in the distribution of risk factors between African Americans, Japanese Americans, Latinos, Native Hawaiians, and whites participating in the Multiethnic Cohort Study (MEC) could explain variations in the age-adjusted incidence of colorectal cancer. Findings from this study may be useful for targeted risk factor interventions among specific ethnic groups, and may suggest directions for future etiological research.
The MEC was designed to investigate the association of dietary, lifestyle, and genetic factors with the incidence of cancer and has been described previously in detail.10 Briefly, over 215,000 men and women 45-75 years of age at recruitment and residing in Hawaii or California (primarily Los Angeles County) were enrolled into the cohort between 1993 and 1996. Drivers' license files, voter registration lists, and Health Care Financing Administration (Medicare) files comprised the primary sampling frame to obtain a multiethnic sample of African Americans, Japanese Americans, Latinos, Native Hawaiians, and whites. At cohort entry, participants completed a self-administered, 26-page baseline questionnaire providing information on demographic characteristics, anthropometric measures, medical history, family history of cancer, reproductive and menstrual history, cancer screening practices, occupational history, physical activity, as well as detailed questions on diet. The study protocol was approved by the institutional review boards of the University of Hawaii and the University of Southern California.
In the present analysis, we excluded participants not in one of the five main ethnic groups recruited into the study (n = 13,991) or who reported implausible dietary values at baseline11 (n = 8,264). Participants identified through either the questionnaire or tumor registry linkage to have had a diagnosis of colon or rectal cancer prior to cohort entry were also excluded (n = 2,561). In addition, participants with missing data on any of the following key covariates were excluded from all analyses: family history of colorectal cancer, history of colorectal polyp, smoking, body mass index (BMI), vigorous physical activity, non-steroidal anti-inflammatory use (NSAID), multivitamin use, history of diabetes, and in women, hormone replacement therapy use (n = 25,291). After exclusions, data from 165,711 participants (n = 76,919 men; n = 88,792 women) were available for analysis.
Incident cases of colorectal cancer were identified through regular linkages of the cohort to the SEER cancer registries for Hawaii (Hawaii Tumor Registry) and California (the Cancer Surveillance Program for Los Angles County and California State Cancer Registry). Deaths were ascertained through routine linkages to the state death certificate files in Hawaii and California, as well as to the National Death Index to identify deaths among emigrants to other parts of the U.S. Case status was determined using the International Classification of Diseases for Oncology (ICD-O-02) codes C18.0-C18.9 and C26.0 (colon), C19.9 (rectosigmoid junction), and C20.9 (rectum) and restricted to invasive carcinomas. Follow-up time began at the date of questionnaire completion and accrued until either diagnosis of colorectal cancer, death, or last follow-up date (December 31, 2004), whichever occurred earlier. Over a median follow-up time of 10.7 years, 2,564 incident cases of colorectal cancer (n = 1,427 men; n = 1137 women) were diagnosed.
Information on demographic characteristics and colorectal cancer risk factors was obtained from the baseline questionnaire. Participants reporting mixed ethnicity were classified using the following hierarchy: African American, Native Hawaiian, Latino, Japanese American, white, and other. More than 95% of all non-Hawaiians reported only one ethnic group. Self-reported height and weight were used to calculate BMI. Pack-years of cigarette smoking was calculated as the product of the average number of packs of 20 cigarettes smoked per day and the number of years smoked. Vigorous physical activity in hours was obtained from the sum of reported hours spent in vigorous work and leisure time physical activity. Participants reporting NSAID use at least two times per week for at least one month were considered NSAID users. Participants taking multivitamins at least once a week during the preceding year were considered multivitamin users. Post-menopausal hormone therapy use was classified as never, past or current use of estrogen and/or progesterone. Information on family history of colorectal cancer, history of colorectal polyp, and history of diabetes was also collected at baseline.
Dietary intake was assessed at baseline using a Quantitative Food Frequency Questionnaire (QFFQ) that obtained frequency and quantity information on food items consumed during the preceding year and was designed for use in this multiethnic population.10 Food items included on the questionnaire were those identified from three-day measured food records as the minimum set that could explain ≥ 85% of intake for nutrients of interest for each ethnic group, as well as foods traditionally consumed. Dietary values were computed from the QFFQ using the food composition database designed specifically for the MEC and maintained by the Cancer Research Center of Hawaii.10
Cox proportional hazard models with attained age as the time metric were used to calculate relative risks (RR) and 95% confidence intervals (CI) for colorectal cancer. In all models, whites served as the referent group, and relative risks were estimated for African Americans, Japanese Americans, Latinos, and Native Hawaiians. Models were constructed to examine the risk of colorectal cancer at any stage, localized stage at time of diagnosis, advanced stage at time of diagnosis, colon cancer, and rectal cancer. In stage-specific and site-specific models, participants were censored on the date of diagnosis if they experienced the alternative event before the event of interest. Estimates were calculated separately for men and women, as results were found to differ by sex. “Age-adjusted models” adjusted for age in the time metric and age at cohort entry in the log-linear model component. “Multivariable models” further adjusted for: family history of colorectal cancer (yes/no), history of colorectal polyp (yes/no), BMI (<25, 25-<30, ≥30), pack-years of cigarette smoking (continuous), multivitamin use (yes/no), NSAID use (yes/no), vigorous physical activity (hours/day), history of diabetes (yes/no), alcohol consumption (grams/day), total energy (log transformed calories/day), red meat (grams per 1000 kcal/day from food), dietary fiber (grams per 1000 kcal/day from food), calcium (mg/day from food and supplements), folate (mcg/day from food and supplements), and vitamin D (IU/day from food and supplements) in the log-linear model component. Adjustment variables were selected if they have been shown consistently in the literature to be associated with colorectal cancer risk or were identified as confounders in our study. The assumption of proportional hazards was found to be satisfied in all models by examining parameter estimates of scaled Schoenfeld residuals against time.
A sensitivity analysis using multiple imputation was conducted to account for missing covariates. Five datasets were created with imputed data values. As results from imputed datasets were similar to those obtained for the complete-case analysis, the complete-case analysis is presented here. We also examined processed meat and educational attainment (as a proxy of socioeconomic status) as potential risk factors. They were not included in the final models as they did not result in a change in the relative risks by more than 10% in any model. To adjust for differences in the distribution of smoking on the relative risk of colorectal cancer between ethnic groups, pack-years of smoking was used in all models as the inclusion of additional smoking variables, as used in some other analyses of MEC data,12 did not contribute further to explaining the risk of colorectal cancer overall or by ethnic group. Ethnic differences in the distributions of colorectal cancer risk factors were tested by the χ2 test of association for categorical variables and F test from ANOVA for continuous variables. All data analyses were performed using SAS 9.2 statistical software (SAS Institute Inc., Cary, NC, USA).
Table 1 shows the distribution of colorectal cancer cases by sex and ethnic group. Among men, there were 1427 cases of colorectal cancer, of which 622 were localized and 742 were at an advanced stage at time of diagnosis. Among women, there were 1137 cases of colorectal cancer, 473 localized and 588 advanced at time of diagnosis. For all sex and ethnic groups except Native Hawaiian men, the majority of colorectal cancer cases were diagnosed at an advanced stage. For men, African Americans (55.4%) and Latinos (54.3%) experienced the highest rates of advanced stage disease with the percentage among women similar across ethnic groups. Among men, a higher percentage of colon cancers were diagnosed among African Americans (81.4%) than for other ethnic groups, while a higher percentage of rectal cancers were diagnosed among Native Hawaiians (30.7%), Japanese Americans (29.4%), and Latinos (28.6%) than for African Americans or whites. Among women, African Americans were diagnosed with a higher percentage (84.3%) of colon cancers and Latinas were diagnosed with a higher percentage (25.1%) of rectal cancers than for other ethnic groups.
Table 2 presents the distribution of colorectal cancer risk factors by ethnic group separately for men and women. At cohort entry, African Americans and Japanese Americans were slightly older than other ethnic groups with Japanese Americans reporting the highest positive family history of colorectal cancer and personal history of colorectal polyps. African Americans, Native Hawaiians, and Latinos reported a higher prevalence of obesity and diabetes than did Japanese Americans and whites. Alcohol consumption and pack-years smoked was highest among whites. Whites, Native Hawaiians, and Latino men reported the highest levels of vigorous physical activity. NSAID use was highest for whites, African Americans, and Latinos with the prevalence of multivitamin use similar across ethnic groups, except for the lower rates of use among Native Hawaiians. For dietary risk factors, African Americans, Native Hawaiians, and Latinos reported the greatest consumption of red meat. Latinos reported the highest intake of dietary fiber and folate with calcium intake highest among Latinos and whites. Whites reported the highest intake of Vitamin D. Among women, post-menopausal HRT use was most prevalent among white and Japanese American women.
Table 3 shows the age-adjusted and multivariable-adjusted relative risk of colorectal cancer for men by ethnic group. The age-adjusted risk of colorectal cancer was higher for African Americans, Native Hawaiians, and Japanese Americans than for whites. After multivariable adjustment, the excess risk among African Americans, Native Hawaiians, and Japanese Americans was reduced by 25%, 51%, and 33%, respectively relative to whites; however, Japanese Americans remained at increased risk of colorectal cancer (RR = 1.27, 95% CI = 1.09-1.48), while the risks for African Americans and Native Hawaiians no longer reached statistical significance.
The multivariable-adjusted risk of localized disease was higher for Native Hawaiians and Japanese Americans, but not for African Americans, compared with whites. The increased risk of advanced stage disease for African Americans and Japanese Americans relative to whites in age-adjusted models was modestly attenuated in multivariable models, with risk factors explaining 25% of the excess risk for African Americans and 36% of the excess risk for Japanese Americans. While no statistically significant differences were observed for advanced stage disease among any ethnic group relative to whites after multivariable adjustment, the point estimates for African Americans and Japanese Americans remained somewhat higher than for other ethnic groups.
For site-specific analyses, the age-adjusted risk of cancer of the colon was higher for African Americans, Native Hawaiians, and Japanese Americans, and the risk of cancer of the rectum was higher for Native Hawaiians and Japanese Americans than for whites. Relative to whites, the excess risk of colon cancer explained by risk factors was 22%, 67%, and 41% for African Americans, Native Hawaiians, and Japanese Americans, respectively; however, African Americans remained at an increased risk (RR = 1.26, 95% CI = 1.03-1.55) while associations were no longer statistically significant for Native Hawaiians and Japanese Americans. Native Hawaiians and Japanese Americans remained at increased risk of rectal cancer compared to whites in multivariable models, despite risk factors accounting for 28% of their excess risk relative to whites. The risk of colorectal cancer by stage and by sub-site did not differ between Latinos and whites with and without multivariable-adjustment.
Table 4 presents the age-adjusted and multivariable-adjusted relative risk of colorectal cancer for women by ethnic group. The age-adjusted risk of colorectal cancer was higher for African Americans and Japanese Americans when compared to whites. After multivariable adjustment, the excess risk among African Americans, Native Hawaiians, and Japanese Americans was reduced by 24%, 33%, and 29%, respectively relative to whites; however, African Americans (RR = 1.48, 95% CI = 1.23-1.79) and Japanese Americans (RR = 1.49, 95% CI = 1.24-1.78) remained at increased risk. Despite modest reductions in the risk of localized and advanced stage disease in multivariable models, the risk remained higher for African Americans and Japanese Americans than for whites after accounting for ethnic differences in the prevalence of risk factors.
For site-specific analyses, the age-adjusted risk of cancer of the colon was higher for African Americans and Japanese Americans; however, no difference in rectal cancer was seen for any ethnic group relative to whites. Relative to whites, the excess risk of colon cancer explained by risk factors was 22% and 41% for African Americans and Japanese Americans, respectively; however, African Americans (RR = 1.63, 95% CI = 1.32-2.00) and Japanese Americans (RR = 1.49, 95% CI = 1.21-1.83) remained at an increased risk of disease. In the multivariable adjusted model, an increased risk of rectal cancer was seen for Japanese Americans compared to whites. The risk of colorectal cancer by stage and by sub-site did not differ between Native Hawaiians or Latinos and whites with and without multivariable-adjustment.
In this large prospective study, we found that ethnic variation in the age-adjusted incidence of colorectal cancer was not fully explained by differences in the prevalence of known/suspected risk factors. Japanese Americans (both sexes) and African American women remained at increased risk of colorectal cancer relative to whites after accounting for differences in the distribution of risk factors. The risk of presenting at an advanced stage at time of diagnosis was somewhat reduced for ethnic groups relative to whites among men, but remained higher for both African American and Japanese American women. Site-specific analyses showed African Americans (both sexes) and Japanese American women to have a higher risk of cancer of the colon, and Japanese Americans (both sexes) and Native Hawaiian men to have a higher risk of cancer of the rectum after adjustment for risk factors. Overall, known colorectal cancer risk factors appeared to explain a greater extent of ethnic variation in disease risk for advanced than for localized stage disease, for colon than for rectal cancer, and for men than for women. Our findings suggest that ethnic disparities in colorectal cancer risk are partially attributable to differences in the distribution of known/suspected risk factors between ethnic groups. However, factors beyond those accounted for in our models are clearly also contributing to variation in disease risk.
The higher risk of colorectal cancer, advanced stage at diagnosis, and colon cancer seen among African Americans in our sample is well established.4, 13-16 Adjusting for differences in the prevalence of risk factors resulted in modest reductions in disease risk, with the risk of colorectal cancer risk at any stage and for advanced stage disease among men no longer statistically significant relative to whites. Post-hoc analyses showed this reduction to be largely attributable to the higher prevalence of obesity and diabetes among African American men. The reduction in risk of advanced stage disease among African Americans shows that differences in known risk factors account for a proportion of their later stage at time of diagnosis. However, that African American women remain at increased risk of advanced disease, coupled with the non-significant increased risk remaining for men, suggests that additional factors not captured in our models are contributing to their later stage at diagnosis.
Japanese Americans were found to be at increased risk for colorectal cancer relative to whites which is consistent with SEER data.8 Only for advanced stage disease and for colon cancer among men was the reduction in risk statistically significant relative to whites, with risk factor adjustment. The higher risk among Japanese Americans is of added interest as modifiable risk factors were generally fewer for Japanese Americans than for other ethnic groups, although they did report the highest prevalence of a positive family history of colorectal cancer. The higher incidence rate for Japanese Americans, coupled with the lower prevalence of modifiable risk factors, suggests a potential interaction between biologic and environmental factors and is consistent with migration studies showing the incidence rate of colorectal cancer among Japanese American migrants to the U.S. to be equal to, or surpass that of, U.S. whites within a single generation.17 Estimates from multivariable models for Japanese American women highlight that disease risk in this population may increase, should the distribution of risk factors among Japanese Americans more closely resemble that of other ethnic groups.
Relative to whites, Native Hawaiian men, but not women, were at increased risk for localized disease and for rectal cancer after adjusting for differences in risk factor prevalence. The increased risk of colorectal cancer at any stage and of colon cancer was largely attenuated and no longer significant relative to whites after multivariate adjustment. In post-hoc analyses, no single risk factor explained a substantial proportion of the reduction in risk. Our results suggest that improvements in risk factor prevalence among Native Hawaiian men may further reduce the incidence of colon cancer and of advanced stage disease at diagnosis in this population, thereby decreasing the disparity in colon cancer mortality. The risk of colorectal cancer for Native Hawaiian women was similar to that of whites, consistent with SEER data collected in Hawaii.8 For Native Hawaiians, some results are based on a small number of cases; additional follow-up should provide more reliable estimates for this population.
The age-adjusted risk of colorectal cancer did not differ between Latinos and whites for any outcome examined or after adjusting for ethnic differences in the distribution of risk factors. Our age-adjusted results differ from those obtained using SEER data that showed the incidence rates to be lower for Latinos than whites.4 Latinos in our population reported a higher prevalence of obesity and diabetes than whites, consistent with Behavioral Risk Factor Surveillance System (BRFSS) data collected in California during this time period;18 however, they also reported the lowest family history of colorectal cancer, colorectal polyps, and smoking, and high consumption of dietary fiber, calcium, and folate which may have mitigated their risk from other factors. Adjusting for differences in the distribution of risk factors did not influence the risk of colorectal cancer for Latinos relative to whites, suggesting that other factors may be responsible for the differences in incidence rates seen in the SEER data. Although the percentage of Latinos in the MEC who reported being foreign born (52%)10 is similar to that obtained for Los Angeles County during the 2000 census (49%),19 it remains plausible that the Latinos who entered our cohort may be more connected with the health care system and therefore more likely to be diagnosed.
Strengths of this study include the prospective design, ethnic diversity of the study population, large number of cases, and detailed information on colorectal cancer risk factors, including diet. There were also study limitations. Data on specific aspects of socioeconomic status (SES) that may potentially confound associations between colorectal cancer incidence and known risk factors were not collected at baseline in the MEC and therefore could not be accounted for in this study. However, attained education was examined as a proxy of SES in preliminary models and found not to impact on the risk of colorectal cancer, stage at diagnosis, or site-specific analyses. Also, specific information was not obtained on colorectal cancer screening practices at baseline. Therefore personal history of colorectal polyp was used a proxy for this measure. Sensitivity analyses were conducted that included history of colonoscopy or sigmoidoscopy (ever/never) to the “multivariable model” among participants who completed a follow-up questionnaire (n = 59,983 men; n = 72,637 women). The addition of screening history had only a modest impact on the reported relative risks for all models and did not appreciably explain ethnic differences in the risk of colorectal cancer. In addition, despite efforts to obtain detailed information on colorectal cancer risk factors, some risk factors, especially dietary, are measured with error and thus may have lead to an underestimation of their contribution to ethnic differences. Lastly, as mentioned, due to the smaller number of Native Hawaiians enrolled in the cohort, some estimates for this population are based on a limited number of cases.
In conclusion, ethnic variation in risk of colorectal cancer could not be fully explained by differences in the prevalence of known/suspected risk factors. Adjusting for differences in the distribution of risk factors between ethnic groups modestly attenuated relative risks for African Americans, Japanese American men, and Native Hawaiian men relative to whites; however, clear differences in risk of disease remained. Study findings suggest that improvements in colorectal cancer risk factors among specific ethnic groups may reduce disparities in disease incidence, especially for African American and Native Hawaiian men, but that factors beyond those included in our models are contributing to ethnic disparities in colorectal cancer risk. These factors likely include differences in additional environmental exposures, as well as, interactions of environmental exposures with commonly inherited low penetrance cancer susceptibly genes. Further research attempting to elucidate additional environmental and biological risk factors for colorectal cancer is warranted and may advance our ability to explain and reduce racial/ethnic disparities in colorectal cancer risk.
We thank all participants in the Multiethnic Cohort Study. The Multiethnic Cohort Study has been supported by grant R37 CA 54281 from the National Cancer Institute. The SEER tumor registries are supported by the National Institutes of Health, Department of Health and Human Services (contracts N01-PC-35137 and N01-PC-35139, respectively). NJO was supported by a postdoctoral fellowship on grant R25 CA 90956.