|Home | About | Journals | Submit | Contact Us | Français|
To examine whether body mass index is associated with reduced colorectal cancer (CRC) screening in a large population of black and white adults.
Cross-sectional data collected at baseline for 9,547 black males, 14,515 black females, 3,519 white males, and 7,245 white females age 50–79 enrolled in the Southern Community Cohort Study from 2002–2009 were used to examine odds ratios (OR) with 95% confidence internals (CI) for use of colonoscopy or sigmoidoscopy in relation to body mass index (BMI) categories (<18.5, 18.5–24.9 (referent), 25–29.9, 30–34.9, 35–39.9, and 40+ (extreme obesity), kg/m2) using logistic regression controlling for age, education, income, health insurance status, last physician visit, cigarette smoking, and alcohol consumption.
Increased BMI was not associated with reduced CRC screening among whites (OR [95% CI] for BMI ≥ 40 = 1.02 [0.71–1.46] for white males and 0.99 [0.83–1.19] for white females), and odds of CRC screening were increased with high BMI among blacks (OR [95% CI] for BMI ≥ 40 = 1.34 [1.03–1.74] for black males and 1.13 [0.98–1.29] for black females). Extreme obesity was associated with reduced odds of CRC screening only among white women in subgroup analyses limited to those with health insurance or income ≥ $25,000/year.
Elevated BMI was not a deterrent to CRC screening overall in this population. In light of low overall screening rates for colorectal cancer nationally, efforts to increase screening in all individuals should remain the focus of public health initiatives.
Colorectal cancer (CRC) is the third most common cancer in the United States with approximately 141,210 new cases and 49,380 deaths expected in 2011 (1). CRC screening is recommended for all adults over the age of 50 in order to reduce CRC incidence through the removal of precancerous lesions and to reduce mortality with early detection (2, 3). Obesity has been consistently linked to increased incidence and mortality of CRC (4, 5), and with the rapidly increasing prevalence of obesity in recent decades (6, 7), screening is of particular importance to prevent substantial morbidity and mortality from this common cancer. However, CRC screening rates are low with only about half of adults over the age of 50 meeting CRC screening guidelines (8). There is evidence to indicate that cancer screening may be lower in overweight or obese individuals and, further, some studies indicate that this association may be modified by race or gender (9–12). However data are limited and have been inconsistent for CRC screening specifically (9), and most studies have not been able to adequately compare associations between obesity and CRC screening across race and gender groups. In this analysis, we examined the association between body mass index (BMI) and CRC screening in the Southern Community Cohort Study which includes large numbers of males and females over a wide range of age and body sizes, and further, includes black and white participants of similar, and often low, socioeconomic status.
The Southern Community Cohort Study (SCCS) is an ongoing prospective cohort study assessing risks of and disparities in cancer and other chronic diseases and their health impact (13, 14). Study enrollment took place at 71 participating community health centers (CHCs) located in twelve southeastern states (Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Virginia, and West Virginia). CHCs are institutions that provide health and preventative services primarily to low income and uninsured persons (15). Potential study participants were CHC patients, friends or family members who were accompanying patients, and users of CHC pharmacies. Participants were recruited to join the study in-person at the CHC by a trained study interviewer. SCCS eligibility requirements included being age 40–79 years, English-speaking, and not having been under treatment for cancer in the past 12 months. Nearly 73,000 participants were recruited via CHCs from 2002 to 2009. The study was approved by Institutional Review Boards at Vanderbilt University and Meharry Medical College.
Participants were administered an in-person baseline interview by a trained interviewer at enrollment. The computer-assisted personal interview contained questions about demographic, medical, familial, lifestyle and other participant characteristics (a copy of the questionnaire is available on the study website www.southerncommunitystudy.org). For participants enrolled from 2002 until September 2007, current weight and height were obtained by self-report. For 24% of these participants, measured height and weight were abstracted from medical records for validation purposes. Starting in October 2007, current height and weight were measured by study interviewers using a SECA 703 digital scale and height rod.
All participants were asked “Have you ever had a colonoscopy?” and “Have you ever had a sigmoidoscopy?” A sigmoidoscopy was described as “[a] short tube inserted into the rectum while you are awake and un-sedated to look for colon or rectal cancer” and a colonoscopy was described as “[a] long tube inserted into the rectum after you are sedated or put to sleep to look for colon or rectal cancer.” Participants who answered yes to these questions were then asked when their last colonoscopy or sigmoidoscopy took place.
Current CRC screening guidelines recommend that screening begin at age 50 (2); therefore this analysis was limited to the 37,731 SCCS participants who were age 50 years or older at the time of enrollment. Study exclusions included 1,613 participants who reported their race as other than White or Black/African-American; 388 participants with missing anthropometric values; 200 participants reporting a prior diagnosis of either colon or rectal cancer; and 704 participants with missing values for sigmoidoscopy or colonoscopy screening. 34,826 participants age 50–79 at enrollment (9,547 black males, 14,515 black females, 3,519 white males, and 7,245 white females) were included in the final analysis.
The primary outcome for analysis was a dichotomous measure of up-to-date colorectal cancer screening by either sigmoidoscopy in the past five years or colonoscopy in the past ten years (hereafter called “CRC screening”). Sigmoidoscopy alone and colonoscopy alone were also examined as outcome measures. BMI was calculated as [weight (kg)]/[height (m)2] and categorized using standard cut-offs (16)as underweight(BMI < 18.5 kg/m2), healthy weight (18.5–24.9), overweight (25.0–29.9), obesity class I (30.0–34.5), obesity class II (35.0–39.9) and obesity class III or extreme obesity (40.0+).
Odds ratios (OR) with 95% confidence intervals (95% CI) for up-to-date CRC screening associated with categories of BMI were estimated using logistic regression. Participants of healthy weight (BMI 18.5–24.9 kg/m2) were used as the referent category. Potential confounders were determined from the literature and included age, educational attainment, annual household income, health insurance coverage, current cigarette smoking status, alcohol consumption, time since last doctor’s visit, and family history of colon or rectal cancer (categorized as shown in Table 1). Physical activity (obtained from a physical activity questionnaire at the baseline interview and including light, moderate, and vigorous occupational and household activity plus moderate and vigorous sports activity, measured as MET-hours/day) and dietary intake (obtained from an 89-item food frequency questionnaire at the baseline interview, measured as total energy intake in kcal/day) were also examined as potential confounders but their inclusion in the multivariate models did not alter the odds ratios between BMI and CRC screening in a meaningful way. Therefore all models presented here do not include adjustment for physical activity or dietary intake.
Logistic regression models were examined separately for males and females. Effect modification by race was evaluated by comparing models with and without interaction terms for race*BMI using the likelihood ratio test (LRT). The LRT was significant (p<0.1, determined a priori) for males (p=0.03) and females (p=0.004). Thus models were stratified by race as well as by gender for this report.
In the SCCS, the prevalence of obesity differed significantly by health insurance status (χ2 p<0.0001). Because CRC screening is also strongly associated with insurance coverage (8), we conducted analyses limited to the 22,220 (64%) individuals with health insurance coverage to reduce the potential for residual confounding. For similar reasons, analyses stratified by categories of annual household income (<$15,000/year, $15,000–$24,999/year, and ≥ $25,000/year) were also conducted within the race and sex strata. The same set of covariates was used in these subgroup analyses as in the main race-sex stratified models.
To assess potential effects of the dependent nature of the data for participants who enrolled in the SCCS at the same CHC, models that used generalized estimating equations (GEE) to produce standard errors and test statistics were compared to models that did not use GEE. In the models that used GEE, an exchangeable working correlation structure was assumed, meaning that a common covariance was assumed among all participants enrolled in a single CHC. The standard errors produced in the models using GEE were negligibly different from those produced in models without the use of GEE, and therefore the results presented are from models that did not use GEE.
All analyses were conducted using SAS/STAT software Version 9.2 of the SAS System for Windows (SAS Institute Inc., Cary, NC).
Overall, the SCCS population is of relatively low socioeconomic status with 61% of participants reporting a household income less than $15,000 per year (63% of blacks and 55% of whites) and approximately one-third of the participants reporting less than a high school education (38% of blacks and 29% of whites). Overweight and obesity were both common in the study population with the highest prevalence among black females (86% overweight and 60% obese) and the lowest in black males (64% overweight and 29% obese). Black participants were on average younger than white participants at enrollment (mean age=57.0 v. 59.0 for males and 58.3 v. 59.3 for females).
Crude CRC screening prevalences were 30% in black males, 37% in black females, 36% in white males, and 44% in white females. Age-adjusted up-to-date CRC screening prevalences are shown in Table 1 by race, sex, and covariate categories. Higher screening prevalence was seen among older participants, and those with a college education, a household income of $50,000 or more, a recent physician visit, and a family history of colorectal cancer while lower screening prevalence was observed among those individuals reporting no health insurance, current smokers, and those reporting more than 2 alcohol drinks per day.
A dichotomous measure of obesity was not associated with decreased CRC in any of the four sex-race groups, with the odds of screening slightly higher among those with BMI over 30 kg/m2 versus 18.5–29.9 kg/m2 (Table 2). In race and sex-stratified multivariate logistic regression models examining categories of BMI, obesity classes I, II, and III were all associated with higher odds of CRC screening among black males compared with those of a healthy weight. A similar trend was observed among black women although the magnitude of the association was smaller and the confidence intervals slightly overlapped the null. Among whites males, being overweight and obese was generally associated with higher odds of CRC screening although not among men with extreme obesity. No association was observed between overweight or obesity and CRC screening among white women. Models including sigmoidoscopy or colonoscopy as individual outcome measures produced very similar results (data not shown).
Analyses limited to SCCS participants reporting any health insurance coverage showed a similar picture to the entire study population for males of both races (data not shown), but the trends for females, while not statistically significant, were somewhat different than among the SCCS population as a whole. Among insured black females, the modest trend towards increased CRC screening with increasing body size was attenuated; neither overweight nor obesity classes I, II, or III were associated with CRC screening in this group (OR [95% CI]=1.04 [0.90–1.20], 0.99 [0.86–1.14], 1.05 [0.89–1.22], and 1.03 [0.88–1.21], respectively). In contrast, among white women, those in obesity classes II and III had lower odds of CRC screening than healthy weight women (0.89 [0.72–1.10] and 0.88 [0.71–1.10]). While household income was strongly positively associated with CRC screening rates overall (Table 1), within strata of household income (<$15,000, $15,000–24,999, and $25,000+), CRC screening rates across BMI categories were similar to the overall population for most race and sex groups (data not shown). White women had a notably different pattern, however, with a CRC screening prevalence among women with household income of $25,000+ of 43% in women with extreme obesity compared with 57% in healthy weight women; these differences in screening prevalence were not seen among white women with household incomes of <$15,000 year (42% versus 38% for extreme obesity versus healthy weight) or $15,000–24,999 (40% versus 39%). In income-stratified multivariate logistic regression models, income was an effect modifier of the BMI-CRC screening relationship only among white women. For white women with income <$15,000 and $15,000–24,999 per year, extreme obesity was not associated with CRC screening (OR=1.14 [0.80–1.61] and 1.20 [0.80–1.80], respectively) but among women with income of $25,000 or more, extreme obesity was associated with reduced odds of being up-to-date for CRC screening (OR=0.66 [0.44–1.00]). Similar trends by income category were seen for obesity classes I and II among white women (data not shown).
In this large population of generally low-income black and white adults, overall we found that BMI was not associated with CRC screening among whites, and, contrary to our initial hypothesis, that increased body size was associated with small increases in screening odds among black men and women. Previous reports examining CRC screening in relation to body size have been inconsistent. Using 2001 Behavioral Risk Factor Surveillance System (BRFSS) data, obese women were found to be less likely to have received a sigmoidoscopy within the past five years while obese males were more likely to have done so (17)while among the 130,000 individuals (97% white) in the Cancer Prevention Study (CPS) II Nutrition Cohort, lower odds of ever having a colonoscopy or sigmoidoscopy were found among overweight and obese males and females (18). Another study reported that overweight and obese males and females were both slightly more likely to have received CRC screening compared to their healthy weight counterparts(19), and two found no association between CRC screening and overweight or obesity (20, 21). None of these studies presented results specifically among African-Americans. Two relatively small studies have focused solely on African American populations with one reporting non-statistically significant increases in CRC screening among obese African Americans (22), and the other reporting decreased CRC screening among obese black women but not obese black men (12).
Studies examining associations between obesity and CRC screening separately by both race and gender are rare, a gap in the literature that this report uniquely fills. In the few studies that have presented separate results by race and gender, it appears that white women may be the most susceptible to lower screening in relation to larger body size. Using 2005 National Health Interview Study (NHIS) data for 4,430 white and 690 black women, Leone et al. observed lower odds of colonoscopy among obese white women (OR for BMI ≥ 30 v. < 30 = 0.66 [0.50–0.85]) and higher odds among obese black women (OR=1.30 [0.83–2.96] (23). With our much larger population of black and white women in the SCCS, we classified women into obese versus non-obese categories, as well as finer categories of BMI. When we grouped women by dichotomous obesity (BMI ≥ 30 versus 18.5–29.9 kg/m2), the OR of 1.05 [0.97–1.14] for obese black women was similar to the OR for of 1.06 [0.96–1.18] for obese white women. However, comparing BMI ≥ 40 vs 18.5–24.9 kg/m2, the OR were 1.13 [0.98–1.29] for black and 0.99 [0.83–1.19] for white women. In addition, among the subset of white women with income >$25,000, the OR for BMI ≥ 40 vs 18.5–24.9 kg/m2 was 0.66 [0.44–1.00], while no parallel decline was seen among black women (OR=1.05 [0.73–1.49]). Hence both the SCCS and NHIS data suggest that racial differences may exist, with lower CRC screening associated with obesity only among moderate to upper income white women.
These differences by race may reflect issues related to negative body image, an issue that has been shown to be more pervasive among white women than black women (24). If white women are more self-conscious about their weight, this may prevent them from receiving certain medical procedures, particularly those that are more invasive such as sigmoidoscopy or colonoscopy. Reasons for the observed difference in screening rates between obese and non-obese individuals in higher but not lower-income women are unclear. It may be that among women of the lowest means, simply meeting the basic needs of food, shelter, and medical care when absolutely necessary is the sole focus while body size is a stronger influence among women with greater means because basic needs have been met. This is an area deserving of further inquiry.
The overall differences observed between the NHIS and CPS populations and the SCCS could likely be related to SES differences. The CPS study population is of higher educational attainment and has more health-conscious behaviors than the general population, and the NHIS is designed to be nationally representative. In contrast, the SCCS has a much higher proportion of low income, low educational attainment, and uninsured individuals than either the NHIS or CPS.
Among black men, we observed higher CRC screening rates among overweight and obese SCCS participants. It has been reported that overweight and obese individuals utilize health care more often than those of healthy weight and thus have more opportunities to discuss cancer screening with their physician (25). It is also possible that medical practitioners are recommending CRC screening more often to patients who are overweight because excess weight is a known risk factor for CRC. However, we did not observe increased CRC screening among overweight white participants and there is not reason to expect that increased medical visits or screening recommendations for obese individuals would be limited to blacks.
One limitation of this analysis was that we were able only to examine CRC screening by sigmoidoscopy or colonoscopy as information on fecal occult blood testing (FOBT) or barium enema was not obtained as part of the SCCS baseline survey. There is some evidence to indicate that FOBT is used more commonly among individuals of low-income, low education, and without health insurance (26); notably, though, the proportion of screening exams done by FOBT is relatively low and has been dropping in recent years (26). In studies that have examined endoscopy alone versus FOBT and endoscopy together in relation to BMI, associations have generally been similar for the two screening measures; ORs for endoscopy alone versus FOBT or endoscopy by categories of BMI were very similar in an analysis of 2000 NHIS data (21)as were adjusted CRC screening rates using BRFSS data (27). In contrast, using 2005 NHIS data, Leone et al. reported an association between obesity and colonoscopy but not FOBT (23). On balance, the relatively low proportion of CRC screening exams done by FOBT coupled with multiple reports that showed similar associations between obesity and CRC screening regardless of screening modality indicate that the results we found for endoscopy are likely reflective of patterns for all CRC screening.
A further limitation is that both the use of sigmoidoscopy and colonoscopy as well as body size were obtained by self-report. However, self-reported endoscopy by either sigmoidoscopy or colonoscopy had high sensitivity (>96%) and specificity (>92%) when compared with health plan records in at least one study (28). As for self-reported body size, while are cent review indicates that among women, height tends to be over-reported and weight under-reported (29), data from the 1999–2004 National Health and Examination Survey show that despite errors in self-report, BMI categories based on self-reported values still generally demonstrate good agreement with BMI categories from measured values (30). Additionally, within the SCCS, BMI values calculated from self-reported height and weight were very highly correlated with BMI values calculated from medical record data overall (Pearson correlation coefficient > 0.95) as well as across strata of race and BMI, indicating that the self-reported values are generally of good quality.
Finally, it should be noted that the SCCS population, due to its unique recruitment through Community Health Centers and the resulting high proportion of low income individuals, is not reflective of the socioeconomic or race distributions of the entire US population. Thus, differences in effects compared with other studies, particular those that are nationally representative such as the NHS, may be due to the socioeconomic status of the underlying study populations. However, this study design does not detract from our ability to make within-cohort comparisons between race groups, and is in fact a major strength of this study because residual confounding due to differences in socioeconomic status between race groups is minimized by design, and additionally because low-income individuals have traditionally been understudied.
In summary, in this large population of black and white adults, being overweight or obese does not appear to play a deterring role in the receipt of CRC screening in any race or sex group, except perhaps among the subset of upper income white women. Given that only half of adults over the age of 50 are currently being screened for colorectal cancer, efforts to increase screening rates in all individuals should remain the focus of public health initiatives.
The Southern Community Cohort Study is funded by grant R01 CA92447 from the National Cancer Institute including special allocations from the American Recovery and Reinvestment Act (3R01 CA092447-08S1).