Change in Body Size and the Risk of Colorectal Adenomas

Colon cancer is the second leading cause of cancer deaths in the United States with ∼146,940 new cancer cases and 56,730 related deaths annually (1). Colorectal adenomatous polyps are recognized as precursor lesions for the majority of invasive colon cancers (2). A substantial literature exists about the association between various body measures and the risk of colorectal adenomas (3). Both body mass index (BMI) and waist-to-hip ratio have been associated with increased risk of adenomas in most (4-14) but not all studies (15-17). Measures of body size (BMI and waist-to-hip ratio) have been associated with an increase adenoma risk in both men (4, 6, 10, 11, 15) and women (5, 12). However, studies that have investigated the association between BMI or visceral adipose tissue and recurrence of colorectal adenomas among people under endoscopic surveillance have reported null findings (18-20). This may be due to the relatively short interval between screenings (0-4 years). When adenomas were not resected, but observed over a 3-year period, higher adiposity was associated with adenoma growth (21), suggesting that adiposity may be associated with factors that promote the growth of existing colorectal adenomas.

Only a few studies have examined body size and/or weight gain at several time points before adenomas were diagnosed (5, 10, 13, 16, 22). Both increased weight gain over various times (10, 22) and weight cycling (13) have been found to be associated with increased adenoma risk, but the association between BMI near the time of screening and the risk of adenomatous polyps seems to be stronger than for BMI measured at earlier times (5, 10, 13).

Although the role for adiposity as a factor for colorectal adenoma risk has been established, many questions remain about the time of weight gain, the importance of adipose tissue distribution, and possible differences between ethnic/racial groups and gender. The majority of previous studies have recruited participants from clinic-based screening programs (4-7, 9, 12-15, 22) and have been conducted largely in white or Japanese populations. To further elucidate the role of adiposity in the development of colorectal adenomas, we assessed adenoma prevalence in relation to body size using prospective data from the multiethnic Insulin Resistance Atherosclerosis Study (IRAS) cohort in which colonoscopies were done on eligible participants regardless of symptoms.

The IRAS is a multicenter, multiethnic prospective cohort study designed to address the association between insulin resistance and atherosclerosis. Details of the design and recruitment of IRAS have been published previously (23). Briefly, individuals were recruited in 1992 to 1994 to obtain a sample of men and women of non-Hispanic White, Hispanic, and African-American ethnicity and different glucose tolerance status (normal, impaired glucose tolerance, and diabetes). The WHO criteria (24) were used to categorize participants' glucose tolerance status based on the values determined using a 2-h, 75-g p.o. glucose tolerance test (Orange-dex, Custom Laboratories, Baltimore, MD). Recruitment was conducted from four clinical centers. Two clinical centers (Los Angeles, CA and Oakland, CA) recruited non-Hispanic Whites and African-Americans from a nonprofit health maintenance organization. Two other sites used ongoing studies for recruitment of non-Hispanic Whites and Hispanics (San Luis Valley, CO and San Antonio, TX; refs. 25-27). Study visits were completed on 1,625 participants between 1992 and 1994. Approximately 5 years after the initial study visit, a second study visit was conducted (1997-1999) using a similar protocol to that used at baseline (1992-1994). The focus of the second study visit was to assess longitudinal changes in glucose tolerance status, insulin resistance, and other risk factors for atherosclerosis.

Between 2002 and 2004, all surviving IRAS cohort participants who were ≥49 years of age and mentally eligible were invited to have a screening colonoscopy, except those with serious concurrent illnesses that would substantially increase the risk from the colonoscopy procedure and/or diminish any potential benefits of the screening. Participants with a prior adenoma were included only if their next surveillance colonoscopy exam was due within the study period. Before enrollment, all participants signed informed consents and the study was approved by the Institutional Review Boards of all collaborating organizations.

Of the initial 1,625 IRAS participants examined at baseline (1992-1994), 85% (n = 1,308) of the population completed the second exam. Of the 1,226 participants remaining in the study in 2002, 999 participants were contacted for eligibility, and 197 were not eligible for the colonoscopy due to recent heart attacks, oxygen-dependent pulmonary disease, renal failure, prosthetic heart valve, dementia with incompetence, colon cancer, or recent colon screening (colonoscopy, sigmoidoscopy, or barium enema). Of the 197 participants, 162 were excluded due to a recent colon screening. Thus, 75.6% (600 of 802) of the eligible study population that was contacted received a colonoscopy. Participants that underwent a colonoscopic exam were on average younger (P < 0.01), less likely to be diabetics (P < 0.01), current smokers (P < 0.01), and to have a baseline BMI of ≥30 kg/m² (P < 0.01) than those who did not have a colonoscopy. Of the 605 participants completing a colonoscopy in this study, participants were dropped from the analyses because they had a histology that was missing or unable to be determined (n = 5). Analyses were done on 600 participants.

Interviewer-administered questionnaires were used to collect information on physical activity, diet, medical history, tobacco use, as well as other demographic factors at all study visits. At baseline (1992-1994) and the second visit (1997-1999), physical activity was measured using a 1-year recall of all physical activity based on a modification of a structured instrument, and total estimate energy expenditure was calculated (28-30). During the colonoscopy study visit (2002-2004), only the self-reported perception of change in physical activity since the previous exam was measured.

Trained staff made anthropometric measurements at each of the three study visits. To avoid weight changes due to the colon prep before the colonoscopy, anthropometric measures were made at the time of study enrollment for colonoscopy rather than on the day of the exam. Height was measured to the nearest 0.1 cm with participants in bare feet or thin socks. Weights were measured to the nearest 0.1 kg with participants wearing loose fitting shirts and shorts. Minimum waist measurements were conducted at the smallest point between the 10th rib and the iliac crest over bare skin with participants standing. An average of two minimum waist and two maximum hip measurements was conducted on each individual over bare skin. Anthropometric measures were obtained in the same way during all three IRAS study visits.

Colonoscopies were done by experienced physicians. The cecum was reached in 96.2% of participants. After the location and size of all visible polyps were recorded, the polyps were removed and standard histologic assessment was done by the local clinical laboratory.

Participants' adenoma status was based on the highest grade lesion identified. Advanced adenomatous polyps (n = 38) were defined as those polyps with villous or mixed tubulovillous features with high-grade dysplasia or >1 cm in diameter. There were no polyps with carcinomas. Nonadvanced adenomatous polyps (n = 98) were those with tubular histology under 1 cm in diameter. The “no adenoma” group included participants with no adenomatous polyps (n = 410) or only hyperplastic polyps (n = 54).

BMI was calculated as weight in kilograms divided by height in meters² (kg/m²). BMI was classified into three categories based on the WHO criteria (31): normal (<25.0 kg/m²), overweight (25.0-29.9 kg/m²), and obese (30+ kg/m²).

Tertiles of waist circumference were categorized separately for men and women based on the distribution of participants' circumferences without an adenoma at the first IRAS exam (1992-1994) to determine the cutpoints for the categories in a nondiseased population. Categories of weight change were created based on two times: weight change occurring in the ∼5-year period before the colonoscopy (between the second study visit and the colonoscopy visit) and weight change occurring over the ∼10-year period before the colonoscopy visit (between the first study visit and the colonoscopy visit). Over both times, weight change was categorized into lost >4 pounds, maintained weight (−4 to 4 pounds), and gained >4 pounds.

Univariate analyses were conducted to assess differences between participants with and without an adenoma for factors of age, sex, study center, race/ethnicity, glucose tolerance status at baseline, smoking status at the time of colonoscopy, physical activity, and self-reported previous polyp removal. χ² analyses were used for categorical variables and two-sample t tests were used for continuous variables. Multivariate analyses were done to obtain estimates of the adjusted odds ratio (OR_adj) and 95% confidence intervals (95% CI) for the separate associations between body size measures (BMI, waist circumference, and weight change) and any adenoma, advanced adenomas, and nonadvanced adenomas. In addition to the body size variables, variables used for recruitment (race/ethnicity and study center), variables that were significant in the univariate analysis (age and previous polyp), and those variables previously shown in the literature to be potential confounders (physical activity and smoking; refs. 5, 17, 18) were included in the multivariate models. The multivariate model of weight change also included baseline BMI. Although glucose tolerance status was a criterion for participant recruitment, this variable is believed to be an intermediate variable in the pathway of obesity and colorectal adenomas and, therefore, was not included in the reported models. Race/ethnicity, study center, previous polyp, and smoking status at colonoscopy were modeled as categorical variables, whereas age and total energy expenditure were continuous variables. Due to the small number of current smokers with an adenoma (n = 12), a smoking variable of never versus ever was used for adjustment in the multivariate models. As detailed measure of physical activity was collected at the first and second visit but not at the time of colonoscopy, we used the physical activity measured at the second visit to adjust for potential confounding. In the multivariate models, missing values for previous polyp history were coded as no previous polyps (n = 2); missing values for total estimated energy expenditure at the second visit were coded using the values from the first visit (n = 29).

To explore the associations between adiposity and colorectal adenomas by the recruitment variables of age, sex, race/ethnicity, and diabetes status (defined as the most severe diagnosis at either of the first two study visits), stratified analyses by each of these variables were done. As a significant difference was observed between sex and any adenoma in the univariate analysis and differences were observed in the risk estimates when stratified by sex, interaction terms for BMI × sex, waist circumference × sex, and weight change × sex were assessed in multivariate logistic models. Because none of the interaction terms were statistically significant (P > 0.10), these interaction terms were not included in the final models. P values for linear trend were calculated by treating categorical measures of body weight and size as continuous variables in multivariate logistic models. Two-sided statistical tests with an α level of 0.05 were used on all analyses except for assessing interaction terms where an α level of 0.10 was used. All statistical tests were done using Intercooled Stata version 9 (Stata Statistical Software: Release 9, StataCorp LP, College Station, TX).

Among the 600 participants in this study, the mean age was 64 years (SD, 8.0). The majority were women (54%) who were overweight (44%) or obese (39%) with normal glucose tolerance status at baseline (56%). The ethnic distribution of the cohort examined was 38.3% non-Hispanic White, 34.8% Hispanic, and 26.8% African-American. Age, sex, study center, and previous polyps all differed between participants with and those without any colorectal adenomas in the univariate analyses (Table 1).

After adjustment of potential confounders, obesity was positively associated with risk of colorectal adenomas at the time of colonoscopy (OR_adj, 2.16; 95% CI, 1.13-4.14; Table 2). Similar trends were observed at ∼5 years before colonoscopy (OR_adj, 1.59; 95% CI, 0.86-2.93) and ∼10 years before colonoscopy (OR_adj, 1.54; 95% CI, 0.87-2.71), but these associations were not statistically significant. Although there is a suggestion of a modestly increased risk in the highest versus the lowest tertile of waist circumference at colonoscopy (OR_adj, 1.49; 95% CI, 0.84-2.64), no statistically significant association was observed between waist circumference and colorectal adenoma risk. No associations were observed between waist-to-hip ratio and the risk of any colorectal adenoma (data not shown). An increase colorectal adenoma risk was observed between those participants who gained >4 pounds compared with those who maintained weight (−4 to 4 pounds) over the ∼5 years (OR_adj, 2.17; 95% CI, 1.15-4.11) and ∼10 years (OR_adj, 2.09; 95% CI, 1.19-3.66) before colonoscopy.

To further investigate the relationship of adiposity and colorectal adenomas, analyses were stratified by sex, age, race/ethnicity, and glucose tolerance status (Table 3). There was a stronger association between BMI at colonoscopy and colorectal adenomas in women (OR_adj, 4.42; 95% CI, 1.53-12.78) than in men (OR_adj, 1.26; 95% CI, 0.52-3.07). A 1.5- to 3.4-fold increase in colorectal adenoma risk was observed within the three age categories among obese compared with normal BMI participants. Among participants who were 60 to 69 years of age, an increased colorectal adenoma risk was observed for obese (OR_adj, 3.36; 95% CI, 1.03-11.02) participants compared with participants with normal BMIs. Similar ∼2-fold increase in risk among obese compared with normal BMI participants was observed among race/ethnicity and glucose tolerance categories with only a significant association observed for non-Hispanic White obese (OR_adj, 2.82; 95% CI, 1.04-7.62).

Analyses of any colorectal adenoma were then stratified by advanced (n = 38) versus nonadvanced adenomas (n = 98) and both were compared with participants without an adenoma (n = 464). At the time of colonoscopy, there was an ∼2-fold increase in risk of both advanced and nonadvanced adenoma among obese participants compared with those participants with normal BMIs (Table 4). Larger waist circumference at colonoscopy was associated with a 2.3-fold increase in risk of advanced adenomas (P for trend = 0.26) in the adjusted model but only a 1.3-fold increase in the risk of nonadvanced adenomas (P for trend = 0.11). Weight gain of >4 pounds was associated with a nonsignificant ∼2-fold increase in advanced and nonadvanced adenoma risk.

Our findings from this multiethnic population of a positive association between adiposity and risk of colorectal adenoma are consistent with previous studies of both diagnosed and prevalent colorectal adenomas and BMI (5-7, 9-13, 15), waist-to-hip ratio (4-6, 11), waist circumference (4, 5), and total weight (15, 21). Among the four studies with female-specific estimates (5, 6, 12, 15), two studies (5, 12) reported increased risk of colorectal adenomas with higher adiposity, consistent with our findings. Whereas previous data on colorectal cancer have suggested that this association with adiposity is stronger among men than women (3), our data on colorectal adenomas showed a stronger association among women than men (OR_adj for women, 4.42; 95% CI, 1.53-12.78 and OR_adj for men, 1.26; 95% CI, 0.52-3.07). Comparable findings with no association in men and an increased association in women were observed by Neugut et al. (12) using data collected on participants undergoing colonoscopies in three practices in New York City.

Similar to our findings, previous studies have found that weight gain (10, 21) or weight cycling (13) is associated with increased colorectal adenoma risk. Interestingly, we did not observe a significant association with change in waist circumference or waist-to-hip measures and the risk of colorectal adenomas (data not shown). The lack of an association may be due to measurement error inherent in the measurement of waist circumference over time. Furthermore, our results and others (5, 10, 12) reported that weight at the time of colonoscopy was more strongly associated with the risk of colorectal adenomas compared with times more distant, suggesting that adiposity somehow is involved in the progression of colorectal neoplasia.

The role of adiposity in the development of carcinogenesis of the colorectum is complex and not fully elucidated. Given that adipose tissue produces numerous secretory factors involved in metabolism, inflammation, sexual hormonal production, and immunity, adipose tissue could have many roles in the carcinogenesis pathway (32, 33). Although obesity increases the risk of colorectal adenomas and cancer, it also is a major determinate of both insulin sensitivity (34) and insulin-like growth factor-I (35).

Lifestyle patterns, such as high intakes of dietary fat and energy, low intake of complex carbohydrates, and low levels of physical activity, are modest risk factors of colorectal neoplasia and insulin resistance (36-38). These same risk factors are important in the development of the “metabolic syndrome” consisting of obesity, type 2 diabetes, hyperlipidemia, and hypertension (39). A current hypothesis that combines these factors together suggests that the development of colon neoplasia is through insulin resistance and insulin-like growth factors (36-38). Prospective studies have reported increased risk of colon cancer associated with type 2 diabetes in both men (40) and women (41) after adjustment for several factors, including BMI and physical activity. Additionally, in the Physician's Health Study, men with higher baseline plasma levels of insulin-like growth factor-I, which is involved in the inhibition of apoptosis, had a 2.5 times higher risk of colon cancer compared with men in the lowest quintile of insulin-like growth factor-I levels measured ∼12 years before diagnosis (42). As obesity influences insulin status, one of the many mechanisms by which obesity may increase the risk of colorectal adenoma and carcinogenesis is through its role on insulin and insulin-like growth factor. Noteworthy in our study, when we included the participants' baseline levels of glucose tolerance in the models along with the other covariates, the findings did not change substantially (data not shown). Although this observation is likely to be due to measurement error inherent to the use of a single glucose tolerance test, it may also suggest that adipose tissue contributes to other pathways in the progression of normal to malignant tissue independent of the insulin pathway.

Several strengths should be noted when reviewing these data. This study was conducted prospectively over an ∼10-year period. Anthropometric and risk factor data from this study were collected prospectively by trained interviewers at multiple time points on a racially and ethnically diverse population of men and women with various glucose tolerances. Unlike many previous studies that have used clinic-based screening programs to recruit participants (4-7, 9, 12-15, 22), colonoscopies from the IRAS cohort were done independent of symptoms on all participants meeting the eligibility criteria and agreeing to participation.

All epidemiologic investigations have inherent limitations. The number of cohort participants with advanced adenoma polyps was relatively small, and thus, we were unable to fully examine the association of adiposity with colorectal adenoma risk specifically for advanced adenomas. Nevertheless, our results and others (7) have reported that advanced adenomas were similar to those observed for nonadvanced adenomas. Stratified analyses of age, sex, race/ethnicity, and glucose tolerance status resulted in small numbers, thus limiting our ability to fully explore these associations. In addition, 76% of our cohort was overweight or obese at the first study visit (1992-1994) and few participants lost weight over the 10-year period. Finally, BMI and waist circumference are only surrogate measures for total body fat and visceral fat. Whereas adipose tissue cannot be distinguished from lean body mass and, therefore, adipose tissue alone and its distributions within the body cannot be fully measured, BMI and waist circumference are highly correlated (43).

In conclusion, findings from this study support a role for adiposity and weight gain in increasing risk for colorectal adenoma. This association is stronger for measures of adiposity taken closer to the time of colon screening and in women. With a growing population of overweight and obese adults, a better understanding of the mechanism by which adiposity influences the progression of colorectal neoplasia is needed.