Calcium, dietary, and lifestyle factors in the prevention of colorectal adenomas
Many studies have suggested a role for calcium in reducing the risk of colorectal adenomas and cancer but its effectiveness may be dependent on interactions with other dietary and/or lifestyle factors. We examined the association between calcium and prevalence of adenomas and assessed whether the association was stronger in biologically plausible subgroups.
Cross-sectional data from 222 cases and 479 adenoma-free controls who underwent colonoscopies and completed food frequency and lifestyle questionnaires were used in the analyses. Multivariable logistic regression was used to estimate the association between calcium and prevalence of adenomas. Stratified analyses and the likelihood ratio test were used to examine effect modification by various demographic, lifestyle, and behavioral factors.
Overall, little association was observed comparing total calcium intake of ≥900 mg/day to <500 mg/day (adjusted odds ratio [OR] = 0.85, 95% confidence interval [CI]: 0.53–1.37). However, stronger associations were observed in patients with lower fat intake and in those who regularly (≥15 times/month) took nonsteroidal antiinflammatory drugs (NSAIDs). Specifically, total calcium intake of ≥900 mg/day was associated with a lower prevalence of adenomas among patients with lower fat intake (OR = 0.47, 95% CI: 0.25–0.91) but not among those with higher fat intake (OR = 1.20, 95% CI: 0.61–2.35; P-value for interaction = .01). For NSAIDs, the associations were OR = 0.37 (95% CI: 0.16–0.86) for regular NSAID users and OR = 1.27 (95% CI: 0.73–2.22) with infrequent or nonuse of NSAIDs, respectively (P = .06).
The data suggest that a lower-fat diet and regular NSAID use may enhance calcium's effectiveness as a colorectal cancer preventive agent. Cancer 2007 © 2007 American Cancer Society.
There is considerable evidence suggesting a role for calcium in reducing the risk of colorectal cancer and its precursor, colorectal adenomas. Calcium supplementation has been shown to reduce the number and incidence of colon tumors in a number of different animal studies.1–5 However, the most promising evidence in humans comes from 2 randomized controlled trials of calcium supplementation to prevent recurrent adenomas in patients previously found to have adenomas during colonoscopy. In those studies, patients taking 1200 mg and 2000 mg calcium supplements were, respectively, 19% (relative risk [RR] = 0.81, 95% confidence interval [CI]: 0.67–0.99) and 44% (RR = 0.66, 95% CI: 0.38–1.17) less likely to develop recurrent adenomas compared with patients taking placebo.6, 7
There are a number of factors that interact with calcium or affect calcium absorption and homeostasis that could plausibly influence theeffectiveness of calcium in the prevention of colorectal adenomas or cancer. For example, previous studies have suggested stronger protective associations with calcium might be found among persons with high fat intake8, 9; however, the results have not been consistent.10, 11 Nonsteroidal antiinflammatory drugs (NSAIDs) have been shown to be associated with a reduced risk of colorectal adenomas and cancer in clinical trials12, 13 and might potentially interact with calcium through similar biologic pathways.14–16 Vitamin D and physical activity affect calcium absorption and homeostasis,17–19 which could modify risk, in part, by affecting the amount of free calcium available within the lumen or cells of the colon.
In this report we used data from a hospital-based cross-sectional study to examine the association between calcium and prevalence of colorectal adenomas while assessing effect modification by fat intake, use of NSAIDs, serum vitamin D, and physical activity in addition to dietary fiber intake, body mass index (BMI), age, race/ethnicity, sex, and family history of colorectal cancer. Further examination of the interaction between calcium and other interrelated factors could help elucidate calcium's potential as a preventive agent.
MATERIALS AND METHODS
The Diet and Health Study III is a cross-sectional study designed to examine potential risk and etiologic factors of colorectal adenomas. Extensive details of the study design, methods, and population have been previously published.20 Briefly, the study recruited consecutive patients from University of North Carolina Hospitals in Chapel Hill who underwent colonoscopies for diagnostic or screening purposes between August 1, 1998, and March 4, 2000. Participants had to be 30 years of age or older, proficient in the English language, and without colitis, previous colonic resection, previous colon cancer or adenoma, and history of familial polyposis. During the study period, 2452 colonoscopies were performed. Based on the eligibility criteria, 926 (38%) of the patients were eligible for the study. Of the eligible patients, 803 (93.4%) consented to participate and 701 completed the necessary questionnaire sections for this analysis.
Dietary and Lifestyle Assessment
Consenting patients were contacted by telephone by a trained interviewer within 12 weeks of the colonoscopy to complete lifestyle and diet questionnaires. The interviewer was blinded to the participant's case status. Lifestyle-related questions included education, smoking history, marital status, occupation, physical activity, use of NSAIDs, bowel habits, family history of cancer, and medical history.
Participants' usual diet during the year before the colonoscopy was assessed with a modified version of the Block food frequency questionnaire.21 The Block questionnaire inquires about foods, portion sizes, and usual frequency for 100 different foods. In the modified version, 29 foods were added to capture foods that are more commonly consumed in North Carolina.22 Information regarding dose (eg, 100, 250, 500, 750, 1000 mg), frequency (days/week), and duration (in years) of supplement use was collected for multivitamins and a number of specific vitamin and mineral supplements, including calcium. Estimated daily supplemental and dietary intakes (mg/day) were calculated by the Block analysis program.
Serum 25-Hydroxyvitamin D Assay
Because there is good indication that vitamin D may be protective against colorectal adenomas and cancer and few studies have used serum measurement, we measured serum for 25-hydroxyvitamin D to examine the association with adenomas and possible interaction with calcium. Fasting blood samples were collected from participants at the time of colonoscopy. Samples were kept at 4°C and serum was separated within 2 hours and frozen at −80°C. Serum 25-hydroxyvitamin D was measured in a subset of patients in which serum was available (adenoma cases n = 111 [51%], adenoma-free controls n = 238 [50%]) using an enzyme immunoassay (ALPCO Diagnostics, Salem, NH). Each sample was assayed in duplicate and the level assigned to each patient was the average of the 2. The intra- and interassay coefficients of variation were 5.3% and 4.3%, respectively. Distribution of demographic characteristics and dietary intakes were not statistically different in the subpopulation with serum 25-hydroxyvitamin D measurement except for race, in which non-Hispanic blacks were slightly less likely to have available serum (24.8% vs 18.1%, P = .03).
All analyses were conducted using SAS v. 8.2 (SAS Institute, Cary, NC). For the purpose of these analyses, patients with at least 1 adenoma were considered cases and patients without were considered controls. Univariate analyses were used to assess the distribution of variables, missing data, and potential outliers. Data points that were more than 4 standard deviations (SD) from the mean were considered outliers. Removal of outliers from variables included in the final models (cases n = 4, controls n = 7) did not appreciably change the point estimates but did increase precision, so only the results excluding outliers are presented.
The variables hypothesized to modify the association between calcium and adenomas were serum 25-hydroxyvitamin D (above vs below median ng/L), NSAID use (regular users, defined as ≥15 times per month in the past 5 years vs <15 times per month), physical activity (above vs below median MET-minutes per day of combined occupational and recreational physical activity),23–25 age (31–50, 51–60, 61–70, >70 years), race (non-Hispanic white, non-Hispanic black), sex, body mass index (BMI, ≥30 kg/m2 vs <30 kg/m2), first-degree family history of colon cancer (yes, no), dietary fiber intake (above vs below median mg/day), and percent energy from dietary fat intake (above vs below median). Percent energy from fat intake was used instead of total fat intake because total energy intake was highly correlated with total fat intake (r = 0.89).
The prevalence odds ratio for calcium intake and presence of adenomas was calculated using unadjusted and multivariable logistic regression. Bivariate associations were first examined stratified by the a priori chosen variables listed above. The Breslow-Day test of homogeneity (α = 0.20) or a 100% difference in stratum-specific odds ratios were used as evidence of effect modification. Variables that showed evidence of effect modification in the stratified analysis were further analyzed with interaction terms in multivariable logistic regression models. Models with interaction terms were analyzed separately. The likelihood ratio test (LRT) and 100% difference in estimates were used as criteria in modeling. Because of the low power of the LRT to assess effect modification, alpha was set at α = 0.10,26 meaning P-values for testing heterogeneity ≤.10 were considered statistically significant if stratified estimates were ≥100% different. Similar methods were used to examine possible effect modification of the association between serum 25-hydroxyvitamin D and adenomas by calcium intake (above vs below the median mg/day).
Covariates assessed as confounders were included in the multivariable models if they changed stratum-specific odds ratios by ≥10%. The variables assessed as potential confounders in this analysis were age, race/ethnicity, sex, BMI, physical activity, family history of colon cancer, smoking status, use of NSAIDs, total energy intake, total vitamin C intake, total fiber intake, and percent energy from fat. Because exposure to sunlight strongly influences the endogenous production of vitamin D, all models that included serum 25-hydroxyvitamin D were adjusted for the time of year that the blood was drawn (February-May, June-September, October-January).
Calcium intake was examined as dietary only and total (diet plus supplements). Calcium intake in this population was lower than that reported in much of the previous literature. Therefore, set cut-points were chosen, using less than 500 mg/day as the reference category, to make our study more comparable. Total and dietary calcium intakes are presented as <500 mg/day, 500–900 mg/day, 900+ mg/day. Covariates determined to be confounders for either of the calcium definitions were included in both models for consistency.
The mean age of patients in the study population was 56.2 years (SD = 11.3), 40% were men, and 20% were non-Hispanic black. The presence of at least 1 adenoma was detected in 31.6% of the patients. Male sex and older age were associated with a higher prevalence of adenomas (Table 1). There were higher mean dietary and total calcium intakes, as well as a higher mean level of 25-hydroxyvitamin D, among controls compared with cases, which were all borderline statistically significant.
Table 1. Comparison of Study Population Characteristics by Adenoma Case Status
| Women||47.2||66.0|| |
| Non-Hispanic white||71.4||76.7||.44|
| Non-Hispanic black||21.2||19.2|| |
| Other||2.9||1.7|| |
|Mean age (SD)||60.1 (10.6)||54.4 (11.1)||<.001|
|Family history of colon cancer, %||23.4||29.8||.08|
|Ever smoker, >100 cigarettes, %||56.0||49.9||.14|
|Dietary calcium intake mg/day (SD)||590.7 (301.2)||639.9 (358.0)||.06|
|Total calcium intake mg/day(SD)||854.4 (517.4)||944.8 (628.3)||.05|
|25-(OH)vitamin D level ng/L (SD)*||27.5 (15.9)||31.4 (20.2)||.05|
|Regular NSAID use, ≥15 times/month, %||20.9||25.2||.22|
|Percentage energy from fat||35.1||34.2||.13|
|Total energy intake, kcal/day (SD)||1510.1 (587.3)||1500.0 (587.3)||.83|
|Dietary fiber intake, mg/day (SD)||13.2 (5.8)||13.3 (6.4)||.88|
|Total vitamin C intake, mg/day (SD)||216.8 (280.3)||257.6 (366.7)||.11|
|Physical activity, mets/day||2361.5 (675.5)||2422.4 (708.0)||.29|
|Body mass index kg/m2 (SD)||28.3 (6.2)||27.6 (6.4)||.17|
Higher dietary and total calcium intakes were associated with a small nonstatistically significant lower prevalence of adenomas (Table 2). There was no apparent effect modification by 25-hydroxyvitamin D, physical activity, BMI, age, race/ethnicity, sex, or family history of colon cancer (Table 3). There was some indication of a stronger association between calcium and decreased prevalence of adenomas among persons with a BMI <30 kg/m2 in unadjusted models; however, after adjusting for covariates, the interaction was not statistically significant (P for interaction: dietary calcium P = .67, total calcium P = .24). Effect modification of the association between total calcium intake and adenomas by dietary fiber was borderline statistically significant (P = .10); however, the pattern differed from dietary calcium, where there was no evidence of effect modification (P = .94). The strongest evidence of effect modification of the association between calcium and prevalence of adenomas was by fat intake and regular use of NSAIDs.
Table 2. Association Between Calcium Intake and Prevalence of Adenomas
| <500 mg/day||66||138||1.0|
| 500–900||74||141||0.99 (0.62–1.61)|
| >900||78||194||0.85 (0.53–1.37)|
| <500 mg/day||102||193||1.0|
| 500–900||84||188||0.71 (0.46–1.10)|
| >900||32||92||0.55 (0.30–1.01)|
Table 3. Prevalence Odds Ratios (OR) and Corresponding 95% Confidence Intervals (CI) Between Levels of Calcium Intake and Prevalence of Adenomas Stratified by A Priori Chosen Variables and Tested for Interaction With the Likelihood Ratio Test
|Percentage fat intake*|
| ≥34.3%||0.94 (0.52–1.72)||1.12 (0.46–2.71)||.07||1.03 (0.53–1.98)||1.20 (0.61–2.35)||.01|
| <34.3%||0.54 (0.29–0.99)||0.38 (0.18–0.83)|| ||0.62 (0.31–1.25)||0.47 (0.25–0.91)|| |
| ≥15 times/mo||0.31 (0.15–0.65)||0.25 (0.07–0.87)||.27||0.56 (0.26–1.23)||0.37 (0.16–0.86)||.06|
| <15 times/mo||0.87 (0.53–1.44)||0.70 (0.36–1.36)|| ||1.28 (0.73–2.25)||1.27 (0.73–2.22)|| |
| ≥27.6 ng/L||0.58 (0.22–1.54)||0.56 (0.18–1.68)||.54||1.12 (0.32–3.93)||0.98 (0.30–3.20)||.86|
| <27.6 ng/L||1.12 (0.51–2.44)||0.70 (0.23–2.18)|| ||1.59 (0.68–3.73)||1.42 (0.60–3.40)|| |
|Dietary fiber intake*|
| ≥12.2 mg/day||0.71 (0.36–1.43)||0.53 (0.24–1.18)||.94||0.93 (0.38–2.25)||0.58 (0.24–1.40)||.10|
| <12.2 mg/day||0.69 (0.38–1.24)||0.63 (0.22–1.76)|| ||0.83 (0.44–1.56)||1.25 (0.68–2.28)|| |
| ≥2179.3 mets/day||1.02 (0.61–1.71)||0.73 (0.38–1.42)||.65||1.12 (0.62–2.04)||0.82 (0.45–1.49)||.94|
| <2179.3 mets/day||0.73 (0.44–1.20)||0.59 (0.30–1.18)|| ||1.12 (0.63–2.00)||0.92 (0.54–1.58)|| |
|Body mass index†|
| ≥30kg/m2||0.88 (0.43–1.77)||0.91 (0.34–2.41)||.67||1.46 (0.67–3.15)||1.54 (0.73–3.28)||.24|
| <30 kg/m2||0.71 (0.42–1.20)||0.56 (0.28–1.13)|| ||0.84 (0.47–1.52)||0.73 (0.40–1.31)|| |
| Yes||0.53 (0.27–1.04)||0.91 (0.37–2.20)||.13||0.80 (0.38–1.70)||0.97 (0.49–1.92)||.46|
| No||0.90 (0.55–1.47)||0.57 (0.28–1.13)|| ||1.18 (0.69–2.03)||0.98 (0.57–1.69)|| |
Increasing calcium intake was associated with a statistically significant lower prevalence of adenomas in patients with a lower-fat diet, whereas there was no association or even a slightly higher prevalence of adenomas in patients with fat intake above the median. Inclusion of an interaction term for calcium and fat intake in a multivariable model was statistically significant (dietary calcium P = .07, total calcium P = .01).
The pattern was similar for increasing calcium intake and use of NSAIDs (Table 3). High dietary and total calcium intake were associated with a lower prevalence of adenomas in patients who regularly used NSAIDs but there was little association in non- and infrequent NSAID users. However, effect modification was statistically significant for total calcium only (dietary calcium P = .27, total calcium P = .06).
Even with the smaller sample size, there was a statistically significant association between the highest tertile of serum 25-hydroxyvitamin D and lower prevalence of adenomas (Table 4). There was also some evidence that calcium modified vitamin D's association with prevalence of adenomas. High serum levels of 25-hydroxyvitamin D among patients with dietary calcium intake above the median was associated with a 68% lower prevalence of adenomas, whereas there was no association for high 25-hydroxyvitamin D levels among those with low calcium intake (P = .30).
Table 4. Association Between Tertiles of 25-Hydroxyvitamin D Levels and Prevalence of Adenomas Adjusted for Age, Sex, Race, and Month of Blood Draw (February-May, June-September, October-January), Presented Overall and Stratified by Median Calcium Intake
|Overall||0.74 (0.40–1.38)||0.51 (0.27–0.98)|| |
|Dietary calcium intake|
| ≤586mg/day||0.89 (0.39–2.01)||0.90 (0.36–2.25)||.30|
| >586mg/day||0.56 (0.22–1.43)||0.32 (0.13–0.82)|| |
|Total calcium intake|
| ≤711 mg/day||1.06 (0.45–2.50)||0.80 (0.28–2.28)||.27|
| >711 mg/day||0.45 (0.19–1.10)||0.31 (0.13–0.74)|| |
We examined data collected from patients who underwent colonoscopies to estimate the association between intake of calcium and prevalence of adenomas and assessed whether the association was modified by a number of a priori chosen exposure variables. Although not statistically significant, the modest overall reduction in prevalence with increasing dietary calcium intake was comparable to previous studies that found an association.9, 27–29 The strongest evidence of effect modification was for lower fat intake and use of NSAIDs.
Because one of the hypothesized mechanisms by which calcium reduces the risk of colorectal cancer involves binding and neutralizing bile acids,30, 31 and free fatty acids and bile acids are more likely to reach the colon with a high-fat Western-style diet,32 we expected calcium to have a stronger effect on adenoma risk among those with higher fat intakes.8, 33 On the contrary, we observed a stronger inverse association between higher calcium intake and prevalence of adenomas in those who had lower fat intakes. Whereas calcium supplementation has been effective in reducing the tumor load and incidence in mice fed high fat / low calcium Western diets,8, 33 anticarcinogenic effects have been observed even with a low-fat diet.5 Additionally, calcium has been shown to directly increase apoptosis in the colorectal epithelium in mice.34 Therefore, our results suggest that calcium may reduce the risk of colorectal adenomas through other cellular functions, independent of its function neutralizing fatty and bile acids.
Alternatively, high fat intake was associated with an increased prevalence of adenomas in this study (OR = 1.44, 95% CI: 1.05–1.99 for fat intake above vs below median and adjusted for total energy intake), and over half of the study population reported consuming greater than 30% of total energy from fat, the upper limit recommended by the American Dietetic Association.35 In addition, the mean intake of dietary fat in the total study population was 59 g/day, and 72 g/day among patients with a percent fat intake above the median. Therefore, another explanation for our results may be that the calcium levels observed in this study were not enough to compensate for the amount of fatty or bile acids that reached the colon with high fat intake.
Studies investigating modification of the association between calcium and colorectal adenomas or cancer by fat intake have been inconsistent. In a randomized clinical trial that found supplemental antioxidants did not prevent recurrent adenomas, the inverse association between dietary calcium and risk of recurrent adenomas was stronger in the high-fat intake group. Conversely, the effect of supplemental calcium was stronger in the low fat intake group.9 A case-control study examining risk factors for colorectal cancer found results similar to ours, in which the association between calcium and risk of cancer appeared stronger in those with lower fat intake.10 Other studies reported no evidence of interaction.21, 36, 37
We also observed a stronger inverse association between calcium intake and prevalence of adenomas in patients who reported regular use of NSAIDs. Calcium and NSAIDs could hypothetically function at different points on the same biologic pathway. For example, secondary bile acids have been shown to activate the transcription of COX-2 in esophageal adenocarcinoma cells14 and increase downstream products of COX-2 in the colon.15 Calcium could therefore reduce the risk of colorectal cancer by indirectly inhibiting COX-2 expression through binding secondary bile acids. NSAIDs have proven to be an important chemopreventive agent of colorectal cancer12, 13 and one proposed mechanism is through direct inhibition of COX-2.16 By acting on different points of the same pathway, the combination of high calcium intake and regular use of NSAIDs could potentially be more effective than each alone.
Other studies examining the potential interaction between calcium and NSAIDs have been conflicting. Our results were consistent with one study38 but in direct contrast to another,39 where the association of dietary calcium was limited to patients not taking NSAIDs. However, that study did find supplemental calcium to be more strongly associated with risk of adenomas in those taking NSAIDs at least once a week. Another study found no evidence of an interaction.21 More laboratory-based data are needed to elucidate whether a true biologic interaction exists.
We found a modest inverse association with high serum vitamin D levels and prevalence of adenomas, which is consistent with two previous studies that used serum measurement of 25-hydroxyvitamin D40, 41 but not another.42 However, whereas stronger inverse associations were observed in patients with high calcium intake in two studies,40, 42 another study found the association limited to patients with low calcium intake.41 Although our results are more consistent with a stronger association for those with higher calcium intake, unfortunately, with our small sample of patients with serum vitamin D measurements, our results provide only limited evidence toward resolving this issue.
Strengths of this study include use of colonoscopy and pathological evaluation of polyps to classify patients to adenoma status, recruitment of consecutive patients, and extensive data on reported risk factors for colon cancer and adenomas. Some past studies of the association between calcium and risk of colorectal adenomas were conducted using sigmoidoscopy only,41, 43 or colonoscopy and sigmoidoscopy.40, 44 Other studies used control populations that were never examined for the presence of adenomas.45–47 Because the presence of adenomas has been found to be as high as 37% in some studies,48 this is not an appropriate control population. Use of colonoscopy to classify adenoma case-status reduces the likelihood of misclassification and allowed us to look at the associations stratified by location.
Although substantial evidence suggests that calcium plays a role in the reduction of colorectal cancer risk, its mechanisms and interactions are still not well understood. This study provides additional evidence that calcium might reduce the risk of colorectal adenomas and that its effectiveness might depend on fat intake and use of NSAIDs. The influence of micronutrients on carcinogenesis does not occur in isolation, and consideration of dietary and lifestyle contexts may be necessary to create effective prevention strategies. Likewise, the true effectiveness of calcium in cancer prevention may only be realized by developing more comprehensive interventions in which calcium is part of a larger ensemble of agents.
We thank Robert Millikan and Andrew Olshan for valuable comments and suggestions on the research and article