Relationships between Serum and Colon Concentrations of Carotenoids and Fatty Acids in Randomized Dietary Intervention Trial

The relative impact of diet on colorectal cancer prevention has been equivocal in dietary intervention studies. Cohort studies have not found much association between dietary carotenoids and colorectal cancer risk, but in the American Association of Retired Persons study, increased fruit and vegetable intakes during adolescence or 10 years before diagnosis were protective of colorectal cancer (1, 2). In the Polyp Prevention Trial, the intervention diet that targeted decreased fat intakes and increased intakes of fiber, fruits, and vegetables had no effect on polyp recurrence and the increase in plasma carotenoids was modest (3). Subjects with excellent compliance, however, did have decreased polyp recurrence and they also exhibited better dietary intakes at baseline (4). In addition, relatively higher concentrations of α-carotene and vitamin A at baseline or averaged over time were protective in that study (5). Similar findings have been observed in the Women's Health Initiative: women with higher serum β-carotene concentrations averaged over ime had lower risk of colon cancer (6). These results indicate that long-term exposures to fruits and vegetables may be necessary for prevention and that blood concentrations are important to measure. Blood concentrations can reflect carotenoid absorption and metabolism in addition to dietary exposures.

A review of the scientific literature completed in 2011 by the American Institute for Cancer Research indicated that there is convincing evidence that colon cancer risk can be affected by fiber-containing foods, red, and processed meat intake (7, 8). The Healthy People 2010 dietary recommendations that were developed by the U.S. Office of Disease Prevention and Health Promotion are consistent with diets that could be useful for colon cancer prevention by encouraging 5 fruits and vegetables/day, whole grains, and decreased saturated fat intakes (9). Mediterranean diets also may be useful for prevention because all the major components of the traditional Greek–Mediterranean diet seem to be protective for colorectal cancer, including olive oil, fish, legumes, whole grains, and fruits and vegetables (10, 11). Epidemiologic data in European populations have found Mediterranean diets to be preventive of colorectal cancer (12). Mediterranean diet scores have generally been used in epidemiologic studies, and the scores are calculated on the basis of relative intakes within the population being studied. In the United states, Mediterranean eating scores have not been consistently preventive and may stem from difference in mean intakes in the United States versus Mediterranean countries (12–14).

In this present study, we sought to evaluate the relative effects of counseling for Mediterranean versus Healthy diets on serum and colon concentrations of micronutrients in persons at increased colon cancer risk. This randomized clinical trial provided 6 months of dietary counseling to each study participant. The primary endpoints of the trial were colon carotenoid and fatty acid concentrations. Secondary endpoints were colonic eicosanoids, epithelial proliferation, and nuclear morphology. This present report is focused on the primary endpoints of the trial.

Blood and colon mucosa samples were obtained before and after dietary change. Although serum concentrations of carotenoids have been shown to be useful markers of fruit and vegetables intakes, there is relatively much less information available on colon concentrations of carotenoids. In addition, it was important to evaluate the effects of dietary change on fatty acids concentrations. Increased omega 3 or fish oil fatty acids and conversely decreased omega 6 fatty acids have been associated with decreased colon cancer in experimental models and humans (15, 16). Increased fruits, vegetables, and omega 3 fats and decreased omega 6 fats could work together to suppress colonic inflammation via fatty acid substrate competition for COX enzymes and inhibition of COXs by phytochemicals from plant-based foods (17). It was therefore important to evaluate the impact that Mediterranean and Healthy diets could have on concentrations of colonic fatty acids and carotenoids.

The Healthy Eating for Colon Cancer Prevention Study was approved by the University of Michigan (Ann Arbor, MI) Institutional Review Board and was registered at the ClinicalTrials.org (NCT00475722). The study recruited subjects at least 21 years of age who were in good health, had body mass index (BMI) between 18.5 and 35 kg/m², and had a dietary pattern that is noncompliant to the United States Department of Agriculture (USDA) recommended guidelines with respect to fat intake as well as fruit and vegetable servings. Eligible subjects had family history of colon cancer in one first-degree or 2 second-degree relatives. Participants were recruited from Ann Arbor, Michigan and vicinity between July 2007 and November 2010 primarily through advertisements in media and flyers. Details of study eligibility and recruitment have been published (18).

Enrolled subjects were randomized to one of 2 dietary regimes: healthy eating or Mediterranean diet. Subjects in each arm received the same frequency of counseling for 6 months. Each subject had 3 study visits during the study period, at baseline, 3, and 6 months, each of which consisted of filling out questionnaires, collection of blood samples, and colon biopsies by flexible sigmoidoscopy as described previously (18). Monetary incentives were offered at each visit.

Dietary intake data for each subject at each time point was based on the average of 2 unannounced recalls and 2 written records. The food records and recalls were analyzed using the Nutrition Data System for Research software (version 2010, Nutrition Coordinating Center, University of Minnesota, Minneapolis, MN).

Anthropometric measures of weight, height, blood pressure, waist, and hip circumference were obtained at each study visit. A study questionnaire captured demographic characteristics and health status. Physical activity was assessed using a validated questionnaire from the Women's Health Initiative (19) that records time spent walking at various speeds and carrying out mild, moderate, and strenuous activities leading to a calculation of metabolic equivalents of energy expenditure.

Both diets were designed to be isocaloric with baseline and were delivered by telephone counseling. The consumption goals for the Healthy Eating diet, following the U.S. Healthy People 2010 recommendations (9), included (i) at least 2 servings per day of fruit; (ii) at least 3 servings per day of vegetables; (iii) at least one serving per day of a dark green or dark orange fruit or vegetable; (iv) at least 3 servings per day from whole grains; and (v) less than 10% of calories from saturated fat. The number of goals was more in the Mediterranean arm. The fat goal was to maintain 30% of calories from fat while achieving a PUFA: SFA: MUFA ratio of 1:2:5, which entailed reducing polyunsaturated fatty acids (PUFA) and saturated fatty acids (SFA) intakes by about 50% and 30%, respectively, and increasing monounsaturated fatty acids (MUFA) intake by about 50%. Subjects in this group were asked to consume foods high in omega 3 fatty acids at least twice a week. The whole grain goal was the same as in the Healthy Eating arm. Fruit and vegetable goals were for at least 7 to 9 U.S. Food and Drug Administration servings per day, depending on energy intake, including culinary herbs and allium vegetables.

Fasting blood sample was obtained from the arm in a coagulation tube for the preparation of serum at baseline and 6 months. Serum was stored at −70°C until analysis. Total serum fatty acids were extracted with Folch reagent and measured as fatty acid methyl ester by gas chromatography with mass spectral detection (20). Carotenoids were extracted with hexane and measured by high-pressure liquid chromatography (20). There was not enough blood for carotenoid analysis from one overweight/obese subject in the Healthy Eating arm at baseline.

Colon mucosal biopsies were obtained circumferentially 15–25 cm above the anal sphincter by flexible sigmoidoscopy without any prior preparation of the bowels. Six biopsies were frozen in liquid nitrogen within 5 seconds of incision and they were stored at −70°C until analysis.

Colon samples were analyzed for carotenoids and fatty acids in a similar way as serum except that a colon tissue homogenate was prepared first. A total of 4 frozen biopsies were homogenized together, using pulverization under liquid nitrogen, a technique described previously (21). A portion of the homogenate that was equivalent to one biopsy (150 μL) was used for carotenoid analysis and an equal portion was used for fatty acid analysis. Samples were diluted with 50 μL PBS before extraction. There was one tissue sample missing at 6 months from an overweight/obese completer in the Mediterranean arm that refused flexible sigmoidoscopy.

Subject characteristics such as gender, age, BMI, marital status were compared across the 2 study arms at baseline by means of χ² or Fisher exact test (for categorical characteristics) and 2-sample unpooled z test (for continuous measures). A similar comparison was carried out between the completers and noncompleters to identify any potential source of bias.

Pairwise correlations between colon micronutrient concentrations, dietary intakes, and serum concentrations of select carotenoids and fatty acids at baseline were obtained. Because of the heavy skewness in the data, Spearman rank correlation was used. Statistical tests to compare the correlations for each of the nutrients were also carried out. Methods are available to compare correlated Pearson correlation coefficients when multivariate data conform to normal distribution, but these methods are not applicable to Spearman correlations. The nonparametric bootstrap method to construct confidence intervals for the pairwise differences in correlation coefficients was therefore used. Briefly, this consisted of resampling with replacement from the pool of 120 subjects and recalculating the correlation coefficients based on the resampled data. The confidence interval for the differences in correlation was calculated using the bias-corrected percentile method (22) based on 2000 bootstrap samples. The difference was deemed statistically significant at 5% level if the 95% bootstrap confidence interval did not include zero. We repeated the procedure with the Spearman correlation coefficient ρ replaced by |$z_\rho$| obtained by Fisher transformation

.

Associations between changes in nutrient concentrations in diet, serum, and colon measurements were investigated in a similar manner through correlations of the percent change from baseline in the nutrient concentrations.

To evaluate and compare average changes over time in fatty acids, carotenoids and other nutrients found in serum across 2 groups, linear mixed regression models were used with time, diet group assignment, and the group-by-time interaction as the primary predictors. Mixed linear regression models provide valid inference in presence of dropouts. Furthermore, because it uses all available data each time point, mixed model analysis is tantamount to an intent-to-treat approach. The models were controlled for baseline age, BMI, and smoking as non–time-dependent covariates. Age was slightly higher in the Mediterranean group than the Healthy group (means of 55 versus 50, respectively). The prevalence of baseline smoking status was slightly different in the 2 study arms (11% in the Healthy arm versus 17% in the Mediterranean arm), although the difference was not statistically significant at 5% level (P = 0.06 based on Fisher exact test). Because smoking status may potentially affect the fatty acid and carotenoid concentrations, it was used as a covariate in the regression models, BMI did not differ appreciably between the 2 groups at baseline, but it is known to affect carotenoid concentrations (23). Furthermore, the samples were analyzed in several analytic batches in the laboratory, which was a potential source of variation, and so batch was used as an additional covariate. Normality of the residuals was assessed through q–q plots and tests for normality, and Box-Cox family of transformations was used to identify a suitably symmetrizing transformation whenever it was necessary (24). SFA was square root transformed. Log transformation was used for all other variables except for MUFA, which required no transformation. Clustering of the pre–post measurements within subjects was accounted for by using a random subject intercept.

Similar models were used for concentrations of fatty acids and carotenoids obtained from colon tissue samples. Apart from baseline age, BMI, and smoking status, the regression models were controlled for variation across laboratory analysis batches. All outcomes required a natural logarithmic transformation with the exception of SFA, which required a square transformation for analysis, and MUFA, which required no transformation.

A total of 120 subjects were enrolled in the study with 61 assigned to the Healthy arm and the remaining 59 in the Mediterranean arm. A total of 94 subjects had available data at 6 months, with number of completers evenly distributed between the Healthy Eating and the Mediterranean diet arms (47 in each).

There were 43 females at baseline in each study arm. The distribution of the participants' age at study entry exhibited a slight left skew with mean and median age of 52.2 and 53.8, respectively. Among the various subject characteristics compared between the 2 groups at baseline, only age turned out to be significantly different between the 2 groups. The mean (SD) of age in years were 50 (14) and 55 (10) in the Healthy Eating and Mediterranean groups, respectively. The P value for this comparison based on an unpooled z test was 0.033. A more complete chart of comparisons between groups was published previously (18). Age also turned out to be the only (borderline) significant demographic characteristics between completers and noncompleters. Noncompleters were slightly younger with a mean age of 47.7 (SD 14) years versus the completers who had a mean age of 53.6 (SD 11.3) years (P = 0.052).

It was of interest to determine whether dietary intakes or serum concentrations of nutrients would be more closely associated with colon concentrations. The correlations between colon and serum nutrients seemed to be strongest, whereas those between diet and colon were the weakest of all (Table 1). All the nutrients measured in colon, except MUFA, SFA, and n6 fatty acids, were significantly associated with serum concentrations at the 1% level of significance. The strongest associations are exhibited between the serum and colon concentrations of β-cryptoxanthin (ρ = 0.58, P < 0.001), α-carotene (ρ = 0.48, P < 0.001), and β-carotene (ρ = 0.45, P < 0.001). The only dietary intake to be significantly associated with both serum and colon concentrations was β-cryptoxanthin (Table 1). The corresponding dietary intake was also significantly associated with serum concentrations of α-carotene (ρ = 0.37, P < 0.001).

Serum fatty acids and carotenoids were measured because they can be useful biomarkers of changes in dietary intakes of fat, fruits, and vegetables. With regard to the major classes of fatty acids, there were no significant changes in serum concentrations of saturated fatty acids, but concentrations of MUFA, N3 PUFA, and the ratio of N3/N6 PUFA changed significantly in the expected directions in the Mediterranean arm (Table 2).

For total serum carotenoids, changes were similar in both diet arms, but the increase in total carotenoids was significant only in the Healthy arm. Specific fruit and vegetable goals in the Healthy arm were for 5 servings per day and including at least one carotenoid-rich dark orange or green vegetable. Increases in lutein were significant in both study arms, β and α-carotene increased significantly only in the Healthy arm, and β-cryptoxanthin increased significantly only in the Mediterranean arm. The fruit and vegetable goals in the Mediterranean arm were to consume at least one serving from each of 7 categories of fruits and vegetables (18). There were no statistically significant changes in lycopene or zeaxanthin concentrations in either arm. Changes that approached statistical significance with 0.05 < P < 0.10 were in β-cryptoxanthin in the Healthy arm and in β and α-carotene in the Mediterranean arm.

Changes in carotenoids and fatty acids in the colon biopsy tissue were in the same direction as in blood, but the changes were smaller and fewer differences were statistically significant (Table 3). Interestingly, concentrations of n3 fatty acids increased in the Healthy arm, but the change was small. Significant increases in several carotenoids were also found in the Healthy arm only (Table 3). Changes in the Mediterranean arm that approached significance with 0.05 < P < 0.10 were for N3 PUFA, n3/n6 ratio, β-cryptoxanthin, and α-carotene.

Because the changes in colon concentrations of nutrients were overall smaller than the changes in serum concentrations, it was of interest to evaluate whether the changes in diet, serum, and colon were correlated with each other. Changes in dietary nutrients were more strongly associated with changes in serum concentrations than with changes in colon concentrations. Analogous to the baseline findings, the correlation between changes in serum and colon measures seem to be stronger than for diet with colon measures, with the largest rank correlation of 0.56 obtained for β-cryptoxanthin (P < 0.001). This correlation was statistically significantly larger than the corresponding ones between diet and colon or diet and serum. Changes in blood serum concentrations of α-carotene and β-carotene were significantly correlated with the corresponding changes in either the diet or colon concentrations. However, there was no statistical difference in the strength of association between these pairs. The correlation between changes in α- and β-carotene levels of colon and diet, was not statistically significant, and was significantly lower than the other 2 correlations in the case of α-carotene.

The results of this dietary intervention study indicate that concentrations of protective nutrients in the human colonic mucosa can be changed by dietary intervention, but that these changes were small and more closely related to changes in serum than to reported changes in dietary intakes (Table 4). Similarly, at baseline, colon concentrations of nutrients were more closely related to serum concentrations than to dietary intakes (Table 1). This could be due to (i) the short time frame of the intervention because tissue stores may require more time to reach equilibrium than blood, (ii) errors inherent in dietary assessment, especially when only 4 days are used at each time point, and (iii) to the role of metabolic factors. Most dietary nutrients are absorbed in the small intestine, and colonic exposures to nutrients therefore are likely to occur at the basolateral, not luminal, side via the systemic circulation. This is especially true for the distal colon that was sampled in this study.

The two dietary interventions used in this study differed in several ways, but our unpublished data indicates that reported fruit and vegetable intakes over 6 months of intervention were unexpectedly similar in the 2 arms. Counseling for fruit and vegetable consumption consistent with Healthy People 2010 goals, namely 2 servings per day of fruit and 3 servings per day of vegetables, with at least one being dark green or orange, was sufficient to increase the concentrations of several serum and colon carotenoids (Table 2). The Mediterranean group with goals for consuming at least 7 servings per day of fruits and vegetables in 7 different categories did not display relatively larger increases in serum carotenoids. One could speculate that subjects in the Healthy arm met their goals with a variety of foods to increase palatability, resulting in a broad spectrum of carotenoid intakes in the Healthy Eating arm. The only serum carotenoid that was increased significantly in the Mediterranean arm, but not the Healthy arm, was β-cryptoxanthin, which is found mainly in fruits. Whether or not other phytochemicals are increased by a Mediterranean diet needs to be investigated. This would include flavonoids such as quercetin, which is high in onions and apples, and phenolic compounds from olives (25, 26).

In experimental models, many individual carotenoids have been shown to be protective of colon cancer including lutein and lycopene (27, 28). A pooled analysis of 11 cohort studies indicated that of the dietary carotenoids, only lutein and zeaxanthin, which were measured together in foods, displayed weak protective effects for colorectal or colon cancer (2). Colon lutein or lycopene concentrations, however, were not significantly increased after 6 months of either intervention (Table 3). Lycopene in both serum and adipose tissue has been previously found to be poorly related to dietary intakes, and supplementation may be a more feasible method to increase concentrations of this carotenoid (29, 30).

Among the carotenoids, relatively more data is available on β-carotene bioavailability in the colon, and this carotenoid is most widely distributed in fruits and vegetables. Several studies have shown that β-carotene supplementation can increase colonic β-carotene concentrations (31–34). In human trials of boiled vegetables or β-carotene supplements, the lag time in the increase in carotenoids in exfoliated cells in the stool was 5–7 days, suggesting that carotenoids might be obtained from the circulating stores at the crypt base (32, 35). This is consistent with our data in the colonic mucosa showing that blood concentrations correlated with colon concentrations more closely than with diet intakes. The previously published plasma–diet correlations of carotenoids have generally varied with r = 0.11 to 0.52 depending on the carotenoid and on gender (36). In this present study at baseline, concentrations of carotenoids in colonic mucosa were more closely associated with serum concentrations than dietary intakes for all the carotenoids and for n3 fatty acids (Table 1). Change in colon concentrations at 6 months was also more strongly associated with change in serum concentrations than with dietary intakes (Table 4).

Given the similarity between the 2 diet arms in reported fruit and vegetable intakes, the main difference between the 2 interventions was in dietary fat intakes. This is potentially important because fatty acid availability is a key determinant of the types of prostaglandins and other eicosanoids that are produced in cells (37). The Mediterranean diet was unique in increasing serum concentrations of MUFA and n-3 fatty acids, but this may reflect recent diet because phospholipids were not isolated. There were few significant changes in colon fatty acids other than a slight increase in n3 PUFA that was significant in the Healthy arm. This indicates the possible importance of metabolic processes in regulating tissue fatty acids concentrations, and these may be genetically determined.

In summary, this study showed that the intervention using Healthy People 2010 goals resulted in larger changes in colon carotenoids than the intervention using Mediterranean goals. Whether or not other phytochemicals are affected to a relatively greater extent with Mediterranean diets remains to be established, but it is encouraging that the modest goals of the Healthy Eating diet were exceeded and resulted in significant beneficial changes in the colon. These data indicate that a relatively simple intervention may be adequate for increasing carotenoids in colon. Colon concentrations of micronutrients were, however, not well correlated with dietary intakes. Dietary assessment is difficult and has many limitations, but metabolic factors also may need to be evaluated as determinants of colon nutrient concentrations.

No potential conflicts of interest were disclosed.

Conception and design: A. Sen, M.T. Ruffin, D.E. Brenner, Z. Djuric

Development of methodology: A. Sen, J. Ren, D.K. Turgeon, D.E. Brenner, M.E. Rapai, M.L. Cornellier, Z. Djuric

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Ren, M.T. Ruffin, M.E. Rapai, M.L. Cornellier, Z. Djuric

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Sen, D.K. Turgeon, D.E. Brenner, E. Sidahmed, Z. Djuric

Writing, review, and/or revision of the manuscript: A. Sen, M.T. Ruffin, D.K. Turgeon, Z. Djuric

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Ren, D.K. Turgeon, M.E. Rapai

Study supervision: M.T. Ruffin, D.K. Turgeon, Z. Djuric

The authors thank Leah Askew, who assisted with data organization and analysis, and all the individuals who volunteered their time to participate in the Healthy Eating Study for Colon Cancer Prevention.

This study was supported by NIH grants RO1 CA120381 and P30 CA130810 and Cancer Center Support Grant P30 CA046592. The study used core resources supported by a Clinical Translational Science Award, NIH grant UL1RR024986 (the Michigan Clinical Research Unit), and by the Michigan Diabetes Research and Training Center funded by NIH grant 5P60 DK20572 (Chemistry Laboratory).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.