Carotenoid Intake and Colorectal Cancer Risk: The Multiethnic Cohort Study

Background A protective effect of fruits and vegetables against colorectal cancer has been supported by many epidemiologic studies. This suggests that the carotenoids frequently found in these foods play a role in the prevention of this common cancer. To examine associations between the intake of individual and total carotenoids and the risk of colorectal cancer, we analyzed prospective data from the Multiethnic Cohort Study. Methods This analysis includes 85 898 men and 105 106 women who completed a quantitative food frequency questionnaire in 1993–1996. The participants were African Americans, Native Hawaiians, Japanese Americans, Latinos, and whites aged 45–75 years at cohort entry. After an average follow-up of 8.2 years, 1292 and 1086 incident cases of colorectal cancer were identified in men and women, respectively. Cox proportional hazard models were used to estimate relative risks of colorectal cancer. Results No significant associations were found between intake of individual and total carotenoids and colorectal cancer risk either in men or women, except for β-cryptoxanthin, which showed a mild protective effect in men. When the associations were investigated separately for colon and rectal cancer, lycopene intake was related to an increased risk of rectal cancer in men. A decreased risk was seen for total β-carotene in male current smokers, but the test for interaction with smoking status was not significant. No association was observed in each ethnic-sex group. Conclusion Overall, our findings do not support a significant association between carotenoid intake and colorectal cancer, although some associations were seen in subgroup analyses.


INTRODUCTION
A protective effect of fruits and vegetables against colorectal cancer has been supported by many epidemiologic studies. [1][2][3] This suggests a potential role for carotenoids, which are frequently found in these foods, in the prevention of this common cancer. Several mechanisms have been proposed, [4][5][6] which are related to the antioxidant properties of carotenoids in cancer prevention. Carotenoids also enhance gap junctional communication and increase immune response. Furthermore, some carotenoids, such as αand β-carotene and βcryptoxanthin, have provitamin A activity.
A number of studies have investigated associations between carotenoid intake and colorectal cancer risk. Results from case-control studies were inconsistent [7][8][9][10][11][12][13][14][15][16][17] and recent cohort studies have reported no clear relation. [18][19][20] In addition, randomized trials of β-carotene supplementation provided no evidence to support an effect of carotenoids on colorectal cancer prevention. 21,22 The carotenoid database for U.S. foods was recently updated to include various foods 23 ; however, few studies in North America have used this updated database to examine intakes of individual carotenoids putatively related to colorectal cancer. 13,15 To further examine associations between the intake of individual and total carotenoids and the risk of colorectal cancer, we analyzed prospective data from a large multiethnic cohort study. In addition to performing overall analyses for colorectal, colon, and rectal cancer, we examined the associations according to smoking status, because cigarette smoking can alter the circulating levels of carotenoids. 24 We also investigated the associations stratified by race/ethnicity and by multivitamin use.

Study population
The design and implementation of the Multiethnic Cohort Study has been described in detail elsewhere. 25 Briefly, the cohort consists of 215 820 men and women living in Hawaii or California who are mainly from 5 racial/ethnic groups: African Americans, Native Hawaiians, Japanese Americans, Latinos, and whites. In 1993-1996, participants aged 45-75 years completed a 26-page mailed questionnaire requesting information on dietary habits, medical history, and lifestyle practices. The study was approved by the review boards of the University of Hawaii and the University of Southern California. Voluntary return of the mailed questionnaire was regarded as an expression of informed consent on the part of the participants. For this analysis, we excluded participants who were not in one of the 5 racial/ethnic groups (n = 13 992) or who provided implausible dietary data (n = 8264). We also excluded participants with a history of colorectal cancer as reported in the baseline questionnaire (n = 2154) or in the cancer registries in Hawaii or California (n = 406). Therefore, a total of 191 004 participants (85 898 men and 105 106 women) remained for the analysis.

Dietary assessment
Dietary intake was assessed by a quantitative food frequency questionnaire (QFFQ) that requested the frequency and amount of consumption during the past year for more than 180 food items. The QFFQ was developed from 3-day measured food records from approximately 60 men and women in each of the main ethnic groups. 25 We performed a calibration study to assess the correlations between nutrient intakes from the QFFQ and 3 repeated 24-hour recalls for a subset of the cohort. 26 Correlations for selected nutrients ranged from 0.57 to 0.74 for nutrient densities. For carotenoid densities, correlations were 0.60 for β-carotene, 0.54 for α-carotene, 0.51 for lycopene, 0.59 for β-cryptoxanthin, and 0.55 for lutein from foods. Daily intake of carotenoids, total energy, and other dietary components from the QFFQ were computed by applying a food composition table that has been developed and maintained at the Cancer Research Center of Hawaii for use in the Multiethnic Cohort Study, 25 and which incorporates carotenoid values from a national database. 23 Total carotenoid intake was calculated by summing the food intake of 5 major carotenoids: β-carotene, α-carotene, lycopene, β-cryptoxanthin, and lutein.
In addition to the QFFQ, the baseline questionnaire asked whether multivitamins or any of 7 single vitamin and mineral supplements had been used at least weekly during the previous year. If a supplement had been taken, details about the duration and frequency for that supplement were also requested. Only regular supplement use, defined as use for >1 year, was considered for this analysis. Among the 5 carotenoids examined in this study, only β-carotene as a single supplement was covered by our questionnaire. Daily supplemental β-carotene intake was calculated based on the following choices of the approximate dosage per tablet: ≤6000 µg, 7000-15 000 µg, ≥16 000 µg, and not known. βcarotene supplement users who did not provide a dosage were assigned to the lowest category (≤6000 µg). Mean daily supplement intake was calculated by multiplying the dose by the frequency of use. Carotenoid intake from multivitamin supplements was not considered because the multivitamin brands most frequently reported in the calibration study did not contain carotenoids at that time. Correlation between supplemental β-carotene intake from the baseline questionnaire and from 3 24-hour recalls was 0.43. 27 Intake of β-carotene was reported from foods and supplements separately, and as "total β-carotene," which was the combined intake from both foods and regularly used supplements. The consumption of other carotenoids was based on foods only.

Identification of colorectal cancer cases
To identify incident colorectal cancer cases, the cohort was linked to the Surveillance, Epidemiology, and End Results (SEER) cancer registries covering Hawaii and California. The cohort was also linked to death certificate files in Hawaii and California and the National Death Index. Cancer identification and death information were complete through December 31, 2002. Colorectal cancer cases were limited to participants diagnosed with invasive adenocarcinoma of the large bowel. A total of 2378 incident cases (1292 men and 1086 women) were identified after an average follow-up of 8.2 years.

Data analysis
We used Cox proportional hazards models with age as the time metric to calculate relative risks (RRs) and 95% confidence intervals (CIs) for colorectal cancer for men and women separately. Observation began at questionnaire completion or at age 45 for the few individuals who were younger than 45 at questionnaire completion, and ended at the earliest of these dates: date of colorectal cancer diagnosis, date of death, and date of closure (December 31, 2002).
We expressed daily carotenoid intakes in terms of density (intake per 1000 kcal) in the analyses, because we found that energy-adjusted intakes led to better correlations between QFFQ values and reference levels. 26 Carotenoid intakes were divided into quintiles determined by the sex-specific intake distribution in the cohort; the lowest quintile was used as the reference group. For β-carotene models, intakes from foods (quintiles) and supplement use (yes/no) were included in the same model so that the result for each was adjusted for the other. "Basic models" were adjusted for race/ethnicity and time since cohort entry-categorized as ≤2 years, 2-5 years, and >5 years-as strata variables and age at cohort entry as a covariate. "Multivariate models" were additionally adjusted for family history of colorectal cancer, history of intestinal polyps, pack-years of cigarette smoking, body mass index (kg/m 2 ), hours of vigorous activity per day, use of nonsteroidal anti-inflammatory drugs, regular multivitamin use, total energy intake (log transformed kcal/day), alcohol intake (% of kcal), red meat intake (g/day), dietary fiber intake (g/1000 kcal/day), total calcium intake (mg/day, foods and supplements), total vitamin D intake (IU/day, foods and supplements), total folate intake (mcg/day, foods and supplements), and use of hormone replacement therapy (in women only). Trend tests were conducted by including a continuous variable in the model assigned the sex and racial/ ethnic-specific median values within the appropriate overall quintiles.
We conducted the analyses for colon and rectal cancer separately. Individuals with synchronous tumors at both sites (14 men and 5 women) were excluded from these analyses. We also investigated whether the association of carotenoid intakes with colorectal cancer risk differed by smoking status at baseline, by race/ethnic group, and by multivitamin use. Tests for interactions were based on the likelihood ratio test, comparing a model with main effects for the subgroup variables and the trend for the carotenoid and one that also includes interactive terms between carotenoids and subgroup membership. All analyses were performed with SAS software (version 9.1), and all P values were 2-sided.

RESULTS
As compared with the entire cohort, colorectal cancer cases were older, more likely to have a family history of colorectal cancer, and less physically active, both in men and women (Table 1). In women, more cases were overweight. In men, cases were more likely to be ever smokers. For intakes of carotenoids as densities, as well as for other dietary components, there was no substantial difference between cases and the entire cohort.
The basic models in men showed inverse associations between colorectal cancer risk and intake of the 5 major carotenoids, except lycopene (Table 2). However, in the multivariate models, the inverse associations were no longer significant except for β-cryptoxanthin (RR of the highest vs. the lowest quintile = 0.89, 95% CI = 0.72-1.09, P for trend = 0.04). Unlike in men, the basic models in women yielded no significant association except for β-cryptoxanthin intake. In multivariate models, this association was attenuated. When we ran the multivariate models without adjusting for dietary fiber intake, which was the main confounder in the association between carotenoids and colorectal cancer, a moderate inverse association was seen for β-carotene intake among men (RR = 0.77, 95% CI = 0.63-0.93, P for trend = 0.05, data not shown). However, the effect of βcarotene was not as strong as that of dietary fiber because the effect of dietary fiber persisted, while that of β-carotene did not, when both of them were in the model. When we examined the associations with colon and rectal cancer risk separately in the multivariate models, lycopene intake in men was significantly associated with higher risk of rectal cancer only (RR = 1.50, 95% CI = 1.04-2.16, P for trend = 0.018) ( Table 3). No association was found between carotenoid intake and colon or rectal cancer risk in women. Table 4 shows the associations between carotenoid intake and colorectal cancer risk by smoking status at baseline. There was an inverse association for total β-carotene intake in male current smokers. However, interaction between carotenoid intake and smoking status was not statistically significant in either men or women. Therefore, the associations did not differ by smoking status.
In accordance with the a priori aim of the Multiethnic Cohort Study, which is to compare dietary effects across racial/ethnic groups, we investigated the associations stratified by race/ethnicity. None of the ethnic-sex groups showed statistically significant associations between carotenoid intakes and colorectal cancer risk (data not shown).

DISCUSSION
In this large Multiethnic Cohort Study, we found no significant associations between intake of individual and total carotenoids and colorectal cancer risk either in men or women, except for β-cryptoxanthin, which showed a mild protective effect in men. β-cryptoxanthin has been shown to decrease colon carcinogenesis in animal models, 28,29 and epidemiologic studies have supported a beneficial effect of βcryptoxanthin on lung cancer, 30,31 but not on colorectal cancer.
Prospective cohort studies have mainly reported the absence of an association. Two female cohort studies-the Canadian National Breast Screening Study (NBSS) 18 and the Shanghai Women's Health Study 19 -did not find significant associations between carotenoid intake and colorectal cancer risk. In male smokers in the Alpha-Tocopherol, Beta-carotene Cancer Prevention (ATBC) Study, no association was seen between dietary and serum carotenoids and colorectal cancer risk. 20 Furthermore, a pooled analysis of 11 cohort studies also did not support an important role for carotenoid intake in the etiology of colorectal cancer. 32 Two randomized clinical trials-the Physicians' Health Study 21 and the ATBC Study 22 -found no beneficial or harmful effect of supplemental β-carotene on colon/colorectal cancer. A metaanalysis 33 of 4 randomized trials, including the ATBC Study, 34 also reported no effect of β-carotene supplementation on primary or secondary prevention of colorectal adenoma, which is the precursor of colorectal cancer.
Findings from case-control studies have been inconsistent. Studies in Switzerland, 10 Italy, 7,8 China, 12 and Japan 17 found inverse associations of colorectal/colon/rectal cancer risk with intake of various carotenoids. However, studies in France, 14 Canada, 15 and Australia 16 reported no association for intakes of different carotenoids. A study in the United Sates reported that high β-carotene intake was associated with a reduced risk for colon cancer in whites. 11 Another study in the United States observed an inverse association between lutein intake and colon cancer risk in both men and women, 9 and found an inverse association of lycopene intake with rectal cancer in women. 13 In the present study, a higher lycopene intake was Used at least once a week for more than 1 year ¶ Intake from foods, multivitamin supplements, and calcium supplements **Intake from foods and multivitamin supplements † † Intake from foods and β-carotene supplements ‡ ‡ Sum of β-carotene, α-carotene, lycopene, β-cryptoxanthin, and lutein intake from foods Adjusted for ethnicity and time since cohort entry as strata variables and the following variables as covariates: age at cohort entry, family history of colorectal cancer, history of intestinal polyps, pack-years of cigarette smoking, body mass index, hours of vigorous activity, use of nonsteroidal antiinflammatory drugs, multivitamin use, total energy intake, alcohol intake, red meat intake, dietary fiber intake, total calcium intake (foods and supplements), total vitamin D intake (foods and supplements), total folate intake (foods and supplements), and use of hormone replacement therapy (for women only). ‡ Intake from foods and β-carotene supplements. § Intakes from foods and supplement use were included in the same model so that results for each were adjusted for the other.

||
Sum of β-carotene, α-carotene, lycopene, β-cryptoxanthin, and lutein intake from foods. *Adjusted for ethnicity and time since cohort entry as strata variables and the following variables as covariates: age at cohort entry, family history of colorectal cancer, history of intestinal polyps, pack-years of cigarette smoking, body mass index, hours of vigorous activity, use of nonsteroidal antiinflammatory drugs, multivitamin use, total energy intake, alcohol intake, red meat intake, dietary fiber intake, total calcium intake (foods and supplements), total vitamin D intake (foods and supplements), total folate intake (foods and supplements), and use of hormone replacement therapy (for women only). † Intake from foods and β-carotene supplements. ‡ Intakes from foods and supplement use were included in the same model so that results for each were adjusted for the other. § Sum of β-carotene, α-carotene, lycopene, β-cryptoxanthin, and lutein intake from foods. related to an increased risk of rectal cancer in women, which was unexpected.
Oxidative stress from tobacco smoke may modify the blood concentrations of potent antioxidants such as certain carotenoids. 24 Only a few studies have investigated associations of carotenoid intake with colorectal cancer risk according to smoking status, and the results are inconsistent. In a large cohort of Canadian women, no association was found in never, past, or current smokers. 18 A pooled analysis of 11 cohort studies reported that null results were not modified by smoking status. 32 A case-control study in Canada observed that a reduced risk of colon cancer was related to high intake of β-carotene among never-smokers and to high intake of lycopene among ever-smokers. 15 In a case-control study conducted in the United States, lutein intake was more strongly associated with decreased risk of colon cancer in current smokers than in never smokers. 9 A case-control study in Europe found the opposite association between β-carotene intake and colorectal adenoma risk in non-smokers and in past/current smokers: a protective effect of β-carotene in nonsmokers, but an adverse effect in smokers. 35 These 3 casecontrol studies and the pooled analysis examined the associations in men and women combined; no sex-specific findings were reported. We found an inverse association for total β-carotene in male current smokers, but there was no interaction between smoking status and intake of carotenoids. Therefore, this finding is likely due to chance.
Although only fewer than 5% of the participants in our study reported using β-carotene supplements regularly, the dosage available in supplements is usually higher than that obtainable from foods. We might have underestimated supplemental β-carotene intake at baseline because we did not record intake from multivitamins. Some multivitamin products might have contained β-carotene, although the brands most frequently used in the calibration study did not. During the follow-up period, these common products began to contain β-carotene and other carotenoids, such as lutein and lycopene. We speculate that subsequent intake of β-carotene and other carotenoids from supplements may have been considerably higher than at baseline. However, the results for multivitamin users and non-users were similar. Lack of information on colorectal cancer screening, including colonoscopy or sigmoidoscopy findings, was also a limitation, although a history of intestinal polyps was adjusted for in the multivariate models.
The strengths of this study include its prospective design, the large number of participants, and detailed dietary information that was validated in the calibration study. A further strength was the ability to control for various factors that might potentially confound the associations between *RR of the highest vs. the lowest quintile; for β-carotene supplement use, yes vs. no. Adjusted for ethnicity and time since cohort entry as strata variables and the following variables as covariates: age at cohort entry, family history of colorectal cancer, history of intestinal polyps, pack-years of cigarette smoking (for former and current smokers only), body mass index, hours of vigorous activity, use of nonsteroidal anti-inflammatory drugs, multivitamin use, total energy intake, alcohol intake, red meat intake, dietary fiber intake, total calcium intake (foods and supplements), total vitamin D intake (foods and supplements), total folate intake (foods and supplements), and use of hormone replacement therapy (in women only). † Intake from foods and β-carotene supplements. ‡ Intakes from foods and supplement use were included in the same model so that results for each were adjusted for the other. § Sum of β-carotene, α-carotene, lycopene, β-cryptoxanthin, and lutein intake from foods.
carotenoid intake and colorectal cancer. Most previous studies on carotenoid intake and colorectal cancer enrolled whites. However, we were able to examine the associations among various ethnic groups with a wide range of intakes from various sources, although we failed to find any difference between ethnic groups.
In conclusion, our findings do not support a significant association between carotenoid intake and colorectal cancer risk.