Meat Consumption, Heterocyclic Amines, and Colorectal Cancer Risk: The Multiethnic Cohort Study

Abstract

Greater consumption of red and processed meat has been associated with an increased risk of colorectal cancer in several recent meta-analyses. Heterocyclic amines (HCAs) have been hypothesized to underlie this association. In this prospective analysis conducted within the Multiethnic Cohort Study, we examined whether greater consumption of total, red, or processed meat was associated with the risk of colorectal cancer among 165,717 participants who completed a detailed food frequency questionnaire at baseline. In addition, we examined whether greater estimated intake of HCAs was associated with the risk of colorectal cancer among 131,763 participants who completed a follow-up questionnaire that included a meat-cooking module. A total of 3,404 and 1,757 invasive colorectal cancers were identified from baseline to the end of follow-up, and from the date of administration of the meat-cooking module to the end of follow-up, respectively. Proportional hazards models were used to estimate basic and multivariable-adjusted relative risks (RRs) and 95% confidence intervals (CIs) for colorectal cancer associated with dietary exposures. In multivariable models, no association with the risk of colorectal cancer was detected for density-adjusted total meat (RR_{Q5 vs Q1}=0.93 [0.83–1.05]), red meat (RR =1.02 [0.91–1.16]), or processed meat intake (RR =1.06 [0.94–1.19]), or for total (RR =0.90 [0.76–1.05]) or specific HCA intake whether comparing quintiles of dietary exposure or using continuous variables. Although our results do not support a role for meat or for HCAs from meat in the etiology of colorectal cancer, we cannot rule out the possibility of a modest effect.

INTRODUCTION

In the U.S., colorectal cancer is the third most frequently diagnosed cancer in both men and women, with an estimated 142,570 new cases and 51,370 deaths occurring annually.¹ International variation and temporal shifts in colorectal cancer rates,^2–4 coupled with migration studies showing that the rates of colorectal cancer among individuals emigrating from countries with a low incidence of colorectal cancer equal or surpass those of U.S.-born whites within a single generation,⁵ provide strong evidence for a role of environmental factors in the etiology of this disease.

To date, several modifiable lifestyle factors, including obesity, sedentary lifestyle, alcohol consumption, smoking and consuming a diet high in red and processed meats, have been associated with an increased risk of colorectal cancer.^6–8 In a recent consensus report, the World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) concluded that the evidence was “convincing” for a role of red and processed meats in the etiology of colorectal cancer and that heme, N-nitroso compounds, heterocyclic amines (HCAs), and polycyclic aromatic hydrocarbons found in cooked meats offer potential underlying mechanisms for these associations.⁹ Similarly, a recent meta-analysis of prospective epidemiological studies found an increased risk of colorectal cancer with greater red and/or processed meat consumption,¹⁰ although risk estimates were not shown to depart significantly from unity in several of the individual studies included.^11–14

HCAs are a class of mutagenic compounds formed from the reaction of creatine or creatinine, amino acids, and sugars during the high-temperature cooking of meat.^15,16 Once ingested and absorbed, HCAs are converted into genotoxic metabolites via hepatic cytochrome P-450 1A2 (CYP1A2), N-acetyltransferase 1 (NAT1), N-acetyltransferase 2 (NAT2), and sulfotransferases.^16–19 HCAs have been shown to be mutagenic in bacterial assays²⁰ and to produce tumors of the gastrointestinal tract in animal models.^21,22 In addition, HCAs have been shown to form HCA-DNA adducts in human colorectal tissue^23,24 providing a plausible mechanism of action by which these compounds may increase the risk of colorectal cancer.

For this report, we examined whether greater consumption of red meat and processed meat was associated with the risk of colorectal cancer in a large, ethnically diverse cohort of older adults participating in the Multiethnic Cohort (MEC) Study. In an attempt to further elucidate potential mechanisms by which cooked meat products may influence the risk of colorectal cancer, we also examined whether intakes of total HCAs, or of the specific HCAs 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP), 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline (MeIQx), and 2-amino-3,4,8-trimethylimidazo[4,5-f ]quinoxaline (DiMeIQx) were associated with disease risk in a sub-sample of participants who completed a detailed meat-cooking module.

MATERIALS AND METHODS

Study Population

The MEC is a longitudinal study designed to examine the associations of dietary, lifestyle, and genetic factors with the incidence of cancer, and has been described previously in detail.²⁵ Briefly, over 215,000 adults who were between 45–75 years of age at recruitment and residing in California or Hawaii entered the cohort from 1993 to 1996. In the primary sampling frame, drivers’ license files, voter registration lists, and Medicare files were used to obtain a multiethnic sample of African Americans, Japanese Americans, Latinos, Native Hawaiians, and whites. At cohort entry, individuals voluntarily completed a self-administered, 26-page mailed questionnaire, indicating their consent to participate in the study. The questionnaire included queries on demographic characteristics, anthropometric measures, medical history, family history of cancer, smoking history, reproductive and menstrual history for women, cancer screening practices, occupational history, and physical activity, as well as detailed questions on diet. In 1999–2000, a follow-up questionnaire was administered to living cohort members that included a module on meat consumption and cooking practices. The study protocol was approved by the institutional review boards of the University of Hawaii and the University of Southern California.

Exclusion criteria

For the present study, data from 165,717 participants were available for the analyses examining associations of meat intake with the risk of colorectal cancer, and data from 131,763 participants were available for the analyses examining associations of HCA exposure with the risk of colorectal cancer. For the analyses examining associations of meat intake, participants were excluded if they 1) were not in one of the five main ethnic groups recruited into the study (n = 13,988); 2) reported implausible dietary values >3 standard deviations from a truncated normal distribution using the middle 80% of logged energy values for variance estimation or >3.5 standard deviations for logged fats, proteins and carbohydrate within each sex–ethnic group (n = 8,260); 3) were identified either through the questionnaire or tumor registry linkage to have had a diagnosis of colorectal cancer prior to cohort entry (n = 2,151); or 4) had missing data on any of the following key covariates: family history of colorectal cancer, history of colorectal polyp, pack-years of cigarette smoking, body mass index (BMI), vigorous physical activity, non-steroidal anti-inflammatory drug (NSAID) use, multivitamin use, history of diabetes, and hormone replacement therapy use (women only) (n = 25,135). In addition, 33,954 eligible participants did not complete the follow-up questionnaire that included the meat consumption and cooking practices module and therefore were not included in analyses examining associations of HCA exposure with the risk of colorectal cancer.

Cohort follow-up and case identification

Follow-up began on the date of questionnaire completion for either the baseline or the follow-up questionnaire and accrued until a diagnosis of colorectal cancer, death, or the last follow-up date for this analysis (December 31, 2007). Routine linkages to the Surveillance, Epidemiology, and End Results (SEER) cancer registries for California and Hawaii were conducted to identify the occurrence of newly diagnosed cancers of the colon and rectum over the follow-up period. Routine linkages to the California and Hawaii death certificate files, as well as to the National Death Index, were conducted to identify deaths over the follow-up period. Case status was classified using the International Classification of Diseases for Oncology, Third Edition (ICD-O-03) codes C18.0-C18.9 (colon), C19.9 (rectosigmoid junction), and C20.9 (rectum). For the analyses examining associations of meat intake with the risk of colorectal cancer, 3,404 incident cases of colorectal cancer were diagnosed over a median follow-up period of 13.6 years. For the analyses examining associations of HCA exposure with the risk of colorectal cancer, 1,757 incident cases of colorectal cancer were diagnosed over a median follow-up period of 8.1 years.

Dietary Assessment

The dietary intake of total meat, red meat, and processed meat was measured at baseline (1993–1996) using a quantitative food frequency questionnaire (QFFQ) that obtained the frequency and quantity of food items consumed during the preceding year.²⁵ Items included on the QFFQ were the minimum set, identified from 3-day measured food records, that could capture ≥ 85% of the intake of key nutrients for each ethnic group, as well as food items traditionally consumed by the populations represented in the MEC. Values for total meat, red meat, and processed meat were calculated using the recipe and food composition tables maintained by the University of Hawaii Cancer Center²⁵ from QFFQ items on individual foods, as well as from mixed dishes. At baseline, participants were also asked about their usual preference regarding meat doneness (rare, medium, well-done). A dietary pattern loading high on meat and fat intake was identified by exploratory factor analysis with a varimax rotation and validated by confirmatory factor analysis in the total MEC study population as previously described.²⁶

The dietary intake of meat “cooked dark brown,” total HCAs, and the specific HCAs PhIP, MeIQx, and DiMeIQx was estimated for the sub-sample of eligible cohort members who completed the meat-cooking module on the follow-up questionnaire (1999–2000). Data on the type of meat, high-temperature cooking method (pan-fried, oven-broiled, grilled or barbequed), frequency of consumption, and degree of doneness were collected and used in conjunction with the CHARRED database (www.charred.cancer.gov) developed by Sinha et al.²⁷ to derive estimates of PhIP, MeIQx, and DiMeIQx exposure. Total HCA exposure was estimated by summing the values for PhIP, MeIQx, and DiMeIQx. We assessed risk associated with the sum of the three HCAs to facilitate comparisons with past studies; however, the meaning of this variable is not clear since specific HCAs demonstrates different mutagenicity levels and may have a varying carcinogenic potential.

Statistical analyses

Cox proportional hazards models with age as the time metric were used to calculate relative risks (RRs) and 95% confidence intervals (CIs) for colorectal cancer. Meats and HCAs were examined as energy densities (per 1,000 kcal) in all models, as correlations between dietary estimates from the QFFQ and 24-hour recalls were found to improve after energy-adjustment in our previous calibration study.²⁸ Dietary exposures were divided into quintiles based on the exposure distribution of eligible participants, with the lowest intake group serving as the referent in all models. Linear trends were tested by entering quintile medians as continuous variables in regression models. The heterogeneity of associations between dietary factors and colorectal cancer by sex and across the ethnic groups included in the MEC was tested by a Wald test of the cross-product terms. Models were also constructed to examine the risk of colon and rectal cancer separately. For analyses stratified by anatomical sub-site (colon/rectum), participants diagnosed with the alternative cancer or with a malignancy of the rectosigmoid junction were censored on the date of diagnosis. The test of heterogeneity by anatomical sub-site was performed using a competing risk technique, where each sub-site was modeled as a different event.²⁹ A Wald test was used to compare the parameter estimates between sites. The assumption of proportional hazards was found to be satisfied in all models by examining the relationship of scaled Schoenfeld residuals with time.

Basic models were adjusted for age as the time metric, ethnicity as a stratum variable, and sex and age at cohort entry in the log-linear model component. Multivariable models were constructed, further adjusting for family history of colorectal cancer (yes/no), history of colorectal polyp (yes/no), body mass index (kg/m²) (BMI), pack-years of cigarette smoking, alcohol consumption, vigorous physical activity, NSAID use (yes/no), multivitamin use (yes/no), history of diabetes (yes/no), total energy (log transformed calories per day), dietary fiber (grams per 1000 kcal per day from food), calcium (mg per day from food and supplements), folate (mcg per day from food and supplements) and vitamin D (IU per day from food and supplements) in the log-linear model component to examine the potential confounding effects of colorectal cancer risk factors on the diet-cancer associations. Covariates selected for inclusion were those that have either shown a consistent association with colorectal cancer risk in the literature or have been found to be associated with colorectal cancer risk in the MEC. We also examined the potential non-linear relations of dietary exposures with the risk of colorectal cancer using restricted cubic splines.³⁰ Tests for non-linearity were performed using the likelihood ratio test (LRT). In no model was the addition of the cubic spline terms found to improve model fit (p ≥ .05 for the LRT) when comparing the model with only the linear term to the model with the linear and the cubic spline terms. Therefore, only the results for the linear models are reported.

Based on our a priori hypothesis of a possible synergistic effect between smoking and HCA exposure on the risk of colorectal cancer, departures from the assumption of additive joint effects for HCAs and pack-years of cigarette smoking were examined by cross-classifying exposures and calculating the relative excess risk due to interaction (RERI) using the method and SAS program described by Li and Chambless for proportional hazards models.³¹ Models were constructed to examine variables cross-classified at their median values and across exposure quintiles. Lag analyses excluding participants diagnosed with colorectal cancer within two years of questionnaire completion and sensitivity analyses restricting case status to include only invasive adenocarcinomas were also conducted. All data analyses were performed using SAS 9.2 statistical software (SAS Institute Inc., Cary, NC, USA).

RESULTS

The characteristics of the study participants according to quintiles of total red meat intake are given in Table 1. Participants in the highest intake quintile for total red meat were more likely to be male (58%), African American (18%), Native Hawaiian (10%), or Latino (28%), report a greater BMI level and a greater number of pack-years of cigarette smoking, take a multivitamin supplement, have a history of diabetes (15%), and never to have used hormone replacement therapy among women (57%). Despite increasing levels of total energy across quintiles of red meat consumption, participants in the highest intake quintile for total red meat also reported lower values for dietary fiber, calcium, folate, and vitamin D.

Table 1.

Baseline characteristics of study participants according to quintiles of total red meat consumption in the Multiethnic Cohort Study 1993 - 2007^*

		Total Red Meat Intake (g/1,000 kcal/day)
		Quintile 1	Quintile 3	Quintile 5
	Total	(Median = 7.4)	(Median = 24.6)	(Median = 48.0)
Total No. at Risk (n)	165,717	33,144	33,144	33,143
Total Cases (n)	3,404	619	680	697
Age at Cohort Entry (years) (mean, SD)	59.9 (8.8)	61.1 (8.9)	60.2 (8.8)	58.2 (8.5)
Sex (% male)	46.4	35.7	46.2	58.1
Ethnicity (%)
African American	15.9	16.2	14.6	18.0
Native Hawaiian	7.3	4.5	7.6	9.6
Japanese American	29.5	28.3	32.2	24.9
Latino	21.3	16.7	20.9	27.8
White	26.0	34.3	24.8	19.8
Family History of Colorectal Cancer (%)	8.1	8.6	8.0	7.7
Positive History of Colorectal Polyps (%)	5.6	5.7	5.9	5.0
Body Mass Index (kg/m2) (mean, SD)	26.5 (5.1)	25.1 (4.6)	26.6 (5.1)	27.9 (5.5)
Pack-years of cigarette smoking (mean, SD)	10.3 (15.0)	7.6 (13.1)	10.2 (15.0)	13.2 (16.5)
Alcohol Consumption (ethanol g/day) (mean, SD)	9.3 (25.3)	8.8 (29.1)	9.7 (25.2)	8.4 (18.5)
Vigorous Physical Activity (hours/day) (mean, SD)	0.39 (0.82)	0.39 (0.79)	0.37 (0.81)	0.42 (0.88)
Non-steroidal Anti-Inflammatory Drug User (%)†	52.5	50.8	52.8	53.8
Multivitamin User (%)‡	49.2	41.4	49.1	57.1
History of Diabetes (%)	11.2	7.9	11.1	15.0
Hormone Replacement Therapy Use (%)§
Never use	53.0	49.3	53.0	59.6
Past use	17.6	18.1	17.9	16.3
Current use	29.4	32.7	29.1	24.1
Dietary (mean, SD)
Energy (kcal/day)	2149 (1029)	1962 (909)	2178 (1020)	2269 (1128)
Dietary Fiber (g/1,000 kcal/day)	11.8 (4.3)	14.8 (5.0)	11.4 (3.8)	9.7 (3.2)
Calcium (mg/1,000/kcal/day) (food and supplements)	1020 (643)	1156 (732)	1014 (616)	885 (549)
Folate (mcg/1,000/kcal/day) (food and supplements)	539 (362)	644 (406)	533 (347)	448 (316)
Vitamin D (iu/1,000/kcal/day) (food and supplements)	346 (349)	424 (395)	343 (338)	271 (303)

Note: Percent values reflect column percents.

Abbreviations: No, number; SD, standard deviation.

^*Median follow-up time 13.6 years.

^†Users defined as those using non-steroidal anti-inflammatory drugs at least twice a week for one month or greater.

^‡Users defined as those using multivitamins at least once a week over the previous year.

^§Values for women only.

The relative risks of colorectal cancer according to intake quintiles for a dietary pattern high in meat and fat, total meat, total red meat, red meat excluding processed red meat, total processed meat, and total and specific HCAs (PhIP, DiMeIQx, MeiQx) are provided in Table 2. In basic models adjusting for age at cohort entry, sex, and race/ethnicity, there was an increased risk of colorectal cancer for participants in the highest intake quintile for a dietary pattern high in meat and fat (RR = 1.20 [1.08 – 1.35]; p for trend < 0.001), density-adjusted total red meat (RR = 1.26 [1.13 – 1.41]; p for trend < 0.001), density-adjusted red meat excluding processed red meat (RR = 1.17 [1.05 – 1.31]; p for trend = 0.004), and density-adjusted total processed meat (RR = 1.25 [1.12 – 1.40]; p for trend < 0.001) when compared to those in the lowest intake quintile. All associations were attenuated in multivariable models adjusting for colorectal cancer risk factors. No statistically significant associations were detected for total meat consumption, total HCAs, or the specific HCAs PhIP, DiMeIQx, or MeiQx in any multivariable model examined. In addition, no statistically significant associations were detected for the preference of meat cooked well-done or for the consumption of total meats prepared by pan-fried, oven-broiled, or grilled/barbequed cooking methods (data not shown). In multivariable models stratified by anatomical sub-site (colon/rectum), restricted to invasive adenocarcinomas only, excluding cases diagnosed within two years of questionnaire completion, or examining dietary exposures as continuous variables, no association with the risk of colorectal cancer was detected (data not shown). No heterogeneity in the associations between dietary factors and the risk of colorectal cancer was detected by sex or ethnicity (p > 0.05).

Table 2.

Relative risks and (95% confidence intervals) for colorectal cancer by quintiles of meat and heterocyclic amine intakes

	Q1	Q2	Q3	Q4	Q5	ptrend‡	pheterogeneity§
Meat and Fat Dietary Pattern (factor score)
Quintile Median	−1.26	−0.53	0.00	0.55	1.33
Cases/No. at Risk	697/33143	659/33144	660/33144	712/33143	676/33143
Basic Model*	1.00	1.00 (0.89–1.11)	1.03 (0.92–1.14)	1.16 (1.04–1.29)	1.20 (1.08–1.35)	<0.001	0.023
Multivariable Model†	1.00	0.95 (0.85–1.07)	0.96 (0.85–1.09)	1.07 (0.93–1.24)	1.12 (0.94–1.33)	0.099	0.123
Total Meat (g/1,000 kcal/day)
Quintile Median	16.18	27.93	37.48	48.58	68.96
Cases/No. at Risk	721/33143	720/33144	699/33143	640/33144	624/33143
Basic Model*	1.00	1.02 (0.92–1.13)	1.05 (0.94–1.16)	1.01 (0.90–1.12)	1.03 (0.92–1.15)	0.696	0.552
Multivariable Model†	1.00	0.98 (0.89–1.09)	0.99 (0.89–1.10)	0.93 (0.83–1.04)	0.93 (0.83–1.05)	0.147	0.624
Total Red Meat (g/1,000 kcal/day)
Quintile Median	7.41	16.56	24.55	33.37	47.99
Cases/No. at Risk	619/33144	697/33143	680/33144	711/33143	697/33143
Basic Model*	1.00	1.12 (1.00–1.24)	1.12 (1.00–1.24)	1.21 (1.08–1.35)	1.26 (1.13–1.41)	<0.001	0.113
Multivariable Model†	1.00	1.04 (0.93–1.16)	0.99 (0.88–1.11)	1.03 (0.92–1.16)	1.02 (0.91–1.16)	0.757	0.208
Red Meat Excluding Processed Red Meat (g/1,000 kcal/day)
Quintile Median	4.59	11.13	16.86	23.40	34.86
Cases/No. at Risk	654/33143	702/33143	712/33144	677/33144	659/33143
Basic Model*	1.00	1.06 (0.96–1.18)	1.12 (1.01–1.25)	1.11 (1.00–1.24)	1.17 (1.05–1.31)	0.004	0.010
Multivariable Model†	1.00	0.99 (0.89–1.11)	1.00 (0.90–1.12)	0.97 (0.87–1.09)	0.98 (0.87–1.10)	0.584	0.064
Processed Meat (g/1,000 kcal/day)
Quintile Median	1.70	4.48	7.28	10.86	17.98
Cases/No. at Risk	599/33144	626/33143	706/33143	704/33144	769/33143
Basic Model*	1.00	1.03 (0.92–1.16)	1.14 (1.02–1.27)	1.14 (1.02–1.27)	1.25 (1.12–1.40)	<0.001	0.047
Multivariable Model†	1.00	0.98 (0.87–1.09)	1.04 (0.93–1.16)	1.00 (0.90–1.13)	1.06 (0.94–1.19)	0.259	0.156
Total Meat Cooked Dark Brown (g/1,000kcal/day)
Quintile Median	0.00	1.74	7.62	18.51	47.23
Cases/No. at Risk	687/52186	278/19894	273/19895	260/19893	259/19895
Basic Model*	1.00	1.07 (0.93–1.23)	1.04 (0.90–1.20)	1.01 (0.87–1.16)	1.03 (0.89–1.19)	0.908	0.680
Multivariable Model†	1.00	1.03 (0.89–1.18)	0.99 (0.86–1.14)	0.95 (0.83–1.10)	0.98 (0.85–1.14)	0.673	0.889
Total HCA (ng/1,000 kcal/day)
Quintile Median	43.82	165.03	321.58	574.69	1237.86
Cases/No. at Risk	358/26353	377/26353	355/26351	346/26353	321/26353
Basic Model*	1.00	1.03 (0.89–1.19)	0.98 (0.84–1.13)	0.99 (0.85–1.15)	0.97 (0.83–1.13)	0.511	0.044
Multivariable Model†	1.00	1.01 (0.87–1.17)	0.94 (0.81–1.09)	0.93 (0.80–1.09)	0.90 (0.76–1.05)	0.122	0.102
PhIP (ng/1,000 kcal/day)
Quintile Median	35.34	130.52	256.05	465.45	1027.30
Cases/No. at Risk	358/26352	375/26354	376/26351	313/26353	335/26353
Basic Model*	1.00	1.03 (0.89–1.19)	1.04 (0.90–1.20)	0.90 (0.77–1.05)	1.01 (0.87–1.18)	0.731	0.042
Multivariable Model†	1.00	1.01 (0.87–1.17)	1.00 (0.86–1.16)	0.85 (0.73–1.00)	0.95 (0.81–1.11)	0.274	0.077
DiMeIQx (ng/1,000 kcal/day)
Quintile Median	0.15	1.38	3.34	6.55	16.75
Cases/No. at Risk	354/26378	356/26315	362/26367	368/26351	317/26352
Basic Model*	1.00	1.00 (0.86–1.16)	1.01 (0.87–1.17)	1.05 (0.90–1.21)	0.94 (0.80–1.09)	0.376	0.283
Multivariable Model†	1.00	0.97 (0.84–1.13)	0.96 (0.83–1.12)	0.99 (0.85–1.15)	0.88 (0.75–1.03)	0.107	0.538
MeIQx (ng/1,000 kcal/day)
Quintile Median	3.09	20.91	49.02	95.04	208.18
Cases/No. at Risk	322/26353	380/26352	346/26352	367/26354	342/26352
Basic Model*	1.00	1.16 (1.00–1.34)	1.04 (0.89–1.21)	1.12 (0.97–1.31)	1.11 (0.95–1.30)	0.416	0.610
Multivariable Model†	1.00	1.13 (0.97–1.31)	0.99 (0.85–1.15)	1.05 (0.90–1.22)	1.01 (0.86–1.19)	0.644	0.813
Pack-Years of Cigarette Smoking
Quintile Median	0.00	2.00	10.20	19.76	39.53
Cases/No. at Risk
Basic Model*	1.00	1.19 (1.08–1.32)	1.22 (1.10–1.36)	1.24 (1.12–1.39)	1.57 (1.41–1.73)	<0.001	0.014
Multivariable Model†	1.00	1.19 (1.07–1.31)	1.20 (1.07–1.34)	1.19 (1.07–1.33)	1.45 (1.31–1.61)	<0.001	0.044

Notes: Relative risks reflect hazard ratios obtained from Cox regression.

Abbreviations: No., number; HCA, heterocyclic amines; DiMeIQx, 2-amino-3,4,8-trimethylimidazo[4,5-f]quinoxaline; MeIQx, 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline; PhIP, 2-amino-1-methyl-6-phenyl-imidazo[4,5-b]pyridine.

^*Adjusted for age as the underlying time metric, ethnicity (strata), and age at cohort entry and sex in the log linear model component.

^†Additionally adjusted for family history of colorectal cancer, history of colorectal polyp, body mass index (kg/m2), pack-years of cigarette smoking, non-steroidal anti-inflammatory use, alcohol consumption, vigorous physical activity, history of diabetes, hormone replacement therapy use (females only), total calories, dietary fiber, calcium, folate, and vitamin D in the log linear model component where appropriate.

^‡p-value for the Wald test when modeling quintile medians as a continuous variable.

^§p-value for the Wald test for heterogeneity of diet and colorectal cancer associations across the ethnic groups included in the MEC.

The main effects for pack-years of cigarette smoking and the joint effects for pack-years of cigarette smoking and HCA exposure on the risk of colorectal cancer are provided in Table 2 and Table 3, respectively. The risk of colorectal cancer was confirmed to be higher among participants reporting a greater number of pack-years of cigarette smoking; however, no departures from additive joint effects were detected for pack-years of cigarette smoking and total or specific HCA exposure. The RERIs ranged from 0.02 – 0.13 and included the null within the 95% confidence interval in all models. The joint effects were similarly null for cross-classified quintiles of pack-years of cigarette smoking and HCAs or when examining smoking status using a classification of never, former, or current smoker (data not shown).

Table 3.

Joint Effects of Cigarette Smoking and HCA Exposure on the Risk of Colorectal Cancer

Pack-Years of Cigarette Smoking	HCA Exposure (ng/1000 kcal/day)	Cases/No. at Risk	RR (95% CI)*	RERI (95% CI)†
Pack-Years ≤ 1.25	Total HCA ≤ 321.58	403/34,854	1.00
	Total HCA > 321.58	356/32,923	0.95 (0.82–1.10)
Pack-Years > 1.25	Total HCA ≤ 321.58	495/31,027	1.31 (1.15–1.51)
	Total HCA > 321.58	503/32,959	1.28 (1.11–1.47)	0.02 (−0.20–0.22)
Pack-Years ≤ 1.25	DiMeIQx ≤ 3.34	396/34,741	1.00
	DiMeIQx > 3.34	363/33,036	0.96 (0.83–1.11)
Pack-Years > 1.25	DiMeIQx ≤ 3.34	486/31,199	1.30 (1.13–1.50)
	DiMeIQx > 3.34	512/32,787	1.30 (1.13–1.49)	0.04 (−0.17–0.25)
Pack-Years ≤ 1.25	MeIQx ≤ 49.02	415/35,542	1.00
	MeIQx > 49.02	344/32,235	0.89 (0.77–1.04)
Pack-Years > 1.25	MeIQx ≤ 49.02	463/30,339	1.25 (1.09–1.43)
	MeIQx > 49.02	535/33,647	1.27 (1.10–1.46)	0.13 (−0.07–0.33)
Pack-Years ≤ 1.25	PhIP ≤ 256.05	406/34,688	1.00
	PhIP > 256.05	353/33,089	0.93 (0.81–1.08)
Pack-Years > 1.25	PhIP ≤ 256.05	502/31,194	1.31 (1.14–1.50)
	PhIP > 256.05	496/32,792	1.26 (1.09–1.45)	0.02 (−0.19–0.23)

Notes: Relative risks reflect hazard ratios obtained from Cox regression. Values for cigarette smoking and HCAs are cross-classified at their median values.

Abbreviations: HCA, heterocyclic amines; DiMeIQx, 2-amino-3,4,8-trimethylimidazo[4,5-f]quinoxaline; MeIQx, 2-amino-3,8-dimethylimidazo[4,5-f] quinoxaline; PhIP, 2-amino-1-methyl-6-phenyl-imidazo[4,5-b]pyridine; RERI, relative excess risk due to interaction; RR, relative risk; No, number.

^*Models adjusted for age as the underlying time metric, ethnicity (strata), age at cohort entry, sex, family history of colorectal cancer, history of colorectal polyp, body mass index (kg/m2), non-steroidal anti-inflammatory use, alcohol consumption, vigorous physical activity, history of diabetes, hormone replacement therapy use (females only), total calories, dietary fiber, calcium, folate, and vitamin D in the log linear model component.

^†RERI assesses departure from additivity in the joint effects (sufficient-cause interaction) calculated as RERI = RR₁₁ - RR₁₀ - RR₀₁ + 1 If no sufficient-cause interaction is present RERI = 0.

DISCUSSION

In a large, ethnically diverse cohort of older adults, we found an increased risk of colorectal cancer among participants consuming a dietary pattern high in meat and fat, and in those consuming the greatest amounts of total red meat, red meat excluding processed red meat, and total processed meat in basic models that adjusted for age, sex, and race/ethnicity; however, these associations were attenuated upon adjustment for established colorectal cancer risk factors. In no model was the preference for meats cooked well done, meats prepared by specific cooking methods, total HCAs, or the specific HCAs PhIP, DiMeIQx, and MeiQx found to increase colorectal cancer risk. The risk estimates for colorectal cancer according to the quintiles of meat and HCA intake were similar for the ethnic groups included in the MEC, as were the risk estimates for the total study population when stratified by anatomical sub-site.

Our findings are consistent with several prospective studies,^11–14 as well as our previous nested case-control study conducted in the MEC,³² that have failed to detect an association between red or processed meat intake and the risk of colorectal cancer. These findings, however, are in contrast to a recent meta-analysis of prospective studies conducted by Chan et al.¹⁰ that reported a summary relative risk of 1.22 (1.11–1.34) for the highest versus the lowest consumers of red and processed meat, as well as with the conclusion of the WCRF/AICR consensus report that the evidence is “convincing” in support of a positive association between red and processed meat intake and the risk of colorectal cancer.⁹ The equivocal findings of individual studies conducted in this area to date may, in part, reflect the limited ability to detect statistically significant associations at the modest levels suggested by recent meta-analyses to be associated with disease risk.^10,33,34 Alternatively, differences in residual confounding due to incomplete covariate assessment and adjustment, differences in the approaches used to classify and define red and processed meat intake,¹² differences in the recipes and the food composition tables used to derive estimates for red and processed meat, and differences in the study populations examined including the range of meat consumption and energy intake could also contribute to the inconstancies between studies; however, a comparison of published values is difficult due to differences in instruments, reporting, and adjustment. In addition, we cannot rule out the possibility that the attenuation in risk estimates between the basic and multivariate models in our sample was not due, in part, to measurement error in the model covariates³⁵ or to over-adjustment given the inherent difficulty in separating out the effects of such highly correlated behaviors.

For HCAs, carcinogenic compounds formed during the high temperature cooking of meat products, associations with the risk of colorectal cancer in epidemiologic studies have also been inconsistent. Several case-control studies,^36,37 as well as a recent, large prospective analysis conducted within the NIH-AARP cohort have reported positive associations of DiMeIQx, MeiQx, and/or PhIP with the risk of colorectal cancer,³⁸ after similar adjustment for multiple colorectal cancer risk factors as in the present study. In line with our findings, though, null associations have also been reported.^32,39,40 These inconsistencies may also be due to the limited ability of questionnaire-based approaches to detect modest diet-disease associations^41,42 and differences in the questionnaires and the dietary databases used to derive estimates of HCA exposure, as well as the potential for differential recall bias in previous case-control studies. Recently, advances have been made in the ability to estimate HCA exposure in epidemiologic studies, including the development of a validated meat-cooking module and a specialized food composition database;²⁷ however, while these meat cooking modules have been shown to rank-order individuals according to HCA intake reasonably well when compared to a 12-day food diary,²⁷ the misclassification error associated with this method when compared to a valid biological marker of HCA exposure remains to be fully determined.

Differences in the genetic susceptibility to compounds in cooked meat products across populations may also partially explain the variation in findings among past studies. HCAs first require N-oxidation by CYP1A2 before conversion to their genotoxic form by N-acetyltransferases and sulfotransferases. Therefore, as CYP1A2 has been shown to be inducible by environmental exposures including smoking,⁴³ we hypothesized a potential synergistic effect between cigarette smoking and HCA exposure in relation to the risk of colorectal cancer. In our study population, while greater cigarette smoking was confirmed to be associated with an increased risk of colorectal cancer, no association with HCA exposure, or joint effect of cigarette smoking and HCA exposure, was detected. In a previous population-based case-control study conducted in Hawaii, we found an increased risk of colorectal cancer among participants who were positive for the “rapid” CYP1A2 and NAT2 phenotype, were smokers, and reported a preference for well done meat (three-way interaction).⁴⁴ Thus, our inability to detect a two-way interaction in the present study may have been due to confounding by “rapid” versus “slow/intermediate” CYP1A2 and NAT2 status, as the potentially deleterious effects of greater HCA exposure may be confined to individuals who can most efficiently metabolize HCAs into their genotoxic form. Most variable across populations is the N-acetyltransferase-2 phenotype. We have shown that in the MEC, the rapid NAT2 genotype interacts with a dietary pattern high in meat and fat to increase the risk of colorectal cancer,³² suggesting the importance of considering genetic susceptibility for studies in this area. Further, a stronger interaction was found with smoking, another source of heterocyclic amines, and the rapid NAT2 genotype.³²

The current study has several strengths, including the prospective design allowing for prediagnostic dietary assessment, the ethnic diversity of the study sample, the administration of a detailed meat-cooking module and use of the CHARRED database to calculate estimates for HCAs, and the population-based sampling frame utilized by the MEC allowing for the generalizability of the study results.

There were also potential limitations. First, findings are based on a single exposure assessment and, therefore, changes in dietary intake or colorectal cancer risk factors over time could not be accounted for. Second, despite adjusting for total energy intake, the inherent measurement error in the QFFQ and meat-cooking module may have introduced an appreciable degree of misclassification error and attenuation of risk estimates.⁴² Third, data on CYP1A2 or NAT2 phenotype were not available for this analysis; however, this approach allowed for the use of the full MEC (n = 165,717) rather than the subset data (n = 1,842).

In conclusion, our results do not support a role for total meat, red meat, processed meat, or HCAs from cooked meat in the etiology of colorectal cancer. However, we cannot rule out the possibility of a modest effect at the levels suggested by recent meta-analyses, given the inherent limitations of our exposure assessment. Additional prospective studies or pooled analyses with large sample sizes, detailed data on HCA exposure, and the ability to examine associations for specific high-risk subgroups (e.g. CYP1A2 and NAT “rapid” phenotypes) are needed to further elucidate the role of HCAs in relation to the risk of colorectal cancer. In addition, further efforts to identify valid biomarkers of meat and HCA consumption for use in large, prospective studies are warranted, and may increase our ability to detect the possibly modest associations with disease risk.^45–47

NOVELTY and IMPACT: We report results from, to the best of our knowledge, the second prospective study to examine the association between heterocyclic amines (HCAs) from cooked meat products and the risk of colorectal cancer. Our findings do not support a role for red or processed meats or for HCAs from meat in the etiology of colorectal cancer; however, we cannot rule out the possibility of a modest effect.

ABBREVIATIONS

MEC

Multiethnic Cohort Study

RR

relative risk

CI

confidence interval

NSAID

nonsteroidal anti-inflammatory drug

SEER

Surveillance Epidemiology and End Results

BMI

body mass index

QFFQ

quantitative food frequency questionnaire

SES

socioeconomic status

HCAs

Heterocyclic amines

WCRF/AICR

World Cancer Research Fund/American Institute for Cancer Research

CYP1A2

cytochrome P-450 1A2

NAT1

N-acetyltransferase 1

NAT2

N-acetyltransferase 2

PhIP

2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine

MeIQx

2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline

DiMeIQx

2-amino-3,4,8-trimethylimidazo[4,5-f]quinoxaline