Meat and heterocyclic amine intake, smoking, NAT1 and NAT2 polymorphisms and colorectal cancer risk in the Multiethnic Cohort Study

Abstract

Background

N-acetyltransferases (NAT) 1 and 2 are polymorphic enzymes catalyzing the metabolic activation of heterocyclic amines (HCA). We investigated the modifying effects of NAT1 and NAT2 polymorphisms on the association of meat consumption, HCA intake, and smoking with colorectal cancer (CRC) risk.

Method

In the Multiethnic Cohort study, participants completed a smoking history and a food frequency questionnaire at recruitment, and a cooked meat module 5 years later to estimate HCA intake (PhIP, MeIQx, DiMeIQx). Blood samples were collected from incident cases and age-, sex- and ethnicity-, frequency-matched controls to determine genotypes. For analysis of meat intake and smoking, data were available for 1,009 cases and 1,522 controls; for HCA intake analyses 398 cases and 1,444 controls were available. Multivariate logistic regression models were used to estimate odds ratios (ORs).

Results

Smoking was associated with an increased CRC risk (OR 1.51 (95% CI: 1.17–1.95) for ≥ 30 pack-years compared to never smokers, p trend=0.0004). The association was stronger with presence of the rapid compared to the slow/intermediate NAT2 genotype (p interaction=0.003). No significant associations were observed for intakes of red meat, processed meat, and HCA, or meat doneness preference, but a dietary pattern high in meat showed a weak positive interaction with the NAT2 genotype (p interaction=0.05).

Conclusion

The enhanced association between smoking and CRC risk in subjects with the NAT2 rapid genotype supports a role for NAT2 and tobacco smoke HCA in the etiology of CRC. This study only provides weak support for a similar association with meat HCA.

Introduction

Colorectal cancer (CRC) is the third most common cancer among men and women in the U.S., and ranks third as a cause of cancer deaths (1). A total of 108,070 new colon cancer cases and 40,740 rectal cancer cases are expected for the year 2008 in the U.S. (2).

There is considerable evidence in support of an association between smoking and colorectal adenomas (3, 4) and a weaker, but suggestive association for CRC (5). Many (6, 7) but not all (8) of the recent studies investigating the effect of smoking on CRC reported a positive association and some, but not all, studies suggested that the effect may be stronger for rectal than for colon cancer (6, 9, 10). A potentially long latency period and inaccurate information on early life smoking behavior has been suggested as one possible explanation for the weaker association of smoking with CRC, compared to adenoma (5). Other well-established risk factors for CRC are red meat and, particularly, processed meat (11, 12). A recent review rated the evidence for a direct association between consumption of red meat or processed meat and CRC risk as “convincing” (13).

One of the hypothesized mechanisms to explain an increased CRC risk with smoking and meat intake is through exposure to heterocyclic amines (HCA) (14). Their carcinogenic properties have been demonstrated in laboratory animals, including non-human primates (15). HCAs are present in tobacco smoke and are formed when meat is cooked at high temperatures (16). HCA formation increases with temperature and duration of cooking, and varies with the type of meat and the cooking method. DNA adducts have been detected in the colon of volunteers at levels of exposure similar to those obtained through the diet (17).

The metabolic activation of HCA is catalyzed by N-acetyltransferases (NAT) 1 or 2 (18, 19) that are coded by genes (NAT1 and NAT2) which are highly polymorphic. HCA can be metabolized more or less efficiently by individuals depending on their NAT genotypes.

Based on the underlying mechanisms, it is hypothesized that rapid NAT1 and/or NAT2 acetylators more readily activate HCAs to their ultimate carcinogenic forms, thereby amplifying the association of cooked meat and smoking with risk of CRC. We sought to investigate this hypothesis in a case-control study nested within the Hawaii-Los Angeles Multiethnic Cohort Study.

Material and Methods

Study design

The Multiethnic Cohort Study was established to investigate lifestyle exposures in relation to various disease outcomes, especially diet and cancer. The respective institutional review boards (University of Hawaii, University of Southern California) approved the study protocol. Recruitment procedures, study design and baseline characteristics have been reported elsewhere (20). In brief, about 215,000 men and women aged 45 to 75 years at cohort creation in 1993, and of five targeted ethnicities (African-American, Japanese-American, Latino, Native Hawaiian, and Caucasian) were enrolled between 1993 and 1996. All study participants initially completed a self-administered comprehensive questionnaire that included a detailed dietary assessment, as well as sections on demographic factors; body weight and height; lifestyle factors other than diet, including smoking history; and family history of cancer. To update selected exposure variables, a four-page questionnaire was administered in 1999–2000, which included a module on types of cooking methods for various meat items (see below).

The Rapid Reporting System of the Hawaii Tumor Registry and quarterly linkage to the Los Angeles County Cancer Surveillance Program were used to identify CRC cases during follow-up. Both registries are members of the Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute. Annual linkages to the State of California’s cancer registry complemented the case ascertainment. A random sample of the cohort stratified by sex and race/ethnicity was selected to serve as potential controls for nested case-control studies. Incident CRC cases occurring since January 1995 and controls were contacted for donation of a blood sample, i.e. for the great majority of cases, the blood was obtained after diagnosis. Controls for this study were frequency matched to cases by sex, ethnicity/race and age. Blood samples were collected at the subjects’ homes, processed within 8 hours, and stored at −80°C. The participation rate among cases was 74% and varied from 70% in African-Americans to 81% in Latinos. The corresponding rate for controls was 66% and varied from 60% in African-Americans to 71% in Caucasians.

Dietary assessment

Usual dietary intake was assessed at baseline using a comprehensive quantitative food frequency questionnaire (QFFQ) especially designed and validated for use in this multiethnic population (20, 21). The QFFQ asks about the consumption of over 180 food items, including over 25 single meat items, and mixed dishes including meat. The QFFQ inquires about the usual frequency, based on 8–9 categories, and amount of food consumed, based on three portion sizes per food item. Prior to calculating food group intake, the food mixtures from the QFFQ were disaggregated to the ingredient level using a customized recipe database. In addition, participants were asked about their usual meat doneness preference.

Approximately five years after baseline, a cooked meat module was administered to assess frequency of consumption and degree of outside brownness for various meat and fish items cooked by three separate high-temperature methods (pan-fried, oven-broiled and grilled or barbecued) during the past year. These data were analyzed in conjunction with the NCI CHARRED database to estimate intakes of three HCA (2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP), 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline (MeIQx), 2-amino-3,4,8-trimethylimidazo[4,5-f]quinoxaline (DiMeIQx)) (22).

Dietary exposures used in this analysis were intakes of red meat, processed red meat, and HCA (PhIP, MeIQx, DiMeIQx), and preference for well-done meat. In addition, we analyzed a dietary pattern variable (“meat and fat pattern”) which loaded heavy on meat, discretionary fat and eggs as an exposure measure in this context. The construction of the dietary patterns in the study population was based on all cohort members and on factor analysis and principal components techniques, and has been reported in detail previously (23).

SNP selection and genotyping

DNA was extracted from blood lymphocytes using a standard method (QIAamp DNA Blood MINI kit, Qiagen, Valencia, CA). NAT2 is encoded by an 870-bp gene (NAT2) which is polymorphic. The reference NAT2*4 allele and at least 24 allelic variants have been described that carry one or several nucleotide substitutions. Seven of these, all located in the coding region, [G191A (R64Q), C282T, T341C (I114T), C481T, G590A (R197Q), A803G (K268R) and G857A (G286T)] occur with a frequency >1% in at least one ethnic group. Genotyping for these seven variants allows for the detection of 26 of the more common alleles (NAT2*4; NAT2*5A,B,C,D,E,G,J; NAT2*6A,B,C,E; NAT2*7A,B; NAT2*11A; NAT2*12A,B,C; NAT2*13; NAT2*14A,B,C,D,E,F,G) (24). Similarly, we considered the allelic variants identified for NAT1. Except for very rare variants (<1%), all (NAT1*3; NAT1*4; NAT1*10; NAT1*11A,B,C; NAT1*14A,B; NAT1*15; NAT1*17; NAT1*19; NAT1*22) can be characterized by genotyping eight SNPs [C97T (R33Stop), C190T (R64W), G445A (V149I), C559T (R187Stop), G560A (R187Q), A752T (D251V), T1088A (3’-UTR)] (25). A consensus listing of variant alleles for NAT2 and NAT1, maintained by an international committee, is published⁴.

Individuals with two “rapid” acetylator alleles (NAT2*4, NAT2*11A, NAT2*12A,B,C and NAT2*13) were predicted to have a “rapid” NAT2acetylator phenotype. Individuals with two “slow” acetylator alleles (all other alleles) were predicted to have a “slow” acetylator phenotype. Subjects with one rapid and one slow allele were predicted to have an intermediate phenotype. For the analyses NAT2 slow and intermediate acetylators were collapsed, because the rapid acetylator phenotype was specifically found to be associated with an increased CRC risk in a previous population-based case-control study by our group (26). Individuals with the NAT1*10 allele were predicted to have the “at risk” NAT1 phenotype.

Genotyping of cases and controls was performed using a fluorescent 5’ endonuclease assay and the ABI PRISM 7900HT Sequence Detection System for allelic discrimination (TaqMan; Applied Biosystems, Foster City, CA). Primers and probes are available on request. For some of the SNPs for which Taqman probes could not be made, the MGB Eclipse Probe System (Nanogen, Bothell, WA) was used. Amplification reactions were carried out in ABI 9700 thermal cyclers and allelic discrimination was determined on the ABI PRISM 7900HT Sequence Detection System. The amplification reaction for Nanogen primers and probes consists of PCR master mix from Sigma (cat. M4693) and JumpStart Taq. Nucleotide-specific PCR primers and fluorogenic probes were designed using Primer Express (Applied Biosystems) and MGB Eclipse Probe Systems (Nanogen). In addition to the quality control performed by the manufacturers, repeat samples were included for 5% of participants. Concordance rates over 98.7% were obtained for the duplicates. For each of the NAT1 and NAT2 SNPs, less than 2.3% of the study samples had undeterminable genotypes. The control distributions within ethnic groups were tested for Hardy-Weinberg equilibrium and all were found to comply with p>0.05.

Statistical analysis

The distributions between cases and controls were statistically compared by the Chi-square test for categorical variables and the Wilcoxon rank-sum test for continuous variables.

Unconditional logistic regression models were used to compute odds ratios (OR) and 95% confidence intervals (CI). After exclusion of participants with missing covariate values, a maximum of 1,009 CRC cases and 1,522 controls were available for the analysis based on the baseline questionnaire. For the analyses with NAT2, 992 cases and 1,493 controls, and for analyses with NAT1, 844 cases and 1,345 controls were available. The analyses of meat doneness preferences were based on a smaller number of participants due to missing information. The analysis of the HCA data was based on 398 incident cases and 1,444 controls. While 90% of the cases and 94% of the controls in the study completed the cooked meat module in the follow-up questionnaire that was administered about 5 years after baseline, the principle of longitudinal analysis allows counting as cases only those colorectal cancer diagnoses occurring after HCA exposure was collected. Indicator variables for quartiles, for main effects, and tertile, for interaction effects, of exposure (meat intake, pack-years of smoking, HCA, factor score for Meat and Fat pattern) were created based on the distribution among controls. Meat intake and HCA intakes were analyzed in terms of densities, i.e. per 1000 kcal per day, because densities have been shown to better correlate with reference dietary intake for nutrients (21, 27). P values for trend were derived from regressions of the quartile (1–4) number as continuous variables. Age-sex-ethnicity adjusted models were calculated along with multivariate adjusted models which contained sex, age at blood draw, race/ethnicity, family history of CRC, BMI, intake (per 1000 kcal per day) of dietary fiber, calcium, vitamin D, folic acid and ethanol, physical activity (metabolic equivalents) and smoking status and pack-years of smoking or meat intake, as appropriate. Further adjustment for aspirin use, educational attainment, and calcium or folic acid from supplements only marginally changed the ORs in the main effects models and these variables were not included.

The likelihood ratio test (LRT) was used to determine the significance of interaction with respect to CRC between genotype (rapid vs slow/intermediate) and exposures represented as one (e.g., cooking preference) or two (e.g., HCA) indicator variable(s). The LRT was computed as the difference in log likelihoods for a main effects, no interaction model and for a fully parameterized model containing all possible interaction terms for the variables of interest and tested as a chi-square statistic with degrees of freedom equal to the number of interaction terms in the latter model. We used polytomous regression and a Wald test to statistically compare the difference in the risk estimates for exposures between colon and rectal cancer.

All statistical analyses were performed using SAS statistical software, version 9 (SAS Institute, Inc., Cary, North Carolina), and all statistical tests were two sided.

Results

The characteristics of the study participants are shown in Table 1. About 55% of cases and 50% of controls were men. Cases were 3 years older than controls on average. Forty-seven percent and 17% of cases were former and current smokers, respectively, compared to 41% and 15% of controls. Number of pack-years of smoking was statistically significantly higher in cases than controls.

Table 1.

Study participants characteristics

Characteristic	Cases	Controls	P*
N	1,009	1,522
Sex
Men	553 (55)†	755 (50)
Women	456 (45)	767 (50)
Age	69 (62–74)‡	66 (60–72)	<0.01
Ethnicity			0.03
African-American	188 (19)	310 (20)
Japanese-American	339 (34)	446 (29)
Native Hawaiian	58 (6)	129 (8)
Latino	225 (22)	344 (23)
Caucasian	199 (20)	293 (19)
BMI	25.9 (23.3–29.2)	25.4 (23.0–28.6)	0.03
Family history of colorectal cancer			<0.01
No	877 (87)	1390 (91)
Yes	132 (13)	132 (9)
Smoking status			<0.01
Never	364 (36)	674 (44)
Former	477 (47)	620 (41)
Current	168 (17)	228 (15)
Ever use of aspirin	381 (39)	563 (38)	0.92
Pack-years of smoking§	3.9 (0–19.8)	2.0 (0–19.8)	<0.01
Red meat intake (g/1000 kcal/d)	18.1 (10.9–25.4)	17.7 (10.4–26.0)	0.73
Processed meat intake (g/1000 kcal/d)	7.2 (4.0–11.1)	6.7 (3.5–11.0)	0.07
Dietary fiber (g/1000kcal/d)	10.9 (8.3–13.8)	11.2 (8.6–14.1)	0.04
Fat (% of energy)	30.4 (25.2–34.7)	30.7 (25.6–35.5)	0.17
Intake of HCA (ng/1000 kcal/d)‖
Total	176.5 (80.0–326.1)	181.2 (84.9–364.0)	0.25
DiMeIQx	1.9 (0.7–4.0)	1.7 (0.6–4.3)	0.74
MeIQx	28.2 (10.4–57.3)	28.1 (10.0–58.6)	0.95
PhIP	140.9 (62.3–267.4)	146.8 (67.3–296.1)	0.15

^*Chi-square test comparing cases and controls for categorical variables, or Wilcoxon ranked-sum test for continuous variables

^†n; % in parentheses (all such values)

^‡median; interquartile range in parentheses (all such values)

^§Pack-years was set to 0 for never smokers.

^‖for analyses with heterocyclic amines 398 cases and 1444 controls were available

Among controls, the frequency for the rapid NAT2 genotype was 9% in African-Americans, 10% in Native Hawaiians, 63% in Japanese-Americans, 13% in Latinos and 6% in Caucasians. For NAT1 the frequencies for the above listed ethnic groups for the *10 allele were 21%, 12%, 33%, 24% and 9%, respectively.

Main effect ORs for the NAT1 and NAT2 phenotype categories, meat intake variables, smoking, HCA exposure, and CRC are shown in Table 2. Neither the NAT2 nor the NAT1 phenotypes were associated with CRC risk. Although a positive association of processed meat intake and CRC risk was observed in the age-sex-ethnicity adjusted models, intake was not significantly associated with CRC risk in the multivariate adjusted analysis. The Meat and Fat pattern was positively associated with CRC in the age-sex-ethnicity adjusted model, but not significantly in the multivariate adjusted model. Red meat intake, doneness preference and HCA intake were not associated with CRC risk.

Table 2.

Main effects of NAT2 and NAT1*10 genotypes, meat intake, doneness preference, pack-years of smoking and HCA intake on colorectal cancer risk^*

	Cases/controls	OR (95% CI)
		Age-sex-ethnicity adjusted	Multivariate adjusted
NAT2
Slow	336/497	1	1
Intermediate	414/652	0.91 (0.75–1.10)	0.92 (0.75–1.12)
Rapid	242/344	0.96 (0.75–1.22)	0.99 (0.77–1.27)
P trend†		0.65	0.83
NAT1
No *10 allele	362/527	1	1
One *10 allele	307/539	0.85 (0.70–1.05)	0.86 (0.70–1.06)
Two *10 allele	175/279	0.98 (0.77–1.26)	1.01 (0.79–1.30)
P trend		0.66	0.83
Red meat intake (g/1000 kcal/d)
0-<10.4	238/380	1	1
10.4-<17.7	249/381	1.04 (0.83–1.31)	1.01 (0.80–1.28)
17.7-<26.0	282/381	1.22 (0.97–1.53)	1.11 (0.88–1.41)
26.0+	240/380	1.07 (0.84–1.35)	0.96 (0.74–1.23)
P trend		0.34	0.67
Processed meat intake (g/1000 kcal/
0-<3.5	222/380	1	1
3.5-<6.7	250/381	1.11 (0.88–1.40)	1.04 (0.82–1.32)
6.7-<11.0	274/381	1.25 (0.99–1.57)	1.13 (0.89–1.44)
110.+	263/380	1.23 (0.97–1.56)	1.08 (0.83–1.39)
P trend		0.05	0.46
Meat and fat pattern (factor score)
Quartile 1	221/380	1	1
Quartile 2	248/381	1.10 (0.87–1.40)	1.03 (0.81–1.31)
Quartile 3	269/381	1.25 (0.99–1.58)	1.11 (0.87–1.43)
Quartile 4	271/380	1.33 (1.04–1.70)	1.13 (0.86–1.47)
P trend		0.01	0.32
Doneness preference
Medium/rare	510/772	1	1
Well-done	492/738	1.03 (0.87–1.22)	1.07 (0.90–1.28)
P		0.76	0.43
Pack-years of smoking
0	364/674	1	1
>0-<30	461/648	1.38 (1.15–1.66)	1.34 (1.11–1.62)
30+	184/200	1.66 (1.29–2.13)	1.51 (1.17–1.95)
P trend		<.0001	0.0004
Total HCA (ng/1000 kcal/d)
0-<217.3	131/481	1	1
217.3-<566.9	146/482	1.17 (0.89–1.54)	1.15 (0.87–1.53)
566.9+	121/481	1.09 (0.81–1.45)	1.03 (0.77–1.39)
P trend		0.55	0.82
DiMeIQx (ng/1000 kcal/d)
0-<1.8	119/481	1	1
1.8-<6.2	147/482	1.29 (0.97–1.70)	1.25 (0.94–1.67)
6.2+	132/481	1.24 (0.93–1.66)	1.18 (0.88–1.59)
P trend		0.14	0.28
MeIQx (ng/1000 kcal/d)
0-<29.8	126/481	1	1
29.8-<93.5	141/482	1.14 (0.86–1.50)	1.05 (0.79–1.40)
93.5+	131/481	1.19 (0.89–1.58)	1.09 (0.81–1.47)
P trend		0.23	0.57
PhIP (ng/1000 kcal/d)
0-<171.1	134/481	1	1
171.1-<460.5	144/482	1.12 (0.85–1.47)	1.11 (0.84–1.47)
460.5+	120/481	1.06 (0.79–1.41)	1.03 (0.77–1.39)
P trend		0.69	0.81

^*number of cases and controls, OR, odds ratio; ORs were calculated by using logistic regression, adjusted for sex, age at blood draw, ethnicity (age-sex-ethnicity adjusted), or with additional adjusted for family history of colorectal cancer, BMI, intake (per 1000 kcal per day) of dietary fiber, calcium, vitamin D, folic acid, ethanol, physical activity (metabolic equivalents), and smoking status pack-years of smoking or meat intake, as appropriate (multivariate adjusted)

^†P for trend is based on Wald test for variable coded as 1–3.

Cigarette smoking was significantly associated with CRC. In the multivariate model, smokers who smoked less than 30 pack-years in their lifetime had a 34% higher risk than never smokers, while smokers of more than 30 pack-years had a 51% higher CRC risk than never smokers. In the subgroup of ever smokers, none of the dietary variables in Table 2 were associated with CRC risk (data not shown). Separate analyses by sex showed a somewhat stronger positive association between pack-years of smoking and CRC in women (449 cases, 750 controls; OR 1.82, 95% CI: 1.16–2.84 for 30+ pack-years of smoking vs. never smokers) than men (543 cases, 743 controls; OR 1.46, 95% CI: 1.05–2.03), but there was no evidence of statistical interaction with sex (p=0.55). All other main effect ORs were similar between men and women (data not shown).

Analyses by ethnicity suggested that the positive association with smoking was strongest for Japanese-Americans (335 cases, 438 controls, OR 2.07, 95% CI: 1.29–3.29) and Caucasians (197 cases, 285 controls; OR 2.10, 95% CI: 1.28–3.44); however, no statistical interaction was present with race/ethnicity (p=0.40).

Sub-site specific analyses showed a stronger positive association with pack-years of smoking for rectal cancer than colon cancer. The multivariate adjusted OR (95% CI) for more than 30 pack-years of smoking was 1.32 (0.99–1.76), with p for trend = 0.02, for colon cancer (697 cases), and 2.12 (1.42–3.14), with p for trend of <0.0001, for rectum cancer (247 cases) (data not shown). The difference between the pack-year risk estimates for colon and rectum cancer was of borderline statistical significance based on a Wald test (p=0.06, df=2).

Table 3 shows age-sex-ethnicity adjusted analyses of interactions between NAT1 or NAT2 genotypes, and meat intake, smoking or HCA exposure. The multivariate adjusted results are generally similar and are not shown. There was a statistically significant increased CRC risk observed for the highest intake of processed meat in NAT2 rapid acetylators compared to slow acetylators (OR=1.48 (1.06–2.07)), but the interaction test did not reach statistical significance (p=0.13) (Table 3). The corresponding OR was no longer statistically significant in the multivariate adjusted model (OR=1.35 (0.96–1.91)). A significant interaction was observed with the Meat and Fat pattern (p=0.047), which remained of borderline significance in the multivariate model (p=0.052); in the latter analysis, however, the OR for individuals in the highest tertile of the Meat and Fat pattern and fast NAT2 genotype compared to those in the lowest tertile and the slow/intermediate genotype was not significant anymore (OR=1.32 (0.93–1.88).

Table 3.

Age-sex-ethnicity adjusted OR (95% CI) for meat intake, doneness preference, pack-years of smoking and HCA intake on colorectal cancer risk by NAT*10 or NAT2 genotype

Meat intake or preference	Acetylator Genotype
	Cases/ controls	NAT2 slow/ intermediate	Cases/ controls	NAT2 rapid	Cases/ controls	NAT1 no *10	Cases/ controls	NAT1 *10
Red meat (g/1000kcal/d)
0-<12.8	244/367	1	83/131	0.89 (0.64–1.24)	120/179	1	162/258	0.98 (0.72–1.34)
12.8-<22.5	260/393	1.01 (0.80–1.27)	72/105	1.01 (0.71–1.44)	122/181	1.03 (0.74–1.43)	152/276	0.88 (0.64–1.21)
22.5-<102.7	246/389	1.00 (0.79–1.27)	87/108	1.20 (0.85–1.69)	120/167	1.16 (0.83–1.63)	168/284	0.99 (0.72–1.35)
P interaction*		0.44				0.77
Processed meat (g/1000kcal/d)
0-<4.7	228/388	1	66/107	0.99 (0.69–1.42)	116/184	1	139/262	0.86 (0.63–1.19)
4.7-<9.4	270/372	1.22 (0.97–1.54)	79/127	1.00 (0.71–1.40)	121/174	1.08 (0.77–1.51)	169/274	1.01 (0.74–1.39)
9.4-<88.3	252/389	1.14 (0.91–1.45)	97/110	1.48 (1.06–2.07)	125/169	1.19 (0.86–1.67)	174/282	1.06 (0.77–1.45)
P interaction		0.13			0.93
Meat and fat pattern (factor score)
Tertile 1	235/372	1	64/126	0.73 (0.51–1.04)	106/179	1	149/252	1.04 (0.76–1.44)
Tertile 2	251/388	1.03 (0.82–1.30)	79/112	1.04 (0.74–1.48)	123/176	1.21 (0.86–1.69)	161/279	1.03 (0.75–1.42)
Tertile 3	264/389	1.14 (0.90–1.45)	99/106	1.50 (1.08–2.10)	133/172	1.43 (1.01–2.01)	172/287	1.15 (0.83–1.58)
P interaction		0.047				0.48
Doneness pref.
Rare/medium	373/563	1	127/194	0.95 (0.72–1.25)	205/263	1	238/411	0.77 (0.60–1.00)
Well-done	371/577	0.99 (0.82–1.21)	114/149	1.11 (0.83–1.48)	156/260	0.77 (0.58–1.02)	242/400	0.81 (0.62–1.06)
P interaction		0.42				0.09
Pack-years
0	278/475	1	82/187	0.70 (0.51–0.96)	119/223	1	183/360	0.98 (0.73–1.32)
>0-<30	336/520	1.14 (0.93–1.42)	113/120	1.61 (1.18–2.20)	165/232	1.39 (1.02–1.89)	223/350	1.30 (0.97–1.75)
≥30	136/154	1.48 (1.11–1.97)	47/37	2.02 (1.26–3.24)	78/72	1.98 (1.33–2.96)	76/108	1.33 (0.91–1.96)
P interaction		0.003				0.34
Total HCA (ng/1000 kcal/d)
0-<217.3	99/363	1	28/111	0.82 (0.50–1.34)	57/202	1	54/228	0.89 (0.58–1.39)
217.3-<566.9	108/354	1.19 (0.87–1.64)	35/114	1.01 (0.63–1.62)	52/163	1.15 (0.75–1.79)	74/265	1.11 (0.73–1.68)
566.9+	96/374	1.13 (0.81–1.57)	24/100	0.88 (0.52–1.48)	45/139	1.36 (0.86–2.15)	53/281	0.82 (0.53–1.27)
P interaction		0.97				0.29
DiMeIQX (ng/1000 kcal/d)
0-<1.8	93/374	1	22/100	0.80 (0.47–1.38)	55/192	1	46/237	0.74 (0.47–1.16)
1.8-<6.2	105/341	1.32 (0.96–1.83)	40/128	1.13 (0.72–1.78)	48/162	1.04 (0.67–1.63)	76/259	1.13 (0.74–1.72)
6.2+	105/376	1.29 (0.93–1.79)	25/97	0.97 (0.58–1.64)	51/150	1.35 (0.86–2.12)	59/278	0.88 (0.57–1.36)
P interaction		0.93				0.22
MeIQX (ng/1000 kcal/d)
0-<29.8	96/356	1	28/117	0.78 (0.48–1.29)	54/198	1	49/229	0.84 (0.54–1.33)
29.8-<93.5	102/367	1.07 (0.77–1.47)	35/102	1.08 (0.67–1.74)	60/161	1.35 (0.88–2.08)	67/260	1.01 (0.66–1.55)
93.5+	105/368	1.22 (0.88–1.70)	24/106	0.83 (0.49–1.40)	40/145	1.18 (0.73–1.91)	65/285	1.01 (0.66–1.54)
P interaction		0.497				0.89
PhIP (ng/1000 kcal/d)
0-<171.1	102/367	1	28/106	0.83 (0.50–1.36)	58/204	1	56/228	0.94 (0.61–1.46)
171.1-<460.5	106/352	1.13 (0.83–1.55)	35/119	0.94 (0.59–1.50)	51/165	1.13 (0.73–1.76)	71/263	1.05 (0.70–1.60)
460.5+	95/372	1.11 (0.80–1.53)	24/100	0.87 (0.51–1.46)	45/135	1.41 (0.89–2.23)	54/283	0.83 (0.54–1.29)
P interaction		0.98				0.25

^*for interaction was based on the likelihood ratio test (df=2) comparing a main effects only model and a model including interaction between genotype and exposure represented as two indicator variables.

A statistically significant interaction was also observed for pack-years of smoking and NAT2 (p=0.003 for both the age-, sex-, and ethnicity adjusted and the multivariate adjusted analysis) on the risk of CRC (Table 3). The rapid NAT2 acetylators who smoked more than 30 pack-years had a higher CRC risk (OR of 2.88 = 2.02 / 0.70 (95% CI: 1.73–4.82)) than slow/intermediate NAT2 aceylators who also smoked more than 30 pack-years (1.48 (1.11–1.97)), compared to individuals in the same NAT2 genotype category who never smoked. The corresponding ORs (95% CIs) for rectum cancer were 5.20 (2.42–11.17) for the NAT2 rapid acetylators and 1.93 (1.24–3.02) for NAT2 slow/intermediate acetylators (p for interaction = 0.02), and for colon cancer, 2.19 (1.23–3.90) and 1.35 (0.98–1.85); p for interaction = 0.04, respectively. None of the remaining associations differed statistically significantly by NAT2 acetylation status, neither in the sex-age-ethnicity adjusted (Table 3), nor in the multivariate adjusted models (data not shown). No statistically significant interaction between pack-years of smoking and NAT2 was present when we collapsed the intermediate and rapid phenotypes and compared to the slow phenotypes (p=0.31).

No statistically significant effect modification by NAT1 genotype was present, but the positive association with pack-years of smoking appeared to be somewhat stronger for the NAT1*10 non-carriers (1.98 (1.33–2.96)) than for the NAT1*10 carriers (1.36 (0.95–1.94), p interaction=0.34) (Table 3).

Combined analysis of NAT1 and NAT2 did not clarify the associations noted above (data not shown). The OR (95% CI) for at least 30 pack-years of smoking with presence of the rapid NAT2 phenotype and NAT1*10 compared to never smokers with slow/intermediate phenotypes in at least one enzyme was 2.15 (1.17–3.95), p for interaction = 0.10.

Discussion

In this nested case-control study, the increased CRC risk associated with smoking was greater among participants with the rapid compared to those with the slow/intermediate NAT2 genotype, and the interaction was statistically significant (p=0.003). Associations with smoking were stronger for rectal cancer than for colon cancer. A weak positive association with a dietary pattern that includes a high intake of fat and meat was also suggested, which appeared to be somewhat more marked among rapid NAT2 acetylators (p for interaction: 0.05). , No clear evidence was found for main effect or interaction with NAT2 or NAT1 genotype for red and processed meat intake, meat doneness preference, or estimated exposure to HCA from meat in a subset of subjects, with regard to CRC risk.

Evidence suggesting that smoking, especially long-term smoking, may be an important risk factor for CRC has emerged in the last 15 years, although the data have not been completely consistent (5, 28). Both recent IARC and U.S. Surgeon General reports concluded that there was insufficient evidence for including CRC among tobacco-related malignancies (29, 30). The present study, based on prospectively collected data, provides further evidence for an association between smoking on CRC risk, particularly among individuals carrying the rapid NAT2 genotype.

Higher risks of CRC and colorectal adenoma associated with smoking have been reported with presence of the NAT2 rapid acetylator genotype (10, 14, 31), but also with the NAT2 slow acetylator genotype (32–34), whereas other studies have reported no modifications by NAT2 genotype (8, 35–37). Fewer studies have investigated the modifying effect of NAT1 genotypes on the smoking and CRC association (32, 37, 38). Consistent with our findings, one case-control study found a slightly elevated OR with smoking among non-carriers of the NAT1*10 allele compared to carriers of the NAT1*10 allele (32). However, two other case-control studies did not observe any effect modification (37, 38). Thus, overall, the findings for NAT1 or NAT2 polymorphisms and smoking on CRC risk are largely inconsistent, with only suggestions of effect modification. However, some studies had low power to formally test for statistical interaction (35–37) and few were prospective in design (8, 10, 35, 37, 38) and none, except for a previous study from our group (26), were able to examine the rapid NAT2 genotype separately. In our data, no increased risk was observed when the rapid acetylator genotype was collapsed with the intermediate genotype and compared to the slow genotype, as most past studies have done. If replicated, these findings would provide support to the role in CRC of specific HCAs that are particularly abundant in tobacco smoke, such as 2-amino-9H-pyridol[2,3-b]indole (39).

The stronger association of smoking with rectal cancer, as compared to colon cancer, in the present study agrees with our earlier findings in a large population based case-control study in Hawaii (9). Current smoking was also reported to be a risk factor for rectal cancer among men, but not among women in a Dutch study (40). However, other studies found higher risks associated with smoking for colon cancer, as opposed to rectal cancer (10), or no difference in risk by anatomical subsite (38).

In the present nested case-control study, only a weak association was observed with processed meats and none for red meat. Although red meat has been associated with CRC in previous studies, the association has been stronger for processed meats (13). We also found that the association with processed meats, and particularly a dietary pattern that includes high intake of fat and meat, appeared somewhat stronger among rapid acetylators. This is consistent with our finding in a recent ecological study based on 27 countries that adjusting for population frequency of the intermediate/fast NAT2 genotype significantly improves the correlation between per capita meat consumption and CRC incidence and that both factors combined explain about 80% of the international variation in CRC incidence (41).

Several other past studies supported the hypothesis that NAT enzymes play a role in the association between CRC or colorectal adenoma and meat intake (32–35, 42, 43). Some studies observed the positive association in conjunction with smoking (33, 35, 42), or with a combination of NAT2 intermediate/rapid and NAT1*10 (32, 43). However, several studies found no or only slight modifications of the associations between meat and CRC by acetylation genotypes/phenotypes (36, 37, 44, 45). Overall, past studies suggested a somewhat elevated CRC risk with meat intake among NAT2 intermediate/rapid acetylator genotypes, but, when performed, tests for gene-environment interactions most often did not reach statistical significance. Thus, the evidence from past studies has been largely inconsistent.

Studies that have attempted to estimate dietary HCA exposure have either reported no associations with intake of MeIQx, DiMeIQx or PhIP (46, 47), or a significantly increased risk, especially for MeIQx (48, 49) with colorectal adenoma. Studies of CRC found significantly positive associations with exposure to dietary HCA (50, 51), or no association (52). Very few studies to-date have investigated the interaction of HCA intake and NAT polymorphisms on CRC risk. A case-control study suggested that high intake of MeIQx in presence of the NAT1*10 allele increased adenoma risk (53). Another case-control study found a positive association of HCA with colon cancer in African-Americans when NAT1*10 allele was present, whereas the non-carriers of NAT1*10 were found to be at higher colon cancer risk with high HCA intake in Caucasians (54). It should be noted that the present study had limited power to validly estimate ORs for HCA intake by NAT genotypes for the different ethnic groups based on the current follow-up.

Meat preparation methods or preferred “brownness” or doneness of meat, often used as surrogates for HCA intake, were not associated with risk of colorectal adenoma in a Dutch case-control study (42) or CRC in several other studies (reviewed in (16)). In contrast, in a population-based case-control study specifically designed to test this hypothesis in Hawaii, preference for well-done meat, in conjunction with the rapid NAT2 genotype and high CYP1A2 enzyme activity, was positively associated with CRC risk (26, 55). This association was especially pronounced in smokers. A mutagen index – calculated as an estimate for exposure to mutagenic or carcinogenic substances based on amount, doneness and method of cooking meat – has been suggested to be associated with rectal cancer in men, but the association was neither statistically significant nor modified by NAT2 imputed phenotype, and was not present in women (44). In a study of colon cancer, a borderline significant positive association was observed with a mutagen index, which was strongest in NAT2 intermediate or rapid acetylators (45).

Consistent with our results, studies investigating the associations of NAT1 or NAT2 polymorphisms on CRC risk provided limited support for a main effect association of NAT1 or NAT2 genotype with CRC risk. A recent meta-analysis of these genes and CRC identified three studies on NAT1, and 10 studies on NAT2 (56). The authors calculated a non-significant elevation in risk with NAT1*10 (OR 1.25 (95% CI: 0.96–1.63)), and no association for the NAT2-rapid acetylator genotype (OR 1.05 (95% CI: 0.94–1.14)). It should be noted, however, that the genotype-phenotype relationship for NAT1 remains unclear at this point. A lack of main effect for the rapid NAT2 genotype is consistent with the fact that CRC incidence in Japanese (who are often rapid acetylator) only increased when they became exposed to a meat-rich western diet, either when they migrated to the US, or more recently in Japan (24).

The present study has several limitations that need consideration. First, we only considered two metabolic enzymes (NAT2 and NAT1). Since additional enzymes are involved in the bioactivation and detoxification of HCA, they may also play a role in modifying the associations of smoking or red meat intake and CRC (26), increasing misclassification for the variables we did measure. Also, we were unable to study the associations of single variant alleles (except NAT1*10) with CRC due to their low frequencies. Furthermore, multiple testing may have increased the risk of chance findings in our study. For the analysis of HCA, power was limited due to a reduced number of incident cases with the available follow-up, especially for analyses of interactions and subgroups. Moreover, although a validation study comparing the HCA intake values derived from a meat module similar to the one used in our study with those from a 12-day food diary obtained deattenuated correlation coefficients between 0.36 and 0.60 (22), there probably remains a considerable amount of measurement error in estimating HCA intake from diet.

The nested case-control study approach is one of the main strengths of this investigation. The assessment of diet and smoking before disease occurrence minimizes recall bias, which is thought to be a serious concern in case-control studies. Furthermore, the study was particularly comprehensive in the number of alleles genotyped and had a large number of subjects with the rapid NAT2 genotype. The study was population-based and multiethnic, suggesting that the findings should be broadly generalizable to the general population.

In conclusion, this study provides only relatively weak support for a role of HCA from meat in CRC. However, smoking was found to be more clearly associated with CRC and its effect was significantly modified by NAT2 genotype, suggesting that HCA in tobacco smoke may play a role in CRC. Finally, the data also suggested that the association with smoking may be stronger for rectal cancer than colon cancer.