Cigarette Smoking, Genetic Variants in Carcinogen-metabolizing Enzymes, and Colorectal Cancer Risk

Abstract

The risk of colorectal cancer associated with smoking is unclear and may be influenced by genetic variation in enzymes that metabolize cigarette carcinogens. The authors examined the colorectal cancer risk associated with smoking and 26 variants in carcinogen metabolism genes in 1,174 colorectal cancer cases and 1,293 population-based controls recruited in Canada by the Ontario Familial Colorectal Cancer Registry from 1997 to 2001. Adjusted odds ratios were calculated by multivariable logistic regression. Smoking for >27 years was associated with a statistically significant increased colorectal cancer risk (adjusted odds ratio (AOR) = 1.25, 95% confidence interval (CI): 1.02, 1.53) in all subjects. Colorectal cancer risk associated with smoking was higher in males for smoking status, duration, and intensity. The CYP1A1-3801-CC (AOR = 0.47, 95% CI: 0.23, 0.94) and CYP2C9-430-CT (AOR = 0.82, 95% CI: 0.68, 0.99) genotypes were associated with decreased risk, and the GSTM1-K173N-CG (AOR = 1.99, 95% CI: 1.21, 3.25) genotype was associated with an increased risk of colorectal cancer. Statistical interactions between smoking and genetic variants were assessed by comparing logistic regression models with and without a multiplicative interaction term. Significant interactions were observed between smoking status and SULT1A1-638 (P = 0.02), NAT2-857 (P = 0.01), and CYP1B1-4390 (P = 0.04) variants and between smoking duration and NAT1-1088 (P = 0.02), SULT1A1-638 (P = 0.04), and NAT1-acetylator (P = 0.03) status. These findings support the hypothesis that prolonged cigarette smoking is associated with increased risk of colorectal cancer and that this risk may be modified by variation in carcinogen metabolism genes.

Colorectal cancer is the third most common malignancy in North America (1). Approximately 20% of patients with colorectal neoplasia have a family history of colorectal cancer implying a significant genetic contribution in this disease (2). Because well-recognized genetic predisposition syndromes account for less than 3% of colorectal cancer (2, 3), low-penetrance genetic factors alone or in combination with environmental factors likely contribute to colorectal cancer development. Furthermore, individual susceptibility to risk factors may be influenced by inherited genetic variation in genes encoding carcinogen-metabolizing enzymes (4).

Cigarette smoke contains a variety of carcinogenic compounds, including polycyclic aromatic hydrocarbons, nitrosamines, and aromatic amines (5–7). To date, cigarette smoking has been linked to increased risk of lung, bladder, gastric, kidney, and pancreatic cancer (8); however, the association with colorectal cancer has been less clear. Colonic mucosal cells may be exposed to mutagenic compounds in cigarette smoke through the circulatory system (9), as well as through direct ingestion (10). Recent cohort studies suggest that prolonged and intense smoking may cause colorectal cancer following a significant lag period (11–16). Case-control studies examining smoking and colorectal cancer (17–24) or adenomas (23, 25) have suggested a probable association, although results have not always been consistent or studies have lacked the statistical power to conclusively demonstrate an association.

The colorectal cancer risk associated with smoking may be modified by the ability to detoxify the carcinogenic compounds derived from cigarette smoke (4). Carcinogens and other environmental contaminants are metabolized by phase I enzymes, which oxidize substrates, and phase II enzymes, which metabolize oxidized substrates into excretable products (26). Genes encoding phase I enzymes, such as the cytochrome P-450 gene (CYP) superfamily, and phase II enzymes, such as the glutathione-S-transferase gene (GST) and the N-acetyltransferase gene (NAT), exhibit significant genetic variability among individuals (27).

To date, there has been limited study of colorectal cancer risk and potential interactions between cigarette smoking and variation in genes encoding carcinogen-metabolizing enzymes. Previous studies either had limited power to study gene-environment interactions or focused on a limited number of genetic variants (17, 28–30). To examine risk associated with cigarette smoking and possible interactions between smoking and a broad panel of variants in carcinogen-metabolism genes, we studied 2,467 colorectal cancer cases and population-based controls in Ontario, Canada.

MATERIALS AND METHODS

All subjects were obtained through the Ontario Familial Colorectal Cancer Registry, 1 of 6 international sites of the US National Cancer Institute-supported Colorectal Cancer Family Registry (31). The design and methodology of the Ontario Familial Colorectal Cancer Registry have been described previously (32–34) and are briefly summarized below.

Study population

All incident, pathology-confirmed (International Classification of Diseases, Ninth Revision, codes 153.0–153.9, 154.0–154.3, 154.8) cases of invasive colorectal cancer who were aged 20–84 years and diagnosed between July 1, 1997, and June 30, 2000, in Ontario, Canada, were identified through the population-based Ontario Cancer Registry and recruited into the Ontario Familial Colorectal Cancer Registry. A reabstraction study was able to link 95% of persons in the Ontario Cancer Registry to population assessment rolls, suggesting that its accuracy and completeness are high (35).

Population-based control subjects were recruited by the Ontario Familial Colorectal Cancer Registry using 2 methods with identical eligibility criteria and were frequency matched to cases by sex and 5-year age group. An age- and sex-stratified random sample of control subjects were identified by using a list of Ontario residential telephone numbers provided by Infodirect (Bell Canada, Montréal, Québec, Canada) (1999–2000) and from the telephone listing of all Ontario residents based on population-based assessment rolls (2001). Controls were contacted by telephone and invited to participate.

Data collection

Consent to approach colorectal cancer patients was requested from physicians and obtained for 90% of cases. Cases were asked to complete a mailed family history questionnaire. Based on the pedigrees constructed, each case was then classified as 1) high familial risk (satisfying hereditary nonpolyposis colorectal cancer (HNPCC) Amsterdam criteria) (36), 2) intermediate familial risk, or 3) low (sporadic) risk (refer to Cotterchio et al. (32) for details). All high and intermediate risk cases and a 25% random sample of low risk cases were selected to participate in phase 2 of the Ontario Familial Colorectal Cancer Registry. Phase 2 subjects were asked to complete a self-administered, mailed, epidemiologic risk factor questionnaire and to provide a blood sample. The response rate for colorectal cancer cases was approximately 61%, and participation was not influenced by gender or stage of colorectal cancer. However, response rates were higher among urban cases than those in rural areas. A detailed description of participants’ details and response characteristics of the Ontario Familial Colorectal Cancer Registry has been provided by Cotterchio et al. (32). Control subjects were mailed the family history and epidemiologic questionnaires and asked to provide a blood sample.

Epidemiologic and smoking information

Smoking exposure was determined on the basis of 7 specific questions in the epidemiologic questionnaire adopted by the Colorectal Cancer Family Registry and the Ontario Familial Colorectal Cancer Registry. Respondents were asked to provide their current smoking status (at the time of questionnaire completion), ages, and years of starting smoking and cessation, total number of years of active smoking, and the average number of cigarettes smoked per day when actively smoking (37). All smoking variables were calculated from a reference point of 2 years prior to colorectal cancer diagnosis for cases to limit the potential effects of colorectal cancer symptoms, diagnosis, and treatment on smoking-related behavior. The reference date for controls was selected as 2 years prior to the midpoint of recruitment. Current smoking was defined as smoking ≥1 cigarettes per day for ≥3 consecutive months at the reference date of 2 years prior to the diagnosis date for cases or the midpoint of recruitment for controls. “Ever” smokers were defined as respondents who reported not actively smoking 2 years prior to the reference/diagnosis date but having smoked ≥1 cigarettes per day for ≥3 consecutive months at some time; “never” smokers were defined as those who did not report ever smoking cigarettes or those whose smoking exposure fell below this threshold.

Response rates/numbers

A total of 1,536 incident colorectal cancer cases were selected to participate in phase 2 of the Ontario Familial Colorectal Cancer Registry; 1,327 cases (86%) provided a blood sample and underwent genotype testing, of whom 1,174 (88%) completed the epidemiologic questionnaire and were included in the analysis. Among the 1,174 colorectal cancer cases included, 4.4% are high (hereditary nonpolyposis colorectal cancer) risk, 34% are intermediate (familial criteria) risk, 14% are intermediate (pathology criteria only) risk, and 47% are low/sporadic risk.

Of the 4,876 eligible controls identified and invited to participate, 2,131 refused (43%), and 2,745 were mailed the questionnaire package. Out of the 1,944 controls (71%) who completed the epidemiology questionnaire, 1,293 (67%) provided a blood (DNA) sample and underwent genotype testing.

DNA preparation and genotyping

The Ontario Familial Colorectal Cancer Registry obtained 40 mL of peripheral blood from participating cases and controls. DNA was extracted from lymphocytes by either phenol-cholorform extraction or spin columns (Qiagen, Inc., Valencia, California) and stored at 4^oC.

Variants in genes encoding phase I and phase II enzymes known to be involved in the metabolism of smoking carcinogens were identified through literature searches and the National Center for Biotechnology Information single nucleotide polymorphism database (http://www.ncbi.nlm.nih.gov/sites/entrez?db = snp). Variants with a minor allele frequency of ≥5% were selected, with preference given to sequence variants with potential impact on enzyme function. The genetic variants tested are listed in Table 1. Genotyping of single nucleotide polymorphisms was performed with a TaqMan 5′ nuclease allele discrimination assay with allele-specific fluorescent probes using the ABI PRISM 7900HT Sequence Detection System (Applied Biosystems, LLC, Foster City, California). The genotype of each sample was determined by using the graphical view of relative fluorescent intensities from the Sequence Detection System, Version 2.1 (Applied Biosystems, LLC). All sample plates tested contained appropriate controls of each genotype. The uridine diphosphate-glucuronosyltransferase genotype (UGT1A1*28) was determined by microsatellite analysis, and the glutathione S-transferase M1 gene (GSTM1) and the glutathione S-transferase T1 gene (GSTT1) locus deletions variants were detected by multiplex polymerase chain reaction (38, 39). Five percent of the samples tested were randomly selected for blinded duplicate analysis.

Table 1.

List of the Genetic Variants Tested in Cases and Controls Recruited by the Ontario Familial Colorectal Cancer Registry, Ontario, Canada, during 1997–2001

Gene	Variant	NCBI dbSNP no.
CYP1A1	c.2455A>G	rs1048943
	c.3801T>C	rs4646903
CYP1A2	c.−163C>A	rs762551
CYP1B1	c.142C>G	rs10012
	c.4326C>G	rs1056836
	c.4390A>G	rs1800440
CYP2C9	c.430C>T	rs1799853
	c.1075A>C	rs1057910
CYP2E1	c.7632T>A	rs6413432
NAT1	c.459G>A	rs4986990
	c.1088T>A	rs8190861
NAT2	c.191G>A	rs1801279
	c.341T>C	rs1801280
	c.590G>A	rs1799930
	c.857G>A	rs1799931
EPHX1	c.17673T>C	rs1051740
COMT	c.472A>G	rs5031015
SULT1A1	c.638G>A	rs9282861
GSTM1	Gene deletion
	c.597G>C (K173N)	rs74837985
GSTM3	delAGG	rs1799735
GSTT1	Gene deletion
UGT1A1	*28 A(TA)6TAA>A(TA)7TAA
UGT1A7	W208R	rs11692021
	N129K	rs17868323
AHR	c.1661G>A	rs2066853

Abbreviations: dbSNP, Single Nucleotide Polymorphism Database; NCBI, National Center for Biotechnology Information.

All sequence alterations, except deletion variants (GSTM1 and GSTT1), were tested for departures from Hardy-Weinberg equilibrium (HWE) in the control group by using PROC ALLELE in SAS, version 9, software (SAS Institute, Inc., Cary, North Carolina). The analysis was performed by using 10,000 permutations to approximate an exact P value for the HWE test as well as 1,000 bootstrap samples to obtain the confidence interval for the allele frequencies and 1-locus HWE coefficients. On the basis of the significance level of 0.05, only CYP1A1-2455 demonstrated a departure from HWE.

Statistical analysis

Associations between smoking and colorectal cancer risk were assessed by computing age-adjusted odds ratios (40) for smoking status, total number of years smoked, and pack-years smoked. The age at which subjects began smoking was also assessed but did not contribute significantly to the study findings. Logistic regression was performed to obtain multivariable-adjusted odds ratio estimates while simultaneously controlling for potential confounders. Variables were considered potential confounders if their inclusion in a multivariate model resulted in a ≥10% change in odds ratio estimates. Potential confounders included the following: sex; inflammatory bowel disease (ulcerative colitis or Crohn's disease); intake of nonsteroidal antiinflammatory drugs; aspirin use; calcium supplement use; body mass index prediagnosis; vegetable, fruit, and red meat consumption; colonic screening (true screening endoscope); alcohol consumption; and first degree family history of colorectal cancer. In addition, cases were stratified on cancer site (colon vs. rectum) and microsatellite instability (MSI) tumor status, and the association between smoking and colon/rectal cancer was assessed within each stratum. Sex was identified as a confounder in our data set, and it modified the main smoking association; as a result, the colorectal cancer risk associated with smoking was stratified by sex, and the final multivariable models contained age and sex.

Interactions between the selected genetic variants and smoking variables were assessed by stratified analyses. Composite variables were derived from combinations of single nucleotide polymorphisms in the same gene based on known or reported consequences of specific variant combinations on enzyme function. Interactions were formally assessed by the statistical significance (P < 0.05) of the likelihood ratio statistic comparing models with and without a multiplicative interaction term (variant × smoking).

Ethics approval

Ethics approval was granted from the Office of Research Services, University of Toronto, and the research ethics boards of Mount Sinai Hospital and the Hospital for Sick Children.

RESULTS

A total of 1,174 colorectal cancer cases and 1,293 controls had complete data on smoking status, potential confounders, and genotyping results for all 26 genetic variants. The distributions and age- and sex-adjusted odds ratios of all colorectal cancer cases and controls by smoking characteristics are shown in Table 2. Among all subjects, smoking for >27 years was associated with a statistically significant increased risk of colorectal cancer (age- and sex-adjusted odds ratio (AOR) = 1.25, 95% confidence interval (CI): 1.02, 1.53), and >22 pack-years smoked was suggestive of increased colorectal cancer risk (AOR = 1.24, 95% CI: 0.99, 1.55) of borderline statistical significance. All other smoking variables were associated with increased colorectal cancer risk that did not reach statistical significance. When the analysis was stratified by sex (Table 3), a significantly increased colorectal cancer risk was observed in men who were former smokers, current smokers, and those who smoked for ≥15 years or ≥9 pack-years. There was no significant colorectal cancer risk associated with smoking in women, and there was evidence of a significant interaction between sex and pack-years smoked (P_interaction = 0.004). No association between cancer site (rectum vs. colon) or MSI status and smoking was observed.

Table 2.

Distribution of Colorectal Cancer Cases and Controls Recruited by the Ontario Familial Colorectal Cancer Registry, Ontario, Canada, during 1997–2001, and Odds Ratio Estimates for Smoking Variables for All Subjects

Variable	All Subjects
	Casesa		Controlsa		ASORb	95% CI
	No.	%	No.	%
Age, years
20–44	54	5	46	4
45–49	80	7	54	4
50–54	153	13	167	13
55–59	204	17	247	19
60–64	262	22	285	22
69–69	272	23	290	22
70–75	149	13	204	16
Sex
Male	476	41	725	56
Female	698	59	568	44
Smoking status
Never smoked	487	42	549	43	1.0
Former smoker	521	45	556	43	1.18	0.99, 1.41
Current smoker	158	14	184	14	1.25	0.96, 1.62
Total number of years smoked
Never smoked	487	43	549	43	1.0
<15	136	12	163	13	0.99	0.76, 1.29
15–27	202	18	226	18	1.20	0.95, 1.52
>27	311	27	330	26	1.25	1.02, 1.53
Pack-yearsc smoked
Never smoked	487	45	549	45	1.0
<9	180	17	201	16	1.04	0.82, 1.32
9–22	179	17	209	17	1.17	0.92, 1.49
>22	237	22	262	21	1.24	0.99, 1.55

Abbreviations: ASOR, age- and sex-adjusted odds ratio; CI, confidence interval.

^aNumbers may not add up to totals because of missing values (<5% of cases).

^b“Age” defined as age at colorectal cancer diagnosis for cases and referent date (June 30, 1999) for controls.

^cAverage packs per day × number of years smoked.

Table 3.

Distribution of Colorectal Cancer Cases and Controls Recruited by the Ontario Familial Colorectal Cancer Registry, Ontario, Canada, during 1997–2001, and Odds Ratio Estimates for Smoking Variables for Males and Females Separately

Variable	Males						Females						Pinteraction
	Casesa		Controlsa		AORb	95% CI	Cases		Controls		AORb	95% CI
	No.	%	No.	%			No.	%	No.	%
Smoking status
Never smoked	140	30	258	36	1.0		347	50	291	51	1.0
Former smoker	250	53	376	52	1.35	1.03, 1.76	244	35	180	32	1.14	0.89, 1.46
Current smoker	82	17	89	12	1.74	1.20, 2.52	103	15	95	17	0.91	0.66, 1.25	0.08
Total no. of years smoked
Never smoked	140	30	258	36	1.0		347	51	291	52	1.0
<15	49	11	97	14	0.97	0.65, 1.46	87	13	66	12	1.09	0.76, 1.56
15–27	113	25	154	22	1.46	1.06, 2.02	89	13	72	13	1.03	0.73, 1.46
>27	158	34	199	28	1.64	1.21, 2.23	154	23	131	23	1.00	0.75, 1.32	0.186
Pack-yearsc smoked
Never smoked	140	32	258	38	1.0		347	54	291	54	1.0
<9	52	12	112	16	0.89	0.60, 1.32	128	20	89	17	1.19	0.87, 1.63
9–22	102	23	141	21	1.50	1.07, 2.10	77	12	68	13	0.95	0.66, 1.37
>22	147	33	171	25	1.76	1.29, 2.40	90	14	91	17	0.82	0.59, 1.14	0.004

Abbreviations: AOR, age-adjusted odds ratio; CI, confidence interval.

^aNumbers may not add up to totals because of missing values (<5% of cases).

^b“Age” defined as age at colorectal cancer diagnosis for cases and referent date (June 30, 1999) for controls.

^cAverage packs per day × number of years smoked.

All cases and controls were genotyped for 26 variants in 15 genes of phase I and phase II carcinogen-metabolizing enzymes; genotype frequencies and 6 derived phenotypic variants are shown in Table 4. The CYP1A1-3801CC and CYP2C9-430CT genotypes were associated with statistically significant decreased colorectal cancer risk, and GSTM1-K173N-CG was associated with an increased colorectal cancer risk. Decreased CYP2C9 activity and the NAT2-341CC genotype were associated with borderline significant reduced colorectal cancer risk.

Table 4.

Distribution of Colorectal Cancer Cases and Controls Recruited by the Ontario Familial Colorectal Cancer Registry, Ontario, Canada, during 1997–2001, and Age- and Sex-adjusted Odds Ratio for Selected Polymorphisms in Carcinogen Metabolism Genes

Gene Variant	Cases (n = 1,174)		Controls (n = 1,293)		ASOR	95% CI
	No.	%	No.	%
CYP1A1_2455a
AA	1,052	90	1,166	90	1.0
AG	98	8	114	9	0.95	0.71, 1.27
GG	10	1	8	1	1.37	0.53, 3.55
CYP1A1_3801
TT	904	77	1,004	78	1.0
TC	247	21	260	20	1.07	0.88, 1.31
CC	12	1	26	2	0.47	0.23, 0.94
CYP1A1 combined variants (derived)b
Wild type	908	77	1,009	78	1.0
Increased activity	249	21	277	21	1.01	0.83, 1.23
CYP1A2_−163
AA	598	51	648	50	1.0
AC	461	39	517	40	0.95	0.80, 1.13
CC	106	9	125	10	0.91	0.68, 1.21
CYP1B1_142
CC	593	51	651	50	1.0
CG	478	41	529	41	0.99	0.83, 1.23
GG	92	8	112	9	0.90	0.67, 1.22
CYP1B1_4326
CC	391	33	424	33	1.0
CG	547	47	617	48	0.96	0.80, 1.15
GG	224	19	250	19	1.00	0.79, 1.26
CYP1B1_4390
AA	775	67	897	69	1.0
AG	354	30	349	27	1.16	0.97, 1.39
GG	34	3	46	4	0.88	0.55, 1.39
GG/AG	388		395		1.13	0.95, 1.34
CYP1B1 combined variants (derived)c
Wild type	337	29	360	28	1.0
Increased activity	821	70	930	72	0.96	0.80, 1.15
CYP2C9_430
CC	886	76	942	73	1.0
CT	255	22	331	26	0.82	0.68, 0.99
TT	23	2	19	1	1.35	0.73, 2.53
CT/TT	278	24	350	27	0.85	0.7, 1.02
CYP2C9_1075
AA	1,014	86	1,115	86	1.0
AC	149	13	167	13	1.02	0.80, 1.30
CC	2	0.2	10	1	0.23	0.05, 1.05
CYP2C9 combined variants (derived)d
Wild type	759	65	791	61	1.0
Decreased activity	403	34	500	39	0.86	0.73, 1.01
CYP2E1_7632
TT	925	79	1,032	80	1.0
AT	226	19	246	19	1.01	0.82, 1.24
AA	14	1	13	1	1.22	0.56, 2.64
NAT1_459
GG	1,106	94	1,227	95	1.0
AG	56	5	60	5	1.0	0.68, 1.46
AA	0		0
NAT1_1088
TT	709	60	790	61	1.0
AT	398	34	430	33	1.04	0.87, 1.24
AA	56	5	70	5	0.90	0.62, 1.30
NAT1 acetylator status (derived)e
Slow acetylator	663	56	737	57	1.0
Fast acetylator	496	42	547	42	1.00	0.85, 1.18
NAT2_191
GG	1,161	99	1,289	99	1.0
AG	5	0.4	3	0.2	1.78	0.41, 7.63
AA	0		0
NAT2_341
TT	392	34	395	31	1.0
CT	562	48	634	49	0.89	0.74, 1.07
CC	209	18	263	20	0.79	0.63, 1.0
NAT2_590
GG	577	49	662	51	1.0
AG	502	43	529	41	1.06	0.90, 1.26
AA	87	7	99	8	1.03	0.75, 1.41
NAT2_857
GG	1,105	94	1,233	95	1.0
AG	58	5	58	4	1.09	0.74, 1.59
AA	2	0.1	1	0	2.57	0.23, 28.8
NAT2 acetylator status (derived)f
Slow acetylator	298	25	363	28	1.0
Fast acetylator	864	74	926	72	1.13	0.94, 1.35
EPHX1_17673
TT	561		625		1.0
CT	502		549		1.01	0.85, 1.20
CC	100		118		0.97	0.72, 1.30
COMT_472
AA	298		350		1.0
AG	586		626		1.11	0.92, 1.35
GG	279		315		1.06	0.84, 1.33
SULT1A1_638
GG	544		598		1.0
GA	502		540		1.02	0.86, 1.22
AA	118		154		0.87	0.66, 1.14
GSTM1 locus deletion
Deletion (no activity)	616		684		1.0
No deletion (active)	550		608		1.00	0.85, 1.17
GSTM1-K173N
CC	320	58	376	62	1.0
CG	48	9	29	5	1.99	1.21, 3.25
GG	185	33	206	34	1.09	0.85, 1.41
GSTM3 delAGG
AGG/AGG	843		924		1.0
AGG/-	281		336		0.91	0.76, 1.10
-/-	40		30		1.45	0.89, 2.38
GSTT1 locus deletion
0 (no activity)	213	18	223	17	1.0
1 (wild-type)	953	81	1,067	83	0.93	0.75, 1.14
UGT1A1*28
TA6/TA6	512	44	593	46	1.0
TA6/TA7	514	44	567	44	1.04	0.87, 1.23
TA7/TA7	127	11	129	10	1.14	0.87, 1.51
UGT1A7_W208R
TT	427	36	479	37	1.0
C/T	556	47	624	48	0.99	0.83, 1.18
C/C	182	16	188	15	1.07	0.84, 1.37
UGT1A7_N129K
G/G	455	39	509	39	1.0
G/T	541	46	599	46	1.01	0.85, 1.21
T/T	145	12	172	13	0.93	0.72, 1.21
UGT1A7 combined variants (derived)g
Wild type	420	36	476	37	1.0
Slightly reduced activity	548	47	616	48	1.00	0.84, 1.20
Very reduced activity	173	15	188	15	1.03	0.80, 1.32
AHR_1661
G/G	922	79	1,019	81	1.0
A/G	219	19	253	20	0.96	0.78, 1.18
A/A	24	2	16	1	1.67	0.87, 3.20

Abbreviations: ASOR, age- and sex-adjusted odds ratio; CI, confidence interval; HWE, Hardy-Weinberg equilibrium.

^aCYP1A1_2455 did not satisfy HWE.

^bCYP1A1 wild type: 2455AA/AG + 3801TT or 2455GG + 3801TC/CC; CYP1A1 increased activity: 2455AA/AG + 3801TC/CC.

^cCYB1B1 wild type: 142CC + 4326CC/CG + 4390AA/AG or 142CG + 4326CC + 4390AA or 42GG + 4326CC + 4390AG; CYP1B1 increased activity: all other combinations.

^dCYP2C9 wild type: 430CC + 1075AA; CYP2C9 decreased activity: 430CC + 1075AC/CC or 430CT + 1075AA/AC or 430TT + 1075AA.

^eNAT1 slow acetylator: 459GG + 1088TT; NAT1 fast acetylator: 459GG + 1088TA/AA or 459GA + 1088TT/TA/AA.

^fNAT2 slow acetylator: 191GG + (341CC + 590GG + 857GG or 341TC + 590GA + 857GG or 341TC + 590GG + 857GA or 341TT + 590GG + 857GA/AA or 341TT + 590GA + 857GA or 341TT + 590AA + 857GG); NAT2 fast acetylator: all other combinations.

^gUGT1A7 wild type: 387TT + 622TT/TC or 387TG/GG + 622TT; UGT1A7 slightly reduced activity: 387TG/GG + 622TC; UGT1A7 very reduced activity: 387GG + 622CC.

We assessed potential interactions between cigarette smoking and all genetic variants listed in Table 4. Each genetic variant was tested for effect modification with 3 smoking variables (smoking status, duration, and pack-years of smoking), and only statistically significant interactions are presented in Table 5. The risk of colorectal cancer associated with smoking duration was modified by the genotypes for NAT1-1088 and sulfotransferase 1a1 (SULT1A1-638), as well as by NAT1 acetylator status. SULT1A1-638, NAT2-857, and CYP1B1-4390 genotypes modified the colorectal cancer risk associated with smoking status. Compared with nonsmokers, carriers of the SULT1A1-638GG genotype who were current or former smokers and those who reported smoking >15 years were at statistically significant increased risk of colorectal cancer, while smoking status or duration did not appear to alter the risk of carriers of the AG or AA genotypes at this locus. Former smokers carrying the CYP1B1-4390AA or the NAT2-857GG genotype were at increased risk of colorectal cancer compared with nonsmoking carriers of those genotypes. For the NAT1-1088 variant, carriers of the TT genotype who smoked 15–27 years and carriers of the AT genotype who smoked >28 years were at statistically significant increased risk of colorectal cancer. Finally, “fast” NAT1 acetylators who were prolonged smokers (>28 years) were at increased risk of colorectal cancer compared with fast-acetylating nonsmokers.

Table 5.

Age- and Sex-adjusted Odds Ratio Estimates and 95% Confidence Intervals for Smoking Status Stratified by Selected Genotypes With a Statistically Significant Interaction (P < 0.05) Between Genotype and Smoking Status Among Colorectal Cancer Cases and Controls Recruited by the Ontario Familial Colorectal Cancer Registry, Ontario, Canada, during 1997–2001

Smoking Variable	Genotype						Pinteraction
	ASOR	95% CI	ASOR	95% CI	ASOR	95% CI
	CYP1B1-4390
	AA (n = 1,672)		AG (n = 703)		GG (n = 80)
Never smoked	1.0		1.0		1.0
Smoking status
Current smoker	1.2	0.9, 1.6	1.34	0.83, 2.16	0.92	0.2, 4.23
Former smoker	1.24	0.99, 1.6	1.01	0.72, 1.42	2.12	0.72, 6.27	0.042
Years smoked
<15	1.02	0.75, 1.4	0.95	0.57, 1.58	0.64	0.09, 4.75
15–27	1.18	0.89, 1.58	1.07	0.69, 1.64	3.66	0.92, 14.55
>27	1.3	1.02, 1.67	1.15	0.78, 1.7	1.32	0.37, 4.63	0.643
Pack-years smoked
<9	0.98	0.73, 1.31	1.13	0.71, 1.8	1.95	0.4, 9.64
9–22	1.23	0.91, 1.61	1.12	0.71, 1.75	1.13	0.25, 5.16
>22	1.29	0.98, 1.71	1.09	0.72, 1.65	1.88	0.55, 6.42	0.389

	NAT1-1088
	TT (n = 1,499)		AT (n = 828)		AA (n = 126)
Never smoked	1.0		1.0		1.0
Smoking status
Current smoker	1.07	0.78, 1.47	1.55	1.03, 2.34	0.81	0.26, 2.51
Former smoker	1.18	0.94, 1.49	1.3	0.95, 1.77	0.81	0.33, 1.95	0.448
Years smoked
<15	0.89	0.64, 1.24	1.28	0.8, 2.06	1.44	0.28, 7.34
15–27	1.49	1.1, 2.03	0.9	0.6, 1.34	0.92	0.32, 2.66
>27	1.09	0.84, 1.42	1.82	1.27, 2.61	0.57	0.21, 1.52	0.015
Pack-years smoked
<9	1.01	0.74, 1.38	1.08	0.72, 1.64	1.42	0.43, 4.7
9–22	1.12	0.82, 1.53	1.58	1.03, 2.42	0.39	0.12, 1.26
>22	1.16	0.87, 1.55	1.49	1.02, 2.18	0.69	0.22, 2.12	0.250

	NAT1-acetylator status
	Fast (n = 1,043)		Slow (n = 1,400)
Never smoked	1.0		1.0
Smoking status
Current smoker	1.4	0.97, 2.02	1.06	0.76, 1.47
Former smoker	1.23	0.93, 1.63	1.2	0.94, 1.52			0.586
Years smoked
<15	1.34	0.88, 2.03	0.85	0.6, 1.21
15–27	0.94	0.66, 1.33	1.5	1.09, 2.06
>27	1.49	1.08, 2.04	1.12	0.85, 1.47			0.027
Pack-years smoked
<9	1.16	0.81, 1.68	0.98	0.71, 1.35
9–22	1.28	0.88, 1.85	1.13	0.82, 1.57
>22	1.31	0.93, 1.85	1.21	0.89, 1.63			0.816

	NAT2-857
	GG (n = 2,337)		AG (n = 116)
Never smoked	1.0		1.0
Smoking status
Current smoker	1.18	0.92, 1.51	3.26	0.75, 14.2
Former smoker	1.24	1.03, 1.5	0.67	0.29, 1.56			0.011
Years smoked
<15	1.02	0.78, 1.34	0.76	0.19, 3.05
15–27	1.22	0.96, 1.56	0.88	0.32, 2.42
>27	1.3	1.05, 1.6	0.8	0.29, 2.17			0.235
Pack-years smoked
<9	1.07	0.83, 1.36	0.72	0.21, 2.56
9–22	1.18	0.92, 1.51	1.65	0.47, 5.79
>22	1.31	1.04, 1.65	0.6	0.21, 1.73			0.199

	SULT1A1-638
	GG (n = 1,142)		GA (n = 1,041)		AA (n = 272)
Never smoked	1.0		1.0		1.0
Smoking status
Current smoker	1.69	1.17, 2.42	0.94	0.65, 1.37	0.73	0.34, 1.6
Former smoker	1.52	1.16, 1.98	1.03	0.78, 1.36	0.82	0.47, 1.44	0.017
Years smoked
<15	1.37	0.94, 2.01	0.8	0.53, 1.21	0.51	0.2, 1.26
15–27	1.64	1.15, 2.33	0.89	0.62, 1.27	1.29	0.62, 2.7
>27	1.62	1.2, 2.18	1.15	0.83, 1.58	0.62	0.33, 1.19	0.038
Pack-years smoked
<9	1.36	0.97, 1.93	0.87	0.6, 1.27	0.64	0.3, 1.39
9–22	1.45	1.02, 2.08	0.99	0.68, 1.45	1.1	0.53, 2.28
>22	1.71	1.23, 2.38	1	0.71, 1.42	0.75	0.36, 1.54	0.119

Abbreviations: ASOR, age- and sex-adjusted odds ratio; CI, confidence interval.

DISCUSSION

Prevention of colorectal cancer can occur through both effective screening (33) and reduction of exposure to modifiable risk factors. Cigarette smoking is a common source of carcinogenic compounds and is thought to increase the risk of colorectal cancer, but the evidence supporting this association has been somewhat inconsistent; as a result, colorectal cancer is not listed among tobacco-related malignancies (8, 18). Early cohort studies through the 1950s–1960s did not demonstrate a relation between smoking and colorectal cancer (41–45), but later studies in the 1970s–1980s and 1980s–1990s began to show a consistent relation between smoking and premalignant colorectal adenomas and invasive cancer, respectively (46–49). Since men and women began smoking heavily in the 1920s and 1940s, respectively, the relation was observed first in studies of males followed later by results in female cohorts (50, 51). These studies confirmed both a dose-response relation supporting a causal association and a significant lag time of 25–30 years (11–14). Furthermore, prolonged and intense smoking has been linked to specific somatic changes in colorectal cancer, with smokers at increased risk for high microsatellite instability (MSI-H) (19, 24), lack of MLH1 expression (17), CpG island methylation, and BRAF mutations (17, 52).

The association between smoking and colorectal cancer may be modified by the ability to detoxify the carcinogenic compounds derived from cigarette burning (27). The activated intermediates of carcinogen metabolism are mutagenic and induce DNA damage through DNA–adduct formation, creation of apurinic sites, and oxidative damage (28, 53). Accumulation of these mutagenic intermediates may be accelerated by increased activity of phase I enzymes, decreased activity of phase II enzymes, or both in combination. Sequence variants of the CYP1A1 gene have been shown to increase the activity of this enzyme, potentially leading to increased activation of carcinogenic metabolites (54, 55). Individuals with homozygous deletions of the GSTM1 or GSTT1 gene do not express the particular enzyme and may be less efficient at metabolizing carcinogen substrates (17).

Previous studies have examined the effects of a limited number of carcinogen-metabolism gene variants on the risk of colorectal neoplasia associated with smoking. Lüchtenborg et al. (17) did not observe an interaction between colorectal cancer-smoking risk and deletion of either GSTM1 or GSTT1 genes. Similarly, Little et al. (30) did not observe any interaction between smoking, GSTM1 and GSTT1 gene deletions, and variants in CYP1A1 and colorectal cancer. Slattery et al. (20) demonstrated an increased risk of rectal cancer in men that was associated with smoking and identified a potential interaction among GSTM1, smoking, and rectal cancer risk, although we did not replicate this finding in our study. Goode et al. (27) studied variants in NAT1, NAT2, SULT1A1, SULT1A2, CYP1A1, and the epoxide hydrolase gene (EPHX1) and smoking in a series of cases with colorectal polyps and controls. Although no individual variants were associated with polyps, Goode et al. did demonstrate a potential gene–gene interaction between NAT1/2 genes and SULT1A1. Their analysis of smoking and genetic variants demonstrated an increased risk of hyperplastic and hyperplastic/adenomatous polyps combined associated with smoking, as well as possible interactions between smoking and SULT1A1-638 variants and potential interactions with CYP1A1-6235 genotypes and the EPHX1 haplotype.

The current study supports the assertion that cigarette smoking is associated with colorectal cancer, and our data are consistent with the hypothesis that prolonged smoking may be required to manifest this risk. Although only people who smoked >27 years were at statistically significant increased risk of colorectal cancer in our series, all other smoking variables were associated with increases in risk that did not reach statistical significance. Furthermore, there is evidence of a dose-response relation for smoking status, duration, and pack-years, implying increasing risk with incremental smoking exposure. We observed an increased colorectal cancer risk associated with smoking predominantly in males with no increased risk seen in females. These findings are consistent with those of other studies that have failed to observe an increased colorectal cancer risk for smoking in females when analyses have been stratified by sex (15, 20, 56–59). Different smoking patterns and less intense smoking in women, as well as protective effects of estrogen, have been suggested to explain the lack of association in women (20, 50). Studies that have demonstrated an increased risk among women have generally shown significant results for women with prolonged or intense smoking histories (13, 14, 60, 61). As a result, the lack of association between smoking and colorectal cancer in women may be due to the lower numbers of women in the highest categories of smoking intensity and duration in our series. The combined effects of smoking and the genetic variants on colorectal cancer risk may differ by sex; however, we did not have a sufficient sample size to adequately test this hypothesis. Future international collaborative studies may be required to achieve sufficient power to perform an adequately powered analysis stratified by sex.

We did not observe any difference in colorectal cancer risk associated with smoking when cases were stratified by MSI status or cancer site (colon vs. rectum), and our study failed to replicate the findings of previous studies linking smoking with MSI-H tumors (17, 19, 21, 24, 62, 63) and rectal cancer (15, 20, 24). Similar to our results, those of Diergaarde et al. (64) did not show an association between MSI-H colorectal cancer and smoking. Alternatively, Samowitz et al. (52) demonstrated that smoking may be more directly associated with CpG island methylation-positive colorectal cancer, which is seen more commonly in MSI-H cancers. The enrichment of familial colorectal cancer cases in the Ontario Familial Colorectal Cancer Registry may increase the proportion of MSI-H CpG island methylation-negative cancers, thus attenuating any association between MSI-H and smoking.

We observed a decreased risk of colorectal cancer associated with the homozygous variant CYP1A1-3801T>C, which was not observed in previous studies (17, 27, 30). This variant is located in the 3′ region of the gene and has been associated with higher inducibility of the CYP1A1 enzyme (65) and increased risk of lung cancer (66), particularly in non-Caucasians (67). Our observation of decreased colorectal cancer risk associated with this allele may be due to different expression patterns in the colonic epithelium; the influence of other variants including CYP1A1-2455A>G, GST, and NAT variants; and the predominantly Caucasian study population. We also observed decreased colorectal cancer risk associated with heterozygous carriers of the CYP2C9-430C>T variant; this missense variant, a R114C substitution, has been associated with decreased colorectal cancer risk (68) and increased polyp risk (69). Consistent with this observation, the current study also noted a decreased odds of colorectal cancer associated with decreased CYP2C9 enzyme activity that approached statistical significance, suggesting a possible role for this enzyme in the metabolism of cigarette carcinogens. We observed an increased risk of colorectal cancer associated with the heterozygous GSTM1-K173N missense variant; interpretation of this finding is difficult because this gene is often deleted. However, because GSTM1 forms dimers in vivo, it is possible that this variant may alter protein–protein interactions within heterodimers. The observation of a protective effect in the heterozygous CYP2C9-430 CT genotype but no altered risk in either homozygous genotype may be related to the low frequency of the homozygous TT genotype.

We observed statistically significant interactions between CYP1B1-4390, NAT1-1088, NAT1 acetylator status, and NAT2-857 and the odds of colorectal cancer associated with smoking, although the pattern of these interactions was not clear. We found statistical interactions between SULT1A1-638 and both smoking status and smoking duration with increased risk of colorectal cancer observed in smokers who carried the 638-GG genotype, strengthening the hypothesis that this enzyme may modify smoking carcinogen metabolism. This nonsynonymous variant leads to a R213H amino acid change, and the variant allele, SULT1A1*2, has been associated with decreased enzyme activity (70). Although associations with this allele and colorectal cancer have been inconsistent, several small studies have suggested increased colorectal cancer risk (71–73) associated with the variant 638A allele but did not show any interaction with smoking. Fan et al. (74) did not detect any interaction among the SULT1A1-638 genotype, smoking, and colorectal cancer risk, although this was a case-only study of only 207 individuals. Data presented by Goode et al. (27) were suggestive of potential interactions among NAT1, NAT2, and SULT1A1 genotypes, as well as a possible interaction between smoking status and SULT1A1 associated with the risk of colorectal polyps. Tiemersma et al. (75) demonstrated a 4-fold increased risk of colorectal adenomas in smokers who carried the SULT1A1-638GG genotype with a significant interaction between this variant and smoking duration (P = 0.03). In vitro studies suggest that SULT1A1 may activate certain procarcinogens found in cigarette smoke (76–78), suggesting that increased activity of these enzymes may enhance carcinogenesis. We previously reported a similar interaction between colorectal cancer risk associated with well-done red meat and the SULT1A1-638 variant, which would further support a role for SULT1A1 in colorectal cancer because well-done red meat and cigarette smoke contain many similar carcinogens (79).

As with most case-control studies, there are several limitations. First, survival bias is possible because deceased cases were not recruited and response rates are lower for cases with later stage disease (32). Although the association between smoking and precancerous polyps is stronger than that seen for invasive colorectal cancer, the risk of colorectal cancer associated with smoking may be stronger among cases with early stage disease (80), thus leading to a potential overestimation of risk in this study. Response bias is possible when high response rates are not achieved (31, 32). Self-reported data regarding smoking may underestimate true smoking exposure, and this bias may not be equal between cases and controls, leading to an underestimation of colorectal cancer risk associated with cigarettes in this series (81). We observed a 14% prevalence of current smoking among controls (defined as ≥1 cigarettes per day for ≥3 months); the Canadian Tobacco Use Monitoring Study demonstrated that 20% of the Ontario population are active smokers, of whom 83% smoke on a daily basis (16.6%). Given the differences in methodology and age of participants, the observed rates of smoking in this study are not dissimilar to other population-based assessments of smoking behavior (82). Furthermore, smoking behavior may change with overall health status, and cases may be affected by their disease prior to diagnosis; to minimize this effect, we defined smoking behavior at a point 2 years before cancer diagnosis for cases and enrollment for controls. Although all but one (CYP1A1-2455) of the genetic variants studied were in HWE and study subjects were predominantly Caucasian, the possible influence of genotyping errors or population stratification on the results of this study cannot be excluded. The genetic variants in this study were selected for testing on the basis of their known or suggested functional impact, as well as their frequency. Additional variation within these genes may be related to smoking or colorectal cancer risk but is not accounted for in this study and may be more adequately addressed in a haplotype or tag single nucleotide polymorphism analysis. Furthermore, we tested for multiplicative interaction between gene variants and smoking and cannot exclude the potential for additive interactions.

The results of this study lend support to the growing body of literature that indicates that extensive cigarette smoking is associated with increased colorectal cancer risk. Furthermore, individual susceptibility to the mutagenic effects of cigarette smoke may be driven by inherited genetic variability in carcinogen-metabolism genes. Specifically, our data suggest that prolonged and/or intense cigarette smoking increases the risk of colorectal cancer, particularly among carriers of specific SULT1A1, CYP1B1, NAT1, and NAT2 variants. Future research in larger study populations is required to validate and further define the risk of colorectal cancer and the interaction between smoking and metabolic gene variations and to explore potential gene–gene interactions. In addition, larger studies would permit analyses to be done separately for males and females.

Glossary

Abbreviations

AOR

adjusted odds ratio

CI

confidence interval

CYP

cytochrome P-450

EPHX

epoxide hydrolase

GSTM1

glutathione S-transferase M1

GSTT1

glutathione S-transferase T1

HWE

Hardy-Weinberg equilibrium

MSI

microsatellite instability

MSI-H

high microsatellite instability

NAT

N-acetyl transferase

SULT

sulfotransferase

UGT

uridine diphosphate-glucuronosyltransferase