Alcohol 45 (2011) 217e225

Alcohol consumption, alcohol dehydrogenase 1C (ADH1C )genotype, and risk of colorectal cancer in the Netherlands Cohort Study

on diet and cancer

Brenda W.C. Bongaertsa,b,*, Anton F.P.M. de Goeijb, Kim A.D. Woutersb,Manon van Engelandb, Ralph W.H. Gottschalkc, Frederik J. Van Schootenc,R. Alexandra Goldbohmd, Piet A. van den Brandta, Matty P. Weijenberga

aGROW e School for Oncology and Developmental Biology, Department of Epidemiology, Maastricht University, 6200 MD Maastricht,

The NetherlandsbDepartment of Pathology, Maastricht University, 6200 MD Maastricht, The Netherlands

cDepartment of Health Risk Analysis and Toxicology, Maastricht University, 6200 MD Maastricht, The NetherlandsdTNO Quality of Life, Department of Prevention and Health, 2301 CE Leiden, The Netherlands

Received 25 August 2009; received in revised form 8 October 2010; accepted 12 October 2010

Abstract

Within the Netherlands Cohort Study (1986), we examined associations between alcohol consumption, the alcohol dehydrogenase 1C(ADH1C ) genotype, and risk of colorectal cancer (CRC). After a follow-up period of 7.3 years, 594 CRC cases with information on geno-type and baseline alcohol intake were available for analyses. Adjusted incidence rate ratios (RRs) and 95% confidence intervals (CIs) wereestimated using Cox proportional hazards models. In subjects who reported to have consumed equal amounts of total alcohol both 5 yearsbefore baseline and at baseline, drinkers of$30 g of alcohol per day with the ADH1C*2/*2 genotype were associateddalthough not statis-tically significantdwith an increased risk of CRC relative to abstainers with the ADH1C*1/*1 genotype (RR: 1.91, 95% CI: 0.68, 5.34).The risk estimate in this exposure group increased slightly when compared with light drinkers of $0.5e!5 g/day with the ADH1C*1/*1genotype (RR: 2.32, 95% CI: 0.80, 6.72). The interaction term however, was not statistically significant (PO .05). In subjects who reportedto have consumed equal amounts of total alcohol both 5 years before baseline and at baseline, drinkers of $30 g of alcohol per day wereassociateddalthough not statistically significantdwith an increased risk of CRC relative to abstainers (RR: 1.38, 95% CI: 0.80, 2.38). Thisrisk estimate for high-level drinkers became stronger when compared with light drinkers (RR: 1.74, 95% CI: 1.01, 2.99). As main effect ofgenotype, we observed that the ADH1C*2/*2 genotype was associated with a 42% increase in risk of CRC when compared with theADH1C*1/*1 genotype. In conclusion, both genotype and alcohol consumption were associated with an increased risk of CRC. Owingto limited statistical power, we found no apparent evidence for the ADH1C genotype as effect modifier of the relationship between alcoholintake and CRC. Nevertheless, the interaction deserves further investigation in larger genetic epidemiologic studies. � 2011 Elsevier Inc.All rights reserved.

Keywords: Alcohol; Alcohol dehydrogenase 1C; Colorectal cancer; Genetic epidemiology; Cohort study; The Netherlands

Introduction

Alcohol consumption is an important risk factor forseveral types of cancer, including cancers of the upper aero-digestive tract (UADT), liver, and breast (World CancerResearch Fund/American Institute for Cancer Research,2007). Additionally, a considerable proportion of colorectalcancers (CRCs) may arise as a result of (excessive) alcohol

* Corresponding author. Department of Pathology, Maastricht Univer-

sity, P.O. Box 616, 6200 MD Maastricht, The Netherlands. Tel.: þ31-43-

387-4633; fax: þ31-43-387-6613.

E-mail address: Brenda.Bongaerts@epid.unimaas.nl (B.W.C. Bongaerts).

0741-8329/$ - see front matter � 2011 Elsevier Inc. All rights reserved.

doi: 10.1016/j.alcohol.2010.10.003

intake (Bagnardi et al., 2001; Boffetta and Hashibe, 2006;Bongaerts et al., 2008; Cho et al., 2004; Corrao et al.,2004; Ferrari et al., 2007; Longnecker et al., 1990;Moskal et al., 2006).

According to current knowledge, mechanisms ofalcohol-associated tumorigenesis are closely related to thealcohol metabolism, in which acetaldehyde is identifiedas the important carcinogen (Seitz and Stickel, 2007). Inthe liver, alcohol dehydrogenase (ADH) enzymes convertalcohol to acetaldehyde, which is then rapidly oxidized toacetate by aldehyde dehydrogenase (ALDH) enzymes(Bosron and Li, 1986). Single nucleotide polymorphisms

218 B.W.C. Bongaerts et al. / Alcohol 45 (2011) 217e225

(SNPs) have been described for ADH isoforms ADH1B andalcohol dehydrogenase 1C (ADH1C ), and ALDH isoformALDH2 (Bosron and Li, 1987; Goedde et al., 1992). Incontrast to SNPs in ADH1B and ALDH2, one functionalSNP in ADH1C is found frequently in Caucasian popula-tions (Borras et al., 2000; Goedde et al., 1992). Its twoallelic variants are present in approximately equalfrequency and homozygotes for the ADH1C*1 allelemetabolize alcohol 2.5 times faster than homozygotes forthe ADH1C*2 allele (Bosron and Li, 1986). Subjects withthe ADH1C*1/*1, ADH1C*1/*2, and ADH1C*2/*2 geno-type are thus classified as being fast, intermediate, and slowalcohol metabolizers, respectively.

Because of the different kinetic properties of theenzymes, the ADH1C polymorphism is speculated tounderlie a genetic susceptibility to alcohol-associatedcancers. In literature, studies have reported an interactionbetween alcohol consumption, ADH1C genotype, and riskof UADT cancer (Coutelle et al., 1997; Harty et al.,1997), breast cancer (Coutelle et al., 2004; Freudenheimet al., 1999; Terry et al., 2006), hepatitis cirrhosis andchronic pancreatitis (Day et al., 1991), and colorectalneoplasia (Chen et al., 2001; Giovannucci et al., 2003;Jung et al., 2008; Tiemersma et al., 2003). Overall, thefast-metabolizing ADH1C*1/*1 genotype has been reportedto be the high-risk genotype in alcohol-related diseases.Three studies on colorectal neoplasia on the other hand,observed ADH1C*2/*2 as the high-risk genotype (Chenet al., 2001; Giovannucci et al., 2003; Jung et al., 2008).

As such, we examined associations between alcoholconsumption, ADH1C genotype, and risk of CRC in a largeprospective cohort study. We hypothesized that the ADH1Cgenotype modifies the association between alcoholconsumption and risk of CRC, that is, the slow-metabolizing ADH1C*2/*2 genotype is related to anincreased risk of alcohol-associated CRC, when comparedwith the fast-metabolizing ADH1C*1/*1 genotype.Namely, slow conversion of alcohol in the liver may resultin alcohol circulating the blood for longer periods of time inwhich ADH activity from both colonic mucosa cells andintestinal bacteria may expose the colorectal tract to rela-tively higherdlocally formeddlevels of the carcinogenicacetaldehyde (Jokelainen et al., 1994, 1996; Pestalozziet al., 1983; Seitz et al., 1996).

Materials and methods

Study population

In September 1986, the Netherlands Cohort Study ondiet and cancer (NLCS) was initiated, as described in detailelsewhere (van den Brandt et al., 1990a). Briefly, the cohortincludes 58,279 men and 62,573 women, aged 55e69 yearsat baseline and originating from 204 municipal populationregistries throughout the Netherlands. All cohort memberscompleted an extensive self-administered questionnaire on

dietary habits during the year preceding the start of thestudy, life style, and other risk factors for cancer. The entirecohort is followed-up for cancer occurrence through annualrecord linkage to the Netherlands Cancer Registry (NCR,comprising nine cancer registries in the Netherlands) andto PALGA, a nationwide database of histo- and cytopathol-ogy reports (Casparie et al., 2007; van den Brandt et al.,1990b).

The NLCS followed the caseecohort approach in whichdata are processed and analyzed only for a random sampleof the cohort plus all incident cancer cases arising each year(Miettinen, 1985; Prentice, 1985). Cases were identified forthe entire cohort, whereas a random sample of the cohort,that is, the subcohort, was used to estimate person yearsat risk accumulating in the cohort (from the date of entryinto the cohort until the date of a CRC diagnosis, deathfrom any cause, leaving the study area or the end of thestudied follow-up period). The subcohort, consisting of5,000 men and women, was followed-up biennially toassess information on vital status and migration to calculateaccumulated person time in the cohort. Prevalent cancercases, other than nonmelanoma skin cancer, were excludedfrom the subcohort, resulting in 4,774 subcohort members.CRCs were classified as comprising all cancers of thecolon, the rectosigmoid, and the rectum (ICD-O codesC18.0 through C20).

Assessment of exposure

Information on daily dietary habits was obtained througha 150-item semiquantitative food frequency questionnaire.Alcohol consumption was measured by six items: (1) beer;(2) red wine; (3) white wine; (4) sherry and other fortifiedwines; (e) liquor types containing on average 16% alcohol;and (f) (Dutch) gin, brandy, and whiskey. Participants wereasked about their usual frequency of consumption and thenumber of glasses consumed at each drinking occasion.Together with the alcoholic content of the various bever-ages, the total amount of mean daily alcohol consumption(g/day) was calculated (Nevo, 1986). In addition, for ‘‘beer’’and ‘‘other alcoholic beverages’’ participants could indicatewhether 5 years ago, they drank (1) more than, (2) equalamounts of, or (3) less than today. The fourth answeringoption was (4) I never use this. This information enabledidentification of subjects that reported to have consumedequal amounts of total alcohol 5 years before baseline,compared with baseline drinking habits.

The questionnaire has been validated against a 9-daydiet record (Goldbohm et al., 1994). The Spearmancorrelation coefficient between mean daily alcohol intakeassessed by the questionnaire and estimated from the9-day diet record was 0.89 for all the subjects and0.85 for users of alcoholic beverages. The absoluteamount of alcohol reported in the questionnaire by usersof alcoholic beverages was, on average, 86% of that re-ported in the 9-day diet record.

219B.W.C. Bongaerts et al. / Alcohol 45 (2011) 217e225

Tissue sample collection and DNA extraction

In August 1999, collection of tumor tissue specimenswas started after approval by the Ethical Review board ofUniversity Maastricht, PALGA and the NCR. Tumor blocksfrom CRC cases that were diagnosed within the first 7.3years of follow-up were collected, using PALGA to identifyand locate the tissue samples in Dutch pathology laborato-ries. Because full coverage of the included municipalitiesby PALGA was not reached until the end of 1988, the first2.3 years of follow-up were excluded from collection toavoid selection bias as a result of this incomplete coverage.As such, from the 925 CRC cases diagnosed betweenJanuary 1989 and January 1994, 815 tissue samples couldbe linked to a PALGA report of the lesion of which 734(90%) could be retrieved from the pathology laboratories.Genomic DNA was extracted from macrodissected tumortissue as described in detail elsewhere (Brink et al., 2003)and used to assess the ADH1C genotype of CRC cases.We considered the use of tumor tissue for correct genotypeassessment justified, as it would be unlikely that a single-point mutation would occur at the exact location of theSNP under study.

In December 2000, all available members of the subco-hort who were still alive at that time point (n5 3,579) werecontacted and asked to collect mouth swabs. In total, 1,929subcohort members (54%) returned the mouth swabs withthe informed consent in 2000. Corresponding with datafrom literature (Meulenbelt et al., 1995), the averageDNA yield per cotton swab was 0.1e10 mg, from whichwe determined the ADH1C genotype of subcohortmembers.

ADH1C genotyping

ADH1C was amplified using the following PCR primers:F50-CTT GTGGCTGACTTTATGGCTA-30 and R50-CTCTTT CCA GAG CGA AGC AG-30. Genotyping was per-formed using the Multiplex SNaPShot single base extension(SBE) reaction (Applied Biosystems, Foster City, CA) asdescribed previously (Knaapen et al., 2004). The SBEprimers were designed to anneal immediately adjacent 50 tothe SNP of interestdthat is, rs698A1047Gdwith a templatespecific part of 20 to 33 bp and a Tm of 60�C (for ADH1C:TTTTCACTGGATGCATTAATAACA-AAT). SBE prod-ucts were analyzed on an ABI Prism 3100 genetic analyzerusing Genmapper Analysis Software (version 4).

Data analysis

Data analysis was based on subcohort members andcases for whom complete information on alcohol consump-tion, ADH1C genotype, and confounding variables waspresent; that is, 1,649 subcohort members (of whom 19were diagnosed with CRC) and 594 CRC cases.

We started with exploring the possibility of bias relatedto the fact that mouth swab collection took place 14 years

after baseline. Because of this, the subcohort might repre-sent a selection of survival-related characteristics, whichin turn might be associated with a certain genetic constitu-tion and/or with environmental exposures, such as diet.Therefore, we compared subcohort members that hadprovided mouth swabs with subcohort members that hadnot. We tested for differences in alcohol consumption andconfounders used in our analyses, using t-tests and Chi-square tests where appropriate. Additionally, because thesubcohort may have suffered from selective loss tofollow-up affecting ADH1C genotype frequency, weapplied analysis of variance to compare the mean age ofsubcohort members across the variant ADH1C genotypes.

To verify that the level of alcohol consumption in ourstudy population was not related to the ADH1C genotype,we used a Chi-square test to test for differences in ADH1Cgenotype frequencies across the different categories ofalcohol intake (0, !30.0, $30.0 g/day) among subcohortmembers. Alcohol consumption categories were chosen inline with previously publications, suggesting a thresholdin alcohol consumption of 30 g/day above which the riskof CRC is increased (Cho et al., 2004; Ferrari et al.,2007). Also, within the NLCS we have confirmed thisthreshold in alcohol consumption (using a follow-up periodof 13.3 yearsdbecause the study did not involve molecularor genetic datadand a total sample of 2,323 CRC cases)(Bongaerts et al., 2008).

Cox proportional hazards models were applied to esti-mate incidence rate ratios (RRs) and corresponding 95%confidence intervals (CIs). The proportional hazardsassumption was tested using the scaled Schoenfeld resid-uals (Schoenfeld, 1982). Standard errors were estimatedusing the robust HubereWhite sandwich estimator toaccount for additional variance introduced by samplingfrom the cohort. We first studied the potential modifyingeffect of the ADH1C genotype on the association betweenalcohol consumption and risk of CRC, including the inter-action term of ADH1C genotype and alcohol consumptionin the multivariable model. To test the interaction, wecomputed a cross-product term using a three-categoryordinal value for ADH1C (based on the genotypesADH1C*1/*1, ADH1C*1/*2, and ADH1C*2/*2) andalcohol (based on intakes of 0, !30.0, and $30.0 g/day).The P value for interaction was based on the Wald testfor this cross-product term added to the model. We consid-ered the abstainers with the ADH1C*1/*1 genotype asreference group. We next examined the main effects ofADH1C genotype and alcohol consumptiondadjusted forconfounding variablesdon CRC risk. In these analyses,individuals with the ADH1C*1/*1 genotype and abstainerswere used as reference group, respectively.

All Cox models were performed on the total study pop-ulation using information on baseline drinking habits.Subsequently, all analyses were restricted to a subgroupof alcohol drinkers that reported to have consumed equalamounts of alcohol both 5 years before baseline and at

Table 1

Baseline characteristics and mean dietary intakes (standard deviation

[S.D.]) for subcohort members and colorectal cancer cases according to

anatomical subsite, the Netherlands Cohort Study on diet and cancer, the

Netherlands, 1986e1993

Baseline information Subcohort

Colorectal

cancer cases

N 1,649 594

Age (years) 60.3 (4.0) 62.8 (4.1)

Men (%) 51 56

Alcohol consumption:

Total alcohol (g/day)a 11.0 (14.0) 11.7 (15.2)

Abstainers (%) 21 24

Stable drinkers (%)b 47 44

ADH1C genotype

*1/*1 36 31

*1/*2 47 47

*2/*2 17 22

Other characteristics

Family history of CRC (% yes) 6 11

Body mass index (kg/m2) 24.8 (2.8) 25.5 (3.2)

Energy intake (kJ/day) 8,233 (2,199) 8,082 (2,014)

Fat intake (g)c 84.1 (15.5) 85.3 (15.1)

Dietary fiber intake (g/day)c 27.7 (6.8) 27.4 (6.7)

Calcium intake (mg)d 936 (286) 922 (271)

Folate intake (mg/day) 214 (68) 213 (71)

Smoking (%)

Never 36 33

Ex-smoker 41 46

Current smoker 23 21

Physical activity (%)

!30 min/day 18 21

30e!60 min/day 34 32

60e!90 min/day 21 21

$90 min/day 27 26

aMean (S.D.) total alcohol consumption for all drinking subcohort

members and colorectal cancer cases.bStable drinkers are those subjects that reported to have consumed

equal amounts of total alcohol 5 years before baseline, compared with

baseline drinking habits.cAdjusted for total energy intake.dAdjusted for energy intake from dairy products.

220 B.W.C. Bongaerts et al. / Alcohol 45 (2011) 217e225

baseline (sample size: 1,065 subcohort members [of whom8 were diagnosed with CRC] and 378 CRC cases). Withthese sensitivity analyses, we attempted to minimize poten-tial misclassification of former (heavy) drinkers as beingabstainer or low-level drinker (Bongaerts et al., 2008).And additionally, we changed the reference group ofabstainers to drinkers of $0.5e!5.0 g of alcohol perday. Abstainers are generally considered to abstain fromalcohol due to health problems and thus to have a relativelypoorer health status than light drinkers. The use ofabstainers as reference group may thus obscure the esti-mated disease risks of drinkers.

All analyses on alcohol, ADH1C genotype, and CRCwere performed for overall CRC only, because smallnumbers of cases did not allow analyses for the anatomicalsubsites separately. Confounding effects of age at baseline(years), family history of CRC (yes/no), body mass index(kg/m2), smoking (never, ex, current), nonoccupationalphysical activity (min/day), energy intake (kJ/day),folate intake (mg/day), meat intake (g/day), calciumintake (mg/day) adjusted for energy intake from dairyproducts and energy-adjusted intakes of fat (g/day), andfiber (g/day) were evaluated for their effect on the relation-ship between alcohol consumption and CRC risk. Thosevariables that altered the estimate for the exposure coeffi-cient between statistical models with and without the poten-tial confounder by more than 10% were included in theanalyses. When studying the main effects of ADH1C geno-type and alcohol consumption, additional adjustment wasmade for total alcohol intake (categorical; 0, !30.0, and$30.0 g/day) and genotype (categorical: ADH1C*1/*1,ADH1C*1/*2, and ADH1C*2/*2), respectively. Becausestratification by gender resulted in small subgroups, allanalyses were performed for men and women combined.According to findings in literature, we further tested forinteractions between alcohol consumption (based onintakes of 0, !30, and $30 g/day) and gender, alcohol,and folate intake (based on the median folate intake inthe subcohort), and alcohol and smoking (based on never,ex, and current smoking), according to the above-mentioned method using a cross-product term.

P values were two-sided and all statistical analyses wereperformed with the STATA Statistical Software Package(version 9) (StataCorp LP, College Station, TX).

Results

The current analyses were based on CRC cases identi-fied after 7.3 years of follow-up of the NLCS, that is, 594CRC cases (403 colon cancers, 62 rectosigmoid cancers,and 129 rectal cancers) and a total of 11,992 person yearsthat accumulated in the subcohort. At baseline, 76% of theCRC cases drank alcohol (Table 1). Cancer cases did notdiffer with respect to daily alcohol consumption, ADH1Cgenotype variants, and other baseline characteristics.

There were no differences in mean daily alcohol consump-tion and confounding variables between subcohort membersthat provided mouth swabs and those that did not, exceptfordas to be expecteddage. Subcohort members withoutmouth swabs were, on average, 2 years older (P! .001)(results not shown). The mean age of subcohort membersdid not differ statistically significantly across the differentADH1C genotypes (PO .05, results not shown). Furthermore,for all subcohort members who were also CRC case (n5 19),theADH1C genotype derived from their mouth swabmatchedthe outcome from the tumor tissue sample.

With respect to allele and genotype frequencies in thesubcohort, no deviation from the HardyeWeinberg equilib-rium was observed (PO .05). Among subcohort members,frequencies of the ADH1C genotype variants were similarwithin the different categories of alcohol intake (PO .05)(Table 2).

Table 2

ADH1C genotype variants and allele frequencies across categories of total

average alcohol intake in subcohort members; the Netherlands Cohort

Study, the Netherlands, 1986e1993

Alcohol consumption

ADH1C Abstainer !30.0 g/day $30.0 g/day

ADH1C genotype

*1/*1 (%) (N5 587) 129 (22%) 393 (67%) 65 (11%)

*1/*2 (%) (N5 775) 163 (21%) 542 (70%) 70 (9%)

*2/*2 (%) (N5 287) 54 (19%) 201 (70%) 32 (11%)

c2 (P value)5 .585

Allele frequency (%)

*1 61 58 60

*2 39 42 40

*5 allele.

221B.W.C. Bongaerts et al. / Alcohol 45 (2011) 217e225

Table 3 presents the results of the interaction betweenalcohol consumption and ADH1C genotype. In drinkerswho reported to have consumed equal amounts of alcoholboth 5 years before baseline and at baseline, consumersof alcohol of $30.0 g/day with either variant of theADH1C genotype were positivelydalthough not statisti-cally significantlydassociated with the risk of CRCcompared with abstainers with the ADH1C*1/*1 genotype.The highest RRs were observed for drinkers with theADH1C*1/*2 genotype (RR: 1.85, 95% CI: 0.85, 3.99)and ADH1C*2/*2 genotype (RR: 1.91, 95% CI: 0.68,5.34). When compared with light drinkers of $0.5e!5.0 g/day with the ADH1C*1/*1 genotype, the RRs fordrinkers of $30.0 g/day were even higher (RR: 2.25,95% CI: 1.00, 5.08 for drinkers with the ADH1C*1/*2

Table 3

Adjusted incidence rate ratios for colorectal cancer according to ADH1C genotyp

1986e1993

Total alcohol

consumption

(g/day)

ADH1C*1/*1 ADH1C*1/*2

Subcohort

(PY) Cases RRa (95% CI)

Subcohort

(PY) Cases

Total study population with complete information on baseline drinking habits an

Abstainers 931 49 1.00 (Reference) 1,179 65

!30.0 2,855 107 0.77 (0.49e1.20) 3,945 180

$30.0 481 27 0.98 (0.51e1.88) 512 33

P for inte

Subgroup of subjects that reported to have consumed equal amounts of total alc

habits (n5 1,435)

Abstainersb 797 37 1.00 (Reference) 975 54

!30.0 1,747 56 0.76 (0.45e1.27) 2,476 105

$30.0 270 15 1.27 (0.55e2.91) 233 21

P for inte

Abstainersb 797 37 1.22 (0.65e2.29) 975 54

$0.5e!5.0 664 24 1.00 (Reference) 919 39

5.0e!30.0 1,083 32 0.87 (0.47e1.64) 1,557 66

$30.0 270 15 1.54 (0.64e3.70) 233 21

P for inte

CI5Confidence interval; PY5 Person years; RR5 Incidence risk ratio.aAdjusted for age, gender, family history of CRC (yes/no), body mass in

intake (kJ/day) and energy-adjusted intakes of fat (g/day), dietary fiber (g/daybAbstainers that reported to have abstained from alcohol consumption 5 yea

genotype and RR: 2.32, 95% CI: 0.80, 6.72 for drinkerswith the ADH1C*2/*2 genotype). The P values for interac-tion however, indicated that there was no modifying effectof the ADH1C genotype on the relationship betweenalcohol consumption and risk of CRC (PO .05).

In addition, interactions between alcohol intake andgender, alcohol and folate intake, and alcohol and smokingwere not statistically significant (PO .05) (results notshown).

In Table 4, the main effects of alcohol consumption andADH1C genotype on risk of CRC are presented. Withinsubjects that reported to have consumed equal amounts ofalcohol both 5 years before baseline and at baseline, a posi-tive associationdalthough not statistically significantdwas observed between consumption of $30.0 g/day andthe risk of CRC, relative to abstaining (RR: 1.38, 95%CI: 0.80, 2.38). Also, CRC risk estimates for consumersof $30.0 g/day increased when light drinkers of $0.5e!5.0 g of alcohol per day were used as reference group(RR: 1.74, 95% CI: 1.01, 2.99). Furthermore, a positiveassociation was observed between ADH1C genotype andrisk of CRC (Table 4). Compared with the ADH1C*1/*1genotype, the ADH1C*2/*2 genotype was associated witha statistically significant 42% increase in risk of CRC.

Discussion

In the present prospective study, we investigated associ-ations between total alcohol consumption, ADH1C geno-type, and the risk of CRC in a large sample of the Dutch

e and alcohol consumption; the Netherlands Cohort Study, the Netherlands,

ADH1C*2/*2

RRa (95% CI)

Subcohort

(PY) Cases RRa (95% CI)

d confounding variables (n5 2,224)

1.06 (0.65e1.73) 401 26 1.29 (0.68e2.43)

0.87 (0.57e1.33) 1,462 95 1.20 (0.75e1.91)

0.98 (0.53e1.83) 226 12 0.98 (0.42e2.32)raction5 .921

ohol 5 years before baseline, compared with baseline drinking

1.17 (0.69e2.00) 328 24 1.60 (0.80e3.20)

0.90 (0.52e1.45) 810 57 1.38 (0.79e2.40)

1.85 (0.85e3.99) 117 9 1.91 (0.68e5.34)raction5 .971

1.43 (0.80e2.58) 328 24 1.96 (0.94e4.06)

1.08 (0.58e2.01) 365 24 1.65 (0.80e3.43)

1.11 (0.62e1.96) 445 33 1.71 (0.86e3.39)2.25 (1.00e5.08) 117 9 2.32 (0.80e6.72)

raction5 .995

dex (kg/m2), nonoccupational physical activity (min/day), total energy

), and calcium (mg/day).

rs before baseline as well.

Table 4

Adjusted incidence rate ratios for colorectal cancer according to alcohol consumption and according to ADH1C genotype; the Netherlands Cohort Study,

the Netherlands, 1986e1993

Alcohol

consumption (g/day) Subcohort (PY) Cases RRa (95% CI)

ADH1C

genotype Subcohort (PY) Cases RRb (95% CI)

Total study population with complete information on baseline drinking habits and confounding variables (n5 2,224)

Abstainers 2,511 140 1.00 (Reference) *1/*1 4,266 183 1.00 (Reference)

!30.0 8,262 382 0.83 (0.63e1.08) *1/*2 5,636 278 1.09 (0.87e1.38)

$30.0 1,219 72 0.91 (0.59e1.41) *2/*2 2,090 133 1.42 (1.06e1.90)

Subgroup of subjects that reported to have consumed equal amounts of total alcohol 5 years before baseline, compared with baseline drinking

habits (n5 1,435)

Abstainersc 2,100 115 1.00 (Reference)

!30.0 5,034 218 0.78 (0.57e1.07)$30.0 620 45 1.38 (0.80e2.38)

Abstainersc 2,100 115 1.26 (0.87e1.82)

$0.5e!5.0 1,948 87 1.00 (Reference)

5.0e!30.0 3,086 131 0.98 (0.68e1.41)

$30.0 620 45 1.74 (1.01e2.99)

CI5Confidence interval; PY5 Person years; RR5 Incidence risk ratio.aAdjusted for age, gender, family history of CRC (yes/no), BMI (kg/m2), nonoccupational physical activity (min/day), total energy intake (kJ/day),

ADH1C genotype (ADH1C*1/*1, ADH1C*1/*2, and ADH1C*2/*2) and energy-adjusted intakes of fat (g/day), dietary fiber (g/day), and calcium (mg/day).bAdjusted for age, gender, family history of CRC (yes/no), BMI (kg/m2), nonoccupational physical activity (min/day), total energy intake (kJ/day), alcohol

consumption (0,!30.0, and$30.0 g/day) and energy-adjusted intakes of fat (g/day), dietary fiber (g/day) and calcium (mg/day).cAbstainers that reported to have abstained from alcohol consumption 5 years before baseline as well.

222 B.W.C. Bongaerts et al. / Alcohol 45 (2011) 217e225

general population. In subjects that reported to haveconsumed equal amounts of alcohol both at 5 years beforebaseline and at baseline, results pointed toward an associa-tion between drinkers of $30.0 g/day with the ADH1C*1/*2 and ADH1C*2/*2 genotype and the risk of CRC.P values for interaction however, were not statisticallysignificant. Our results further showed a positive relation-ship between alcohol drinkers of $30.0 g/day and the riskof CRC, compared with abstaining. Also, individuals withthe ADH1C*2/*2 genotype were associated with anincreased risk of CRC, when compared with individualswith the ADH1C*1/*1 genotype.

In alcohol-induced (colorectal) carcinogenesis, acetalde-hydeda well-known mutagen and carcinogendis believedto be the important cancer-causing agent. Most acetalde-hyde in the body originates from the hepatic alcohol metab-olism involving ADH and ALDH enzymes. Albeit ata lower level than in the liver, ADH genes are expressedin a variety of extrahepatic tissues, including the colorectalmucosa (Estonius et al., 1996; Pestalozzi et al., 1983; Yinet al., 1994). It has often been speculated that variation inADH enzyme activity, and thus variation in alcohol break-down and subsequent acetaldehyde formation, may influ-ence colorectal carcinogenesis. Still, only a limitednumber of studies have investigated the relationshipbetween alcohol consumption, the ADH1C genotype, andrisk of colorectal neoplasia (Chen et al., 2001;Giovannucci et al., 2003; Jung et al., 2008; Tiemersmaet al., 2003). Nested caseecontrol studies performed withinthe Health Professionals’ Follow-Up Study (Giovannucciet al., 2003) and the Physicians’ Health Study (Chenet al., 2001) reported that drinkers in the highest alcohol

consumption category with the ADH1C*2/*2 genotypewere associated with an increased risk of colorectaladenomas and carcinomas, respectively, compared withdrinkers in the lowest category with the ADH1C*1/*1genotype (RR: 2.94, 95% CI: 1.24, 6.92 for the HealthProfessionals’ Follow-Up Study and RR:1.63, 95% CI:0.60, 6.30 for the Physicians’ Health Study). In both studieshowever, the interaction term between genotype andalcohol consumption was not statistically significant. Noassociations were observed between genotype and risk ofcolorectal neoplasia. Recently, a caseecontrol study oncolorectal adenomas reported an association between theADH1C*2/*2 genotype and risk of adenomas for high-level drinkers, compared with abstainers with theADH1C*1/*1 genotype (RR: 1.95, 95% CI: 0.60, 6.30)(Jung et al., 2008). The interaction between alcohol intakeand ADH1C genotype was statistically significant(P! .05), although the authors did not observe a relation-ship between ADH1C genotype and risk of colorectaladenomas. Contradictory results have been described ina Dutch caseecontrol study on colorectal adenomas(Tiemersma et al., 2003). The authors considered theADH1C*1/*1 genotype as high-risk genotype and conse-quently used low-level drinkers with the ADH1C*1/*2and ADH1C*2/*2 genotypes combined as the referencegroup. Compared with this reference, subjects with theADH1C*1/*1 genotype in the highest consumption cate-gory were found to be associated with an increased riskof colorectal adenomas (RR: 1.76, 95% CI: 1.00, 3.11).The test for interaction did not reach statistical significanceand the authors had not examined associations betweenADH1C genotype and risk of colorectal adenomas. In

223B.W.C. Bongaerts et al. / Alcohol 45 (2011) 217e225

contrast to the above-mentioned studies, we observed a rela-tionship between genotype and risk of CRC. In line withthree of the previous studies (Chen et al., 2001;Giovannucci et al., 2003; Jung et al., 2008) was our findingof a positive association between high alcohol consumptionlevels and CRC in individuals with the ADH1C*2/*2 geno-type however, our P values for interaction provided noevidence of effect modification. Still, comparing our resultswith the previously published ones should be done withcaution. Several differences exist with regard to studydesign, outcome variables, and studied levels of (high)alcohol consumption that hamper a direct comparison. Amajor limitation of this study is the number of cases inthe genotype and alcohol consumption categories, whichis relatively small for genetic epidemiologic analyses.Because power and sample size calculations for geneeenvironment interactions in caseecohort studies are notreadily available, we used adopted methods developed forcaseecontrol studies (Gauderman, 2002). These samplesize calculations showed that for the detection of a statisti-cally significant interaction between alcohol consumptionand ADH1C genotype, the number of cases needed to beat least as high as 459, instead of the current 378 cases.Another issue that may have hampered the detection ofstatistically significant risk estimates is the limited rangein alcohol consumption. For example, in the subgroup ofalcohol consumers drinking $30 g/day, over 50% of thissubgroup consumed 30e40 g of alcohol per day and only20% consumed $50 g/day.

As mentioned before, we observed a positive associationbetween ADH1C genotype and risk of CRC. Speculationsabout the underlying mechanism involve ADH activity ofthe intestinal bacteria. It has been shown that the ability ofcolonic mircoflora to breakdown alcohol contributes signif-icantly to intestinal acetaldehyde levels (Jokelainen et al.,1994; Jokelainen et al., 1996; Seitz et al., 1990). Thuspossibly, compared with drinkers with the fast-metabolizing ADH1C*1/*1 genotype, the large intestine ofdrinkers with the slow-metabolizing ADH1C*2/*2 genotypewill be exposed to circulating levels of alcohol for a rela-tively longer period of time. Intestinal bacterial ADHactivity may consequently expose the bowel to locallyformed acetaldehyde for a relatively longer period of time,implying a higher risk of carcinogenesis.

Another interesting observation involved the elevateddalthough not statistically significantdrisk estimates fordeveloping CRC in abstainers who reported to have ab-stained from alcohol already 5 years before baseline, whencompared with light drinkers (Tables 3 and 4). The RRsconsistently pointed toward an increased risk of CRC, inagreement with the general belief that abstainers are notper definition healthy individuals. Instead, abstainers areoften reported to have a poorer health than light drinkers(Poikolainen et al., 1996; Rehm, 2000; San Jose et al.,1999). Because of consisting of lifetime abstainers andformer drinkers, the overall health status of the group

may become negatively influenced by former drinkers thathave stopped drinking because of underlying health prob-lems (Shaper, 1995). Within the NLCS, we were unableto distinguish lifetime abstainers from former drinkers.Still, our approach of restricting the analyses to subjectsthat reported to have had similar drinking habits both 5years before baseline and at baseline excluded at least someof the recently quitted drinkers and thereby their unfavor-able contribution to the overall health status of the abstainergroup.

The restriction approach has implications for the variousalcohol consumption categories also. In studies on alcoholconsumption and disease in which only baseline drinkinghabits are measured and used, a certain number of former(heavy) alcohol drinkers will be misclassified. Drinkinghabits at baseline do not necessarily reflect habits of the(recent) past and will thus not reveal former heavy drinkerswho cut down on drinking because of preclinical diseasesymptoms or other health problems, that is, individuals witha potential higher risk of alcohol-related conditions. Wehave previously shown that associations between baselinealcohol consumption and CRC risk became stronger whenrestricted to subjects that reported to have consumed equalamounts of alcohol both 5 years before baseline and atbaseline (Bongaerts et al., 2008). Although relativelyunderpowered, the present results also suggested that stabledrinking habits reflect alcohol exposure more accuratelyand minimize misclassification of former drinkers. As such,measuring alcohol use in the recent past or repeatingmeasurements over time may help future studies on alcoholconsumption and disease to model alcohol exposure moreaccurately.

In conclusion, both the ADH1C genotype and highalcohol consumption were related to an increased risk ofCRC. Results however, did not support clear evidence fora modifying effect of the ADH1C genotype on the associa-tion between alcohol consumption and CRC. Large geneticepidemiologic studies with sufficient statistical power areneeded to lend support to the current findings and toprovide clues for underlying mechanisms of disease.Further, we highly recommend to assess and analyzelong-term alcohol exposure in addition to baseline drinkinghabits for the benefit of future research on alcohol intakeand chronic diseases.

Acknowledgments

We are indebted to the participants of this study andfurther wish to thank the regional cancer registries (IKA,IKL, IKMN, IKN, IKO, IKR, IKST, IKW, IKZ, and VIKC),and the Netherlands nationwide registry of pathology (PAL-GA). We also thank all the Dutch pathology laboratories fortheir cooperation in providing the tissue blocks, Dr. M.Brink for the collection of the tissue samples and Dr. L.van de Vijver for the collection of buccal swabs; Prof A.de Bru€ıne for expert pathological advice; Dr. S. de Vogel

224 B.W.C. Bongaerts et al. / Alcohol 45 (2011) 217e225

for molecular analyses; S. van de Crommert, H. Brants, J.Nelissen, C. de Zwart, M. Moll, W. van Dijk, and A. Pistersfor data management; and H. van Montfort, T. van Moer-gastel, L. van den Bosch, and R. Schmeitz for programmingassistance. Finally, we would like to thank Dr. A. Volovics,Dr. A. Kester, and Dr. V Lima Passos for statistical advice.

Funding: This work was supported by a grant from theEuropean Research Advisory Board (ERAB) and the DutchCancer Society (grant number UM2004-3171).

References

Nevo Table (1986). Dutch food composition table 1986–1987, Voorlich-

tingsbureau voor de voeding, The Hague, The Netherlands.

World Cancer Research Fund / American Institute for Cancer Research

(2007). Food, nutrition, physical activity, and the prevention of cancer:

a global perspective. Washington, DC: AICR.

Bagnardi, V., Blangiardo, M., La Vecchia, C., and Corrao, G. (2001). A

meta-analysis of alcohol drinking and cancer risk. Br. J. Cancer 85,

1700–1705.

Boffetta, P., and Hashibe, M. (2006). Alcohol and cancer. Lancet Oncol. 7,

149–156.

Bongaerts, B. W. C., van den Brandt, P. A., Goldbohm, R. A., de

Goeij, A. F. P. M., and Weijenberg, M. P. (2008). Alcohol consumption,

type of alcoholic beverage and risk of colorectal cancer at specific

subsites. Int. J. Cancer 123, 2411–2417.

Borras, E., Coutelle, C., Rosell, A., Fernandez-Muixi, F., Broch, M.,

Crosas, B., et al. (2000). Genetic polymorphism of alcohol dehydroge-

nase in Europeans: the ADH2*2 allele decreases the risk for alco-

holism and is associated with ADH3*1. Hepatology 31, 984–989.

Bosron, W. F., and Li, T. K. (1986). Genetic polymorphism of human liver

alcohol and aldehyde dehydrogenases, and their relationship to alcohol

metabolism and alcoholism. Hepatology 6, 502–510.

Bosron, W. F., and Li, T. K. (1987). Catalytic properties of human liver

alcohol dehydrogenase isoenzymes. Enzyme 37, 19–28.

van den Brandt, P. A., Goldbohm, R. A., van ’t Veer, P., Volovics, A.,

Hermus, R. J., and Sturmans, F. (1990a). A large-scale prospective

cohort study on diet and cancer in The Netherlands. J. Clin. Epidemiol.

43, 285–295.

van den Brandt, P. A., Schouten, L. J., Goldbohm, R. A., Dorant, E., and

Hunen, P. M. (1990b). Development of a record linkage protocol for

use in the Dutch Cancer Registry for Epidemiological Research. Int.

J. Epidemiol. 19, 553–558.

Brink, M., de Goeij, A. F., Weijenberg, M. P., Roemen, G. M.,

Lentjes, M. H., Pachen, M. M., et al. (2003). K-ras oncogene mutations

in sporadic colorectal cancer in The Netherlands Cohort Study. Carci-

nogenesis 24, 703–710.

Casparie, M., Tiebosch, A. T., Burger, G., Blauwgeers, H., van de Pol, A.,

van Krieken, J. H., et al. (2007). Pathology databanking and biobank-

ing in The Netherlands, a central role for PALGA, the nationwide

histopathology and cytopathology data network and archive. Cell On-

col. 29, 19–24.

Chen, J., Ma, J., Stampfer, M. J., Hines, L. M., Selhub, J., and

Hunter, D. J. (2001). Alcohol dehydrogenase 3 genotype is not

predictive for risk of colorectal cancer. Cancer Epidemiol.

Biomarkers Prev. 10, 1303–1304.

Cho, E., Smith-Warner, S. A., Ritz, J., van den Brandt, P. A.,

Colditz, G. A., Folsom, A. R., et al. (2004). Alcohol intake and colo-

rectal cancer: a pooled analysis of 8 cohort studies. Ann. Intern. Med.

140, 603–613.

Corrao, G., Bagnardi, V., Zambon, A., and La Vecchia, C. (2004). A meta-

analysis of alcohol consumption and the risk of 15 diseases. Prev. Med.

38, 613–619.

Coutelle, C., Hohn, B., Benesova, M., Oneta, C. M., Quattrochi, P.,

Roth, H. J., et al. (2004). Risk factors in alcohol associated breast

cancer: alcohol dehydrogenase polymorphism and estrogens. Int. J.

Oncol. 25, 1127–1132.

Coutelle, C., Ward, P. J., Fleury, B., Quattrocchi, P., Chambrin, H.,

Iron, A., et al. (1997). Laryngeal and oropharyngeal cancer, and

alcohol dehydrogenase 3 and glutathione S-transferase M1 polymor-

phisms. Hum. Genet. 99, 319–325.

Day, C. P., Bashir, R., James, O. F., Bassendine, M. F., Crabb, D. W.,

Thomasson, H. R., et al. (1991). Investigation of the role of polymor-

phisms at the alcohol and aldehyde dehydrogenase loci in genetic predis-

position to alcohol-related end-organ damage. Hepatology 14, 798–801.

Estonius, M., Svensson, S., and Hoog, J. O. (1996). Alcohol dehydroge-

nase in human tissues: localisation of transcripts coding for five classes

of the enzyme. FEBS Lett. 397, 338–342.

Ferrari, P., Jenab, M., Norat, T., Moskal, A., Slimani, N., Olsen, A., et al.

(2007). Lifetime and baseline alcohol intake and risk of colon and

rectal cancers in the European prospective investigation into cancer

and nutrition (EPIC). Int. J. Cancer 121, 2065–2072.

Freudenheim, J. L., Ambrosone, C. B., Moysich, K. B., Vena, J. E.,

Graham, S., Marshall, J. R., et al. (1999). Alcohol dehydrogenase 3

genotype modification of the association of alcohol consumption with

breast cancer risk. Cancer Causes Control 10, 369–377.

Gauderman, W. J. (2002). Sample size requirements for matched case-

control studies of gene-environment interaction. Stat. Med. 21, 35–50.

Giovannucci, E., Chen, J., Smith-Warner, S. A., Rimm, E. B., Fuchs, C. S.,

Palomeque, C., et al. (2003). Methylenetetrahydrofolate reductase,

alcohol dehydrogenase, diet, and risk of colorectal adenomas. Cancer

Epidemiol. Biomarkers Prev. 12, 970–979.

Goedde, H. W., Agarwal, D. P., Fritze, G., Meier-Tackmann, D., Singh, S.,

Beckmann, G., et al. (1992). Distribution of ADH2 and ALDH2 geno-

types in different populations. Hum. Genet. 88, 344–346.

Goldbohm, R. A., van den Brandt, P. A., Brants, H. A., van’t Veer, P.,

Al, M., Sturmans, F., et al. (1994). Validation of a dietary questionnaire

used in a large-scale prospective cohort study on diet and cancer. Eur.

J. Clin. Nutr. 48, 253–265.

Harty, L. C., Caporaso, N. E., Hayes, R. B., Winn, D. M., Bravo-Otero, E.,

Blot, W. J., et al. (1997). Alcohol dehydrogenase 3 genotype and risk of

oral cavity and pharyngeal cancers. J. Natl. Cancer Inst. 89, 1698–1705.

Jokelainen, K., Roine, R. P., Vaananen, H., Farkkila, M., and Salaspuro, M.

(1994). In vitro acetaldehyde formation by human colonic bacteria. Gut

35, 1271–1274.

Jokelainen, K., Siitonen, A., Jousimies-Somer, H., Nosova, T., Heine, R.,

and Salaspuro, M. (1996). In vitro alcohol dehydrogenase-mediated

acetaldehyde production by aerobic bacteria representing the normal

colonic flora in man. Alcohol. Clin. Exp. Res. 20, 967–972.

Jung, A. Y., Poole, E. M., Bigler, J., Whitton, J., Potter, J. D., and

Ulrich, C. M. (2008). DNA methyltransferase and alcohol dehydroge-

nase: gene-nutrient interactions in relation to risk of colorectal polyps.

Cancer Epidemiol. Biomarkers Prev. 17, 330–338.

Knaapen, A. M., Ketelslegers, H. B., Gottschalk, R. W., Janssen, R. G.,

Paulussen, A. D., Smeets, H. J., et al. (2004). Simultaneous genotyping

of nine polymorphisms in xenobiotic-metabolizing enzymes by multi-

plex PCR amplification and single base extension. Clin. Chem. 50,

1664–1668.

Longnecker, M. P., Orza,M. J., Adams,M. E., Vioque, J., and Chalmers, T. C.

(1990). A meta-analysis of alcoholic beverage consumption in relation to

risk of colorectal cancer. Cancer Causes Control 1, 59–68.

Meulenbelt, I., Droog, S., Trommelen, G. J., Boomsma, D. I., and

Slagboom, P. E. (1995). High-yield noninvasive human genomic

DNA isolation method for genetic studies in geographically dispersed

families and populations. Am. J. Hum. Genet. 57, 1252–1254.

Miettinen, O. S. (1985). Theoretical Epidemiology. Principles of Occur-

rence Research in Medicine. New York: Wiley.

Moskal, A., Norat, T., Ferrari, P., and Riboli, E. (2006). Alcohol intake and

colorectal cancer risk: a dose-response meta-analysis of published

cohort studies. Int. J. Cancer 120, 664–671.

225B.W.C. Bongaerts et al. / Alcohol 45 (2011) 217e225

Pestalozzi, D. M., Buhler, R., von Wartburg, J. P., and Hess, M.

(1983). Immunohistochemical localization of alcohol dehydroge-

nase in the human gastrointestinal tract. Gastroenterology 85,

1011–1016.

Poikolainen, K., Vartiainen, E., and Korhonen, H. J. (1996). Alcohol intake

and subjective health. Am. J. Epidemiol. 144, 346–350.

Prentice, R. L. (1985). A case-cohort design for epidemiologic cohort

studies and disease prevention trails. Biometrika 73, 1–12.

Rehm, J. (2000). Alcohol consumption and mortality. What do we know

and where should we go? Addiction 95, 989–995.

San Jose, B., van de Mheen, H., van Oers, J. A., Mackenbach, J. P., and

Garretsen, H. F. (1999). The U-shaped curve: various health measures

and alcohol drinking patterns. J. Stud. Alcohol 60, 725–731.

Schoenfeld, D. (1982). Partial residuals for the proportional hazards

regression model. Biometrika 69, 239–241.

Seitz, H. K., Egerer, G., Oneta, C., Kramer, S., Sieg, A., Klee, F., et al.

(1996). Alcohol dehydrogenase in the human colon and rectum. Diges-

tion 57, 105–108.

Seitz, H. K., Simanowski, U. A., Garzon, F. T., Rideout, J. M., Peters, T. J.,

Koch, A., et al. (1990). Possible role of acetaldehyde in ethanol-related

rectal cocarcinogenesis in the rat. Gastroenterology 98, 406–413.

Seitz, H. K., and Stickel, F. (2007). Molecular mechanisms of alcohol-

mediated carcinogenesis. Nat. Rev. Cancer 7, 599–612.

Shaper, A. G. (1995). Mortality and alcohol consumption. Non-drinkers

shouldn’t be used as baseline. Br. Med. J. 310, 325.

Terry, M. B., Gammon, M. D., Zhang, F. F., Knight, J. A., Wang, Q.,

Britton, J. A., et al. (2006). ADH3 genotype, alcohol intake and breast

cancer risk. Carcinogenesis 27, 840–847.

Tiemersma, E. W., Wark, P. A., Ocke, M. C., Bunschoten, A., Otten, M. H.,

Kok, F. J., et al. (2003). Alcohol consumption, alcohol dehydrogenase 3

polymorphism, and colorectal adenomas. Cancer Epidemiol. Biomarkers

Prev. 12, 419–425.

Yin, S. J., Liao, C. S., Lee, Y. C., Wu, C. W., and Jao, S. W. (1994).

Genetic polymorphism and activities of human colon alcohol and alde-

hyde dehydrogenases: no gender and age differences. Alcohol. Clin.

Exp. Res. 18, 1256–1260.