Alcohol metabolism genes and risks of site‐specific cancers in Chinese adults: An 11‐year prospective study

Abstract Two genetic variants that alter alcohol metabolism, ALDH2‐rs671 and ADH1B‐rs1229984, can modify oesophageal cancer risk associated with alcohol consumption in East Asians, but their associations with other cancers remain uncertain. ALDH2‐rs671 G>A and ADH1B‐rs1229984 G>A were genotyped in 150 722 adults, enrolled from 10 areas in China during 2004 to 2008. After 11 years' follow‐up, 9339 individuals developed cancer. Cox regression was used to estimate hazard ratios (HRs) for site‐specific cancers associated with these genotypes, and their potential interactions with alcohol consumption. Overall, the A‐allele frequency was 0.21 for ALDH2‐rs671 and 0.69 for ADH1B‐rs1229984, with A‐alleles strongly associated with lower alcohol consumption. Among men, ALDH2‐rs671 AA genotype was associated with HR of 0.69 (95% confidence interval: 0.53‐0.90) for IARC alcohol‐related cancers (n = 1900), compared to GG genotype. For ADH1B‐rs1229984, the HRs of AG and AA vs GG genotype were 0.80 (0.69‐0.93) and 0.75 (0.64‐0.87) for IARC alcohol‐related cancers, 0.61 (0.39‐0.96) and 0.61 (0.39‐0.94) for head and neck cancer (n = 196) and 0.68 (0.53‐0.88) and 0.60 (0.46‐0.78) for oesophageal cancer (n = 546). There were no significant associations of these genotypes with risks of liver (n = 651), colorectal (n = 556), stomach (n = 725) or lung (n = 1135) cancers. Among male drinkers, the risks associated with higher alcohol consumption were greater among ALDH2‐rs671 AG than GG carriers for head and neck, oesophageal and lung cancers (P interaction < .02). Among women, only 2% drank alcohol regularly, with no comparable associations observed between genotype and cancer. These findings support the causal effects of alcohol consumption on upper aerodigestive tract cancers, with ALDH2‐rs671 AG genotype further exacerbating the risks.


| INTRODUCTION
Cancer is a leading cause of premature mortality and disability globally, accounting for an estimated 19.3 million new cancer cases and 10 million deaths in 2020. 1 Worldwide, 24% of total cancer cases, including 37% of lung cancer and 47% of digestive tract (oesophagus, stomach and liver) cancers, occurred in China. 1 Based mainly on observational studies, the International Agency for Research on Cancer (IARC) reported that there is sufficient evidence that alcohol consumption is causally related to development of cancers in the head and neck, oesophagus, liver, colon-rectum and female breast, but causal evidence remains inconclusive for other cancer sites including lung and stomach due to other possible confounders (eg, smoking, diet). 2 Worldwide, it was estimated that 3 million deaths could be attributed to alcohol consumption, including >0.4 million from cancer. 3 Alcohol consumption has been increasing over recent decades in China, almost exclusively in men, 4 and is a major contributor to the total cancer burden in Chinese men. 5,6 In China and other East Asian populations, two common genetic variants affect alcohol tolerability and are strongly associated with lower alcohol intake. 7

An East
Asian-specific loss-of-function variant in the aldehyde dehydrogenase 2 (ALDH2) gene (rs671 G>A) substantially decreases the breakdown of acetaldehyde, which is a Group 1 human carcinogen classified by IARC and a toxic metabolite produced during alcohol metabolism, causing the characteristic East Asian alcohol flushing response. 2,8 Another variant in the alcohol dehydrogenase 1B (ADH1B) gene (rs1229984 G>A) accelerates acetaldehyde formation from alcohol. 9 These genetic variants, which are randomly allocated at conception and usually independent of other lifestyle exposures, can be used as instruments for alcohol intake to help assess the likely causal effects of alcohol consumption on disease risks. 7,10 Importantly, appropriate understanding of the interplay between these genetic variants and between genetic variants and alcohol consumption may provide insight into the involvement of alcohol-derived acetaldehyde in the carcinogenesis of certain site-specific cancers.
Previous studies have shown that the ALDH2-rs671 AA and ADH1B-rs1229984 AA and AG genotypes were associated with lower oesophageal cancer risk compared to the GG genotype, 11,12 and that ALDH2-rs671 genotype may modify the relationship between alcohol intake and oesophageal cancer risk. [12][13][14] However, there is limited evidence on the associations of these genotypes with risk of cancer at other sites, 11,15 and for the potential interactions between genotype and alcohol intake on cancer risks. [16][17][18][19][20][21][22][23] A comprehensive assessment of the interplay between ALDH2-rs671, ADH1B-rs1229984 and alcohol consumption on risks of different cancer types in a large-scale population-based cohort study may provide valuable insights into the aetiological role of alcohol on different cancers.
Using data from the prospective China Kadoorie Biobank (CKB), we investigated the associations of ALDH2-rs671 and ADH1B-rs1229984 with total and common site-specific cancers in 151 000 Chinese adults. In addition, we investigated possible gene-alcohol and gene-gene interactions on cancer risks.

| Study population
Details of the CKB study design and methods have been previously reported. 24 Briefly, 512 726 adults aged 30 to 79 years were recruited from 10 rural and urban areas across China during 2004 to 2008. Trained health workers administered a laptop-based questionnaire recording sociodemographic factors, lifestyles (eg, alcohol drinking, smoking, diet, physical activity) and medical history; undertook physical measurements (eg, blood pressure, anthropometry); and collected a blood sample for long-term storage. Two resurveys of~5% randomly selected surviving participants were conducted using similar procedures in 2008 and 2013 to 2014.

| Assessment of alcohol consumption
Detailed questionnaire assessment of alcohol consumption has been described previously. [25][26][27] Based on their past and current drinking history, participants were classified into: abstainers; ex-regular drinkers; occasional drinkers and current regular drinkers (ie, had drunk alcohol in most weeks in the past year). Current regular drinkers were asked further questions about their drinking patterns including drinking frequency, beverage type and amount consumed for each type on a typical drinking day, age started drinking and experience of alcohol flushing response after drinking. Level of alcohol consumption was calculated as grams (g) of alcohol per week based on frequency, beverage type and amount consumed. Further details of alcohol assessment are reported in the Supporting Information Methods.

| Follow-up and main outcome measures
The vital status of participants was obtained periodically from local death registries, supplemented by annual active confirmation through local residential, health insurance and administrative records. Incident cancers were collected through linkage with cancer registries and the national health insurance system (>98% coverage across the 10 study areas), supplemented by active follow-up approach (see Table S1 and Supporting Information Methods for further details on completeness and quality of cancer outcome measures). 28 All events were coded with International Classification of Diseases, 10th Revision (ICD-10), blinded to the baseline information.
The main cancer outcomes investigated in our study were total cancer; IARC alcohol-related cancers (defined as cancers with convincing causal relevance with alcohol as concluded by IARC 2 ) which were cancers of the head and neck (included cancers of the lip and oral cavity, pharynx and larynx; ICD-10: C00-C14, C32), oesophagus (C15), colon-rectum (C18-C20), liver (C22) and female breast (C50); and certain other site-specific cancers including lung cancer (C33-C34) and stomach cancer (C16). Other cancers, apart from illdefined neoplasms (C76-C80, C97), were combined as "other cancers of known sites." Upper aerodigestive tract (UADT) cancers were defined as cancers of the head and neck and oesophagus. By 1 January 2018, 49 459 (9.7%) deaths were recorded among the 512 726 CKB participants, with 5302 (1.0%) lost to follow-up.

| Statistical analysis
Participants with missing data on genomic principal components (n = 313) were excluded from the analyses, leaving 150 722 participants in the study (see Figure S1). Means and percentages of baseline characteristics were calculated by genotype, standardised to the age and study area structure of the genotyped study population.
Cox proportional hazard models, stratified by age-at-risk and study area and adjusted for 12 genomic principal components, were used to estimate hazard ratios (HRs) for cancers reported during follow-up associated with ALDH2-rs671 and ADH1B-rs1229984 genotypes, in men and women separately. The genotypic associations with cancer risk were further examined separately by drinking status.
Potential effect modification of the associations between amount of alcohol consumption and cancer risks by genotype were investigated among current regular drinkers. The joint effects of alcohol consumption and ALDH2-rs671 were assessed by estimating the HRs associated with four categories defined by baseline alcohol intake (<280, 280+ g/wk in men; <70, 70+ g/wk in women) and genotype (GG, AG), excluding AA individuals as few of them drank. The models were stratified by age-at-risk and study area and adjusted for 12 genomic principal components, education, household income, smoking, fruit intake, physical activity, body mass index (BMI) and family history of cancer. Likelihood ratio test was used to test for interaction between alcohol consumption and genotype by comparing two models with and without the interaction term. As tobacco smoking also produces acetaldehyde, subgroup analyses by smoking status were conducted to assess potential residual confounding from smoking. The joint effects of alcohol and ADH1B-rs1229984 were assessed using similar methods. Alcohol intake was also modelled as a continuous variable to estimate adjusted HRs of cancers associated with a 280 g/wk higher usual alcohol intake by genotype, with heterogeneity in effect sizes assessed by χ 2 tests.
The joint effects of ALDH2-rs671 and ADH1B-rs1229984 were examined by estimating the HRs associated with the nine groups defined by the combination of genotypes for both variants (from GG/ GG to AA/AA [ALDH2-rs671/ADH1B-rs1229984], which represented the highest to lowest mean alcohol intake) in men, stratified by ageat-risk and study area and adjusted for 12 genomic principal components.
Various sensitivity analyses were performed, including: (a) additional adjustments for socioeconomic status and major lifestyle risk factors for cancer; (b) area-stratified analysis by combining within-area genotypic effects using inverse-variance-weighted meta-analysis to investigate potential residual confounding by population stratification (given the differences in allele frequencies between study areas) and (c) excluding individuals with a prior history of cancer at baseline from the genealcohol interaction analysis to reduce potential reverse causation due to changes in drinking habits for health reasons.
As few women drank alcohol regularly in CKB, 25,26 the main analyses were focused among men, with genotypic analyses among women conducted to assess the presence of pleiotropic effects (ie, genotypic associations that are not mediated by alcohol consumption).
For analyses involving more than two exposure categories, the floating absolute risk method was used to compute the groupspecific 95% confidence intervals (CIs) derived from the variance of the log hazard of each category, such that each HR (including the one for the reference group) has a group-specific 95% CI that facilitates comparisons between any two categories, as described previously. 7,29,30 For comparisons of two groups (ie, an exposure category with the reference group), conventional 95% CIs were reported. Repeat alcohol measures for participants who attended both subsequent resurveys were used to correct for regression dilution bias. 31    The overall mean alcohol intake was calculated across all categories of drinking status. Calculations assigned an intake of 0 g/wk to baseline nondrinkers and 5 g/wk to baseline occasional drinkers.

| RESULTS
Among the 150 722 study participants, the mean age was 52.1 (SD 10.7) years, 40% were men and 56% lived in rural areas. The overall A-allele frequency was 0.21 (range by area from 0.13 to 0.29) for ALDH2-rs671 and 0.69 (from 0.64 to 0.74) for ADH1B-rs1229984, with a generally higher frequency in southern than northern areas for both variants (Table S2).
Among men, ALDH2-rs671 was strongly associated with the prevalence of current regular drinking (46%, 17% and 1% for GG, AG and AA, respectively) and mean alcohol intake (143, 35, 2 g/wk, respectively; all P trend < .0001) ( Table 1). ADH1B-rs1229984 genotype was also associated with current regular drinking prevalence (43%, 34% and 32%) and mean alcohol intake (146, 99 and 91 g/wk). Among male current regular drinkers, ALDH2-rs671 was strongly associated with the alcohol flushing response (11%, 56% and 62%) and age at drinking onset (28, 32 and 40 years); the correlations with the alcohol flushing response were consistent directionally for A alleles of both variants, but the effects were weaker with ADH1B-rs1229984 (15%, 18% and 20%) (P trend < .0001 for all above). In women, similar patterns of associations between drinking patterns and genotype were observed as in men, but with very low prevalence of regular drinking (2%) the differences were small in magnitude (Table S3). There were no material effects of these genotypes on smoking or other lifestyle characteristics in men or women, except a slightly higher prevalence of daily fresh fruit intake and lower physical activity in male ALDH2-rs671 A-allele carriers, and  F I G U R E 1 Associations of genotypes for ALDH2-rs671 (A) and ADH1B-rs1229984 (B) with risks of total and selected site-specific cancers, in men. Cox models were stratified by age-at-risk and study area, and adjusted for 12 genomic principal components. Each solid square represents HR with the area inversely proportional to the variance of the group-specific log hazard. but not in never-regular drinkers. A higher risk of lung cancer was observed in those with AG vs GG genotype among male neverregular drinkers, but not among ever-regular drinkers (Table S4).    Figure 2 (P interaction < .0001); for IARC alcohol-related cancers, the corresponding HRs were more extreme (P interaction = .0002). Similar significant interactions between alcohol and ALDH2-rs671 were also observed for site-specific cancers, especially oesophageal cancer (P interaction < .0001) and head and neck cancer (P interaction < .01) and less so for lung cancer (P interaction = .016) (Figure 2). The associations between these cancers and joint alcohol-ALDH2-rs671 groups were broadly similar in never-regular smokers and in ever-regular smokers ( Figure 3). There were no clear interactions between alcohol consumption and ALDH2-rs671 for liver or colorectal cancers, but the dose-response association between alcohol and stomach cancer appeared stronger in ALDH2-rs671 AG drinkers than in GG drinkers (HR = 3.36 [1.73-6.54] vs 1.02 [0.73-1.41], per 280 g/wk; P heterogeneity = .002; Figure S2).
For ADH1B-rs1229984, no clear interactions with alcohol consumption on cancer risks were observed among male current regular drinkers (Figures S3 and S4).

Examination of the joint effects of the two genetic variants in men
showed that the risks of IARC alcohol-related cancers and of UADT cancers were highest for the combination of ALDH2-rs671 AG with ADH1B-rs1229984 GG genotypes, followed by the combination of GG/GG genotypes, and were lowest for the combinations of ALDH2-rs671 AA with ADH1B-rs1229984 AG or AA genotypes (Table S6).
However, these modest gene-gene interactions were not significant.
In men the genotypic associations with cancers were unaltered with additional adjustment for other cancer risk factors (education, household income, smoking, fresh fruit intake, physical activity, BMI, family cancer history, hepatitis B virus [HBV] infection status; Figure S5) or in area-stratified as opposed to area-adjusted analyses ( Figure S6). Similar ALDH2-rs671-alcohol interactions were observed as in the main analyses after excluding individuals with prior cancer or further adjusting for HBV infection status ( Figures S7 and S8).
Among women, there were no clear associations of these two genetic variants with risks of overall or IARC alcohol-related cancers ( Figure 4). For site-specific cancers, compared to ALDH2-rs671 GG genotype, AG genotype was associated with a lower oesophageal cancer risk, while AA genotype was associated with a higher liver cancer risk, but the numbers of cases involved were small. When comparing the associations of these genotypes with cancer risks between men and women, heterogeneity of the associations was seen for several IARC alcohol-related cancers (Table S7).   Figure 1 Among women, no clear gene-alcohol interactions were observed (Tables S8-S10).

| DISCUSSION
In this large genetic study of Chinese adults, two common genetic variants, ALDH2-rs671 G>A and ADH1B-rs1229984 G>A, which strongly reduced alcohol consumption, were associated with lower risks of overall and IARC alcohol-related cancers, especially UADT cancers, in men among whom over a third drank alcohol regularly. Among male drinkers, ALDH2-rs671 genotype significantly modified the effects of alcohol consumption on certain cancers, with greater excess risks in men with the AG than GG genotype for a given level of alcohol consumption, especially for UADT cancers and potentially for lung cancer, regardless of smoking status. Among women, very few drank alcohol regularly and these variants were not associated with overall or IARC alcohol-related cancer risk.
Previous case-control studies and meta-analyses in East Asian populations have reported associations of ALDH2-rs671 genotypes with risks of oesophageal cancer 10,12,32 and head and neck cancer, 33 and that the relationships may be modified by alcohol consumption.
Compared to GG genotype, AA genotype was associated with an overall lower risk, while AG genotype was associated with a higher risk among drinkers but not among never drinkers. 10,12,33 The increased risks among ALDH2-rs671 AG drinkers, who have markedly elevated acetaldehyde levels after consuming alcohol, but not among never drinkers suggest that acetaldehyde may be the underlying mechanism through which alcohol consumption increases UADT cancer risk. In the present prospective study of Chinese population, we observed broadly similar effect modifications of ALDH2-rs671 genotypes on risks for IARC alcohol-related and oesophageal cancers associated with alcohol consumption, as in previous studies. 10 [12][13][14]20,33 there was also suggestive evidence of an ALDH2-rs671-alcohol interaction on lung cancer risk, which was concordant with a previous Japanese case-control study (505 cases) 36 and our previous report of a stronger dose-response association of alcohol with lung cancer among male drinkers reporting the alcohol flushing response. 5 While we found no clear evidence of interactions between ALDH2-rs671 and alcohol consumption for colorectal or liver cancers, the somewhat stronger dose-response association of alcohol intake with stomach cancer among male ALDH2-rs671 AG drinkers was consistent with case-control studies in Japan (1375 cases). 19 In contrast to the ALDH2-rs671-alcohol interactions observed, we found no clear interactions between alcohol consumption and ADH1B-rs1229984, or between the two genetic variants, on cancer risks, which were largely consistent with previous studies. [16][17][18][19][20][21]44,[49][50][51] The biological pathways via which alcohol consumption may cause cancers are not fully understood and likely vary by cancer site.
A major pathway proposed is via local exposure to alcohol-formed acetaldehyde, especially in the upper gastrointestinal tract where, in contrast to the liver, the mucosa has limited capacity to eliminate acetaldehyde. 2,52 After drinking, local acetaldehyde exposure in the upper digestive tract mucosa starts instantly, mainly due to microbial acetaldehyde formation from alcohol in saliva, followed by long-term acetaldehyde formation from alcohol that is diffused back to saliva from blood circulation. 53 Particularly in ALDH2-deficient individuals, excess salivary acetaldehyde may be produced through human ethanol metabolism in the salivary glands, resulting in excess long-term acetaldehyde exposure in the upper digestive tract mucosa. 52,53 This is supported by our findings of increased UADT cancer risks only among ALDH2-rs671 AG drinkers but not among ALDH2-rs671 AG never-regular drinkers, and the greater excess risks in ALDH2-rs671 AG drinkers than GG drinkers for a given amount of alcohol consumed. For a given level of alcohol consumption, ALDH2-deficient individuals were reported to be exposed to 2-fold to 3-fold (salivary) and 5-fold to 6-fold (gastric juice) higher acetaldehyde concentrations than those with active ALDH2 enzyme, 52 supporting the putative involvement of local alcohol-derived acetaldehyde in upper gastrointestinal tract carcinogenesis. In addition to increased alcohol consumption, the ADH1B-rs1229984 GG genotype is associated with slower ethanol oxidation such that ethanol remains in the blood and saliva for longer, which may result in prolonged exposure to salivary acetaldehyde due to oral microbial acetaldehyde production from ethanol and consequently increased risks of UADT cancers. 54,55 This prolonged salivary acetaldehyde exposure would be greater in the presence of ALDH2-rs671 AG genotype. 53 Alcohol may also increase risks of cancers of the upper digestive and respiratory tract by acting as a solvent for tobacco carcinogens. 40 Moreover, smoking and heavy drinking combined may also modify the oral microflora to produce higher acetaldehyde levels in saliva. 53 It is possible that the observed ALDH2-rs671-alcohol interactions might be partly related to acetaldehyde from smoking rather than from alcohol consumption alone, especially for lung cancer for which no causality of alcohol has been inferred by genotypic associations. This is however unlikely, as similar The chief strengths of our study include the prospective study design, large community-based study population, reliable alcohol consumption data 5,7 and extensive information on lifestyle risk factors, and reasonably large numbers of incident events for various common cancer sites traced via comprehensive and complete follow-up. We were also able to minimise population stratification bias with adjustments for study area and genomic principal components. Nevertheless, several limitations also warrant consideration. Although the two genetic variants were strong instruments for alcohol intake and were not associated with smoking, they were weakly associated with other cancer risk factors (eg, fresh fruit intake, physical activity, BMI) in CKB. However, the differences were extremely small in magnitude and might have been the consequences of alcohol consumption.
Importantly, additional adjustments for these risk factors did not alter the main findings. Also, the two enzymes affected by the studied genetic variants are involved in many biochemical pathways, 59,60 which might potentially affect carcinogenesis independent of alcohol consumption. Nonetheless, among women who rarely drank alcohol despite their genotype, there were no clear genotypic associations with IARC alcohol-related cancers or most site-specific cancers.
Although ALDH2-rs671 was associated with oesophageal cancer and liver cancer among women, which might be partly related to acetaldehyde exposure from other sources (eg, air pollution, cooking oil fumes, passive smoking) or endogenous aldehyde exposure, 2,59 the associations were not directionally consistent to those observed in men.
These findings suggest the genotypic results in men were likely to be driven chiefly by alcohol consumption rather than by pleiotropic pathways. While further adjustment for HBV infection status did not materially alter our findings, other major risk factors, for example, Helicobacter pylori infection for stomach cancer and hepatitis C infection for liver cancer were not available to be included in our analysis. Although these infectious agents are unlikely to be confounding factors in the associations between the studied genetic variants and cancer risks, whether they may interact with alcohol consumption and ALDH2 deficiency remains to be elucidated.
Finally, our study may be underpowered to detect any weak causal effects of alcohol intake on site-specific cancers other than UADT cancers.
In conclusion, in Chinese men ALDH2-rs671 G>A and ADH1B-rs1229984 G>A genotypes were associated with lower risks of overall and IARC alcohol-related cancers, mainly UADT cancers. Furthermore, ALDH2-rs671 genotype may modify the effects of alcohol consumption on certain cancers, especially UADT cancers. These findings support the causal role of alcohol consumption in the aetiology of UADT cancers, which is exacerbated in individuals with inherited low alcohol tolerability. The study reinforces the need to lower population-levels of alcohol consumption for cancer prevention, especially in China where alcohol consumption is increasing despite the low alcohol tolerability among a subset of the population.