Alcohol consumption and risk of gastric cardia adenocarcinoma and gastric non-cardia adenocarcinoma: a 16-year prospective analysis from the NIH-AARP Diet and Health cohort

Abstract

The role of alcoholic beverages in the etiology of gastric cancer is unclear. Recent summaries showed a positive association between higher alcohol intake and gastric cancer risk, but the magnitude of association is small, there is moderate heterogeneity among studies, and most cases were from Asian populations. We prospectively investigated the associations of alcohol consumption with gastric cardia adenocarcinoma (GCA) and gastric non-cardia adenocarcinoma (GNCA) in 490,605 adults, aged 50-71 years at baseline who participated in the NIH-AARP Diet and Health Study. Alcohol consumption in the past year was assessed at baseline by questionnaire and defined as total grams of ethanol intake per day or as a categorical variable: non-drinker, up to or including one drink/day, one to three drinks/day, and greater than three drinks/day. We used multivariable-adjusted Cox proportional hazards regression to calculate the hazard ratios (HR) and 95% confidence intervals (CI) for associations between alcohol intake and risk of gastric adenocarcinomas. Through 2011, 662 incident cases of GCA and 713 of GNCA occurred. We found no association between higher alcohol consumption and GCA or GNCA, when examined as total alcoholic beverage intake individual beverage types of beer, wine and liquor. Furthermore, we observed no association by stratum of sex, ethnic group, educational level, or smoking status. We did, however, observe lower risk of GNCA among participants who drank up to one drink per day (HR=0.81, 95%CIs:0.67-0.97) compared to nondrinkers. In conclusion, alcohol consumption was not associated with increased risk of GCA or GNCA in this large US cohort.

Introduction

Alcohol consumption has been causally linked to cancers of the oral cavity, pharynx, larynx, esophagus, liver, colon, rectum, and female breast, and was classified by the International Agency for Research on Cancer (IARC) as a Group 1 carcinogen ^1-3. However, evidence linking alcohol consumption to gastric adenocarcinomas is inconsistent.

In 2009, an IARC working group concluded that there was inadequate evidence of an association between alcohol consumption and gastric adenocarcinoma ⁴. In 2016, with the accrual of more studies, the World Cancer Research Fund International (WCRF) Continuous Update Project (CUP) found “strong” evidence that consuming three or more alcoholic drinks per day (i.e., ≥ 45 grams of ethanol/day) is a probable a cause of stomach cancer⁵. However, the risk estimates for the 23 studies included in the CUP meta-analysis were inconsistent, resulting in a weak, but statistically significant association, between alcohol consumption and gastric cancer. Notably most cases (70%) were from Asian studies, with only 401 cases from North American studies, raising the question of generalizability given genetic differences in ethanol metabolism across populations. East Asian populations have certain polymorphisms in acetaldehyde dehydrogenase (ADH) and aldehyde dehydrogenase (ALDH) that alter the formation of the carcinogen acetaldehyde from ethanol, cause a flushing response, and have been demonstrated to modify associations between alcohol intake and cancer ^6-8.

We previously examined the association between alcohol intake and gastric adenocarcinomas (n=375, 1995/1996-2000) in the NIH-AARP cohort and found little evidence of association ⁹. However, this initial analysis lacked the statistical power to detect modest associations. With extended follow-up and the ascertainment of 1000 additional incident cases, we have updated and expanded upon the prior investigation by conducting analyses stratified by sublocation of the gastric cancer and potential effect modifiers such as sex and smoking.

Materials and Methods

Study population

The NIH-AARP Diet and Health Study is a prospective cohort that was established in 1995-1996 with the mailing of questionnaires to 3.5 million AARP members, aged 50-71 years, living in one of six US states (California, Florida, Louisiana, New Jersey, North Carolina, and Pennsylvania) or one of two metropolitan areas (Atlanta, Georgia, and Detroit, Michigan) ¹⁰. The baseline questionnaire queried about demographics, alcohol use, tobacco use, diet, anthropometrics, and medical history. Overall, 17.6% (n= 617,119) of potential participants returned the questionnaire. Of the 566,398 respondents who satisfactorily completed the baseline questionnaire, we further excluded: proxy respondents (n= 15,760), subjects with a history of cancer (except nonmelanoma skin cancer) at baseline (n=51,346), those with only information on cancer death (n=4,268), those with energy intake more than two interquartile ranges above or below the median, which we have used as a marker for questionnaire completeness and accuracy (n=4,387), and those who died or were diagnosed with cancer on the first day of follow-up (n =32). Our final analytic sample included 490,605 participants of whom 454,038 (92.6%) self-reported their race/ethnicity as non-Hispanic white. The study was approved by the Institutional Review Board of the US National Cancer Institute.

Cohort follow-up and outcomes

Vital status was obtained by linkage to the National Death Index and cancer diagnoses were updated via linkage to the state cancer registries. Case ascertainment using these methods was shown to have a high sensitivity (89.2%) and specificity (99.5%) ¹¹. Incident gastric cardia adenocarcinoma (GCA) and gastric non-cardia adenocarcinoma (GNCA) were identified by anatomic site and histologic codes of the International Classification of Diseases for Oncology, Third Edition (ICD-O-3)¹². GCA was defined with site code C16.0, which presented the cardioesophageal junction, the esophagogastric junction, and the gastroesophageal junction. GNCA was classified with site codes C16.1-C16.9. Both types of adenocarcinomas excluded histologic codes of 8070, 8072, 8240, 8246, 8560, ≥8800, 8083, 8050, 8245, and 8512.

Exposure assessment

We assessed alcohol consumption in the past year using data from the baseline 124-item food-frequency questionnaire (FFQ). Self-reported alcohol consumption, as measured by a questionnaire, was previously shown to have high reproducibility (Spearman r=0.84 after 4 years) and validity in relation to 1-week diet records (Spearman r=0.86 for men and 0.90 for women) ¹³. We calculated the number of alcoholic drinks consumed per day by combining data on usual frequency (0–6 times per day) and portion size of wine, liquor, and beer consumption over the past 12 months. One drink corresponded to one serving size according to the US Department of Agriculture’s food guide pyramid: one 12-fluid ounce beer, one 5-fluid ounce glass of wine, or one 1.5-ounce shot of liquor (each approximately 13-14 g of ethanol) ^{14, 15}. We further categorized total alcohol consumption as: non-drinker, up to or including one drinks/day, greater than one to three drinks/day, and greater than three drinks/day. The categories were combined as non-drinker, up to or including one drinks/day, greater than one drinks/day for the beer, wine, and liquor consumption analyses to allocate enough cases in each subgroup. We restricted our analysis to alcohol consumed as a beverage and did not include alcohol used in cooking.

Statistical analysis

We used Cox proportional hazards regression to calculate the hazard ratios (HR) and 95% confidence intervals (CI) for the association between alcohol consumption and risk of gastric adenocarcinomas. Follow-up time in person-years was used as the underlying time metric and was calculated from baseline to the date of diagnosis of gastric, esophageal, or head and neck cancer (to avoid possible detection bias due to surveillance for second cancers); the date of death; the date moved out of catchment area; or the end of the follow-up period (31 December 2011), whichever occurred first.

Potential confounders, including age, sex, race/ethnicity, health status, education level, body mass index (BMI), smoking status, self-reported diabetes, and the Healthy Eating Index 2010 (HEI-2010) score, were tabulated by categories of alcohol consumption and were included in multivariable-adjusted models. We defined BMI as a categorical variable according to WHO definitions¹⁶: underweight (<18.5), normal weight (18.5–24.9), overweight (25.0–29.9), and obese (≥30.0); we defined smoking status as never, former, and current (participants who reported quitting within the past year were considered as current smokers). To adjust for overall diet quality and compliance with the U.S. dietary guidelines, we used the HEI-2010 score that was estimated for each participant using data from the baseline food frequency questionnaire ¹⁰. Unknown values for adjustment variables were replaced with the most common value for a given variable. We used likelihood ratio tests to assess the potential interactive effects of sex, educational level, and smoking status on the association between total alcohol intake and risk of gastric adenocarcinoma. Linear trend tests were assessed in multivariable-adjusted models by fitting the defined category of alcohol intake as a continuous variable. In addition, we investigated the associations of beer, wine, or liquor consumption on risk of gastric adenocarcinoma, simultaneously adjusting for the other two types of alcohol as categorical variables to account for the total alcohol consumption. We also ran models stratified by sex, education level, and smoking status.

In sensitivity analyses, we excluded people who self-reported poor or fair health status as they may have recently altered their alcohol drinking due to ill-health; we excluded those with ICD site codes C16.8 (n=82) and C16.9 (n=169) which refer to overlapping lesions or not otherwise specified lesions of stomach to minimize the misclassification for risk assessment; we excluded cases diagnosed in the first 2 years to assess potential reverse causation; and analyzed the association restricted to alcohol drinkers only.

Finally, we ran models using restricted cubic splines (trimming for the highest 1% of alcohol consumption) to examine possible nonlinear relationships between alcohol consumption (defined as continuous variable, grams/ day) and risks of GCA and GNCA, in the total cohort and in alcohol drinkers only. Analyses were performed with SAS, version 9.3, software (SAS Institute, Inc., Cary, North Carolina) and we used separate tests for an overall association with disease risk and for evidence on nonlinearity. An alpha level of less than 0.05 was considered significant, and all tests were two sided. Figures were produced using the R Hmisc package (version 3.3.1).

Results

During 6.6 million person-years of follow-up (median = 15.5 years, IQR: 13.7-15.6 years), 662 GCA and 713 GNCA cases were identified. Table 1 shows the distributions of demographic and health-related factors by category of alcohol consumption. The mean age at baseline was 62 years, and a majority were non-Hispanic White (92.6%) and male (59.6%). Alcohol drinkers were more likely to be male, have obtained a college or graduate education, and were more likely to be former or current smokers than nondrinkers. People who reported consuming >3 drinks/day had a lower HEI-2010 score than those consuming less or no alcohol. In contrast, a higher percentage of alcohol nondrinkers than drinkers had self-reported diabetes.

Table 1.

Baseline characteristics of different levels of alcohol drink consumption in 490,605 members of NIH-AARP Diet and Health Study Cohort.

Characteristics	Total	Total alcohol (drinks/ day)
		Non-drinker	>0-1	>1-3	>3
Total Cohort	490,605	119,883	259,512	74,488	36,722
Age at baseline (means ± SD)	62.0±5.4	62.3±5.3	61.8±5.4	62.3±5.3	62.1±5.3
Sex (%)
Male	59.6	51.0	56.1	73.0	85.7
Female	40.4	49.0	43.9	27.0	14.3
Race/ethnicity (%)
Non-Hispanic white	92.6	89.2	92.8	95.5	95.5
Non-Hispanic black	3.9	6.3	3.5	2.0	2.5
Hispanic	1.9	1.9	2.1	1.6	1.3
Asian/Pacific Islander/American Indian/Native Alaskan	1.7	2.6	1.6	0.9	0.7
Education (%)
Less than or high school graduate1	35.3	45.4	33.9	26.4	30.9
Some college	23.2	21.5	24.0	22.7	24.1
College graduate	41.5	33.1	42.2	50.9	45.0
Smoking status (%)
Never smoker	36.6	45.2	38.4	26.3	16.6
Former smoker	49.3	42.1	48.3	59.0	59.4
Current smoker	14.2	12.8	13.3	14.7	24.0
Health Status (%)
Good/excellent health	53.9	44.4	55.4	63.3	55.2
Good health	34.0	36.2	34.3	29.5	34.6
Poor/fair health	12.1	19.4	10.3	7.2	10.3
BMI (%)
18.5-<25	33.6	30.6	33.9	38.3	31.4
<18.5	1.0	1.2	0.9	1.1	1.0
25-<30	43.9	42.0	43.6	45.8	48.9
>=30	21.5	26.2	21.6	14.9	18.7
Self-reported diabetes (%)	9.0	15.7	7.7	4.5	5.6
Healthy eating index 2010 (means ± SD)	66.0±10.7	65.5±11.1	66.7±10.3	67.8±9.7	58.7±11.5
Gastric cardia adenocarcinoma	662	155	312	113	82
Gastric non-cardia adenocarcinoma	713	202	344	116	51

¹including less than high school, completed high school, and some post-high school training.

In Table 2, we present risk estimates for GCA and GNCA by category of alcohol consumption overall and for pre-specified subgroups. Overall, we found no evidence of a consistent association between categories of total daily alcohol consumption and risks of GCA or GNCA. We did, however, find a statistically significant lower risk of GNCA among those who drank one or less drinks per day (HR=0.81, 95%CIs=0.67-0.97) compared to non-drinkers, but this association was somewhat attenuated in the later sensitivity analyses by excluding people who were in poor/ fair health status; who were diagnosed with gastric adenocarcinomas within the first 2 years; and who were diagnosed with GNCA without a clear anatomic site in the stomach. Furthermore, we observed no consistent evidence of an association in analyses stratified by self-reported diabetes, or analyses restricted to alcohol drinkers. We found no evidence of a dose-response in either GCA or GNCA, and tests for linear trend across categories of alcohol drinks were not statistically significant (P>0.05).

Table 2.

Hazard ratios and 95% CIs of different numbers of total alcohol drinks /day for gastric adenocarcinoma in the NIH-AARP Diet and Health Study Cohort (1995-2011)

Groups	Variable	Gastric cardia adenocarcinoma					Gastric non-cardia adenocarcinoma
		Total alcohol (drinks/ day)					Total alcohol (drinks/ day)
		never	>0-1	>1-3	>3	Ptrend	never	>0-1	>1-3	>3	Ptrend
All participants1	No. of cases	155	312	113	82		202	344	116	51
	HR (95%CI)	1.00	0.86(0.70-1.05)	0.91(0.70-1.17)	1.06(0.80-1.42)	0.8	1.00	0.81(0.67-0.97)	0.96(0.76-1.23)	0.73(0.53-1.01)	0.2
Exclude poor health2	No. of cases	111	275	98	78		160	313	106	45
	HR (95%CI)	1.00	0.96(0.76-1.20)	0.94(0.71-1.26)	1.25(0.92-1.71)	0.3	1.00	0.85(0.70-1.04)	1.00(0.77-1.29)	0.75(0.53-1.06)	0.3
Exclude first 2-year cases3	No. of cases	134	271	97	72		170	296	94	46
	HR (95%CI)	1.00	0.86(0.69-1.07)	0.89(0.68-1.18)	1.07(0.79-1.46)	0.8	1.00	0.83(0.68-1.01)	0.94(0.72-1.22)	0.80(0.57-1.13)	0.3
Drinkers only4	No. of cases		312	113	82			344	116	51
	HR (95%CI)	-	1.00	1.03 (0.82-1.29)	1.25 (0.96-1.63)	0.1	-	1.00	1.19 (0.96-1.49)	0.88 (0.65-1.20)	1.0
Stratification Model5
Sex
Male	No. of cases	128	280	96	78		124	242	98	47
	HR (95%CI)	1.00	0.90(0.72-1.12)	0.84(0.64-1.11)	1.07(0.79-1.44)	1.0	1.00	0.84(0.67-1.06)	1.01(0.77-1.33)	0.75(0.53-1.06)	0.3
Female	No. of cases	27	32	17	4		78	102	18	4
	HR (95%CI)	1.00	0.60(0.34-1.04)	1.63(0.83-3.20)	1.30(0.43-3.88)	0.3	1.00	0.76(0.56-1.05)	0.88(0.51-1.50)	0.66(0.24-1.83)	0.2
Education level
Less than or high school graduate6	No. of cases	77	114	33	18		93	141	43	17
	HR (95%CI)	1.00	0.83(0.61-1.12)	0.85(0.55-1.30)	0.63(0.36-1.10)	0.1	1.00	0.92(0.70-1.21)	1.20(0.83-1.76)	0.71(0.41-1.20)	0.6
Some college	No. of cases	35	67	31	24		41	84	24	15
	HR (95%CI)	1.00	0.77(0.50-1.17)	1.08(0.65-1.79)	1.04(0.58-1.84)	0.5	1.00	0.78(0.53-1.15)	0.77(0.45-1.30)	0.84(0.46-1.56)	0.5
College graduate	No. of cases	43	131	49	40		68	119	49	19
	HR (95%CI)	1.00	1.01(0.71-1.46)	0.94(0.61-1.45)	1.60(1.02-2.53)	0.1	1.00	0.70(0.51-0.95)	0.88(0.60-1.31)	0.69(0.40-1.17)	0.3
Smoking status
Never smoker	No. of cases	29	57	18	7		82	98	24	6
	HR (95%CI)	1.00	1.03(0.65-1.63)	1.36(0.74-2.49)	1.36(0.58-3.17)	0.3	1.00	0.74(0.55-1.00)	0.91(0.56-1.46)	0.63(0.27-1.45)	0.2
Former smoker	No. of cases	91	190	64	50		82	172	75	33
	HR (95%CI)	1.00	0.85(0.66-1.09)	0.73(0.53-1.02)	1.03(0.72-1.47)	0.6	1.00	0.90(0.69-1.18)	1.12(0.81-1.55)	0.85(0.56-1.29)	0.9
Current smoker	No. of cases	26	42	25	19		28	49	15	12
	HR (95%CI)	1.00	0.71(0.43-1.16)	1.24(0.70-2.18)	0.97(0.53-1.79)	0.5	1.00	0.78(0.49-1.25)	0.72(0.38-1.36)	0.60(0.30-1.21)	0.1

¹HRs come from Cox models adjusted for age, sex, ethnic group, health status, educational level, BMI, smoking status, self-reported diabetes, and healthy eating index 2010.

²sensitivity analysis, excluding participants self-reported as poor/fair health status.

³sensitivity analysis, excluding gastric adenocarcinomas identified in the first 2 years.

⁴sensitivity analysis, excluding non-drinkers.

⁵HRs come from Cox models adjusted for age, sex, ethnic group, health status, educational level, BMI, smoking status, self-reported diabetes, and healthy eating index 2010, except for the stratification variable in each subgroup analysis.

⁶including less than high school, completed high school, and some post-high school training.

As shown in Table 1, alcohol drinkers were more likely than non-drinkers to be male, current or former smokers, and college graduates. Therefore, we explored the possible heterogeneity of effect among strata of sex, smoking status, and educational level. HR estimates were generally similar in stratified analyses (Table 2) and corresponding p-values for interaction tests were not statistically significant (P>0.05). In addition, we investigated the association between alcohol intake and gastric adenocarcinomas by alcohol type (i.e., beer, wine, or liquor), simultaneously adjusting for the other two types of alcohol as categorical variables to account for the total alcohol consumption, but found no evidence for an association between any type of beverage and gastric adenocarcinomas (Table 3).

Table 3.

Hazard ratios and 95% CIs for the associations between beer, wine, or liquor drinks/day with the risk of gastric cardia adenocarcinoma and non-cardia adenocarcinoma in the NIH-AARP Diet and Health Study cohort (1995-2011)

Cancer type	Alcohol consumption	All1			Male2			Female2
		Case No.	HR(95%CI)	Ptrend	Case No.	HR(95%CI)	Ptrend	Case No.	HR(95%CI)	Ptrend
Gastric cardia adenocarcinoma	Beer (drinks/ day)
	never	257	1.00	0.04	213	1.00	0.02	44	1.00	0.3
	>0-1	352	0.91(0.74-1.11)		317	0.83(0.67-1.03)		35	1.47(0.87-2.50)
	>1	53	0.74(0.53-1.04)		52	0.71(0.50-0.99)		1	0.92(0.13-6.76)
	Wine (drinks/ day)
	never	281	1.00	0.9	250	1.00	0.8	31	1.00	0.4
	>0-1	327	0.95(0.78-1.15)		288	0.94(0.76-1.16)		39	1.01(0.57-1.78)
	>1	54	1.16(0.84-1.61)		44	1.05(0.74-1.50)		10	2.10(0.91-4.83)
	Liquor (drinks/ day)
	never	293	1.00	0.2	250	1.00	0.1	43	1.00	0.7
	>0-1	271	0.98(0.80-1.20)		244	1.06(0.85-1.32)		27	0.59(0.33-1.04)
	>1	98	1.27(0.97-1.66)		88	1.3(0.98-1.73)		10	1.39(0.65-2.96)
Gastric non-cardia adenocarcinoma	Beer (drinks/ day)
	never	321	1.00	0.3	182	1.00	0.3	139	1.00	1.0
	>0-1	333	1.03(0.85-1.25)		272	1.03(0.81-1.29)		61	1.03(0.73-1.47)
	>1	59	1.22(0.90-1.65)		57	1.21(0.88-1.68)		2	0.80(0.20-3.24)
	Wine (drinks/ day)
	never	330	1.00	0.2	228	1.00	0.5	102	1.00	0.2
	>0-1	343	0.88(0.73-1.06)		251	0.93(0.75-1.16)		92	0.80(0.57-1.14)
	>1	40	0.90(0.64-1.28)		32	0.96(0.65-1.43)		8	0.79(0.37-1.68)
	Liquor (drinks/ day)
	never	364	1.00	0.06	243	1.00	0.06	121	1.00	0.7
	>0-1	285	0.89(0.74-1.08)		215	0.90(0.72-1.12)		70	0.89(0.63-1.27)
	>1	64	0.80(0.60-1.07)		53	0.76(0.55-1.06)		11	1.04(0.54-2.01)

¹HRs come from Cox models adjusted for age, sex, ethnic group, health status, educational level, BMI, smoking status, self-reported diabetes, and healthy eating index 2010, and other types of alcohol drinking.

²HRs come from Cox models adjusted for age, ethnic group, health status, educational level, BMI, smoking status, self-reported diabetes, and healthy eating index 2010 and other types of alcohol drinking.

The restricted cubic splines modeling ethanol intake (continuous variable, grams/ day) and risk of gastric adenocarcinomas are shown in Figure 1. Consistent with findings in the categorical analyses, we observed no associations between ethanol intake and GCA or GNCA and found no evidence of a dose-response association for either endpoint.

Figure. Cubic spline of ethanol intake (grams/ day) and gastric cardia adenocarcinoma and gastric non-cardia adenocarcinoma in the NIH-AARP Diet and Health Study cohort (1995/1996-2011).

A) Total ethanol intake per day and gastric cardia adenocarcinoma in total population

B) Total ethanol intake per day and gastric non-cardia adenocarcinoma in total population

C) Total ethanol intake per day and gastric cardia adenocarcinoma in drinkers

D) Total ethanol intake per day and gastric non-cardia adenocarcinoma in drinkers

Note: cubic spline analysis, data were trimming at the highest 1% of alcohol drinks, and adjusted for age at baseline, sex, ethnic group, health status, education level, BMI, smoking status, self-reported diabetes, and healthy eating index 2010.

Discussion

In this large prospective cohort, we found no evidence of an association between higher alcohol consumption and gastric cardia adenocarcinoma or gastric non-cardia adenocarcinomas, and similar findings were observed for beer, wine, and liquor. There was no evidence for a dose-response between alcohol intake and GCA or GNCA. Furthermore, we observed no consistent evidence of an association in analyses stratified by sex, educational level, or smoking status. We did, however, find that those who consumed one or fewer drinks per day had a lower risk of GNCA than non-drinkers (HR=0.81, 95%CIs:0.67-0.97), but this association was attenuated in our sensitivity analyses and should be interpreted with caution.

The lack of statistically significant positive associations in our study is at odds with several recent reports that conclude that heavier alcohol consumption increases risk of gastric cancer. Study design and geographic differences may contribute to discrepancies between these prior reports and our study. In 2017, the “Stomach cancer pooling Project” pooled results from 23 case-control studies and found a higher risk of gastric cancer for more than 50 grams of ethanol intake (approximately 4 alcoholic drinks) per day ¹⁷; however, case control studies are more susceptible to recall and selection bias than prospective studies. The 2016 WCRF CUP report summarized results from 23 prospective cohort studies and found that consuming ≥45 grams of ethanol per day increased gastric cancer risk ⁵; however, in an analysis stratified by continent, a 10 gram increase in ethanol intake per day was positively associated with gastric cancer for studies conducted in Asia (accounting for 70% of the total cases in CUP report, RR=1.03, 95%CI=1.01-1.04), but not for those conducted in North America or Europe ⁵.

It seems possible that differences between studies conducted in different populations may have an underlying biological basis. For example, the SNP rs671 in ALDH2 is a common variant in East Asians, which is associated with the “alcohol flushing response,” and is caused by rapid accumulation of the carcinogen acetaldehyde. This variant is common in East Asians but rare in Europeans, Africans, and Americans ^{18, 19}. In addition, different population characteristics may also contribute to these discrepancies. For example, the NIH-AARP Diet and Health Study members tend to be light or moderate drinkers, approximately 95% drank less than 55 grams of ethanol per day and 75% drank less than 11.5 grams of ethanol per day (data not shown in table), which may limit our ability to evaluate the impact of heavy drinking.

Few studies that have been conducted in North American populations, and they have had a small number of cases ^{5, 17, 20}, which limited their ability to conduct analyses stratified by cancer-type, and explore associations with beverage-type. However, similar to our study, a meta-analysis including 3 cohort studies from North America found no association between alcohol intake and gastric cancer ²⁰. With more than 1300 cases, a major advantage of our study was the ability to evaluate associations by location within the stomach, because of the known differences in risk for tumors in the cardia compared to noncardia stomach. We also analyzed alcoholic beverage-type and potential effect modifiers such as sex and smoking. In the sublocation analysis, we observed no association of alcohol intake with GCA and GNCA. This finding was consistent with the subtype analysis in the 2016 WCRF CUP report as well as several other meta-analyses ^{5, 20, 21}. Previous studies also found null associations between alcohol intake and other upper gastrointestinal adenocarcinomas, i.e., esophageal adenocarcinoma. The Barrett’s and Esophageal Adenocarcinoma Consortium (BEACON) found higher alcohol consumption was not associated with increased risk of esophageal adenocarcinoma (EA) or adjacent tumors of the esophagogastric junction(EGJA)²², even at 7 drinks/day. This suggests that adenocarcinomas of the upper gastrointestinal tract may not be influenced by alcohol beverage intake.

We found a lower risk of GNCA among people who consumed one or fewer alcoholic drinks per day. Because our data are observational, this should be interpreted with caution and could be due to unmeasured confounding. A previous meta-analysis also found a lower risk for GNCA among people who consumed low amounts of alcohol (i.e., 10 grams ethanol per day)²¹, and a second meta-analysis of cohort studies found a lower risk of gastric cancer among female light drinkers ²³. A Swedish study found a decreased risk of gastric cancer in alcohol use disorders (AUD) compared to those without AUD, and the effect was strongest for tumors located in the corpus²⁴. In addition, somewhat similar inverse associations were found in BEACON study between moderate alcohol intake and EA and EGJA ²². Even for esophageal squamous cell carcinoma, a cancer with strong evidence for an adverse association with alcohol, a J-shaped rather than linear dose-response relationship was found in many studies^{25, 26}. Previous mechanistic studies suggest that antioxidants found in alcoholic beverages may favorably impact cancer risk ²⁷; alcohol beverage may also favor suppression or eradiation of H. pylori^{28, 29}, providing some biological plausibility for a potential protective effect. . However, we caution that the apparent inverse association observed with modest alcohol consumption and GNCA should be evaluated in future prospective studies

This study has several strengths, including its prospective design, extensive information on potential confounders, and large sample size, which allowed us to conduct analyses stratified by major risk factors for different gastric adenocarcinoma subtypes. Furthermore, we were able to consider total alcoholic beverage intake and type of beverage. This study also has a number of limitations, including a lack of data on H. pylori status, which is a well-known risk factor for gastric cancer^{5, 30}. However, we did analyze sublocation separately, and since nearly all GNCA cases but few GCA cases are likely to be H. pylori positive, the results would likely be similar. Furthermore, a previous pooled analysis found that the association between alcohol intake and gastric adenocarcinomas was independent of H. pylori infection ¹⁷. Our study also lacked data on duration, lifetime drinking history, binge drinking or weekend drinking, which may result in exposure misclassification and bias the results. Although, a previous study found that daily alcohol intake, not duration of alcohol use, was associated with cancer ¹, detailed data on lifetime alcohol exposure and patterns of alcohol use would allow for better understanding of this potentially important risk factor. Finally, the NIH-AARP Diet and Health Study population is well-educated and predominantly white, which may limit the generalizability of our results to the other populations.

In conclusion, we found that alcohol consumption was not associated with increased risk of gastric cardia adenocarcinoma or gastric non-cardia adenocarcinoma in this large US prospective cohort. The lower risk of GNCA observed among light drinkers warrants further investigation and should be interpreted with caution given the potential for residual confounding.

Novelty and Impact.

Recent summaries showed a positive association between higher alcohol intake and gastric cancer risk, but most cases were from Asia. We provide new robust evidence from North Americans for no association between higher alcohol consumption and gastric adenocarcinomas in a large US cohort, when examined as total alcohol intake, individual beverage type, or by sublocation of the stomach. This study raises questions of genetic and ethnic difference when assessing the associations in different populations.

Abbreviations:

ADH

acetaldehyde dehydrogenase

ALDH

aldehyde dehydrogenase

BEACON

Barrett’s and Esophageal Adenocarcinoma Consortium

BMI

body mass index

CI

confidence intervals

CUP

Continuous Update Project

EA

esophageal adenocarcinoma

EGJA

adjacent tumors of the esophagogastric junction

GCA

gastric cardia adenocarcinoma

GNCA

gastric non-cardia adenocarcinoma

HEI-2010

Healthy Eating Index 2010

HR

hazard ratios

IARC

International Agency for Research on Cancer

WCRF

World Cancer Research Fund International