Alcohol Intake, Beverage Choice, and Cancer: A Cohort Study in a Large Kaiser Permanente PopulationArthur L Klatsky, MD; Yan Li, MD, PhD; H Nicole Tran, MD, PhD; David Baer, MD, PhD; Perm J 2015 Spring; 19(2):28-34 [Full Citation] https://doi.org/10.7812/TPP/14-189AbstractContext: Heavy intake of alcoholic beverages is associated with an increased risk of developing several types of cancers at specific body sites. However, evidence is conflicting regarding alcohol-associated cancers in other sites of the body as well as the role played by choice of wine, liquor, or beer. IntroductionIn observational studies1-6 alcohol intake has been consistently associated with increased risk of cancers of the oral cavity, pharynx, larynx, esophagus, liver, large intestine, and female breast. Reports conflict about associations between alcohol consumption and risk of malignancies of the stomach, pancreas, lung, bladder, prostate, endometrium, ovary, cervix, and skin. Limited data suggest possible inverse (ie, protective) associations between alcohol consumption and the risk of renal cell carcinoma7,8 and several types of hematologic malignancy.9-11 It seems clear that a comprehensive study of the role of drinking in cancer risk should examine all cancer plus individual cancer types. Although cancer risk is clearest for heavier intake of alcohol, some reports1-5 suggest that light-moderate (light-to-moderate) drinking is also linked to increased cancer risk. Among specific cancer types, increased risk of light-moderate drinkers has most consistently been found in studies of female breast cancer.12-14 The associations of lighter alcohol intake with cancer are relatively weak, and there are plausible confounders. Thus, the role of light-moderate drinking in cancer risk remains unresolved. Because drinking and smoking are associated habits in many populations,15,16 tobacco use is a potential confounder of some alcohol-cancer associations. Incomplete control for smoking could produce spurious alcohol associations with tobacco-related cancer types.17 Underreporting of heavier intake is another possible source of spurious linkage of light-moderate drinking with increased risk of cancer.18,19 Other potential confounders include chronic infections, diet, adiposity, exercise, air pollution, radiation exposure, and various chemicals.17,19 However, most of these probably do not have sufficiently strong associations with drinking to explain the observed alcohol-cancer associations. Independent roles for nonalcohol ingredients in alcoholic beverages have long been of interest. Early examples include reports of promotion of esophageal cancer by congeners in calvados (apple brandy)20 and a possible role for beer in the risk of rectal cancer.21 The latter was later considered confounded.22 In recent years, interest about this aspect has focused on hypothetical protective effects of phenolic compounds in red wine, especially resveratrol.23,24 Hoping to cast light on several of these uncertainties, we have performed a cohort analysis in a large, multiethnic Northern California population. We report here data about risks for all cancer and specific cancer types associated with light-moderate and heavier drinking. The analysis is controlled for smoking and several other confounders. We also present data about risk of cancer among a strata of persons who drink preponderantly wine, liquor, or beer. Materials and MethodsStudy Population and DataThe study protocols were approved by the Kaiser Permanente (KP) Northern California institutional review board. Baseline data were obtained from questionnaires given at health examinations from 1978 through 1985. The examinees were a multiethnic cohort of 124,193 men and women (mean baseline age = 41.0 years) who had no history of cancer and who were members of a comprehensive prepaid health plan in the San Francisco Bay Area. Usually taken as a voluntary routine health appraisal, the examination25 included health measurements, self-classified ethnicity, and queries about sociodemographic status, habits, medical history, and symptoms. Data about alcohol consumption were supplied during the examination on a special check-sheet questionnaire. The study cohort comprised 79.8% of all examinees; the remainder included persons who took the examination during absences of a special research clerk and persons who declined, largely because of lack of fluency in English. Lifelong abstainers were defined as persons who reported drinking no alcohol during the past year and "never or almost never" before the past year. Ex-drinkers were nondrinkers during the previous year who indicated prior alcohol drinking. Current drinkers described usual drinking as less than one drink per month ("special occasions only"), more than one drink per month but less than one drink per day, or as daily number of drinks, one to two, three to five, six to eight, and nine or more. Drinkers received separate questions about the number of days per week that they drank wine, liquor, or beer. Wine, liquor, or beer "preponderance" was defined among persons reporting more than one drink per month as exclusive intake of one beverage type or drinking the beverage type more days per week than either of the other two; preponderance was "none" for persons reporting more than one type with equal frequency. Subjects with CancerOccurrence of cancer was ascertained from the KP Cancer Registry, which covers all subscribers and contributes to the local Surveillance, Epidemiology, and End Results (SEER) program. Codes were translated into International Classification of Diseases, Ninth Revision (ICD-9) codes and the composite incidence of codes 140 to 209 (N = 18,637) was studied as "all cancer." Endpoints studied included any cancer and multiple specific cancer types. We report here data about the 15 types of cancer with 150 or more incident cases plus the following 3 composites: 1) Upper airway digestive (UAD) cancers included 552 subjects with 1 or more of codes 140 (lip), 141 (tongue), 143 (gum), 144 to 145 (mouth), 146 (oropharynx), 148 (hypopharynx), 149 (other oral cavity), 150 (esophagus), and 161 (larynx). 2) Hematologic malignancies included 1639 subjects with 1 or more of codes 201 (Hodgkin disease), 202 (non-Hodgkin lymphoma), 203 (multiple myeloma), 204 (lymphocytic leukemia), 205 (myelocytic leukemia), and 206 to 208 (other leukemia). 3) "Alcohol-related" malignancies included 9246 subjects with 1 or more of the following types: UAD, codes 153-154 (colorectal), 155 (liver), 162 (lung), 172 (melanoma), and 174 (breast). Table 1 presents selected distributions in the study population and subjects with cancer. Analytic MethodsSubjects were followed until December 31, 2012, or until a diagnosis of cancer, their death, or other termination of Health Plan membership. The mean follow-up was 17.8 years, yielding an estimated 2,216,631 person-years of follow-up. Multivariate models used the Cox proportional hazards model. Alcohol was studied categorically. In most models, lifelong abstainers were the referent with 4 other categories: ex-drinkers, less than 1 drink per day (2 categories combined), 1 to 2 drinks per day, and 3 or more drinks per day (3 categories combined). Models for all persons and for subgroup strata included as covariates age (continuous), race or ethnicity (white referent, black, Asian, Hispanic, other), education (no college referent, some college, college graduate), body mass index (< 25 kg/m2 referent, 25-29 kg/m2, ≥ 30 kg/m2), marital status (married referent, never married, formerly married), and cigarette smoking (never smoked referent, ex-smoker, current < 1 pack per day, current > 1 pack per day). Beverage choice was studied among persons reporting more than one drink per day. These drinkers were stratified into persons defined as reporting preponderance as wine, liquor, beer, or none. Models limited to each of these preponderance beverage groups ascertained the role of total alcohol in each stratum. These models used less than one drink per day in each preponderance as referent, plus one to two and three or more drinks per day as the other categories. Among light-moderate drinkers, we studied strata of persons inferentially suspected or not suspected of underreporting alcohol intake. These strata were derived from subjects with at least two computer-stored examinations (index measurement and at least one other examination before or after). Suspected underreporters either reported heavier intake at another time or had an alcohol-related diagnosis (death certificate, hospitalization, or outpatient) at some time. A more detailed description of this method has been published.9 We report hazard ratios (HRs), 95% confidence intervals (CIs), and associated p values. The term significant is used in this article for p values < 0.05. We refer to persons reporting 3 or more drinks per day as "heavy" drinkers and to those reporting less than 1 or 1 to 2 drinks per day as "light-moderate" drinkers. ResultsAll CancerThe risk of any cancer was increased by 10% to 20% in daily drinkers (Table 2). The progressive nature of this relationship was partially masked by rounding; for example, for all persons the HR was 1.10 at less than 1 drink per day, 1.15 at 1 to 2 drinks per day, and 1.23 at 3 or more drinks per day. The magnitude of the increased risk was generally similar for the sexes, race/ethnic groups, and smoking strata but did appear to weaken after 20 years. Ex-drinkers also had a higher risk than lifelong abstainers, but not for cancers diagnosed after 20 years. Specific CancersTable 3 presents data for specific cancer types and several composites. Heavy drinkers had a significantly increased risk of the following cancer types: UAD, colorectal, lung, melanoma, and breast. Light-moderate drinkers had an increased risk of all of these except lung cancer. The following types of cancer were unrelated to baseline alcohol intake: stomach, pancreas, liver, kidney, brain, thyroid, prostate, bladder, cervix, ovary, uterine body, and hematologic system. Ex-drinkers were at increased risk of UAD, liver, breast, and brain cancers. Table 3 includes data about the relationships of total alcohol intake to the alcohol-related composite; as expected, the HRs were larger than those for all cancer and were progressive with increasing alcohol intake. Among UAD subgroups, heavier drinkers had an increased risk of each type. For example, for esophageal cancer the HR was 2.2 (CI = 1.0 to 4.9, p < 0.05), for laryngeal cancer it was 1.9 (CI = 1.2 to 3.0, p < 0.01) and for the remaining UAD cancers it was 2.6 (CI = 1.4 to 4.5, p < 0.001). The lower risk of heavy drinkers for hematologic malignancies was of borderline significance (p = 0.06). Among hematologic subtypes, all had inverse alcohol relationships, but p was < 0.05 only for lymphatic leukemia (HR = 0.5; CI = 0.3 to 0.98). Alcohol-Associated Risk in Never SmokersThe data in Table 2 show generally similar alcohol-associated cancer risk among smoking strata, including never smokers. For the alcohol-related composite, the HR among never smokers reporting 3 or more drinks per day was 1.3 (CI = 1.1 to 1.6, p < 0.001). Three types of cancer had significant relationships among never smokers reporting 3 or more drinks per day, with these HRs: UAD = 1.3 (CI = 1.1 to 1.6), liver = 4.2 (CI = 2.2 to 2.8), and melanoma = 1.8 (CI = 1.2 to 2.8), all with p < 0.001. Covariate AssociationsTable 4 presents selected covariate data for all cancer and for the alcohol-related composite. For all cancer, age, male sex, black race, obesity, and smoking were associated with increased risk. For the composite, women had higher risk, driven by the large number with breast cancer, and black persons had slightly lower risk than whites, driven by the virtually total absence of melanoma. Smoking was slightly more strongly related to the composite, driven by the large number of lung cancer cases. Beverage Preponderance StrataWith drinkers of less than 1 per day as referent, there was a significantly increased risk among heavy drinkers in the groups with a beer preponderance and with no beverage preponderance (Table 5); these findings were similar for men and women (data not shown). Analyses in the alcohol-associated composite showed that heavy drinkers in all preponderance groups evidenced significant associations, with a slightly larger HR for the beer preponderance group; these HRs were similar for men and women (data not shown) except for liquor preponderance (HR for men = 1.7, p < 0.001; for women, HR = 1.1, not significant). Relationships of heavy drinkers in preponderance groups differed for the individual cancers except that all groups showed significant associations with UAD; only beer was significantly associated with breast cancer; only wine was significantly related to melanoma. The no preponderance group had a significantly higher risk of liver cancer. Suspected Underreporter StrataAmong light-moderate drinkers, increased alcohol-associated cancer risk was concentrated in the stratum suspected of underreporting. For example, among persons reporting 1 to 2 drinks per day on the index examination but considered likely underreporters, the HR for any cancer was 1.4 (CI = 1.3 to 1.7, p < 0.001), whereas among those considered unlikely to be underreporters, it was 1.1 (CI = 0.9 to 1.2). More details have been published.19 DiscussionDisparate Alcohol-Cancer RelationshipsOur analyses confirmed the presence of increased risk in drinkers for most of the cancer types consistently related to alcohol in prior reports. The exception was liver cancer, possibly because there were relatively small numbers of this type. Because never smokers, ex-drinkers, and heavy drinkers with a beer preponderance had an increased risk of liver cancer, we left this type of cancer in the alcohol-related composite. We also found that drinkers had an increased risk of lung cancer and of melanoma. We found no support for an alcohol association with stomach cancer, pancreatic cancer, or any type of genitourinary cancer in either sex. A slight inverse relationship with hematologic malignancies was of borderline significance. A weakening of this inverse association since our previous report9 may be caused by a reduction in alcohol intake with increasing age. These disparities in alcohol-cancer associations are masked in the models for all cancer because the positive associations dominate. Except for lung cancer, the increased cancer risk of baseline drinkers was present in both light-moderate drinkers and heavy drinkers, with progressive associations for all. The strongest alcohol-cancer associations among light-moderate drinkers were those with melanoma and breast cancer. Hypothetical MechanismsAlthough ethyl alcohol is not a mutagenic carcinogen, its first metabolite acetaldehyde probably is.26,27 Hypothetically plausible mechanisms for alcohol-cancer associations vary with the sites. Genetic factors influencing alcohol metabolism may modulate alcohol-associated risk.5,26,27 For UAD cancers it has been proposed that alcohol may operate as a promoter or facilitator of smoking-associated risk.5,28 For UAD cancers, liver cancer, and breast cancer, a role has been proposed for acetaldehyde, which might damage DNA and thus act as a carcinogen.26,27 Chronic liver disease, usually cirrhosis, is intermediary between heavy alcohol intake and hepatocellular cancer, analogous to the situation for chronic viral hepatitis and cancer.26,27 There is evidence that an alcohol-estrogen interaction is involved in the breast cancer association.12-14 Estrogenic hormones are an established risk factor for breast cancer, and several analyses, including ours,14 show the increased breast cancer risk concentrated in women with estrogen-sensitive tumors. It has been suggested that relative folate deficiency may be involved in the relationships between alcohol and colorectal cancer, breast cancer, and others.29 A more detailed discussion of these and other potential mechanisms is beyond the scope of this report. Lung CancerThe powerful relationship of smoking to lung cancer complicates the study of alcohol relations.17,30,31 Published details of a KP analysis17 show that the alcohol association primarily involved adenocarcinoma in heavy-drinking women. In that report the adenocarcinoma HR for 3 or more drinks per day among women was 2.1 (CI = 1.4 to 3.1, p = 0.0002), vs men with HR = 1.0 (CI = 0.7 to 1.5). The HR for squamous cell carcinoma among women reporting 3 or more drinks per day was 1.2 (CI = 0.7 to 2.0). The smoking-lung cancer association was stronger for squamous cell cancer than for adenocarcinoma, a fact lessening the likelihood that residual confounding by smoking was involved. We cannot explain the cell type specificity. MelanomaAlthough there are previous reports of a possible increased risk of melanoma in drinkers,32 the association of drinking with an increased risk of melanoma in this analysis is noteworthy for its strength in both heavy and light drinkers. We have presented data showing that the alcohol-associated risk is similar for men and women.33 In our data34 and several other reports,35 smoking is inversely related to melanoma, so residual confounding by smoking is not a plausible explanation for this finding. A noteworthy feature in our melanoma analyses is that the alcohol association is stronger for noninvasive than invasive disease,34 suggesting an earlier diagnosis in drinkers. Earlier diagnosis could, in turn, be related to higher socioeconomic status and more recreational sun exposure. Among light-moderate drinkers wine preponderance is related to increased melanoma risk, another possible indicator of higher socioeconomic status.36 We hope that further work by others will help to sort out the alcohol/smoking/melanoma puzzle. Beverage ChoiceAlthough the beverage type data suggest a slightly stronger association with beer than with wine or liquor, the differences were not large and the CIs overlapped. In this population, female heavy drinkers of beer have unfavorable lifestyle traits,36 which may account for the slightly increased breast cancer risk in these women. We interpret our data as suggesting little, if any, disparity related to specific beverage types. LimitationsAs with all reported analyses of alcohol intake and cancer, this study is observational, leaving uncontrolled confounders (eg, dietary habits, cigar and pipe smoking, environmental smoke exposure, occupational factors, or exercise) as potential factors in the findings. Despite known relative stability of drinking in this population,37 determination of alcohol habits only at baseline is another limitation. Changes in habits probably accounted for weakening alcohol associations with the passage of time. The beverage choice data are limited by the fact that the primary queries about wine, liquor, and beer inquired about number of days per week rather than the usual number of drinks. Thus, we do not know the actual proportion of each person's intake represented. We do know from more detailed data in a 1984 to 1985 subset of examinees that there was good correlation of these values to the total number of drinks per week of the beverage type, with an average of 80% to 90% of alcoholic beverages consumed as the preponderant type.36 Public Health ConsiderationShould these findings influence medical advice from health care professionals to patients? Increased cancer risk is clearly one of multiple medical reasons to avoid heavy drinking. The more important issue is the possible cancer risk of light-moderate alcohol intake. This issue remains clouded by uncertainty about whether the findings are confounded by underreporting and other traits. However, at present, a possible increased cancer risk at moderate intake should enter into individual estimations of the overall risk-benefit equation for alcohol drinking, especially for young persons. For most persons older than age 50 years, the overall benefits of lighter drinking, especially the reduced risk of atherothrombotic disease,38,39 outweigh possible cancer risk. This is best evidenced by lower total mortality risk among middle-aged and older light-moderate drinkers.39 For younger persons, especially young women with no coronary disease risk factors, the breast cancer data suggest the wisdom of limiting alcohol intake to very modest amounts. As always in medical practice, advice needs to be individualized.40 Disclosure StatementAll authors have participated actively in the execution of the study and/or preparation of the manuscript. The authors(s) have no conflicts of interest to disclose. AcknowledgmentThe research was performed at the Kaiser Permanente Northern California Division of Research with support by grants from the Kaiser Foundation Community Budget Program to Dr Li as Principal Investigator. Data collection in 1978 to 1985 was supported by a grant to Dr Klatsky from the Alcoholic Beverage Medical Research Foundation of Baltimore, MD. We are grateful to Cynthia Landy for assistance with data collection in 1978 to 1985. Kathleen Louden, ELS, of Louden Health Communications provided editorial assistance. References 1. Boffetta P, Hashibe M. Alcohol and cancer. Lancet Oncol 2006 Feb;7(2):149-56. DOI: https://doi.org/10.1016/S1470-2045(06)70577-0. |
ETOC
Click here to join the eTOC list or text ETOC to 22828. You will receive an email notice with the Table of Contents of The Permanente Journal.
CIRCULATION
2 million page views of TPJ articles in PubMed from a broad international readership.
Indexing
Indexed in MEDLINE, PubMed Central, EMBASE, EBSCO Academic Search Complete, and CrossRef.