Alcohol Use is Associated With Hepatic Steatosis Among Persons With Presumed Nonalcoholic Fatty Liver Disease

Journal Pre-proof Alcohol Use is Associated with Hepatic Steatosis Among Persons with Presumed Non-alcoholic Fatty Liver Disease Michelle T. Long, MD, MSc, Joseph M. Massaro, PhD, Udo Hoffmann, MD, Emelia J. Benjamin, MD, ScM, Timothy S. Naimi, MD, MPH

PII: DOI: Reference:

S1542-3565(19)31309-6 https://doi.org/10.1016/j.cgh.2019.11.022 YJCGH 56862

To appear in: Clinical Gastroenterology and Hepatology Accepted Date: 8 November 2019 Please cite this article as: Long MT, Massaro JM, Hoffmann U, Benjamin EJ, Naimi TS, Alcohol Use is Associated with Hepatic Steatosis Among Persons with Presumed Non-alcoholic Fatty Liver Disease Clinical Gastroenterology and Hepatology (2019), doi: https://doi.org/10.1016/j.cgh.2019.11.022. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 by the AGA Institute

Long et al. Alcohol use in NAFLD Title: Alcohol Use is Associated with Hepatic Steatosis Among Persons with Presumed Non-alcoholic Fatty Liver Disease

Short title: Alcohol use in NAFLD Authors: Michelle T. Long, MD, MSc1, Joseph M. Massaro, PhD2, Udo Hoffmann, MD3, Emelia J. Benjamin, MD, ScM4,5, Timothy S. Naimi, MD, MPH6

1

Section of Gastroenterology, Boston Medical Center, Boston University School of Medicine, Boston, MA

2

Department of Mathematics and Statistics, Boston University, Boston, MA, United States

3

Department of Radiology, Massachusetts General Hospital, Boston, MA.

4

National Heart, Lung, and Blood Institute’s and Boston University’s Framingham Heart Study, Framingham, MA 5

Division of Cardiology, Boston Medical Center, Boston University School of Medicine, Boston, MA

6

Section of General Internal Medicine, Boston Medical Center, Boston University School of Medicine, Boston, MA

Grant Support: The Framingham Heart Study is supported by HHSN268201500001I; N01-HC 25195; 75N92019D00031. Dr. Long is supported in part by the National Institute of Diabetes and Digestive and Kidney Diseases K23 DK113252, the Boston University School of Medicine Department of Medicine Career Investment Award, the Doris Duke Foundation, and the Boston University Clinical Translational Science Institute UL1 TR001430. Dr. Benjamin is supported by R01HL128914; 2R01 HL092577; 1R01 HL141434 01A1, 2U54HL120163 and American Heart Association, 18SFRN34110082; Robert Wood Johnson 74624.

Abbreviations: ALD, alcoholic liver disease; CT, computed tomography; CI, confidence interval; NAFLD, nonalcoholic fatty liver disease; SD; standard deviation

Page 1 of 24

Long et al. Alcohol use in NAFLD Correspondence: Michelle T. Long, MD Boston University School of Medicine, Section of Gastroenterology 85 East Concord Street 7th Floor, Boston MA 02118 Email: [email protected] Tel: (617) 638-8392 Fax: (617) 638-6529

Disclosures: The other authors have no conflicts to report.

Author contributions Study concept and design: MTL, EJB, TSN; acquisition of data: JMM, UH; analysis and interpretation of data: MTL, JMM, EJB, TSN; drafting of the manuscript: MTL; critical revision of the manuscript for important intellectual content: MTL, JMM, UH, EJB, TSN. The authors had access to all study data, take responsibility for the accuracy of the analysis, and had authority over manuscript preparation and the decision to submit the manuscript for publication. All authors approved the manuscript.

Word count: 3920

Page 2 of 24

Long et al. Alcohol use in NAFLD Abstract

Background & Aims: Many individuals presumed to have non-alcoholic fatty liver disease (NAFLD) consume moderate amounts of alcohol. Little known about patterns of alcohol use in patients with NAFLD or how drinking behaviors affect liver fat. Methods: We conducted a cross-sectional study of 2475 participants of the Framingham Heart Study with hepatic steatosis, determined by computed tomography. We performed multivariableadjusted logistic regression models to evaluate the association between alcohol drinking patterns and hepatic steatosis. Models were adjusted for sociodemographic factors, diet, and the components of the metabolic syndrome. We excluded heavy alcohol users, defined as women who consume more than 14 alcohol drinks/week and men who consume more than 21 alcohol drinks/week. Results: In our sample (mean age, 49.8±10.2; 50.3% women), the prevalence of hepatic steatosis was 17.5%. The total number of alcohol drinks/week and the maximum drinks consumed per drinking day were each associated with hepatic steatosis (adjusted-odds ratio [aOR] 1.15; 95% CI, 1.02–1.29 and aOR 1.15; 95% CI 1.02–1.30). Binge drinking occurred in 25.4% of individuals with presumed NAFLD and was associated with an increased odds of hepatic steatosis (aOR, 1.45; 95% CI, 1.06–1.98) among alcohol users. In a beverage-specific analysis, alcohol use patterns were associated with hepatic steatosis among beer drinkers, but not among wine drinkers. Conclusions: In a cross-sectional study of participants of the Framingham Heart Study with hepatic steatosis, we observed an association between alcohol use and liver fat, even after excluding heavy alcohol users from our analysis. Alcohol use therefore appears to be a risk factor for NAFLD. Prospective studies are needed to validate these findings and determine if alcohol use should be a focus for research, prevention, and treatment of presumed NAFLD. Keywords: epidemiology, CT, social, lifestyle factor

Page 3 of 24

Long et al. Alcohol use in NAFLD Introduction Non-alcoholic fatty liver disease (NAFLD), which is characterized by the presence of hepatic steatosis in the absence of secondary causes, is the most common cause of chronic liver disease worldwide.1 Pathologically, NAFLD and alcoholic liver disease (ALD) appear identical and the only distinguishing feature is the presence or absence of significant alcohol exposure. The association of alcohol consumption and NAFLD is controversial, as the findings in the literature are mixed. Whereas studies demonstrate that excessive alcohol consumption is a risk factor for fatty liver, steatohepatitis, and cirrhosis, some studies have reported a beneficial effect of ‘moderate’ alcohol consumption in NAFLD2-6; while others, have reported either no protection or detrimental effects on liver fat.7-13 The extent to which these discrepant results may be attributed to different alcohol consumption patterns in different study populations, even among persons with similar average daily consumption, is unknown. Also, those with chronic diseases may be more likely to abstain from alcohol so the presence of co-existing metabolic disease may not only influence the pattern of alcohol use but also confound relationships with moderate alcohol use. Recent evidence suggests that the pattern of alcohol consumption may be an important predictor of the health effects of alcohol.14 However, most of the studies evaluating alcohol use among individuals with NAFLD focus instead on average daily or weekly alcohol consumption, which obscures differences in alcohol use behavior. Currently the American Association for the Study of Liver Disease guidelines lack details regarding recommendations regarding alcohol consumption among those with NAFLD. Thus, in this study we examined the association between different alcohol drinking patterns and liver fat defined on multi-detector computed tomography (MDCT), among participants of the Framingham Heart Study, after adjusting for covariates. Since our focus was on participants with presumed NAFLD, we excluded heavy alcohol users. We also determined if associations between alcohol use pattern and liver fat differ by type of alcohol consumed.

Page 4 of 24

Long et al. Alcohol use in NAFLD

Methods Study sample We conducted a cross-sectional, secondary analysis in the Framingham Heart Study (FHS).15, 16 The study sample was drawn from 3,180 participants in the Offspring and Third Generation Cohorts who participated in the MDCT sub-study to evaluate adipose tissue deposits between 2002 and 2005. For the Offspring Cohort participants, alcohol use and most covariates were measured at the closest examination period to the MDCT examinations (seventh examination; 1998–2001), unless noted. We drew income data from the third examination visit when the Offspring participants were in prime working age and education data from the second examination visit. For the Third Generation Cohort participants, alcohol use behavior and all covariates were assessed at the first examination visit conducted contemporaneously with the MDCT sub-study. We excluded individuals with missing alcohol use data or significant alcohol use, defined as >14 drinks per week for women and >21 drinks per week for men (143 participants)17, missing covariate data including smoking status, metabolic syndrome components, and physical activity (157 participants), missing education data (116 participants), or missing income data (287 participants). Our final study sample was 2,475 participants. For the analysis among current drinkers, we additionally excluded 473 former or never drinkers, resulting in 2002 participants. For the beverage-specific analysis, we excluded non-drinking participants or participants with mixed beverage type, resulting in 1633 participants. The MDCT sub-study was approved by the institutional review boards of Boston University School of Medicine and the Massachusetts General Hospital and all participants provided written informed consent.

Alcohol use behaviors

Page 5 of 24

Long et al. Alcohol use in NAFLD We defined a number of exposures to measure various alcohol use behaviors including: pastweek number of alcoholic drinks consumed, risky weekly drinking (≥8 drinks/week for women and ≥15 drinks/week for men), frequency of alcohol use (# drinking days), usual quantity consumed per drinking day, usual consumption above US Dietary Guidelines suggested limits, past-month maximum drinks consumed in 24 hours, and binge drinking (≥4 drinks/day for women and ≥5 drinks/day for men) as described in the online supplement.

Computed Tomography Liver Fat Measurement The primary dependent variable was hepatic steatosis, defined on MDCT scan, which we described in detail previously and in the online supplement.18-20 Liver attenuation in Hounsfield Units was measured in three areas from the liver and from an external phantom and averaged to compute the liver phantom ratio (LPR). We defined hepatic steatosis as a LPR ≤0.33, which sensitive and specific for hepatic steatosis.18 As a secondary dependent variable we considered the LPR as a continuous variable and modeled more liver fat as a decreasing LPR (see online supplement).

Covariates Primary covariates include age, sex, cohort (Offspring vs Third Generation), current smoking, physical activity, diet, education, and income. Secondary covariates included the components of the metabolic syndrome. Definitions of primary and secondary covariates are available in the online supplement.

Statistical analysis We conducted the analyses with SAS 9.4 (SAS Institute, Cary, NC) using a two-sided significance level of alpha =0.05. We performed multiple imputations to impute missing values for the Mediterranean-Style Diet Score for 228 participants.21 We used logistic regression

Page 6 of 24

Long et al. Alcohol use in NAFLD models to obtain multivariable-adjusted odds ratios (aOR) for the association between various drinking patterns and hepatic steatosis (defined by the dichotomous LPR ≤0.33). We modeled the ORs per standard deviation (SD) increase in the continuous exposure. In model 1, we accounted for the primary covariates, including age, sex, cohort, smoking, diet, physical activity, education, and income. We evaluated for effect modification by age and sex. In Model 2, we added adjustment for the components of the metabolic syndrome (waist circumference, low HDL-c, high triglycerides, elevated blood pressure, or impaired fasting glucose) to Model 1. We repeated the primary analysis after excluding non-drinkers, recognizing that non-drinkers represent a heterogeneous group of former and never drinkers. We conducted a sensitivity analysis by limiting the sample to Third Generation Cohort participants. Finally, we performed a number of pre-specified secondary analyses. We performed a beverage-specific analysis by repeating our analyses stratifying among those who drank primarily beer, wine, or liquor/spirits. We also considered our dependent variable as a continuous variable (continuous LPR).

Results Study sample characteristics We present the characteristics of the study sample (mean age±SD 49.8±10.2, 50.3% women) in Table 1. A total of 433 participants, 17.5% of the sample, had hepatic steatosis. Compared to those without hepatic steatosis, participants with hepatic steatosis were more often men, had a lower proportion reporting the top income and education categories, and had a higher prevalence of the components of the metabolic syndrome. We depict alcohol use patterns among participants with hepatic steatosis in Table 1 and Figure 1. Risky weekly drinking occurred at a relatively low frequency and was equally common among women and men. Binge drinking was common, particularly among men, and drinking in excess of the US Dietary Guidelines limits for usual consumption was common among both women and men. The prevalence of hepatic steatosis also increased from 15.3% to 54.3%

Page 7 of 24

Long et al. Alcohol use in NAFLD across increasing categories of alcohol use as measured by participant reported alcoholic drinks per week (Figure 2).

Multivariable-adjusted regression models for the association between alcohol consumption measures and hepatic steatosis (dichotomous LPR) In adjusted models, the odds of hepatic steatosis increased by 15% for each SD increase in the number of alcoholic drinks consumed per week (Model 2: aOR 1.15; 95% confidence interval (CI) 1.02,1.29) (Table 2). Risky weekly drinking was also associated with a 54% increased odds of hepatic steatosis (Model 2: aOR 1.54; 95% CI 1.07,2.21). The odds of hepatic steatosis increased by 16% for each SD increase in the maximum drinks per week (Model 2: aOR 1.15; 95% CI 1.023,1.30). There was no significant association between the other drinking patterns evaluated and hepatic steatosis. We did not observe evidence of effect modification by age or sex. Because non-drinkers included a mix of former and never drinkers, we repeated the analysis among drinkers exclusively. The associations between alcohol drinks per week, risky weekly drinking and maximum drinks in 24 hours and hepatic steatosis became stronger (Table 2). We additionally observed associations between frequency of alcohol use and hepatic steatosis (Model 2: aOR 1.07; 95% CI 1.01,1.14) and binge drinking and hepatic steatosis (Model 2: aOR 1.45; 95% CI 1.06,1.98) among drinkers in adjusted models. Results were similar when limiting our sample to the Third Generation Cohort participants (see online supplement).

Characteristics of the study sample, by beverage type Over half of the participants reported either primary beer drinking (n=575) or primary wine drinking (n=809) (Table 3). Beer drinkers were more likely to be younger, male, smokers, less educated, have more cardiometabolic risk factors, and have hepatic steatosis. In addition, beer

Page 8 of 24

Long et al. Alcohol use in NAFLD drinkers consumed more alcoholic drinks per week, higher maximum consumption, and greater likelihood of binge drinking (49.6% versus 17.3% for wine drinkers).

Multivariable-adjusted logistic regression models for the association between various drinking patterns and hepatic steatosis, stratified by beverage type When stratifying by alcoholic beverage type consumed, among beer drinkers we observed a significant association between alcohol drinks per week, maximum drinks in 24 hours, and binge drinking behavior and hepatic steatosis (Table 4). In multivariable-adjusted models, for beer drinkers, the odds of hepatic steatosis increased by 28% per SD increase in the number of beer drinks per week (Model 2: aOR 1.28; 95% CI 1.0,1.63) and by 44% per SD increase in the maximum number of drinks consumed (Model 2 aOR: 1.44; 95% CI 1.11,1.86). Additionally, beer drinkers who reported binge drinking had 2.52 times the adjusted odds of hepatic steatosis (95% CI 1.41,4.51) compared to beer drinkers who did not report binge drinking. For liquor or spirit drinkers (n=249) the odds of hepatic steatosis increased by 24% per SD increase in the frequency of alcohol consumed (aOR 1.24; 95% CI 1.04,1.47) or by 50% per SD increase in the maximum number of drinks consumed (aOR 1.50; 95% CI 1.02,2.20). For wine drinkers (n=809), we observed no associations between alcohol use patterns and hepatic steatosis.

Discussion Despite excluding participants with alcohol consumption exceeding current criteria for NAFLD17, there was a substantial prevalence of binge drinking and consumption in excess of the US Dietary Guidelines limits 22 in the FHS. Moreover, a higher number of alcoholic drinks consumed per week including risky weekly drinking, and a higher maximum number of drinks consumed were all associated with hepatic steatosis, after adjusting for components of the metabolic

Page 9 of 24

Long et al. Alcohol use in NAFLD syndrome. In analyses restricted to drinkers, the frequency of alcohol use and binge drinking were also associated with hepatic steatosis. If we considered the alcoholic beverage type usually consumed, beer drinkers had an increased odds of hepatic steatosis, whereas there were no associations between alcohol use patterns and hepatic steatosis among primary wine drinkers. There are currently no recommendations regarding specific drinking patterns and how drinking behaviors may impact liver fat in patients with NAFLD. The in-depth examination of associations between various alcohol use patterns with objectively measured hepatic steatosis allow us to offer insights that prior studies have been unable to provide. With regard to NAFLD, our findings support consumption within the limits suggested by US Dietary Guidelines recommendations22. Most adults in the US who drink do not drink every day, so the calculating the average number of drinks consumed over a week does not account for the number of alcoholic drinks consumed on drinking days or the maximum number of alcoholic drinks consumed in 1 day.23, 24 We observed frequent binge drinking behavior in our sample. Similarly, in the Coronary Artery Risk Development in Young Adults Study, binge drinking behavior occurring in 27.7% participants with NAFLD who consumed alcohol.25 Among alcohol users in our sample, binge drinking was associated with hepatic steatosis, even after accounting for metabolic syndrome components. Several prior observational studies suggest a protective effect of light or moderate alcohol use on hepatic steatosis or more advanced NAFLD histology.2, 3, 26-29 In contrast, we found the odds of hepatic steatosis increased per SD increase in the weekly average alcoholic drinks consumed after accounting for multiple covariates. In the absence of a randomized controlled study, any protective association between alcohol use and NAFLD histology needs to be interpreted with caution for a number of reasons.14, 30 First, alcohol use is often self-reported and former or infrequent drinkers may be misclassified as non-drinkers, which often serves as the reference group. This misclassification may bias away from the null since the reference group

Page 10 of 24

Long et al. Alcohol use in NAFLD may be composed of former heavy alcohol users with advanced liver histology. Additionally, confounding from socio-demographic differences between non-drinkers or moderate alcohol drinkers may also partially or completely account for the beneficial effects of alcohol. A recent Mendelian randomization study demonstrated no beneficial effect of moderate alcohol use on the histologic severity of NAFLD.7 Two subsequent prospective studies have observed less improvement in histologically-defined NAFLD8 or a worsening of non-invasive markers of histology9 for moderate alcohol users compared to non-drinkers. The eligibility threshold for NAFLD clinical trial candidates, which defines significant alcohol as >14 drinks per week for women and >21 drinks per week for men over a 2-year period preceding baseline histologic examination, is based on consensus recommendations.17, 31 Current guidelines recommend using standardized assessments of alcohol use during the clinical trial31, such as The Alcohol Use Disorders Identification Test32; however, these tools are rarely used to determine study eligibility. Excess alcohol is a well-established risk factor for liverrelated morbidity and mortality in numerous studies. In the Copenhagen City Heart Study, the increased risk of alcohol-related liver disease was significant for women consuming 7 to 13 drinks per week and men consuming 14 to 27 drinks per week33, though the risk of alcoholrelated liver disease occurred at a higher alcohol threshold in other studies.34, 35 Our results suggest that alcohol use is associated with hepatic steatosis even among individuals meeting the typical NAFLD clinical trial eligibility thresholds of alcohol use. Alcohol use may represent a source of heterogeneity in NAFLD populations and could represent a source of bias towards the null for studies of therapies specifically directed towards NAFLD. Additional studies are required to determine if more stringent alcohol use thresholds or if criteria for defining binge drinking behavior should be adopted into NAFLD clinical trial design. We observed many differences between participants based on the typical beverage type consumed. In adjusted models, we observed associations between multiple alcohol use patterns among participants who report primary-beer drinking but no associations among

Page 11 of 24

Long et al. Alcohol use in NAFLD primary wine drinkers. Much of the presumed health benefits of alcohol are attributed to polyphenols, which are present in wine and, to a lesser extent, beer.36 However, although we accounted for socio-demographic factors and metabolic disease in our adjusted models, it is likely that differences in odds of hepatic steatosis between beer and wine drinkers is, at least in part, accounted for by residual confounding by other factors. In our study, the prevalence of hepatic steatosis (17%) was lower compared to other studies, which report a prevalence of 20-30%;37 however, these apparently discrepant results may be explained by differences in the prevalence of hepatic steatosis over time or differing definitions of hepatic steatosis. There are a number of potential limitations to consider. First, moderate alcohol use is not randomly distributed and there will be numerous differences between those with hepatic steatosis who consume modest alcohol vs those that abstain. These factors may impact the pattern of alcohol use and the severity of liver disease and serve as confounders of the association between alcohol and liver fat. Second, by virtue of being a cross-sectional, observational study we cannot account for residual confounding, or establish temporality of relations. Additionally, alcohol use behavior is self-reported and subject to recall bias and nondifferential misclassification with systematic under-reporting of alcohol use by heavy alcohol users. Individuals with CT-defined hepatic steatosis in our sample were community-dwellers who may not have received alcohol use counseling so our findings may not be generalizable other populations. Whereas our study assessed hepatic steatosis, it was not able to assess for degree of liver fibrosis, which may be more closely associated with liver-related complications and death. Finally, since FHS participants are predominately white, the generalizability to nonwhite races is not known. Alcohol-related liver fat may be present among individuals presumed to have NAFLD suggesting significant overlap between NAFLD and ALD. Additional studies are needed to further investigate the relationships and interactions between alcohol consumption and the

Page 12 of 24

Long et al. Alcohol use in NAFLD components of the metabolic syndrome. Finally, prospective studies are needed to validate our findings and to determine if more comprehensive alcohol use screening tools should be used in practice or clinical trial settings.

Page 13 of 24

Long et al. Alcohol use in NAFLD References:

1. Younossi ZM. Non-alcoholic fatty liver disease - A global public health perspective. Journal of hepatology. 2018. Epub 2018/11/12. 2. Dunn W, Sanyal AJ, Brunt EM, et al. Modest alcohol consumption is associated with decreased prevalence of steatohepatitis in patients with non-alcoholic fatty liver disease (NAFLD). Journal of hepatology. 2012;57(2):384-91. Epub 2012/04/24. 3. Dunn W, Xu R, Schwimmer JB. Modest wine drinking and decreased prevalence of suspected nonalcoholic fatty liver disease. Hepatology. 2008;47(6):1947-54. 4. Kwon HK, Greenson JK, Conjeevaram HS. Effect of lifetime alcohol consumption on the histological severity of non-alcoholic fatty liver disease. Liver international : official journal of the International Association for the Study of the Liver. 2014;34(1):129-35. Epub 2013/07/03. 5. Mitchell T, Jeffrey GP, de Boer B, et al. Type and Pattern of Alcohol Consumption is Associated With Liver Fibrosis in Patients With Non-alcoholic Fatty Liver Disease. The American journal of gastroenterology. 2018. Epub 2018/06/15. 6. Yamada K, Mizukoshi E, Seike T, et al. Light alcohol consumption has the potential to suppress hepatocellular injury and liver fibrosis in non-alcoholic fatty liver disease. PloS one. 2018;13(1):e0191026. Epub 2018/01/18. 7. Sookoian S, Flichman D, Castano GO, et al. Mendelian randomisation suggests no beneficial effect of moderate alcohol consumption on the severity of nonalcoholic fatty liver disease. Alimentary pharmacology & therapeutics. 2016;44(11-12):1224-34. Epub 2016/10/26. 8. Ajmera V, Belt P, Wilson LA, et al. Among Patients With Nonalcoholic Fatty Liver Disease, Modest Alcohol Use Is Associated With Less Improvement in Histologic Steatosis and Steatohepatitis. Clin Gastroenterol Hepatol. 2018;16(9):1511-20.e5. Epub 2018/01/30. 9. Chang Y, Cho YK, Kim Y, et al. Non-heavy drinking and worsening of non-invasive fibrosis markers in nonalcoholic fatty liver disease: A cohort study. Hepatology. 2018. Epub 2018/07/19. 10. Ekstedt M, Franzen LE, Holmqvist M, et al. Alcohol consumption is associated with progression of hepatic fibrosis in non-alcoholic fatty liver disease. Scandinavian journal of gastroenterology. 2009;44(3):366-74. Epub 2008/11/20. 11. Kimura T, Tanaka N, Fujimori N, et al. Mild drinking habit is a risk factor for hepatocarcinogenesis in non-alcoholic fatty liver disease with advanced fibrosis. World J Gastroenterol. 2018;24(13):1440-50. Epub 2018/04/11. 12. Liu P, Xu Y, Tang Y, et al. Independent and joint effects of moderate alcohol consumption and smoking on the risks of non-alcoholic fatty liver disease in elderly Chinese men. PloS one. 2017;12(7):e0181497. Epub 2017/07/21. 13. Younossi ZM, Stepanova M, Ong J, et al. Effects of Alcohol Consumption and Metabolic Syndrome on Mortality in Patients with Non-alcoholic and Alcohol-Related Fatty Liver Disease. Clin Gastroenterol Hepatol. 2018. Epub 2018/11/27. 14. Naimi TS, Xuan Z, Brown DW, et al. Confounding and studies of 'moderate' alcohol consumption: the case of drinking frequency and implications for low-risk drinking guidelines. Addiction. 2013;108(9):1534-43. 15. Kannel WB, Feinleib M, McNamara PM, et al. An investigation of coronary heart disease in families. The Framingham offspring study. American journal of epidemiology. 1979;110(3):281-90.

Page 14 of 24

Long et al. Alcohol use in NAFLD 16. Splansky GL, Corey D, Yang Q, et al. The Third Generation Cohort of the National Heart, Lung, and Blood Institute's Framingham Heart Study: design, recruitment, and initial examination. American journal of epidemiology. 2007;165(11):1328-35. 17. Chalasani N, Younossi Z, Lavine JE, et al. The diagnosis and management of nonalcoholic fatty liver disease: Practice guidance from the American Association for the Study of Liver Diseases. Hepatology. 2018;67(1):328-57. Epub 2017/07/18. 18. Speliotes EK, Massaro JM, Hoffmann U, et al. Fatty liver is associated with dyslipidemia and dysglycemia independent of visceral fat: the Framingham Heart Study. Hepatology. 2010;51(6):1979-87. 19. Fox CS, Massaro JM, Hoffmann U, et al. Abdominal visceral and subcutaneous adipose tissue compartments: association with metabolic risk factors in the Framingham Heart Study. Circulation. 2007;116(1):39-48. 20. Speliotes EK, Massaro JM, Hoffmann U, et al. Liver fat is reproducibly measured using computed tomography in the Framingham Heart Study. Journal of gastroenterology and hepatology. 2008;23(6):894-9. 21. Barnard J, Meng XL. Applications of multiple imputation in medical studies: from AIDS to NHANES. Stat Methods Med Res. 1999;8(1):17-36. Epub 1999/05/29. 22. U.S. Department of Agriculture USDoHaH, Services. Nutrition and your health: dietary guidelines for Americans. 8th ed. (www.healthierus.gov/dietaryguidelines). Goverment Printing Office; 2015. 23. Naimi TS. "Gray area" alcohol consumption and the U.S. Dietary Guidelines: a comment on Dawson and Grant (2011). Journal of studies on alcohol and drugs. 2011;72(4):687. Epub 2011/06/21. 24. Dawson DA, Grant BF. The "gray area" of consumption between moderate and risk drinking. Journal of studies on alcohol and drugs. 2011;72(3):453-8. Epub 2011/04/26. 25. VanWagner LB, Ning H, Allen NB, et al. Alcohol Use and Cardiovascular Disease Risk in Patients With Nonalcoholic Fatty Liver Disease. Gastroenterology. 2017;153(5):1260-72.e3. Epub 2017/08/15. 26. Hamaguchi M, Kojima T, Ohbora A, et al. Protective effect of alcohol consumption for fatty liver but not metabolic syndrome. World J Gastroenterol. 2012;18(2):156-67. Epub 2012/01/19. 27. Hashimoto Y, Hamaguchi M, Kojima T, et al. Modest alcohol consumption reduces the incidence of fatty liver in men: a population-based large-scale cohort study. Journal of gastroenterology and hepatology. 2015;30(3):546-52. Epub 2014/09/23. 28. Lau K, Baumeister SE, Lieb W, et al. The combined effects of alcohol consumption and body mass index on hepatic steatosis in a general population sample of European men and women. Alimentary pharmacology & therapeutics. 2015;41(5):467-76. Epub 2015/01/16. 29. Sookoian S, Castano GO, Pirola CJ. Modest alcohol consumption decreases the risk of non-alcoholic fatty liver disease: a meta-analysis of 43 175 individuals. Gut. 2014;63(3):530-2. 30. Naimi TS, Stockwell T, Zhao J, et al. Selection biases in observational studies affect associations between 'moderate' alcohol consumption and mortality. Addiction. 2017;112(2):207-14. Epub 2016/06/19. 31. Sanyal AJ, Brunt EM, Kleiner DE, et al. Endpoints and clinical trial design for nonalcoholic steatohepatitis. Hepatology. 2011;54(1):344-53. 32. Reinert DF, Allen JP. The Alcohol Use Disorders Identification Test (AUDIT): a review of recent research. Alcohol Clin Exp Res. 2002;26(2):272-9. Epub 2002/04/20.

Page 15 of 24

Long et al. Alcohol use in NAFLD 33. Becker U, Deis A, Sorensen TI, et al. Prediction of risk of liver disease by alcohol intake, sex, and age: a prospective population study. Hepatology. 1996;23(5):1025-9. Epub 1996/05/01. 34. Fuchs CS, Stampfer MJ, Colditz GA, et al. Alcohol consumption and mortality among women. The New England journal of medicine. 1995;332(19):1245-50. Epub 1995/05/11. 35. Thun MJ, Peto R, Lopez AD, et al. Alcohol consumption and mortality among middleaged and elderly U.S. adults. The New England journal of medicine. 1997;337(24):1705-14. Epub 1997/12/11. 36. Arranz S, Chiva-Blanch G, Valderas-Martinez P, et al. Wine, beer, alcohol and polyphenols on cardiovascular disease and cancer. Nutrients. 2012;4(7):759-81. Epub 2012/08/02. 37. Younossi Z, Tacke F, Arrese M, et al. Global Perspectives on Non-alcoholic Fatty Liver Disease and Non-alcoholic Steatohepatitis. Hepatology. 2018. Epub 2018/09/05.

Page 16 of 24

Long et al. Alcohol use in NAFLD Figure 1: Prevalence of alcohol consumption patterns among participants with presumed NAFLD (n=433) and the entire cohort (n=2475). Risky weekly drinking was defined as >8 drinks per week for women and >15 drinks per week for men. Usual consumption above US Dietary Guidelines was defined as >2 drinks for women or >3 drinks for men. Binge drinking was defined as > 4 drinks for women or > 5 drinks for men in 24 hours.

Figure 2: Prevalence of hepatic steatosis by alcoholic drink per week category. The unadjusted prevalence of hepatic steatosis increased from 15.3% (n=159/1037) to 16.3% (n=73/447) to 25.0% (n=36/144) to 54.3% (n=6/11; men only) across increasing categories of alcohol use as measured by reported drinks per week. We excluded women who drink >14 drinks per week and men who drink >21 drinks per week from the analysis.

Page 17 of 24

Long et al. Alcohol use in NAFLD Table 1: Characteristics of Study Sample, by hepatic steatosis Hepatic steatosis No Hepatic (n=433) steatosis (n=2042) Socio-demographic factors Age (years), mean (SD) Women, No. (%) Third Generation, No. (%) Current smoking use, No. (%) Physical activity index, mean (SD) Mediterranean Diet Score, mean (SD) Education, No. (%) Some high school High school graduate Some college College graduate Graduate degree a Income , No. (%) <$12,000/yr $12,000-24,999/yr $25,000-49,999/yr $50,000-74,999/yr $75,000-100,000/yr >$100,000/yr Metabolic and liver fat measures Waist circumference (inches), mean (SD) Body Mass Index, mean (SD) Low HDL cholesterol, No. (%) High Triglycerides, No. (%) Elevated blood pressure, No. (%) Impaired Fasting glucose, No. (%) b Liver phantom ratio , mean (SD) Alcohol use patterns Alcohol drinks/week, mean (SD) c Risky weekly drinking , No. (%) Frequency of drinking (#drinking days/week), mean (SD) Usual quantity consumed c,d (drinks/drinking days) , mean (SD) (n=1687) Usual consumption in excess of dietary c,d guidelines , No. (%) Maximum drinks in 24 hours in last month, mean (SD) Binge drinkingc, No. (%) Non-drinkers, No. (%) Never drinkers Former drinkers c Primary beverage consumed , No. (%) Beer

Total (n=2475)

51.2 (10.2) 177 (40.9) 275 (63.5) 49 (11.3) 37.0 (7.2) 11.8±4.2

49.5 (10.1) 1068 (52.3) 1412 (69.2) 237 (11.6) 37.6±7.2 12.4±4.1

49.8 (10.2) 1245 (50.3) 1687 (68.2) 286 (11.6) 37.5±7.2 12.3±4.1

8 (1.9) 103 (23.8) 146 (33.7) 99 (22.9) 77 (17.8)

29 (1.4) 399 (19.5) 609 (29.8) 594 (29.1) 411 (20.1)

37 (1.5) 502 (20.3) 755 (30.5) 693 (28) 488 (19.7)

7 (1.6) 13 (3.0) 81 (18.7) 111 (25.6) 93 (21.5) 128 (29.6)

16 (0.8) 63 (3.1) 373 (18.3) 450 (22.0) 472 (23.1) 668 (32.7)

23 (0.9) 76 (3.1) 454 (18.3) 561 (22.7) 565 (22.8) 796 (32.2)

41.8 (5.6)

37.0 (5.3)

37.8 (5.6)

31.2 (5.7) 210 (48.5) 221 (51.0) 195 (45.0) 113 (26.1) 0.27±0.06

26.6 (4.8) 524 (25.7) 419 (20.5) 479 (23.5) 173 (8.5) 0.37±0.02

27.4 (5.3) 734 (29.7) 640 (25.9) 674 (27.2) 286 (11.6) 0.36±0.05

4.5 (5.4) 50 (11.6)

4.0 (4.5) 203 (9.9)

4.1 (4.7) 253 (10.2)

2.2 (2.4)

2.1 (2.1)

2.1 (2.2)

1.9 (1.4)

1.7 (1.3)

1.7 (1.3)

99/275 (36.0)

493/1412 (34.9)

592/1687 (35.1)

2.6 (3.0)

2.4 (2.5)

2.4 (2.6)

110 (25.4)

467 (22.9)

577 (23.3)

16 (3.7) 81 (18.7)

42 (2.1) 334 (16.4)

58 (2.3) 415 (16.8)

102 (23.6)

473 (23.2)

575 (23.2)

Page 18 of 24

Long et al. Alcohol use in NAFLD

Wine Liquor/Spirits Beer and wine Liquor/Spirits and beer or wine Non-drinker or unknown

Hepatic steatosis No Hepatic (n=433) steatosis (n=2042) 123 (28.4) 686 (33.6) 59 (13.6) 190 (9.3) 13 (3.0) 99 (4.9) 10 (2.3) 84 (4.1) 126 (29.1) 510 (25.0)

Total (n=2475) 809 (32.7) 249 (10.1) 112 (4.5) 94 (3.8) 636 (25.7)

Data are expressed as mean ± standard deviation or number (percentage) unless otherwise noted. a Income indexed to 2003 dollars b Liver phantom ratio is derived from computed tomography scans as the average liver Hounsfield Units divided by a radiopaque phantom (lower liver phantom ratio is consistent with more liver fat) c Risky weekly drinking was defined as ≥ 8 drinks per week for women and ≥ 15 drinks per week for men. Usual consumption above US Dietary Guidelines was defined as ≥2 drink/drinking day for women or ≥ 3 drinks/drinking day for men. Binge drinking was defined as > 4 drinks for women or > 5 drinks for men in 24 hours. Primary beverage type is the reported highest amount of an alcoholic beverage type consumed in a week. If the participant reported consuming more than one beverage type in equal amounts they were considered to be in a mixed beverage type category. d Third Generation Participants only

Page 19 of 24

Long et al. Alcohol use in NAFLD

Table 2: Multivariable-adjusted logistic regression models for the association between various drinking patterns and hepatic steatosis (liver phantom ratio ≤ 0.33), overall and for drinkers only in the Framingham Heart Study.

Alcohol use pattern Alcohol drinks per week Risky weekly drinkinga Yes No Frequency (# drinking days/week) Usual quantityb (# drinks /drinking day) Usual consumption in excess of dietary guidelinesa,b Yes No Max drinks in 24 hours in last month Binge drinkinga Yes No Non-drinkers Former drinkers Never drinkers

Overall (n=2475) Model 1 Model 2 aOR, (95% CI); p value aOR, (95% CI); p value 1.09 (0.98, 1.21); 0.10 1.15 (1.02, 1.28); 0.02

Drinkers only (n=2002) Model 1 aOR, (95% CI); p value 1.14 (1.01, 1.28); 0.04

Model 2 aOR, (95% CI); p value 1.19 (1.05, 1.35); 0.009

1.34 (0.96, 1.88); 0.09 reference 1.02 (0.97, 1.07); 0.40

1.54 (1.07 2.21); 0.02 reference 1.05 (1.0, 1.10); 0.08

1.39 (0.99, 1.96); 0.06 reference 1.02 (0.99, 1.10); 0.15

1.56 (1.08, 2.26); 0.02 reference 1.07 (1.01, 1.14); 0.03

1.04 (0.95, 1.14); 0.40

1.04 (0.93, 1.15); 0.51

1.08 (0.96, 1.22); 0.21

1.06 (0.93, 1.20); 0.41

1.13 (0.85, 1.50); 0.41 reference 1.11 (1.0, 1.25); 0.05

1.14 (0.84, 1.54); 0.42 reference 1.15 (1.02, 1.30); 0.02

1.14 (0.84, 1.54); 0.41 reference 1.16 (1.02, 1.32); 0.02

1.14 (0.82, 1.58); 0.345 reference 1.20 (1.05, 1.37); 0.007

1.16 (0.89,1.53); 0.28 reference

1.23 (0.92, 1.65); 0.16 reference

1.36 (1.01, 1.82); 0.04 reference

1.45 (1.06, 1.98); 0.02 reference

0.61 (0.31,1.05); 0.28 reference

0.69 (0.35, 1.38); 0.42 reference

For continuous alcohol use behaviors, odds ratios are modeled per standard deviation increase in the continuous independent variable. Model 1 is adjusted for age, sex, cohort, smoking, physical activity, diet, education, and income. Model 2 is adjusted for Model 1 and the components of the metabolic syndrome (waist circumference, low HDL-c, high triglycerides, elevated blood pressure, and impaired fasting glucose). a Risky weekly drinking was defined as ≥ 8 drinks per week for women and ≥ 15 drinks per week for men. Usual consumption above US Dietary Guidelines was defined as ≥2 drink/drinking day for women or ≥ 3 drinks/drinking day for men. Binge drinking was defined as > 4 drinks for women or > 5 drinks for men in 24 hours. b Third generation participants only

Page 20 of 24

Long et al. Alcohol use in NAFLD Table 3: Characteristics of Study Sample, by beverage type consumeda

Socio-demographic factors Age (years), mean (SD) Women, No. (%) Third Generation, No. (%) Current smoking use, No. (%) Physical activity index, mean (SD) Mediterranean Diet Score, mean (SD) Education, No. (%) Some high school High school graduate Some college College graduate Graduate degree b Income , No. (%) <$12,000/yr $12,000-24,999/yr $25,000-49,999/yr $50,000-74,999/yr $75,000-100,000/yr >$100,000/yr Metabolic and liver fat measures Waist circumference (inches), mean (SD) Body Mass Index, mean (SD) Low HDL cholesterol, No. (%) High Triglycerides, No. (%) Elevated blood pressure, No. (%) Impaired Fasting glucose, No. (%) c Liver phantom ratio , mean (SD) d Hepatic steatosis , No. (%) Alcohol use patterns Alcohol drinks/week, mean (SD)

Beer (n=575)

Wine (n=809)

Liquor/Spirit (n=249)

Beer/Wine (n=112)

Mix of beer/wine/spirit Unknown or (n=94) Nondrinkers (n=636)

44.8 (7.8) 102 (17.7) 488 (84.9) 82 (14.3) 38.5 (8.6) 11.8±3.9

50.5 (9.3) 544 (67.2) 541 (66.9) 62 (7.7) 36.7 (5.9) 13.0±4.3

52.1 (11.5) 120 (51.8) 149 (59.8) 54 (21.7) 37.4 (7.7) 11.6±3.8

47.5 (8.8) 51 (45.5) 80 (71.4) 7 (6.3) 36.9 (5.4) 12.8±4.0

49.7 (9.5) 51 (54.3) 61 (64.9) 7 (7.5) 36.9 (76.4) 11.2±3.4

52.9 (11.0) 377 (59.3) 368 (57.8) 74 (11.6) 37.9 (7.3) 12.2±4.1

6 (1) 107 (18.6) 172 (29.9) 198 (34.4) 92 (16)

9 (1.1) 111 (13.7) 230 (28.4) 252 (31.2) 207 (25.6)

4 (1.6) 53 (21.3) 99 (39.8) 48 (19.3) 45 (18.1)

0 21 (18.8) 28 (25.0) 40 (35.7) 23 (20.5)

3 (3.2) 17 (18.1) 28 (29.8) 27 (28.7) 19 (20.2)

15 (2.4) 193 (30.4) 198 (31.1) 128 (20.1) 102 (16.0)

2 (0.4) 16 (2.8) 93 (16.2) 128 (22.3) 147 (25.6) 189 (32.9)

5 (0.6) 14 (1.7) 116 (14.3) 164 (20.3) 173 (21.4) 337 (41.7)

0 3 (2.7) 16 (14.3) 24 (21.4) 31 (27.7) 38 (33.9)

14 (1.9) 36 (4.8) 166 (22.2) 193 (25.8) 172 (23.0) 167 (22.3)

0 4 (4.3) 20 (21.3) 15 (16.0) 23 (24.5) 32 (34.0)

14 (2.2) 33 (5.2) 150 (23.6) 169 (26.6) 141 (22.2) 129 (20.3)

38.3 (4.9)

36.5 (5.4)

39.3 (6.0)

36.7 (5.7)

38.1 (5.1)

38.6 (6.1)

27.7±4.8

26.3±4.9

28.4±5.5

26.8±5.0

27.7±4.7

28.3±5.8

158 (27.5) 155 (27.0) 133 (23.1) 57 (9.9) 0.36 (0.06) 102 (17.7)

202 (25.0) 165 (20.4) 214 (26.5) 59 (7.3) 0.36 (0.05) 123 (15.2)

31 (27.7) 30 (26.8) 15 (13.4) 9 (8.0) 0.35 (0.06) 13 (11.6)

267 (35.7) 210 (28.1) 219 (29.3) 121 (16.2) 0.36 (0.04) 139 (18.6)

21 (22.3) 25 (26.6) 18 (19.1) 8 (8.5) 0.36 (0.04) 10 (10.6)

236 (37.1) 180 (28.3) 204 (32.1) 112 (17.6) 0.35 (0.05) 126 (19.8)

6.5 (4.9)

5.1 (4.4)

5.5 (5.2)

3.9 (5.6)

3.5 (3.3)

0

Page 21 of 24

Long et al. Alcohol use in NAFLD

Beer (n=575) a

Wine (n=809)

Risky weekly drinking , No. (%) 71 (12.4) 131 (16.2) Frequency of drinking (#drinking 3.0 (2.0) 2.8 (2.2) days/week), mean (SD) Usual quantity consumed a,e 2.5 (2.0) 1.7 (0.7) (drinks/drinking days) , mean (SD) (n=1687) Usual consumption in excess of dietary 230/488 (47.1) 205/541 (37.9) a,e guidelines , No. (%) Maximum drinks in 24 hours in last 4.7 (3.1) 2.5 (1.8) month, mean (SD) Binge drinkinga, No. (%) 285 (49.6) 140 (17.3) Non-drinkers, No. (%) Never drinkers 0 0 Former drinkers 0 0

Liquor/Spirit (n=249) 37 (14.9)

Beer/Wine (n=112) 8 (7.1)

Mix of beer/wine/spirit Unknown or (n=94) Nondrinkers (n=636) 6 (6.4) 0

3.0 (2.4)

1.8 (1.6)

1.8 (1.4)

0.2 (0.5)

2.0 (1.1)

1.6 (0.7)

1.9 (1.0)

0.7 (1.4)

72/149 (48.3)

23/80 (28.8)

22/61 (36.1)

40/368 (10.9)

2.6 (2.3)

2.5 (1.6)

2.9 (2.7)

0.18 (0.06)

24 (21.4)

82 (11)

25 (26.6)

58 (9.1)

0 0

0 0

0 0

58 (9.1) 415 (65.3)

Data are expressed as mean ± standard deviation or number (percentage) unless otherwise noted. a Risky weekly drinking was defined as ≥ 8 drinks per week for women and ≥ 15 drinks per week for men. Usual consumption above US Dietary Guidelines was defined as ≥2 drink/drinking day for women or ≥ 3 drinks/drinking day for men. Binge drinking was defined as > 4 drinks for women or > 5 drinks for men in 24 hours. Primary beverage type is the reported highest amount of an alcoholic beverage type consumed in a week. If the participant reported consuming more than one beverage type in equal amounts they were considered to be in a mixed beverage type category. b Income indexed to 2003 dollars c Liver phantom ratio is derived from computed tomography scans as the average liver Hounsfield Units divided by a radiopaque phantom (lower liver phantom ratio is consistent with more liver fat) d Hepatic steatosis defined as liver phantom ratio≤ 0.33 e Third Generation Participants only

Page 22 of 24

Long et al. Alcohol use in NAFLD Table 4: Multivariable-adjusted logistic regression models for the association between various drinking patterns and hepatic steatosis (liver phantom ratio ≤0.33), by alcohol beverage typea in the Framingham Heart Study. Beer drinkers (n=575) Wine drinkers (n=809) Spirit drinkers (n=249) Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 aOR, (95% CI); aOR, (95% CI); aOR, (95% CI); aOR, (95% CI); aOR, (95% CI); aOR, (95% CI); Alcohol use pattern p value p value p value p value p value p value 1.20 (0.97, 1.50); 1.28 (1.0, 1.63); 1.10 (0.90, 1.34); 1.08 (0.87, 1.34); 1.12 (0.79, 1.59); 1.34 (0.89, 2.02); Alcohol drinks per week 0.10 0.05 0.36 0.48 0.53 0.16 Risky weekly drinkingb 1.63 (0.87, 3.07); 1.70 (0.84, 3.45); 1.37 (0.82, 2.29); 1.53 (0.88, 2.65); 1.19 (0.48, 2.93); 1.45 (0.52, 4.02); Yes 0.13 0.14 0.24 0.13 0.71 0.48 No reference reference reference reference reference reference Frequency (# drinking 1.07(0.95, 1.20); 1.13 (0.99, 1.28); 0.98 (0.90, 1.07); 0.97 (0.88, 1.07); 1.13 (0.97, 1.30); 1.24 (1.04, 1.47); days/week) 0.26 0.06 0.68 0.55 0.12 0.02 Usual quantity 1.06 (0.90, 1.25); 1.05 (0.88, 1.26); 1.26 (0.90, 1.76); 1.22 (0.85, 1.74); 1.20 (0.80 1.79); 1.13 (0.67, 1.92); 0.50 0.60 0.18 0.28 0.38 0.65 (drinks/drinking day)b,c b,c Usual consumption in excess of dietary guidelines 1.30 (0.78, 2.19); 1.47 (0.82, 2.62); 1.42 (0.79, 2.54); 1.32 (0.71, 2.44); 1.52 (0.61, 3.80); 148 (0.45, 4.79); Yes 0.34 0.19 0.24 0.38 0.37 0.52 No reference reference reference reference reference reference Max drinks in 24 hours 1.41 (1.12, 1.77); 1.44 (1.11, 1.86); 1.12 (0.91, 1.38); 1.15 (0.92, 1.43); 1.28 (0.90, 1.81); 1.50 (1.02, 2.20); in last month 0.003 0.006 0.29 0.23 0.16 0.04 Binge drinkingb 2.17 (1.30, 3.63); 2.52 (1.41, 4.51); 1.48 (0.87, 2.54); 1.49 (0.85, 2.60); 0.85 (0.34, 2.13); 1.06 (0.38, 2.91); Yes 0.003 0.002 0.15 0.17 0.73 0.91 No reference reference reference reference reference reference For continuous alcohol use behaviors, odds ratios are modeled per standard deviation increase in the continuous independent variable. Model 1 is adjusted for age, sex, cohort, smoking, physical activity, diet, education, and income. Model 2 is adjusted for Model 1 and the components of the metabolic syndrome (waist circumference, low HDL-c, high triglycerides, elevated blood pressure, or impaired fasting glucose). a

Beer, wine, and spirit drinkers were defined as reporting consuming beer, wine, or spirits in the highest amount of any alcoholic beverage type consumed in a week. Participants consuming equal amounts of beer, wine, or spirits were excluded from this analysis. weekly drinking was defined as ≥ 8 drinks per week for women and ≥ 15 drinks per week for men. Usual consumption above US Dietary Guidelines was defined as ≥2 drink/drinking day for women or ≥ 3 drinks/drinking day for men. Binge drinking was defined as > 4 drinks for women or > 5 drinks for men in 24 hours.

Page 23 of 24

Long et al. Alcohol use in NAFLD c

Third Generation participants only (n=1687)

Page 24 of 24

Online Supplement Contents: Methods Results References eTables 1-4

Methods Alcohol use behaviors We defined average alcohol consumption as the number of alcoholic beverages of any type consumed per week as derived from the physician-directed questionnaire at the index examination. We defined risky weekly drinking as ≥ 8 drinks per week for women and ≥ 15 drinks per week for men1. Alcohol use frequency was assessed by asking: “Over the past year, on average, on how many days per week did you drink an alcoholic beverage of any type?” for Third Generation participants. We asked Offspring participants the average number of days per week they drank either beer, red wine, white wine, or liquor. We considered the most number of drinking days of any beverage type as the drinking frequency. We assessed the usual quantity consumed by asking: “Over the past year, on a typical day when you drink, how many drinks do you have” for Third Generation participants only (not asked in Offspring participants). Women who consumed more than one drink/drinking day and men who consumed more than two drinks per drinking day were classified as drinking in excess of the U.S. dietary guidelines.2 We recorded the maximum drinks in last month by asking: “what was the maximum number of drinks you had in a 24 hour period during the past month” for the Third Generation participants. We asked Offspring participants their usual maximum limit of alcohol by alcohol type. We considered the maximum drinks as the highest maximal limit of alcohol across beverage types. For Third Generation participants we defined binge drinking as drinking 4 or more (for women) or 5 or more (for men) alcoholic beverages in a 24-hour period in the last month or answering yes to the question “since your last exam, has there been a time when you drank 5 or more alcoholic drinks of any kind almost daily.” For Offspring participants, binge drinking was defined as a maximum

limit of 4 or more drinks for women or 5 or more drinks for men. Responses were grouped into 2 categories (no vs yes binge drinking). We divided non-drinkers into former drinkers and never drinkers. For Third Generation participants, we identified former drinkers with the question “at what age did you stop drinking” and any participant that reported an age less than the current age was considered a former drinker. For Offspring participants, we identified former drinkers as any participant who reported consuming at least 1 drink/week on any prior exam (exam 1-6). For a secondary analysis, we considered the specific type of alcoholic beverage consumed, specifically wine, beer, liquor/spirit drink use, or combinations thereof. A respondent’s primary beverage type was on the basis of the largest number of drinks consumed during the past week on a beverage-specific basis.

Computed Tomography Liver Fat Measurement In brief, 25 contiguous 5-mm-thick slices (120 kVp, 400mA, gantry rotation time 500 ms, and table feed 3:1) in the abdominal region (covering 125 mm above S1) were obtained from participants in the supine position using an 8-slice CT scanner (LightSpeed Ultra, General Electric, Milwaukee, WI). A calibration phantom (Image Analysis, Lexington, KY) was placed under each participant and present for all images obtained.

Covariates We defined current smoking use as smoking at least one cigarette per day in the year preceding the FHS examination. We measured physical activity by the physical activity index score.3 We assessed education as a categorical variable (<12 years, high school graduate, some college, college graduate, or graduate degree). We accounted

for diet quality using a modified Mediterranean-Style Diet Score (MDS), which we have previously utilized in the FHS.4 The MDS consists of 9 components including: vegetables, fruits, nuts, legumes, whole grains, fish, red meat, ratio of monounsaturated fatty acids to saturated fatty acids, and alcohol.5 We modified the MDS to use just the first 8 components and eliminate the alcohol component given our study design. The scores for each component of the MDS range from 0-3 from lowest to highest quartile (except for red meat, which ranges from 0-3 from highest to lowest quartile). We summed all the score components (range 0-24), with higher scores indicating a healthier diet pattern. For Offspring Cohort participants, education was assessed at the second examination visit (1979–1983). We assessed income as a categorical variable of household income/year. For Offspring Cohort participants, income data was obtained from the third examination (1983–1987) and adjusted for inflation to 2003 (correction factor of 71%). Offspring Cohort participants were at a mean age of 48 years when income information was collected which is similar to the age of the Third Generation Cohort participants at their first examination (2002–2005) (mean age 40 years). Trained technicians used standard protocols for measuring heart rate, blood pressure, height, weight, and waist circumference (in inches) at the examination visits. We measured serum total and high density lipoprotein cholesterol (HDL-c), triglycerides, and glucose on fasting morning blood samples.6 We defined low HDL-c as <50 mg/dL for women and <40 mg/dL for men. We defined elevated triglycerides as triglycerides ≥150 mg/dL. We defined elevated blood as a systolic blood pressure ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, or the use of anti-hypertensive medication. We defined impaired fasting glucose as a fasting glucose ≥110mg/dL or treatment with a hypoglycemic agent or insulin.

Results Continuous Liver Phantom Ratio as Outcome Measure Overall, results were similar when we considered continuous LPR as the dependent variable though frequency of alcohol use was no longer associated with liver fat among drinkers in Model 2 and maximum drinks in 24 hours was no longer associated with liver fat among drinkers in Model 1 (eTable 1). For the beverage type specific analyses, results were overall similar when we considered continuous LPR as the dependent variable, though the associations with liver fat for liquor/spirit drinkers was no longer significant (eTable 2).

Sensitivity Analysis We performed a sensitivity limiting the sample to Third Generation Cohort participants only (n=1687). Overall, the magnitude and direction of the associations were similar among Third Generation Cohort participants for the associations between alcohol use patterns and both hepatic steatosis (eTable 3) and continuous liver fat (eTable 4) though most associations were no longer statistically significant.

References

1.

2.

3. 4.

5.

6.

Alcoholism NIoAAa. Rethinking Drinking: Alcohol and your health. In. Vol NIH Publication N0. 15-3770. RethinkingDrinking.niaaa.nih.gov: National Institute of Health; 2016:1-15. U.S. Department of Agriculture USDoHaH, Services. Nutrition and your health: dietary guidelines for Americans. In. 8th ed. (www.healthierus.gov/dietaryguidelines). Goverment Printing Office; 2015. Kannel WB, Sorlie P. Some health benefits of physical activity. The Framingham Study. Arch Intern Med. 1979;139(8):857-861. Ma J, Hennein R, Liu C, et al. Improved Diet Quality Associates With Reduction in Liver Fat, Particularly in Individuals With High Genetic Risk Scores for Nonalcoholic Fatty Liver Disease. Gastroenterology. 2018;155(1):107-117. Fung TT, Rexrode KM, Mantzoros CS, Manson JE, Willett WC, Hu FB. Mediterranean diet and incidence of and mortality from coronary heart disease and stroke in women. Circulation. 2009;119(8):1093-1100. Porter SA, Pedley A, Massaro JM, Vasan RS, Hoffmann U, Fox CS. Aminotransferase levels are associated with cardiometabolic risk above and beyond visceral fat and insulin resistance: the Framingham Heart Study. Arteriosclerosis, thrombosis, and vascular biology. 2013;33(1):139-146.

eTable 1: Multivariable-adjusted linear regression models for the association between various drinking patterns and more liver fat (-LPR), overall and among drinkers only Overall (n=2475) Model 1

Model 2

Drinkers only (n=2002) Model 1

Model 2

Alcohol use pattern*

β, 95% CI, p value

β, 95% CI, p value

β (95% CI); p value

β (95% CI); p value

Alcohol drinks per week Risky weekly drinking† Yes No Frequency (# drinking days/week) Usual quantity** (# drinks /drinking day) Usual consumption in excess of dietary guidelines†,** Yes No Max drinks in 24 hours in last month Binge drinking† Yes No Non-drinkers Former drinkers Never drinkers

0.002 (-2.7E-05, 0.004); 0.05

0.003 (-0.00003, 0.004); 0.05 0.002 (0.00004, 0.005); 0.05 0.003 (0.001, 0.005); 0.007

0.006 (-0.0003, 0.013); 0.06 reference

0.008 (0.002, 0.015); 0.01 reference

0.006 (-0.0003, 0.013); 0.06 reference

0.008 (0.002, 0.014); 0.01 reference

0.0007 (-0.001, 0.003); 0.52

0.002 (-0.0003, 0.004); 0.10

0.001 (-0.001, 0.003); 0.42

0.002 (-0.0004, 0.004);0.10

-0.00004 (-0.002, 0.002); 0.97 0.001 (-0.002, 0.003); 0.33

0.002 (-0.001, 0.004); 0.27

0.0008 (-0.002, 0.003); 0.29

0.002 (-0.003, 0.007); 0.40 reference

0.002 (-0.003, 0.007); 0.44 reference

0.002 (-0.003, 0.008); 0.41 reference

0.002 (-0.003, 0.007); 0.48 reference

0.002 (-0.0002, 0.004); 0.07

0.003 (0.0004, 0.005); 0.04

0.002 (-0.0003, 0.005); 0.08

0.003 (0.0002, 0.005); 0.03

0.004 (-0.001, 0.009); 0.15 reference

0.005 (-0.0002, 0.01); 0.06 reference

0.006( 0.0002, 0.011); 0.04 reference

0.006( 0.001, 0.012); 0.02 reference

reference -0.005 (-0.02, 0.009); 0.76

-0.002 (-0.02, 0.01); 0.75

*For continuous alcohol use behaviors, betas are modeled per standard deviation increase in the continuous independent variable. Model 1 is adjusted for age, sex, cohort, smoking, physical activity, education, and income. Model 2 is adjusted for Model 1 and the components of the metabolic syndrome (waist circumference, low HDL-c, high triglycerides, elevated blood pressure, or impaired fasting glucose). † Risky weekly drinking was defined as ≥ 8 drinks per week for women and ≥ 15 drinks per week for men. Usual consumption above US Dietary Guidelines was defined as ≥2 drink/drinking day for women or ≥ 3 drinks/drinking day for men. Binge drinking was defined as > 4 drinks for women or > 5 drinks for men in 24 hours. **Third Generation Participants only

eTable 2: Multivariable-adjusted linear regression models for the association between various drinking patterns and more liver fat (-LPR), by alcohol type* Beer drinkers Wine drinkers Spirit drinkers Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 β, 95% CI, p β, 95% CI, p β, 95% CI, p β, 95% CI, p Alcohol use pattern** β, 95% CI, p value β, 95% CI, p value value value value value 0.006 (0.001, 0.006 (0.002, 0.002 (-0.002, 0.002 (-0.002, -0.001 (-0.010, 0.0008 (-0.007, Alcohol drinks per week 0.011); 0.014 0.01); 0.005 0.005); 0.33 0.005); 0.37 0.007); 0.76 0.009); 0.85 Risky weekly drinking† 0.020 (0.006, 0.019 (0.006, 0.003 (-0.006, 0.005 (-0.004, -0.008 (-0.029, -0.007 (-0.027, Yes 0.034); 0.005 0.033); 0.004 0.012); 0.51 0.013); 0.3 0.013); 0.46 0.013); 0.49 No reference reference reference reference reference reference Frequency (# drinking 0.002 (-0.003, 0.004 (-0.0011, -0.0003 (-0.004, -0.0003 (-0.004, 0.004 (-0.005, 0.006 (-0.002, 0.01); days/week) 0.007); 0.40 0.008); 0.13 0.003); 0.87 0.003); 0.88 0.01); 0.39 0.16 Usual quantity*** (# 0.003 (-0.002, 0.002 (-0.004, -0.0003 (-0.004, -0.001 (-0.005, 0.002 (-0.008, -0.0004 (-0.010, drinks /drinking day) 0.008);0.28 0.007); 0.54 0.004); 0.89 0.003); 0.63 0.013); 0.66 0.009); 0.94 Usual consumption in excess of dietary guidelines†,*** 0.008 (-0.003, 0.019); 0.17 No reference Max drinks in 24 hours in 0.009 (0.004, last month 0.014); 0.001 Binge drinking† 0.019 (0.009, Yes 0.029);<0.001 No reference Yes

0.007 (-0.003, 0.017); 0.17 reference 0.008(0.003, 0.012); 0.001

0.0002 (-0.008, 0.008); 0.95 reference 0.001 (-0.002, 0.005); 0.47

-0.001 (-0.009, 0.007); 0.79 reference 0.001 (-0.002, 0.005); 0.43

0.009 (-0.011, 0.029); 0.37 reference -0.003 (-0.012, 0.005); 0.46

0.004 (-0.015, 0.023); 0.66 reference -0.002 (-0.010, 0.006); 0.67

0.018 (0.008, 0.027);<0.001 reference

0.002 (-0.008, 0.011); 0.74 reference

0.001 (-0.007, 0.01); 0.76 reference

-0.02 (-0.038, 0.004); 0.11 reference

-0.01 (-0.034, 0.006); 0.18 reference

For continuous alcohol use behaviors, betas are modeled per standard deviation increase in the continuous independent variable. *Beer, wine, and spirit drinkers were defined as reporting consuming beer, wine, or spirits in the highest amount of any alcoholic beverage type consumed in a week. Participants consuming equal amounts of beer, wine, or spirits were excluded from this analysis. **Model 1 is adjusted for age, sex, cohort, smoking, physical activity, education, and income.

Model 2 is adjusted for Model 1 and the components of the metabolic syndrome (waist circumference, low HDL-c, high triglycerides, elevated blood pressure, or impaired fasting glucose). †

Risky weekly drinking was defined as ≥ 8 drinks per week for women and ≥ 15 drinks per week for men. Usual consumption above US Dietary Guidelines was defined as ≥2 drink/drinking day for women or ≥ 3 drinks/drinking day for men. Binge drinking was defined as > 4 drinks for women or > 5 drinks for men in 24 hours. ***Third Generation participants only (n=1687)

eTable 3: Multivariable-adjusted logistic regression models for the association between various drinking patterns and hepatic steatosis (LPR ≤ 0.33) (Third Generation only) Model 1 Model 2 Alcohol use pattern*

AOR, 95% CI

Alcohol drinks per week

p value AOR, 95% CI

p value

1.06 (0.93,1.20)

0.41

1.13 (0.97, 1.30)

0.11

Yes

1.12 (0.73, 1.71)

0.60

1.28 (0.81, 2.04)

0.29

No

reference

Risky weekly drinking

†

reference

Frequency (#drinking days/week)

1.00 (0.94, 1.07)

1.0

1.04 (0.97, 1.12)

0.24

Maximum drinks in 24 hours in the last month

1.14 (1.00, 1.30)

0.05

1.18 (1.02, 1.36

0.03

Yes

1.12 (0.84, 1.50)

0.43

1.19 (0.87, 1.63)

0.27

No

reference

Binge drinking† reference

Non-drinkers Former drinkers

0.41 (0.16, 1.01)

Never drinkers

reference

0.08

0.45 (0.17, 1.22)

0.14

reference

*For continuous alcohol use behaviors, odds ratios are modeled per standard deviation increase in the continuous independent variable. Model 1 is adjusted for age, sex, cohort, smoking, physical activity, education, and income. Model 2 is adjusted for Model 1 and the components of the metabolic syndrome (waist circumference, low HDL-c, high triglycerides, elevated blood pressure, or impaired fasting glucose). † Risky weekly drinking was defined as ≥ 8 drinks per week for women and ≥ 15 drinks per week for men. Binge drinking was defined as > 4 drinks for women or > 5 drinks for men in 24 hours.

eTable 4: Multivariable-adjusted linear regression models for the association between various drinking patterns and more liver fat (-LPR) (Third generation only) Model 1 Alcohol use pattern*

Model 2

β (95% CI)

p value β (95% CI)

p value

0.002 (-0.0008, 0.004)

0.17

0.003 (0.001, 0.005)

0.02

Yes

0.005 (-0.003, 0.012)

0.23

0.007 (-6.32E-05, 0.014)

0.05

No

reference

Frequency (#drinking days/week)

0.00002 (-0.002, 0.002)

0.98

0.002 (-0.001, 0.004)

0.17

Maximum drinks in 24 hours in the last month

0.002(-0.0002,0.005)

0.08

0.003 (0.0004, 0.005)

0.02

Yes

0.004 (-0.001,0.009)

0.15

0.005 (0.0001,0.01)

0.04

No

reference

Alcohol drinks per week Risky weekly drinking†

reference

Binge drinking† reference

Non-drinkers Former drinkers

-0.013 (-0.033, 0.007)

Never drinkers

reference

0.22

-0.008 (-0.03, 0.01)

0.41

reference

*For continuous alcohol use behaviors, betas are modeled per standard deviation increase in the continuous independent variable. Model 1 is adjusted for age, sex, cohort, smoking, physical activity, education, and income. Model 2 is adjusted for Model 1 and the components of the metabolic syndrome (waist circumference, low HDL-c, high triglycerides, elevated blood pressure, or impaired fasting glucose). † Risky weekly drinking was defined as ≥ 8 drinks per week for women and ≥ 15 drinks per week for men. Binge drinking was defined as > 4 drinks for women or > 5 drinks for men in 24 hours.

Page 32 of 11

What you need to know Background Many individuals presumed to have non-alcoholic fatty liver disease (NAFLD) consume moderate amounts of alcohol. Little is known about patterns of alcohol use in patients with NAFLD or how drinking behaviors affect liver fat. Findings In a cross-sectional study of participants of the Framingham Heart Study with hepatic steatosis, we observed an association between multiple alcohol use patterns and liver fat, even after excluding heavy alcohol users from our analysis. Alcohol use therefore appears to be a risk factor for NAFLD. Prospective studies are needed to validate these findings and determine if alcohol use should be a focus for research, prevention, and treatment of presumed NAFLD. Implications for patient care For people who meet the definition of non-alcoholic fatty liver disease, alcohol consumption appears to be a risk factor. Specific patterns of drinking might increase risk for liver fatty disease and other health outcomes.