Binge Drinking and the Risk of Liver Events: a Population-Based Cohort Study

Received: 10 January 2017 | Accepted: 26 February 2017 DOI: 10.1111/liv.13408

GENETIC AND METABOLIC LIVER DISEASE

Binge drinking and the risk of liver events: A population-based cohort study

Fredrik Åberg1 | Jaana Helenius-Hietala2 | Pauli Puukka3 | Antti Jula3

1Transplantation and Liver Surgery Clinic, Helsinki University Hospital, Helsinki Abstract University, Helsinki, Finland Background & Aims: Binge drinking or heavy episodic drinking is increasingly preva- 2 Department of Oral and Maxillofacial lent, but the health effects are incompletely understood. We investigated whether Diseases, Helsinki University Hospital, Helsinki University, Helsinki, Finland binge drinking increases the risk for liver disease above and beyond the risk due to 3Department of Health, National Institute for average alcohol consumption. Health and Welfare, Turku, Finland Methods: 6366 subjects without baseline liver disease who participated in the Finnish Correspondence population-based Health 2000 Study (2000-2001), a nationally representative cohort. Fredrik Åberg, MD, PhD, Transplantation Follow-up data from national registers until 2013 were analysed for liver-related ad- and Liver Surgery Clinic, Helsinki University Hospital, Helsinki University, Helsinki, Finland. missions, mortality and liver cancer. Binge drinking (≥5 drinks per occasion, standard Email: [email protected] drink 12 g ethanol) was categorised as weekly, monthly, or as less often or none. Funding information Multiple confounders were considered. FÅ received research grants from Wilhelm and Else Stockmanns Foundation, Liv och Hälsa, Results: Eighty-four subjects developed decompensated liver disease. Binge drinking and Finska Läkaresällskapet. frequency showed a direct association with liver-disease risk after adjustment for aver-

Handling Editor: Helena Cortez-Pinto age daily alcohol intake and age. After adjustment, the hazard ratios (HRs) for weekly and monthly binge drinking were 3.45 (P=.001) and 2.26 (P=.007) and were higher after excluding regular heavy drinkers. The HR for weekly binging was 6.82 (P=.02) in women; 2.34 (P=.03) in men; and 4.29 (P=.001) in subjects with the metabolic syndrome. Weekly binge drinking and the metabolic syndrome produced supra-additive increases in the risk of decompensated liver disease. Weekly, and to a lesser extent monthly, binging retained significance in sequential multivariate models that additionally adjusted for beverage preference and lifestyle, metabolic, and socioeconomic factors. Conclusions: Binge drinking is associated with an increased risk for liver disease independently of average alcohol intake and confounders. The rising prevalence of binge drinking and the metabolic syndrome is particularly concerning.

KEYWORDS alcohol, chronic liver disease, cirrhosis, heavy drinking

See Editorial on Page 1281

1 | INTRODUCTION liver conditions, and there are 170 000 annual deaths from liver cirrhosis plus another 47 000 annual deaths from liver cancer.2 Chronic liver disease represents a major health burden worldwide.1,2 Alcohol is the strongest risk factor for liver cirrhosis. At the popula- In Europe alone, an estimated 29 million people suffer from chronic tion level, there is a clear dose-response relationship between alcohol consumption and the incidence of liver cirrhosis.3 However, only 15%- 20% of heavy alcohol drinkers develop liver cirrhosis.4-6 Therefore, aver-

Abbreviations: GGT, gamma-glutamyl transferase; HR, hazards ratio. age alcohol exposure alone is insufficient to explain alcohol use as a risk

Liver International. 2017;37:1373–1381. wileyonlinelibrary.com/journal/liv © 2017 John Wiley & Sons A/S. | 1373 Published by John Wiley & Sons Ltd 1374 | ÅBERG et al. factor for liver disease. Individual susceptibility to alcoholic liver disease and the effect of alcohol in the context of co-existent risk factors for liver Key points disease remain incompletely understood. The pattern of alcohol intake • At the population level, alcohol has a dose-related asso- and spirit type has been suggested to modify the risk of liver disease. ciation with risk of chronic liver disease, but individual The phenomenon of drinking too much too fast, termed binge drink- risk factors, such as binge drinking, remains unclear. ing, is growing in Western countries, especially in the UK and northern • We found that binge drinking (≥5 alcohol drinks per occa- 7 Europe. Although there is no standard definition of binge drinking, the sion) at least once monthly is associated with an elevated most widely used definition is the consumption of five or more drinks risk of incident liver disease independently of average on a single occasion or day, or sometimes modified to ≥5 drinks for men alcohol consumption. 7-10 and ≥4 drinks for women in about 2 hours. Around 20% of adults in • Binge drinking seems particularly harmful when the meta- 9 Europe and 17% in USA report binge drinking up to once a week. bolic syndrome co-exists. Most epidemiological studies of liver outcomes analyse average • Binge drinking should be considered in the assessment of alcohol intake; fewer specifically address binge drinking, and few of alcohol use and in the assessment of patients with liver these have been conducted in countries with a high prevalence of disease. binge drinking. It is widely acknowledged that the common quantity- frequency measures of alcohol intake used in most studies fail to capture a habit of weekend binging superimposed on top of regular low intake on weekdays.7,8,11 informed consent for future registry linkage. The Epidemiology Ethics Animal studies suggest that binge drinking alone induces liver Committee of the Helsinki and Uusimaa Hospital Region approved the damage and increases liver injury in chronically alcohol-exposed liv- Health 2000 Study protocol, and all participants provided signed in- ers,9,12,13 but human data are limited. Several authors have noted the formed consent. need for further study to define the impact of binge drinking on the Respondents were asked to report how often they consumed risk of liver disease,9,10,12,14-16 and recent guidelines prioritise this alcoholic beverages during the previous 12 months and the average specific research topic.8 amount they consumed per week during the previous month. There Finland has a drinking culture that is characterised by irregular were separate questions for consumption of beer, wine and spirits. heavy drinking, typically concentrated on weekends and outside of Average alcohol consumption (g/d) and beverage-specific consump- mealtimes, and it has lower proportional wine consumption than in tion were calculated based on these data. Participants were also asked southern European countries.17 Metabolic risk factors are common if they had been abstinent their entire life (lifetime abstainer) or had in Finland, but the prevalence of chronic viral hepatitis is very low. used alcohol earlier and then stopped (current abstainer). When a Finland has one of the highest liver-mortality rates in Europe, and liver subject reported that a specific beverage, that is, beer, wine or spir- deaths are rising.2 There is a paucity of longitudinal Finnish studies its, constituted at least 50% of their total alcohol consumption, it was that look at liver-related outcomes. considered the predominant beverage type. We analysed the impact of binge drinking on the risk of liver- Binge drinking was defined as drinking five or more alcohol drinks related outcomes in a Finnish cohort that is representative of the per occasion, in line with previous studies.16,19 Respondents reported entire Finnish population with long-term follow-up. the number of times during the last 12 months that they consumed five or more drinks per occasion. A standard drink in Finland is considered to have 12 g ethanol. Educational level was classified as basic, 2 | MATERIALS AND METHODS secondary or higher. The question about exercise asked how often the subject performs moderate-intensity or high-intensity physical exer- Baseline data were extracted from the Health 2000 Study, which was cise for at least half an hour. a multidisciplinary epidemiologic survey in Finland that involved a re- Metabolic syndrome was defined according to the Joint Interim gional two-stage stratified cluster sampling procedure to ensure that Statement criteria,20 which require three out of five metabolic com- the sample was representative of the entire Finnish population. The ponents for a diagnosis. These components include: waist circum- Health 2000 Study was conducted in 2000-2001, was coordinated ference ≥94 cm for men and ≥80 cm for women; serum triglycerides by the National Public Health Institute (now known as the National ≥1.7 mmol/L; serum HDL-cholesterol <1.0 mmol/L for men and Institute for Health and Welfare), and originally comprised 8028 <1.3 mmol/L for women; blood pressure ≥130/85 mm Hg or anti- adults aged 30 years and over. Data were collected via structured hypertensive treatment; and fasting serum glucose ≥5.6 mmol/L or home or telephone interviews, self-report questionnaires, clinical diabetes medication. Diabetes was defined either by fasting serum glu- measurements (height, weight, blood pressure), blood tests, and by cose ≥7.0 mmol/L, taking diabetes medication, or a diabetes diagnosis. clinical examination by a physician. The Health-2000-Study protocol Follow-up data for hospitalisations were obtained from the is described in detail elsewhere.18 National Hospital Discharge Register, which covers all hospitalisa- Of the original sample of 8028 adults, 6771 subjects (84%) were tions in Finland starting in 1969. Data for cancers were obtained interviewed at their homes or at an institution and provided signed from the Finnish Cancer Registry, and vital status and cause of death ÅBERG et al. | 1375 data were obtained from Statistics Finland, which systematically col- and the metabolic syndrome were tested by a Cox model. Finally, we lects data about the deaths of all Finnish citizens. In Finland, each performed Cox modelling using alternate outcome variables, namely person who dies is by law assigned a cause of death (in accordance liver-related mortality, overall mortality, or a combined endpoint of with the International Classification of Diseases) to the official any liver event or death from any cause. P values <.05 were consid- death certificate, issued by the treating physician based on medi- ered statistically significant. Data were analysed with R software ver- cal or autopsy evidence, or forensic evidence when necessary; the sion 3.2.5 or SPSS version 23 (Armonk, NY, USA). death codes are then verified by medical experts at the register and recorded according to systematic coding principle. One or several ICD-diagnoses are assigned to each hospitalisation at discharge; 3 | RESULTS these diagnosis codes are systematically recorded in the National Hospital Discharge Register. Data collection to all these registries is The study cohort comprised 6366 participants that were followed-up obligatory and general quality is consistent and complete. Linkage until death or until the end of the study. The mean follow-up time was was performed using the unique personal identifiers assigned to all 11.4 years (SD 3.3, range 0-13 years; Table 1). Of the participants, Finnish residents. 4856 (76%) reported alcohol use (any) at baseline, 1156 (18%) were The primary end point event of this study was a first hospitalisa- lifetime abstainers, and 354 (6%) were current abstainers. Of the cur- tion due to liver disease or liver-related death or a diagnosis of (pri- rent alcohol users, at baseline 320 (7%) reported weekly binge drink- mary) liver cancer, whichever came first. In line with previous studies, ing, and 889 (18%) reported monthly binging. The average number of liver disease was defined as ICD8/9 codes 570-573 and ICD10 codes annual binge drinking occasions was 12.6 (SD 31.7), with an average K70-K77 and C22.21,22 Liver-related death was defined as having one of 100 occasions (SD 73.9) for weekly binge drinkers, 24.7 (SD 11.5) of these as the underlying cause of death. Patients were followed-up for monthly binge drinkers, and 1.9 (SD 2.8) for subjects who reported for deaths and hospitalisations until December 2013 and for cancers less frequent binging. until December 2012. All of the baseline demographic variables we tested showed statisti- We excluded participants with records showing that one of the cally significant differences among the groups that were based on binge study endpoints (n=37) occurred before study baseline and partic- drinking frequency (Table 1). However, these differences were generally ipants with missing information for alcohol intake or binge drinking fairly small, except for male predominance, smoking and heavy smoking, (n=366). In this latter group, there were four liver events during fol- amount of alcohol consumed, and gamma-glutamyl transferase (GGT) low-up. The final study cohort therefore comprised 6366 subjects. levels, which were notably different for the frequent binge drinkers compared to the other groups (Table 1). Average alcohol consumption was 44.4 g/d for weekly binge drinkers, 25.5 g/d for monthly bingers, and 2.1 | Statistical analysis 8.3 g/d for subjects who reported less frequent binging. In the group For comparing groups, we used chi-square, Mann-Whitney, or Kruskall- of current drinkers, there was a moderate correlation between average Wallis tests as appropriate. We calculated Spearman’s correlation daily alcohol consumption and annual binge drinking frequency (r=.60, coefficient to examine the magnitude of rank correlation between P<.001), both for men (r=.56, P<.001) and women (r=.51, P<.001). average daily alcohol consumption and binge drinking frequency. During follow-up, 84 people experienced a liver decompensation We used Cox proportional hazards models to test the relationship event, with the first liver event occurring a mean of 6.4 years (SD 3.7, of binge drinking frequency with the risk of liver disease. We first ex- range 0-12 years) after baseline. The first recorded liver event was amined this relationship in subjects who reported current drinking at hospitalisation in 60 cases, liver cancer in 9 and liver-related death baseline using the penalised spline smoothing method23 and adjusted in 15. Of the liver events, 15 occurred in weekly binge drinkers, 20 for age and average daily alcohol consumption. We then catego- in monthly binge drinkers, 33 in subjects who reported binge drink- rised binge drinking frequency as “weekly” (52 or more annual binge ing less often, 11 in lifetime abstainers, and 5 in current abstainers. drinking episodes), “monthly” (12-51 annual binge drinking episodes), Unadjusted cumulative incidence of liver events by binge-drink sub- and “less often” (1-11 annual binge drinking episodes). We included group is shown in Fig. S1. lifetime abstainers for comparison as a separate subgroup. We then used sequential Cox models with varying levels of control to examine 3.1 | Binge drinking and liver events the role of various confounders in attenuating the hazard from binge drinking frequency. All analyses were adjusted for age and average Figure 1 shows the association between the number of annual binge daily alcohol consumption. The proportional hazard assumptions of drinking episodes and the relative risk of liver events for current drink- the Cox model were tested for each covariate by visually examining ers at baseline after adjustment for average daily alcohol consumption the parallelism of stratified survival curves, but no violations were de- and age. When considered as categorical variables, both weekly and tected. Subgroup analyses were performed according to gender, age, monthly binge drinking at baseline were significantly associated with average daily alcohol consumption (moderate drinking: 20-50 g/d), future liver events compared to less frequent binging, both in univari- predominant beverage type, smoking status and presence of the ate analysis and after adjusting for average daily alcohol intake, age metabolic syndrome. Interactions between weekly binge drinking and gender (Table 2). 1376 | ÅBERG et al.

TABLE 1 Baseline demographics of the study cohort and according to binge drinking category

Binge drinking frequency at baseline No alcohol use at baseline

All Lifetime Current P among individuals Weekly Monthly Less often abstainers abstainers groups

Persons 6366 320 889 3647 1156 354 Age (y), mean (SD) 53 (15) 47 (11) 45 (10) 51 (14) 67 (16) 60 (16) <.001 Women, n (%) 3497 (55) 56 (18) 233 (26) 2126 (58) 946 (82) 136 (38) <.001 Marital status, n (%) Married/partnership 4335 (68) 206 (65) 648 (73) 2700 (74) 569 (49) 212 (60) <.001 Single 2017 (32) 112 (35) 236 (27) 941 (26) 586 (51) 142 (40) Education, n (%) Basic 2544 (40) 95 (30) 240 (27) 1227 (34) 771 (67) 212 (60) <.001 Secondary 2001 (32) 130 (41) 366 (41) 1182 (33) 237 (21) 86 (24) Higher 1799 (28) 94 (29) 278 (32) 1229 (33) 143 (12) 55 (16) Employment status, n (%) Part- or full-time employed 3510 (55) 219 (69) 687 (78) 2239 (62) 266 (23) 99 (28) <.001 Retired 2170 (34) 55 (17) 91 (10) 1012 (28) 806 (70) 206 (58) Other 673 (11) 44 (14) 106 (12) 391 (10) 83 (7) 49 (14) Diabetes, n (%) 620 (10) 18 (6) 57 (6) 270 (7) 204 (18) 71 (20) <.001 Hyperglycaemia or diabetes, n (%) 2284 (36) 128 (40) 312 (35) 1155 (32) 541 (48) 148 (43) <.001 Elevated triglycerides (>1.7 mmol/L), n (%) 2135 (34) 139 (43) 330 (37) 1051 (29) 472 (42) 143 (41) <.001 Reduced HDL cholesterola, n (%) 2194 (35) 80 (25) 231 (26) 1176 (32) 559 (49) 148 (43) <.001 Increased waist circumferenceb, n (%) 4227 (68) 228 (72) 555 (63) 2344 (65) 854 (79) 246 (73) <.001 Elevated blood pressurec, n (%) 3995 (63) 217 (68) 533 (60) 2152 (59) 879 (78) 147 (49) <.001 Metabolic syndromed 2857 (45) 154 (48) 386 (43) 1435 (40) 704 (62) 178 (51) <.001 Body mass index, n (%) < 25 kg/m2 2392 (38) 89 (28) 307 (34) 1498 (41) 372 (33) 126 (36) <.001 25-29.9 kg/m2 2529 (40) 161 (50) 389 (44) 1407 (39) 450 (40) 122 (35) ≥30 kg/m2 1410 (22) 70 (22) 193 (22) 733 (20) 313 (27) 101 (29) Exercise, n (%) At least 2 times a week 3614 (58) 159 (50) 506 (57) 2184 (60) 576 (53) 189 (55) <.001 2-4 times a month 1695 (27) 109 (34) 279 (32) 1012 (28) 216 (20) 79 (23) Less often 954 (15) 52 (16) 101 (11) 429 (12) 295 (27) 77 (22) Smoking, n (%) Current 1662 (26) 178 (56) 418 (47) 888 (24) 69 (6) 109 (31) <.001 Former 1387 (22) 69 (22) 211 (24) 853 (24) 114 (10) 140 (40) Never 3288 (52) 71 (22) 255 (29) 1896 (52) 964 (84) 102 (29) Cigarettes per day, n (%) None 3476 (72) 85 (34) 312 (48) 1990 (73) 972 (94) 117 (54) <.001 1-9 286 (6) 16 (7) 55 (9) 182 (7) 19 (2) 14 (7) 10-19 532 (11) 37 (15) 122 (19) 313 (11) 26 (3) 34 (16) ≥20 563 (11) 107 (44) 159 (24) 234 (9) 13 (1) 50 (23) Alcohol consumption (g/d), mean (SD) 10.7 (20.9) 44.4 (37.3) 25.5 (26.8) 8.3 (15.8) 0 0 <.001 Predominant beverage type (>50% of total alcohol consumption), n (%) Wine 925 (22) 61 (19) 102 (12) 723 (25) <.001 Beer 1913 (46) 147 (47) 414 (49) 1315 (46) .047

(Continues) ÅBERG et al. | 1377

TABLE 1 (Continued)

Binge drinking frequency at baseline No alcohol use at baseline

All Lifetime Current P among individuals Weekly Monthly Less often abstainers abstainers groups

Spirits 1104 (26) 80 (25) 254 (30) 748 (26) .050 Gamma-glutamyltransferase (U/L), mean 37 (47) 65 (85) 48 (47) 34 (45) 30 (32) 33 (33) <.001 (SD)

Statistical differences between groups were calculated with the chi square test or Kruskall-Wallis test as appropriate. a<1.0 mmol/L for men and <1.3 mmol/L for women. b≥94 cm for men and ≥80 cm for women. cBlood pressure ≥130/85 mm Hg or antihypertensive medication. dAccording to the Joint Interim Statement Criteria,20 defined in the presence of three of the following five metabolic syndrome components: hyperglycaemia, elevated triglycerides, reduced HDL, increased waist circumference, and elevated blood pressure.

In subgroup analyses that were adjusted for age and average daily alcohol intake, weekly binge drinking was significantly associated with liver events for the following subgroups: men, women, people aged 30-55 years at baseline, people who consumed 20-50 g of alcohol/d, predominantly spirit drinkers, predominantly beer drinkers, current smokers, non-smokers or irregular smokers, and those with the metabolic syndrome (Table 3). Similarly, monthly binge drinking was significantly associated with liver events for women, people aged 30-55 years at baseline, people who consumed 20-50 g of alcohol/d, non-smokers or irregular smokers, and people with the metabolic syndrome (Table 3). Binge drinking status was not significantly associated with liver events for predominantly wine drinkers, but the number of liver events in this subgroup was too small (n=2) to draw conclusions.

3.3 | Interactions between weekly binge drinking and the metabolic syndrome FIGURE 1 The association between number of binge drinking (≥5 Supra-additive interactions emerged between binge drinking and the drinks per occasion, standard drink 12 g ethanol) episodes per year metabolic syndrome for the risk of liver events after adjustment for and risk for decompensated liver disease in comparison to subjects reporting to consume minimal amounts of alcohol. Current and age and average alcohol consumption (Figure 2). lifetime abstainers at baseline were excluded. Dashed lines represent 95% confidence intervals to the relative risk (hazards ratio) estimate. Analysis is adjusted for average alcohol consumption (g/d) and age 3.4 | The effects of binge drinking on mortality and combined endpoints

Weekly binge drinking was significantly associated with liver-related mortality, overall mortality, and the combined endpoint of a liver event 3.2 | Multivariate and subgroup analyses or death from any cause. Monthly binge drinking was significantly In Cox regression analyses that were adjusted for age and average associated only with the combined endpoint, but showed a trend to- daily alcohol intake and that used binge drinking less often than once wards a significant association (P=.051) with overall mortality (Table 4). monthly as a reference, the effects of weekly and monthly binge drinking on liver events remained significant when separately adjusted for any predominant beverage type (wine, beer or spirits), frequency of 4 | DISCUSSION alcohol intake, smoking status, smoking amount, frequency of exercise, waist circumference, diabetes, metabolic syndrome, triglycerides, The primary finding of this study is that a pattern of binge drinking HDL, blood pressure, GGT, or education level (Tables 2 and S1). In alcohol was associated with an elevated risk of liver disease independ- these analyses, the HRs for weekly binge drinking ranged from 3.0 to ent of average alcohol intake. Furthermore, the risk of liver events 4.5, and the HRs for monthly binge drinking ranged from 2.0 to 3.0 increased as a function of binge drinking frequency, with the clearest (Table S1) compared to binging less often. risk seen for subjects who binge drink at least weekly. 1378 | ÅBERG et al.

In our series of multivariate and subgroup analyses that were designed to reduce the possibility of confounding, binge drinking retained .001 .032 .16 .011 P <.001 <.001 its significant association with an elevated risk of liver disease, especially when binging occurring at least weekly. Average alcohol intake tended

1.002 to be higher for frequent binge drinkers compared to non-binge drink- 7.28 3.88 1.07 3.24 0.66 ers or rare binge drinkers, but the harmful impact of binge drinking re- 95% CI 1.66- 1.06- 1.001- 1.02- 0.83- 0.04- mained significant after excluding regular heavy drinkers. Binge drinking frequency was also associated with metabolic and cardiovascular risk factors at baseline, but variation was fairly small, so residual confounding n.c. Model 4 HR 3.47 2.03 Ref 1.002 1.04 Ref 1.64 0.16 should be minimal after adjusting for these factors in regression analyses. At least three mechanisms could explain the epidemiological as- .002 .027 .73 .084 P <.001 <.001 sociation between binge drinking and elevated liver risk. Firstly, binge drinking itself could harm the liver above and beyond the harm due to average alcohol intake. Secondly, binge drinkers may have a higher 1.003 6.11 3.65 1.93 1.06 2.80 propensity to develop alcohol-use disorder later on and escalate their alcohol consumption.24 Thirdly, binge drinking might simply reflect 95% CI 1.48- 1.08- 0.39- 1.001- 1.02- 0.94- higher alcohol exposure that standard quantity-frequency approaches to measuring alcohol consumption fail to capture.11 Indeed, most Model 3 HR 3.01 1.99 Ref 0.87 1.002 1.04 Ref 1.62 studies measure alcohol consumption using a quantity-frequency approach in which the respondent reports the frequency of drinking .001 .007 .47 during a specific period and, separately, the typical amount of alcohol P <.001 <.001 consumed on the days that they drank. This approach distinguishes between two drinking patterns that result in equivalent total con-

1.003 sumption, that is, frequently drinking small amounts vs infrequently 6.94 4.10 1.64 1.06 drinking larger quantities. However, this approach may fail to capture 95% CI 1.71- 1.25- 0.34- 1.001- 1.02- variations in drinking around a reported typical quantity, such as with irregular binge drinking, and therefore underestimate the true total alcohol consumption over a specific time period. For example, a person Model 2 HR 3.45 2.26 Ref 0.75 1.002 1.04 who drinks two drinks daily and a person who drinks the same amount plus ten additional drinks on Saturday might report the same typical .003 .044 .55 P <.001 quantity of two drinks per occasion. Thus, there is a need to explicitly address the issue of binge drinking. The study strengths include its prospective design, the nationally 1.003 5.61 3.21 2.63 representative cohort, complete follow-up data, high validity of liver 95% CI 1.41- 1.02- 0.60- 1.001- diagnoses, and the ability to control for several potential confounders. Study outcomes represented clinically relevant liver disease. We chose a holistic approach in terms of liver outcomes by not restricting the Model 1 HR 2.81 1.81 Ref 1.25 1.002 outcomes to alcoholic cirrhosis, which eventually depends on subjective classification. Moreover, with the rising rates of central obesity .002 .48 and the metabolic syndrome, it seems increasingly likely that there are P <.001 mixed aetiologies underlying chronic liver disease.21,25 We therefore believe that our approach presents an improved picture of the true 9.87 4.28 2.53 burden of liver disease related to alcohol consumption. 95% CI 2.91- 1.41- 0.65- There is a lack of consensus with regards to the definition of binge drinking in terms of the minimum amount of alcohol intake per epi- Unadjusted HR 5.36 2.46 Ref 1.28 sode and the period of time that is considered a single event. Some propose a threshold of four drinks for women10; we were unable to make this distinction. UK guidelines propose a threshold of 8 units for men and 6 units for women.26 Given that a standard alcohol unit in the UK equals 8 g ethanol, our five drinks of 12 g each is actually the equivalent of 7-8 UK units. Women Men From the initial dataset, 366 subjects were excluded due to miss- consumption drinker Binge drinking frequency Weekly Monthly Less often Lifetime abstainers Average daily alcohol Age Gender Predominant wine TABLE 2 Risk of liver events according to binge drinking frequency at baseline by multivariate Cox regression analyses n.c., not calculable. ing data for relevant alcohol measures. In this group, there were four ÅBERG et al. | 1379

TABLE 3 Risk of liver events according to binge drinking frequency at baseline in various subgroups. Analyses are adjusted for age and average daily alcohol consumption

Binge drinking frequency

Weekly Monthly Less often Lifetime abstainers

Subgroup HR 95% CI P HR 95% CI P HR HR 95% CI P

Men 2.34 1.08-5.06 .03 1.53 0.77-3.04 .22 Ref 0.36 0.05-2.69 .32 Women 6.82 1.31-35.6 .023 3.84 1.14-12.9 .030 Ref 1.19 0.42-3.39 .74 Ages 30-55 y at baseline 3.44 1.48-7.99 .004 2.13 1.06-4.29 .034 Ref 1.63 0.55-4.82 .38 Average alcohol consump- 5.69 1.58-20.5 .008 4.25 1.04-17.3 .044 Ref tion 20-50 g/d Predominant wine drinker n.c. .99 1.65 0.07-38.7 .76 Ref Predominant spirit drinker 3.44 1.11-10.6 .032 2.33 0.94-5.81 .069 Ref Predominant beer drinker 5.62 2.14-14.7 <.001 1.97 0.73-5.31 .18 Ref Current smoker 2.60 1.02-6.62 .045 0.72 0.25-2.06 .54 Ref 0.83 0.11-6.45 .86 Non-smoker or irregular 5.20 1.48-18.3 .010 4.79 1.78-12.9 .002 Ref 1.16 0.42-3.23 .78 smoker Metabolic syndrome 4.29 1.80-10.2 .001 3.17 1.56-6.47 .001 Ref 0.64 0.23-1.79 .39 No metabolic syndrome 2.12 0.66-6.86 .21 0.75 0.21-2.70 .66 Ref 1.09 0.33-3.55 .89 n.c., not calculable.

FIGURE 2 Relative risks (hazards ratios) for the contribution of weekly binge drinking, the metabolic syndrome (MetS), and their combination (interaction) to the risk for decompensated liver disease in comparison to subjects without the metabolic syndrome who reported no binge drinking or binge drinking less often than once monthly (baseline risk). Analysis was by Cox regression and adjusted for age and average alcohol consumption liver events (1%), which is exactly the same as in the study cohort. intention was to detect serious liver events (decompensation, HCC or Therefore, we do not consider these exclusions a source of bias. death) that require hospital contact. We consider the lack of informa- However, one major limitation of our study is that it assumed that al- tion on hepatitis C status a minor limitation because of its low preva- cohol drinking habits and other lifestyle factors remained the same lence in Finland (0.3%). over time. Animal studies suggest that binge drinking can induce liver injury Although the registries used for the study outcome data are con- by several mechanisms (reviewed in refs. [9,12,13]). However, human sidered of high standard internationally, registry data were not a priori studies are scarce. A 2010 meta-analysis that addressed alcohol as a intended for the specific study purpose. Limitations with regards to risk factor for liver cirrhosis failed to draw any conclusions regarding completeness of data cannot be excluded. We acknowledge that the the issue of binge drinking due to a lack of such studies.14 study does not capture compensated cirrhosis without hospitalisa- Three case-control studies that attempted to retrospectively as- tions, as baseline and follow-up imaging and liver-specific laboratory sess previous alcohol drinking habits for patients with alcoholic cirrho- parameters (other than GGT) were unavailable. However, our specific sis generally reported that the patients had consumption patterns that 1380 | ÅBERG et al.

TABLE 4 Association between binge drinking frequency and hepatic steatosis. Furthermore, animal studies show that binge drink- various outcomes by Cox regression analysis adjusted for average ing aggravates hepatic oxidative stress and promotes the development alcohol use, age, and gender of steatohepatitis from simple steatosis.9 Binge drinking might thus be Binge drinking frequency HR 95% CI P particularly harmful for people with a fatty liver. Information about binge drinking is important for health education Liver-related mortality (n=36) purposes and for planning future programs that aim to prevent liver Weekly 3.00 1.11-8.15 .031 disease. Although the individual-level harm associated with chronic Monthly 1.31 0.50-3.41 .59 heavy drinking is probably greater than the harm associated with binge Less often Ref drinking, there are more binge drinkers, suggesting that the attribut- Lifetime abstainers n.c. able risks associated with binging may exceed those associated with Any liver event or death from any cause (n=1288) chronic heavy drinking. Weekly 1.82 1.34-2.45 <.001 At the level of the individual, the failure to account for binge drink- Monthly 1.36 1.06-1.75 .015 ing when assessing alcohol consumption may lead to underestimation Less often Ref of the true level of alcohol exposure and alcohol-related harm. The Lifetime abstainers 1.39 1.20-1.61 <.001 currently recommended “safe limits” for weekly or daily alcohol con- Overall mortality (n=1257) sumption may not apply to people who binge drink. Weekly 1.73 1.26-2.37 .001 Our findings together with previous evidence seem to suggest particular harm from binge drinking for people with the metabolic syn- Monthly 1.30 1.00-1.68 .051 drome. Binge drinking might drive progression of fatty liver disease, even Less often Ref if average alcohol intake is low. If this proves to be correct, it would blur Lifetime abstainers 1.37 1.18-1.58 <.001 current distinction between alcoholic and non-alcoholic liver disease. This n.c., not calculable. is of concern given that the prevalence of both the metabolic syndrome and binge drinking are rising. Interventions may need to be designed to by current definition could be regarded as daily or near-daily binge target the behaviour associated with engaging in binge drinking. drinking.27-29 A Danish prospective study of heavy drinkers found a We observed reduced all-cause mortality among light drinkers higher risk of alcoholic cirrhosis for daily drinkers than periodic drink- compared to lifetime abstainers, who exhibited a risk of death compa- ers (gaps of at least 1 day in which there was no alcohol consump- rable to monthly binge drinkers. All-cause mortality was heavily driven tion).30 However, even subjects showing the lowest tier of drinking, by cardiovascular disease, and a beneficial effect of light drinking on 5-9 drinks per day, can by the current definition be considered daily cardiovascular risk parallels previous evidence.33 However, intermit- or near-daily binge drinkers. These studies comprised selected cohorts tent binge drinking eliminates any such benefit.33 and relied on a subjective classification of cirrhosis into alcoholic cir- In conclusion, a binge-drinking pattern of alcohol intake seems rhosis. No conclusions can be drawn from the studies regarding the to exacerbate the dose-dependent risk of alcohol-associated liver impact of binge drinking on the overall burden of liver morbidity and disease. This is of particular concern in people with the metabolic mortality in the general population. syndrome, as both binge drinking and the metabolic syndrome are In agreement with our findings, a large prospective US study increasingly prevalent. comprising a nationally representative cohort found a progressive increase in the risk of liver disease with an increasing frequency of ACKNOWLEDGEMENT binge drinking (men≥5 drinks/d, women≥4 drinks/d).31 The effect became significant for weekly binge drinking, although monthly and We would like to thank all participants of, and persons involved in the less-than- monthly binging also showed elevated but non-significant conduct of the Health 2000 study. risk estimates. However, the study relied on self-reporting of a diagnosis of any type of liver disease, and the follow-up was relatively short (approximately 3 years). CONFLICT OF INTEREST A Swedish study of 71 patients with biopsy-confirmed non- The authors do not have any disclosures to report. alcoholic fatty liver disease (weekly alcohol intake <140 g/d), found that binge drinking (men 60 g per occasion, women 48 g) as infrequently as once a month was associated with liver-fibrosis progres- REFERENCES 32 sion. In this regard, the effect of binge drinking was stronger than the 1. Rehm J, Samokhvalov AV, Shield KD. Global burden of alcoholic liver effect of average alcohol consumption. This parallels our findings. We diseases. J Hepatol. 2013;59:160‐168. found the impact of binge drinking on incident liver disease greater 2. Blachier M, Leleu H, Peck-Radosavljevic M, Valla DC, Roudot-Thoraval F. The burden of liver disease in Europe: a review of available epidemi- for subjects with the metabolic syndrome, the majority of whom also ological data. J Hepatol. 2013;58:593‐608. 16 have a fatty liver. Recently, Lau et al. reported that binge drinking 3. Norstrom T, Ramstedt M. Mortality and population drinking: a review and overweight/obesity showed supra-additive likelihood of having of the literature. Drug Alcohol Rev. 2005;24:537‐547. ÅBERG et al. | 1381

4. Gramenzi A, Caputo F, Biselli M, et al. Review article: alcoholic 21. Hart CL, Morrison DS, Batty GD, Mitchell RJ, Davey Smith G. Effect of liver disease–pathophysiological aspects and risk factors. Aliment body mass index and alcohol consumption on liver disease: analysis of Pharmacol Ther. 2006;24:1151‐1161. data from two prospective cohort studies. BMJ. 2010;340:c1240. 5. Mann RE, Smart RG, Govoni R. The epidemiology of alcoholic liver 22. Unalp-Arida A, Ruhl CE. Noninvasive fatty liver markers pre- disease. Alcohol Res Health. 2003;27:209‐219. dict liver disease mortality in the U.S. population. Hepatology. 6. Mills SJ, Harrison SA. Comparison of the natural history of alco- 2016;63:1170‐1183. holic and nonalcoholic fatty liver disease. Curr Gastroenterol Rep. 23. Roshani D, Ghaderi E. Comparing smoothing techniques for fitting 2005;7:32‐36. the nonlinear effect of covariate in cox models. Acta Inform Med. 7. Mathurin P, Bataller R. Trends in the management and burden of alco- 2016;24:38‐41. holic liver disease. J Hepatol. 2015;62:S38‐S46. 24. Piasecki TM, Robertson BM, Epler AJ. Hangover and risk for alcohol 8. European Association for the Study of Liver. EASL clinical practi- use disorders: existing evidence and potential mechanisms. Curr Drug cal guidelines: management of alcoholic liver disease. J Hepatol. Abuse Rev. 2010;3:92‐102. 2012;57:399‐420. 25. Ruhl CE, Everhart JE. Joint effects of body weight and alcohol on el- 9. Llerena S, Arias-Loste MT, Puente A, Cabezas J, Crespo J, Fabrega E. evated serum alanine aminotransferase in the United States popula- Binge drinking: burden of liver disease and beyond. World J Hepatol. tion. Clin Gastroenterol Hepatol. 2005;3:1260‐1268. 2015;7:2703‐2715. 26. NHS Choices. http://www.nhs.uk/Livewell/alcohol/Pages/ 10. Zakhari S, Li TK. Determinants of alcohol use and abuse: impact Bingedrinking.aspx. Accessed February 1, 2017. of quantity and frequency patterns on liver disease. Hepatology. 27. Hatton J, Burton A, Nash H, Munn E, Burgoyne L, Sheron N. Drinking 2007;46:2032‐2039. patterns, dependency and life-time drinking history in alcohol-related 11. Stahre M, Naimi T, Brewer R, Holt J. Measuring average alcohol con- liver disease. Addiction. 2009;104:587‐592. sumption: the impact of including binge drinks in quantity-frequency 28. Stokkeland K, Hilm G, Spak F, Franck J, Hultcrantz R. Different drink- calculations. Addiction. 2006;101:1711‐1718. ing patterns for women and men with alcohol dependence with and 12. Mathurin P, Deltenre P. Effect of binge drinking on the liver: an alarm- without alcoholic cirrhosis. Alcohol Alcohol. 2008;43:39‐45. ing public health issue? Gut. 2009;58:613‐617. 29. Parrish KM, Dufour MC, Stinson FS, Harford TC. Average daily alcohol 13. Shukla SD, Pruett SB, Szabo G, Arteel GE. Binge ethanol consumption during adult life among decedents with and without cir- and liver: new molecular developments. Alcohol Clin Exp Res. rhosis: the 1986 National Mortality Followback Survey. J Stud Alcohol. 2013;37:550‐557. 1993;54:450‐456. 14. Rehm J, Taylor B, Mohapatra S, et al. Alcohol as a risk factor for liver 30. Kamper-Jorgensen M, Gronbaek M, Tolstrup J, Becker U. Alcohol and cirrhosis: a systematic review and meta-analysis. Drug Alcohol Rev. cirrhosis: dose–response or threshold effect? J Hepatol. 2004;41:25‐30. 2010;29:437‐445. 31. Dawson DA, Li TK, Grant BF. A prospective study of risk drinking: at 15. Brandish E, Sheron N. Drinking patterns and the risk of serious liver risk for what? Drug Alcohol Depend. 2008;95:62‐72. disease. Expert Rev Gastroenterol Hepatol. 2010;4:249‐252. 32. Ekstedt M, Franzen LE, Holmqvist M, et al. Alcohol consumption is 16. Lau K, Baumeister SE, Lieb W, et al. The combined effects of alcohol associated with progression of hepatic fibrosis in non-alcoholic fatty consumption and body mass index on hepatic steatosis in a general liver disease. Scand J Gastroenterol. 2009;44:366‐374. population sample of European men and women. Aliment Pharmacol 33. Roerecke M, Rehm J. Alcohol consumption, drinking patterns, and Ther. 2015;41:467‐476. ischemic heart disease: a narrative review of meta-analyses and a 17. Makela P, Tigerstedt C, Mustonen H. The Finnish drinking cul- systematic review and meta-analysis of the impact of heavy drinking ture: change and continuity in the past 40 years. Drug Alcohol Rev. occasions on risk for moderate drinkers. BMC Med. 2014;12:182. 2012;31:831‐840. 18. Aromaa A, Koskinen S. Health and Functional Capacity in Finland. Baseline results of the Health 2000 health examination survey. SUPPORTING INFORMATION Publications of National Public Health Institute, Series B 12/2004. Helsinki, Finland: NPHI; 2004:71. ISBN 951-740-299-6 (print), ISBN Additional Supporting Information may be found online in the sup- 951-740-300-6 (pdf). porting information tab for this article. 19. Naimi TS, Brewer RD, Mokdad A, Denny C, Serdula MK, Marks JS. Binge drinking among US adults. JAMA. 2003;289:70‐75. 20. Alberti KG, Eckel RH, Grundy SM, et al. Harmonizing the metabolic How to cite this article: Åberg F, Helenius-Hietala J, Puukka P, syndrome: a joint interim statement of the International Diabetes Jula A. Binge drinking and the risk of liver events: Federation Task Force on Epidemiology and Prevention; National A population-based cohort study. Liver Int. 2017;37:1373–1381. Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; https://doi.org/10.1111/liv.13408 and International Association for the Study of Obesity. Circulation. 2009;120:1640‐1645.