Alcohol consumption appears to protect against non-alcoholic fatty liver disease

Authors


Dr Y. Iwasaki, Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, 2-5-1 Shikata-cho, Kita-ku, Okayama 700-8558, Japan.
E-mail: yiwasaki@cc.okayama-u.ac.jp

Abstract

Aliment Pharmacol Ther 2011; 33: 378–388

Summary

Background  Moderate alcohol consumption may have certain beneficial effects against non-alcoholic fatty liver disease, which is associated with metabolic syndrome.

Aim  To determine the association between drinking pattern and fatty liver in Japanese men and women.

Methods  A cross-sectional study was performed with health checkup data including information concerning alcohol consumption and ultrasonographic assessment of fatty liver.

Results  We analysed 4957 men and 2155 women without reported liver diseases (median age, 49 years). In men, 40% of nondrinkers and 28% of drinkers had fatty liver. Alcohol consumption was inversely associated with fatty liver (adjusted odds ratio, 0.54; 95% confidence interval, 0.46–0.63). The prevalence of fatty liver in each category of drinking frequency was 38% (1–3 days/week), 29% (4–6 days/week), and 24% (daily drinking); there was a significant inverse correlation between drinking frequency and the prevalence of fatty liver (< 0.001). In women, 16% of nondrinkers and 10% of drinkers had fatty liver. Drinking less than 20 g on 1–3 days/week was associated with low prevalence of fatty liver (adjusted odds ratio, 0.47; 95% confidence interval, 0.23–0.96).

Conclusions  Alcohol consumption appears to protect against non-alcoholic fatty liver disease.

Introduction

Many studies have demonstrated that moderate alcohol consumption promotes health. For example, alcohol consumption of a certain amount and frequency prevents the development of diabetes mellitus,1, 2 reduces the risk of coronary heart disease3, 4 and elevates high-density lipoprotein cholesterol and adiponectin levels.5–7 These healthy contributions of alcohol consumption appear equivalent to a reduction in the risk of developing metabolic syndrome. Improved insulin sensitivity by alcohol consumption5, 8, 9 may be responsible for ameliorating metabolic syndrome, since insulin resistance is a key component of metabolic syndrome.10

Non-alcoholic fatty liver disease (NAFLD) is also a feature of metabolic syndrome in the liver.11–16 Recent studies have suggested an inverse association between light to moderate alcohol consumption and the likelihood of NAFLD. An analysis of a limited number of morbidly obese subjects suggested that moderate alcohol consumption reduced the likelihood of non-alcoholic steatohepatitis.17 Other studies performed on a large, general population have also revealed an inverse association between alcohol18 or wine19 consumption and unexplained hypertransaminasemia. Furthermore, a cross-sectional study on a general population revealed an inverse association of light to moderate alcohol consumption with ultrasonographically determined fatty liver.20 These studies indicate that light to moderate alcohol consumption may have a beneficial effect on NAFLD.

On the other hand, excessive alcohol consumption causes alcoholic liver injury,21 and the molecular mechanisms of alcoholic fatty liver have been indicated.22 Therefore, the beneficial effect of alcohol consumption on the liver has not been sufficiently studied and the details remain unclear. Whether the amount (light to moderate), frequency or both of alcohol consumption are responsible for the beneficial effect on NAFLD development is unknown, although the association of drinking pattern with the risk of disease progression has been reported in other metabolic syndrome-related diseases, such as type 2 diabetes2 or coronary heart disease.3, 4 In addition, no studies have assessed the association between alcohol consumption and fatty liver in women alone. The previous reports on the inverse association between alcohol consumption and NAFLD involved only men18, 20 or a combination of men and women,17, 19 therefore those results may not apply to women alone due to sex differences in alcohol pharmacokinetics.23 Actually, there is a sex difference in the association of drinking pattern and the risk of coronary heart disease.4 The current study investigated the association of drinking pattern with fatty liver in men and in women, separately.

Materials and methods

Study population

The clinical and demographic characteristics and laboratory data were collected from 7894 Japanese subjects (5548 men and 2346 women) who underwent ultrasonography as a part of systematic health checkups at Junpukai Health Maintenance Center in 2003. The health checkups were periodically organised by employers of various workplaces, intended for all of their employees and conducted in a standardised manner.

The data included the age, sex, height, body weight, blood pressure, aspartate aminotransferase (AST, n = 7893), alanine aminotransferase (ALT, n = 7893), γ-glutamyltransferase (GGT, n = 7867), total cholesterol (n = 7893), triglycerides (n = 7838), high-density lipoprotein cholesterol (n = 7740), fasting plasma glucose (n = 7514), uric acid (n = 7739) and ultrasonographic assessment of fatty liver that were measured or recorded at the time of the health checkup. The data also included information on the alcohol consumption (n = 7864), smoking history (n = 7874), physical exercise (n = 7874) and medical history that were obtained by a questionnaire.

The study excluded 782 individuals who had any missing components of data (n = 630), reported liver diseases under treatment or follow-up (n = 141), or both (n = 11). Finally, 7112 eligible subjects were analysed (4957 men and 2155 women), with a median age of 49 years (interquartile range, 42–54).

Questionnaire

Current lifestyle-related information (alcohol consumption, smoking history and physical exercise) and medical history were obtained by a questionnaire. The drinking frequency was determined by the number of drinking days per week. The amount of beverage consumed per drinking day was recorded in units of sake, which is a traditional Japanese alcoholic beverage and contains 22 g of alcohol per unit (180 mL). Each amount of other beverages consumed was converted into units of sake (500 mL of beer, 240 mL of wine and 60 mL of liquor were equal to a unit of sake) and added up; the values were calculated to the first decimal place. The questionnaire also recorded the current smoking status, number of cigarettes smoked per day and period of smoking. The assessment also included the average exercise duration at one time (including sports, walking commute, bicycle commuting and farm work) and frequency of exercise per week. The complete medical history of each subject was recorded. In particular, the questionnaire addressed any treatment for hypertension, atherogenic dyslipidemia, diabetes mellitus, hyperuricemia and liver diseases (including viral hepatitis). Public health nurses interviewed subjects individually and confirmed each item on the questionnaire.

Assessment of fatty liver

Experienced technicians performed real-time ultrasonography to detect vascular blurring, deep attenuation and increased liver echotexture in comparison to the kidneys (liver-kidney contrast) to assess fatty infiltration.24 Specialists subsequently confirmed the validity of the findings and judged the presence of one or more findings as indicative of a fatty liver.

Variables

Obesity was defined as a body mass index ≥25 kg/m2, which is commonly used to define obesity in the Japanese population.25 Hypertension was defined as blood pressure at/above 130/85 mmHg or use of medications for hypertension. Atherogenic dyslipidemia was defined as triglycerides ≥1.7 mmol/L, high-density lipoprotein cholesterol <1.03 mmol/L, or use of medications for dyslipidemia. Glucose intolerance was defined as fasting plasma glucose ≥5.6 mmol/L or use of medications for diabetes mellitus. Hyperuricemia was defined as uric acid >450 μmol/L for men and >390 μmol/L for women or use of medications for hyperuricemia. Based on the suggested healthy range of ALT levels,26 we defined ALT elevation as ALT ≥30 IU/L for men and ≥19 IU/L for women. Smoking was divided into current smoking and current nonsmoking (including previous smoking). Regular exercise was defined as exercise performed at least once per week, regardless of exercise duration. A drinker was defined as an individual who reported current alcohol consumption. These parameters were treated as dichotomous variables.

Drinking frequency was classified into four categories: none, 1–3 drinking days per week, 4–6 drinking days per week and daily drinking. The amounts of beverages consumed per drinking day were converted into grammes of alcohol (22 g per unit of sake) and classified into five categories: none, 0.1–19.9 g, 20.0–39.9 g, 40.0–59.9 g and 60.0 g or more. Total amount of alcohol consumed per week was classified into five categories: none, 0.1–69.9 g, 70.0–139.9 g, 140.0–279.9 g and 280.0 g or more.

Statistical analysis

Continuous variables were expressed as the median with interquartile range. Categorical variables were compared using Fisher’s exact test. The Cochran–Armitage test for trend was used to confirm the linear trend of categorised groups. Logistic regression analyses were used to estimate the association between fatty liver and the presence of obesity, hypertension, atherogenic dyslipidemia, glucose intolerance, hyperuricemia, current smoking, regular exercise, alcohol consumption and each category of drinking. Multivariable logistic regression analyses were used to adjust the influence of alcohol consumption for other confounders that exhibited statistical significance in the individual analyses. Differences with < 0.05 were considered to be significant. The statistical analysis was performed with StatView for Windows version 5.0 (SAS Institute Inc., Cary, NC, USA) or R version 2.10.1 (R Foundation for Statistical Computing, Vienna, Austria).27

Ethical issues

The institutional review board of Okayama University Graduate School of Medicine, Dentistry, and Pharmaceutical Sciences approved this study.

Results

Association of alcohol consumption with fatty liver

Table 1 presents the subjects’ clinical and demographic characteristics and laboratory data. Table 2 presents the prevalence of metabolic syndrome-related factors (Table S3). Fatty liver was observed in 31% of men and 15% of women; alcohol consumption was reported in 74% of men and 29% of women. The prevalence of fatty liver was significantly lower in drinkers than in nondrinkers [28% vs. 40% in men (< 0.001) and 10% vs. 16% in women (< 0.001)]; alcohol consumption was inversely associated with fatty liver in men [odds ratio (OR), 0.59; 95% confidence interval (CI), 0.52–0.68] and in women (OR, 0.60; 95% CI, 0.45–0.80).

Table 1.   Subjects’ clinical and demographic characteristics and laboratory data
 Men (n = 4957)Women (n = 2155)
  1. AST, aspartate aminotransferase; ALT, alanine aminotransferase; GGT, γ-glutamyltransferase.

  2. All variables are expressed as median (interquartile range).

Age (years)49 (42–55)48 (41–53)
Body mass index (kg/m2)23.2 (21.4–25.1)21.3 (19.5–23.3)
Systolic blood pressure (mmHg)119 (108–133)114 (103–128)
Diastolic blood pressure (mmHg)76 (68–84)70 (63–80)
Amount of alcohol consumed in drinkers (g/drinking day)33.0 (22.0–44.0)17.6 (11.0–22.0)
Drinking frequency in drinkers (drinking days/week)7 (4–7)4 (2–6)
Total amount of alcohol consumed in drinkers (g/week)154.0 (99.0–308.0)55.0 (22.0–108.9)
AST (IU/L)21 (17–26)18 (15–22)
ALT (IU/L)22 (16–33)14 (11–19)
GGT (IU/L)40 (25–68)16 (12–23)
Total cholesterol (mmol/L)5.25 (4.71–5.85)5.33 (4.73–5.92)
Triglyceride (mmol/L)1.29 (0.92–1.86)0.80 (0.60–1.12)
High-density lipoprotein cholesterol (mmol/L)1.37 (1.16–1.63)1.71 (1.47–1.97)
Fasting plasma glucose (mmol/L)5.50 (5.22–5.94)5.16 (4.94–5.50)
Uric acid (μmol/L)360 (320–410)250 (210–290)
Table 2.   Prevalence of metabolic syndrome-related factors
Metabolic syndrome-related factorsPrevalence
Men (%)Women (%)
  1. * Triglyceride ≥1.7 mmol/L, high-density lipoprotein cholesterol <1.03 mmol/L, or use of medications for dyslipidemia.

  2. † Fasting plasma glucose ≥5.6 mmol/L or use of medications for diabetes mellitus.

  3. ‡ Uric acid >450 μmol/L for men, uric acid >390 μmol/L for women, or use of medications for hyperuricemia.

  4. § Blood pressure at/above 130/85 mmHg or use of medications for hypertension.

Obesity (body mass index ≥25 kg/m2)2613
Atherogenic dyslipidemia*3613
Glucose intolerance†5023
Hyperuricemia‡142
Hypertension§3626
Fatty liver3115
Regular exercise (at least once a week)4744
Current smoking (excluding previous smoking)487
Alcohol consumption7429

In men, obesity, atherogenic dyslipidemia, glucose intolerance, hyperuricemia, hypertension and regular exercise were significant factors associated with fatty liver, whereas current smoking was not. Multivariable analyses found that alcohol consumption was an independently significant negative predictor of fatty liver (OR, 0.54; 95% CI, 0.46–0.63). In women, obesity, atherogenic dyslipidemia, glucose intolerance, hyperuricemia, hypertension and current smoking were significant factors, whereas regular exercise was not. After adjusting for these significant confounders, alcohol consumption was not a significant factor (OR, 0.80; 95% CI, 0.57–1.12) (Table 3) (Table S2).

Table 3.   Individual and multivariable analyses of factors associated with fatty liver
Metabolic syndrome-related factorsOR (95% CI) in menOR (95% CI) in women
Individual analysisMultivariable analysisIndividual analysisMultivariable analysis
  1. OR, odds ratio; CI, confidence interval.

  2. * Triglyceride ≥1.7 mmol/L, high-density lipoprotein cholesterol <1.03 mmol/L, or use of medications for dyslipidemia.

  3. † Fasting plasma glucose ≥5.6 mmol/L or use of medications for diabetes mellitus.

  4. ‡ Uric acid >450 μmol/L for men, uric acid >390 μmol/L for women, or use of medications for hyperuricemia.

  5. § Blood pressure at/above 130/85 mmHg or use of medications for hypertension.

Obesity (body mass index ≥25 kg/m2)7.28 (6.34–8.38)5.95 (5.13–6.89)15.91 (11.91–21.26)10.34 (7.57–14.11)
Atherogenic dyslipidemia*3.57 (3.14–4.05)2.74 (2.38–3.15)4.96 (3.73–6.58)2.73 (1.93–3.87)
Glucose intolerance†1.64 (1.46–1.86)1.31 (1.14–1.51)3.40 (2.65–4.36)1.88 (1.39–2.54)
Hyperuricemia‡2.17 (1.84–2.55)1.52 (1.26–1.84)5.14 (2.84–9.31)1.73 (0.78–3.80)
Hypertension§1.71 (1.51–1.93)1.18 (1.02–1.36)3.92 (3.06–5.01)1.79 (1.32–2.42)
Regular exercise (at least once a week)0.80 (0.71–0.91)0.82 (0.72–0.94)1.06 (0.84–1.35) 
Current smoking (excluding previous smoking)0.95 (0.84–1.07) 0.43 (0.23–0.81)0.67 (0.33–1.36)
Alcohol consumption0.59 (0.52–0.68)0.54 (0.46–0.63)0.60 (0.45–0.80)0.80 (0.57–1.12)

Association of total amount of alcohol consumed per week with fatty liver

Table 4 demonstrates the association of fatty liver with alcohol consumption stratified by the total amount of alcohol consumed per week.

Table 4.   Prevalence and odds ratio of fatty liver in each category of the total amount of alcohol consumed per week
   nPrevalenceOR (95% CI)Adjusted OR*
  1. OR, odds ratio; CI, confidence interval.

  2. * Adjusted OR (95% CI) for obesity, atherogenic dyslipidemia, glucose intolerance, hyperuricemia, hypertension and regular exercise in men; adjusted OR (95% CI) for obesity, atherogenic dyslipidemia, glucose intolerance, hyperuricemia, hypertension and current smoking in women.

  3. † Analysis in drinkers using the Cochran–Armitage test for trend.

MenTotal amount of alcohol consumed per weekNondrinker126840%1.001.00
0.1–69.9 g68936%0.83 (0.69–1.01)0.75 (0.60–0.94)
70.0–139.9 g70530%0.63 (0.52–0.77)0.62 (0.49–0.77)
140.0–279.9 g124826%0.52 (0.44–0.62)0.50 (0.41–0.61)
≥280.0 g104726%0.52 (0.43–0.62)0.40 (0.32–0.49)
P for trend†<0.001
WomenTotal amount of alcohol consumed per weekNondrinker153516%1.001.00
0.1–69.9 g3699%0.50 (0.34–0.73)0.68 (0.44–1.05)
70.0–139.9 g14713%0.76 (0.46–1.25)0.85 (0.48–1.51)
140.0–279.9 g8212%0.71 (0.36–1.39)1.26 (0.60–2.65)
≥280.0 g2214%0.80 (0.24–2.74)1.13 (0.25–5.19)
P for trend†0.189

In men, there was a decreasing prevalence of fatty liver with an increase in the total amount of alcohol consumed per week (< 0.001). Alcohol consumption in each drinking category except alcohol consumption 0.1 to 69.9 g/week was inversely associated with prevalence of fatty liver. A significant inverse association was observed in all drinking categories after adjustment for other significant confounders (obesity, atherogenic dyslipidemia, glucose intolerance, hyperuricemia, hypertension and regular exercise).

In women, the increase in the total amount of alcohol consumption per week seemed to directly associate with an increase in prevalence of fatty liver, although it was not significant (= 0.189). A significant inverse association between fatty liver and alcohol consumption was found only in those who consumed 0.1 to 69.9 g alcohol per week. The association did not become statistically significant after adjustment for other significant confounders (obesity, atherogenic dyslipidemia, glucose intolerance, hyperuricemia, hypertension and current smoking).

Association of drinking frequency with fatty liver

Table 5 shows the prevalence of fatty liver and OR in comparison to nondrinkers according to the drinking frequency.

Table 5.   Prevalence and odds ratio of fatty liver in each category of drinking frequency
  Drinking frequency (drinking days/week)P for trend‡
Nondrinker1–34–67
  1. OR, odds ratio; CI, confidence interval.

  2. * Adjusted OR (95% CI) for obesity, atherogenic dyslipidemia, glucose intolerance, hyperuricemia, hypertension and regular exercise.

  3. † Adjusted OR (95% CI) for obesity, atherogenic dyslipidemia, glucose intolerance, hyperuricemia, hypertension and current smoking.

  4. ‡ Analysis in drinkers using the Cochran–Armitage test for trend.

Menn12688288322029<0.001
Prevalence40%38%29%24%
OR (95% CI)1.000.92 (0.76–1.10)0.61 (0.51–0.74)0.48 (0.41–0.55)
Adjusted OR*1.000.80 (0.65–0.98)0.52 (0.42–0.65)0.44 (0.37–0.53)
Womenn15353061641500.592
Prevalence16%10%10%12%
OR (95% CI)1.000.57 (0.39–0.85)0.55 (0.32–0.94)0.69 (0.42–1.16)
Adjusted OR†1.000.77 (0.49–1.23)0.70 (0.38–1.27)0.99 (0.55–1.77)

In men, the prevalence of fatty liver was 38% for drinking 1–3 days/week, 29% for drinking 4–6 days/week and 24% for daily drinking. The trend towards an inverse association between the prevalence of fatty liver and drinking frequency was significant (< 0.001). The ORs were 0.80 (95% CI, 0.65–0.98) for drinking 1–3 days/week, 0.52 (95% CI, 0.42–0.65) for drinking 4–6 days/week and 0.44 (95% CI, 0.37–0.53) for daily drinking after adjusting for other significant confounders (obesity, atherogenic dyslipidemia, glucose intolerance, hyperuricemia, hypertension and regular exercise).

In women, there was no trend towards an association between the prevalence of fatty liver and drinking frequency (= 0.592). There were inverse associations between alcohol consumption and fatty liver in drinking 1–3 days/week (OR, 0.57; 95% CI, 0.39–0.85) and in drinking 4–6 days/week (OR, 0.55; 95% CI, 0.32–0.94) but not in daily drinking (OR, 0.69; 95% CI, 0.42–1.16). The associations between alcohol consumption and fatty liver were not statistically significant in any category of drinking frequency after adjusting for other significant confounders (obesity, atherogenic dyslipidemia, glucose intolerance, hyperuricemia, hypertension and current smoking).

Association of drinking pattern (frequency and amount) with fatty liver

Drinking patterns were classified according to both drinking frequency and drinking amount, i.e., amount of alcohol consumed per drinking day and total amount of alcohol consumed per week to address the question whether the amount, frequency or both in alcohol consumption are responsible for the low risk of fatty liver. We then estimated the influence of alcohol consumption on the presence of fatty liver in comparison to nondrinkers for each category (Table 6).

Table 6.   Trend regarding the prevalence of fatty liver in each category of drinking frequency and amount of alcohol consumed
   Drinking frequency (drinking days/week)P for trend*
Nondrinker1–34–67
MenAmount of alcohol consumed per drinking dayNondrinker40% (n = 1268)
0.1–19.9 g36% (n = 168)30% (n = 114)20% (n = 161)<0.001
20.0–39.9 g37% (n = 402)28% (n = 427)24% (n = 898)<0.001
40.0–59.9 g37% (n = 162)29% (n = 211)24% (n = 731)<0.001
≥60.0 g46% (n = 96)31% (n = 80)29% (n = 239)0.005
P for trend* 0.1840.8130.081 
Total amount of alcohol consumed per weekNondrinker40% (n = 1268)
0.1–69.9 g36% (n = 631)26% (n = 53)60% (n = 5)0.541
70.0–139.9 g42% (n = 146)29% (n = 403)19% (n = 156)<0.001
140.0–279.9 g47% (n = 47)28% (n = 303)24% (n = 898)0.001
≥280.0 g50% (n = 4)33% (n = 73)25% (n = 970)0.067
P for trend* 0.0580.6590.268 
  1. * Analysis in drinkers using the Cochran–Armitage test for trend.

WomenAmount of alcohol consumed per drinking dayNondrinker16% (n = 1535)
0.1–19.9 g7% (n = 160)7% (n = 97)8% (n = 60)0.724
20.0–39.9 g15% (n = 123)16% (n = 58)15% (n = 69)0.870
40.0–59.9 g5% (n = 19)0% (n = 6)19% (n = 16)0.190
≥60.0 g0% (n = 4)0% (n = 3)0% (n = 5)
P for trend*0.3260.6050.496 
Total amount of alcohol consumed per weekNondrinker16% (n = 1535)
0.1–69.9 g10% (n = 291)4% (n = 71)0% (n = 7)0.070
70.0–139.9 g8% (n = 13)16% (n = 81)9% (n = 53)0.639
140.0–279.9 g0% (n = 2)0% (n = 11)15% (n = 69)0.166
≥280.0 g0% (n = 1)14% (n = 21)
P for trend*0.5930.2680.248 

The drinking frequency in men was inversely correlated to the prevalence of fatty liver in each category of the amount of alcohol consumed per drinking day. There was a significant inverse correlation between drinking frequency and prevalence of fatty liver in the categories of drinking 70.0 to 139.9 g/week and 140.0 to 279.9 g/week, when the drinking amount was classified by the total amount of alcohol consumed per week. On the other hand, there was no trend towards an inverse association between the amount of alcohol consumed and the prevalence of fatty liver in each category of drinking frequency. The amount of alcohol consumed tended to be directly correlated to the prevalence of fatty liver in some categories of drinking frequency, although that was not statistically significant. The proportion of daily drinkers was higher in the category of the larger total amount of alcohol consumed (< 0.001), which explains the inverse correlation between the total amount of alcohol consumed per week and the prevalence of fatty liver as shown in Table 4. Therefore, the drinking frequency but not amount of alcohol consumed was an essential factor for the inverse correlation between alcohol consumption and fatty liver in men.

In women, there was neither any correlation of the amount of alcohol consumed in each category of drinking frequency nor any correlation of drinking frequency in each category of amount of alcohol consumed with the prevalence of fatty liver. Drinking 0.1–19.9 g for 1–3 days/week and drinking 0.1–19.9 g for 4–6 days/week were associated with a low prevalence of fatty liver. Moreover, drinking 0.1–19.9 g for 1–3 days/week remained statistically significant after adjusting for other confounders (OR, 0.47; 95% CI, 0.23–0.96).

Association of alcohol consumption with fatty liver in metabolic syndrome cases

We attempted to determine whether the presence of metabolic syndrome affected the association between alcohol consumption and fatty liver. Because of the absence of waist circumference data, we substituted obesity (body mass index ≥25 kg/m2) for an elevated waist circumference and identified the patients with probable metabolic syndrome by the presence of obesity and two or more of the following: hypertension, atherogenic dyslipidemia and glucose intolerance.

In men with the probable metabolic syndrome (n = 722), the prevalence of fatty liver was significantly lower in drinkers than in nondrinkers [64% vs. 82% (< 0.001)]. The prevalence of fatty liver was inversely correlated with the total amount of alcohol consumed per week (= 0.008), and inversely correlated with drinking frequency (< 0.001). In the analyses stratified by both drinking frequency and amount of alcohol consumed, the inverse association between alcohol consumption and fatty liver was significant in almost half of the categorised drinking; the inverse correlation between drinking frequency and the prevalence of fatty liver was significant in the subjects who consumed 20.0–39.9 g of alcohol per drinking day (= 0.009).

In women with probable metabolic syndrome (n = 122), the prevalence of fatty liver was lower in drinkers than in nondrinkers, although the difference was not statistically significant [57% vs. 71% (= 0.208)]. In the stratified analyses, none of the drinking categories showed significant inverse association with fatty liver; no significant linear trend was found (Table S4).

Association of alcohol consumption with ALT elevation

This study also analysed the association of the drinking pattern with the prevalence of ALT elevation. The prevalence of ALT elevation was significantly lower in drinkers than in nondrinkers [30% vs. 34% in men (= 0.007) and 22% vs. 29% in women (= 0.002)]. The present study also confirmed the prevalence of ALT elevation to be significantly higher in cases with fatty liver than in those without [59% vs. 18% in men (< 0.001) and 62% vs. 21% in women (< 0.001)].

In men, the analyses in drinkers did not demonstrate a significant trend towards an association between the total amount of alcohol consumed per week and the prevalence of ALT elevation (= 0.291). The drinking frequency was inversely correlated to the prevalence of ALT elevation (< 0.001). Within similar categories of amount of alcohol consumed, the drinking frequency was also inversely correlated to ALT elevation, and the inverse correlation was significant in the category of drinking 0.1–19.9 g/drinking day (< 0.001), 20.0–39.9 g/drinking day (< 0.001), ≥60.0 g/drinking day (= 0.020), 70.0–139.9 g/week (< 0.001), 140.0–279.9 g/week (< 0.001) and ≥280.0 g/week (= 0.034). On the other hand, the analysis in each category of drinking frequency showed that the amount of alcohol consumed per drinking day was directly correlated to the ALT elevation, and the direct correlation was significant in drinking 4–6 days/week (= 0.027) and in daily drinking (< 0.001). The total amount of alcohol consumed per week was also directly correlated to ALT elevation in drinking 1–3 days/week (= 0.023), in drinking 4–6 days/week (= 0.013) and in daily drinking (< 0.001) (Table S1).

The linear trends that were observed in men, however, were not found in women.

Discussion

The present study explored the association between drinking pattern and the prevalence of ultrasonographically determined fatty liver. The results demonstrated an inverse association between the drinking frequency, but not the amount consumed and the prevalence of fatty liver in men. This finding was also supported by similar analysis of alterations in ALT instead of ultrasonographically determined fatty liver. Recent studies suggested an inverse association between light to moderate alcohol consumption and the prevalence of fatty liver.17–20 To the best of our knowledge, however, the importance of drinking frequency on this association has never been fully investigated, although an inverse association between drinking frequency and type 2 diabetes or coronary heart disease has been reported. We believe that this is the first report to demonstrate an inverse association between drinking frequency and prevalence of fatty liver in men.

The mechanisms that explain the inverse association between alcohol consumption and fatty liver are not known. Dixon et al.17 suggested that moderate alcohol consumption reduced the risk of non-alcoholic steatohepatitis, possibly due to reduced insulin resistance. In contrast, in the present study, the significant inverse association between alcohol consumption and fatty liver was independent of metabolic syndrome-related factors including glucose intolerance. Alcohol consumption was not associated with low prevalence of the glucose intolerance (data not shown). Although the discrepancy between the previous reports1, 2, 5, 8, 9, 17 and our observation cannot be discussed because of the lack of information on serum insulin levels in the subjects, the effect of alcohol consumption on fatty liver does not seem to be entirely associated with reduced insulin resistance. Dunn et al.19 reported a low prevalence of elevated ALT in modest wine drinkers and suggested that both alcohol and non-alcoholic components of wine are possibly associated with the low prevalence of metabolic syndrome features. The liver protective effects of modest wine drinking are supported by other studies.28, 29 Degrace et al.30 showed that the triglyceride content of liver was significantly decreased by beer or ethanol-free beer treatment in mice. Other studies reported an increase in circulating adiponectin due to alcohol consumption7 or combined consumption of alcohol and saturated fat31 and suggested the prevention of hepatic steatosis mediated through adenosine monophosphate-activated protein kinase and peroxisome proliferator-activated receptor α enhanced by activation of adiponectin receptor 2.31 Alcoholic beverages certainly seem to have a beneficial effect on the liver, and the current observations suggest that frequent drinking enhances the effect. Further studies are required to elucidate the underlying mechanism(s) of the association between fatty liver and alcohol consumption, particularly in terms of frequent intake.

Moreover, to the best of our knowledge, this is also the first report to find an inverse association between alcohol consumption and presence of fatty liver in women. The association in women was less robust in comparison to that in men and was observed only in light and infrequent drinking. This may be partially explained by the fact that the number of women drinkers was insufficient and the analysis for women was therefore underpowered. It was also suggested that even one serving of beverage per day could be finally an excessive total amount for women with daily drinking. Women have a significantly higher relative risk of developing an alcohol-related liver disease than men.32 Women may therefore be more vulnerable to alcohol than men, and an enhanced alcoholic hepatotoxicity in women may be responsible for the cancellation of the beneficial effect of frequent drinking. Therefore, the inverse association between alcohol consumption and prevalence of fatty liver in women was statistically significant only in a limited drinking pattern.

The present study used ultrasonography as the diagnostic modality for fatty liver. There is a discrepancy between ultrasonographic assessment and histological examination33 and the latter is the gold standard for the diagnosis of fatty liver. However, conducting liver biopsies in a large epidemiological study is not feasible, with regard to safety, compliance and cost. Ultrasonography was considered to be a practical option for the diagnosis of fatty liver in a survey such as this.

The association between fatty liver and ALT elevation is well established.34 In the present study, ALT elevation was directly associated with ultrasonographically determined fatty liver; the association between ALT elevation and the drinking pattern was consistent with that between fatty liver and the drinking pattern, which confirmed our interpretation. The direct correlation between the prevalence of ALT elevation and the amount of alcohol consumed within similar categories of drinking frequency suggested that there was an increased prevalence of fatty liver and/or alcoholic liver injury as the amount of alcohol consumed increased.

There are some limitations associated with the current study. It is impossible to conclude whether the drinking pattern directly contributed to the development of fatty liver or abnormalities on liver function tests because this was a cross-sectional study. There was no way to distinguish abstinent alcoholics or ex-drinkers from lifetime abstainers, since the questionnaire only distinguished current alcohol abuse. In addition, previously undiagnosed patients with viral hepatitis were not excluded because viral markers were not examined at the time of the health checkups. Furthermore, the association with meals, total caloric intake and the types of beverages consumed, such as beer and wine, were not evaluated. A certain number of participants were under treatment for hypertension, diabetes mellitus, atherogenic dyslipidemia or hyperuricemia and the details of medication were unknown. Finally, intervention based on the results of previous health checkups may possibly have modified the lifestyle of some of the subjects. Further prospective studies are required to address these issues.

In conclusion, the drinking frequency was inversely correlated to the prevalence of ultrasonographically determined fatty liver and ALT elevation in men, whereas light and infrequent alcohol consumption was associated with a low risk of fatty liver in women. However, these observations do not encourage alcohol consumption with any frequency; proper alcohol consumption or abstinence should be determined for each person individually with due consideration of his or her total health and not based only on the health of the liver.

Acknowledgement

Declaration of personal and funding interests: None.

Ancillary