Body fat distribution, relative weight, and liver enzyme levels: A population-based study

Authors


Abstract

Regional body fat distribution may represent an independent risk factor for several conditions, especially metabolic and cardiovascular diseases; recent findings have shown that abdominal fat accumulation can be an independent predictor of hepatic steatosis. Very few studies, mostly using selected clinical samples, have focused on the relationship between indices of abdominal visceral fat accumulation and the most commonly used biochemical liver tests, such as alanine aminotransferase (ALT), aspartate aminotransferase (AST), and gamma-glutamyltransferase (GGT). The aim of the present study was to evaluate the relation between central fat accumulation, as assessed by abdominal height, relative weight, as determined by body mass index (BMI), and liver function tests (ALT, AST, and GGT) in a random sample of 2,704 residents of Erie and Niagara Counties in New York State, 35–80 years of age and free from known hepatic disease. Multiple linear regression models were used, with liver enzymes as dependent variables with abdominal height and BMI as independent variables, and the inclusion of several covariates (age, race, education, smoking status, pack-years of smoking, drinking status, and total ounces of ethanol in the past 30 days). Abdominal height was consistently a better correlate of ALT and GGT levels than BMI in both sexes. In addition, abdominal height was the most powerful independent predictor of ALT in both sexes as well as of GGT among women. In conclusion, these findings support a role for central adiposity independent from BMI in predicting increased levels of hepatic enzymes, likely as a result of unrecognized fatty liver. (HEPATOLOGY 2004;39:754–763.)

Obesity is an important predictor of several diseases,1 as well as one of the risk factors most frequently associated with increased liver enzymes.2–7 In the last several years, many epidemiological studies strongly indicated that regional body fat distribution, with abdominal accumulation, irrespective of total body fat quantity, as assessed by body mass index (BMI), may represent a major independent risk factor for several conditions, especially metabolic and cardiovascular diseases.8, 9 Furthermore, recent findings have shown that central adiposity can be an independent predictor of hepatic steatosis (fatty liver),10, 11 a common clinical and histological condition frequently associated with alcohol consumption and excessive body weight. In addition, obesity seems to represent a better predictor of fatty liver than heavy drinking and the coexistence of both conditions (obesity and heavy drinking) may produce a very high risk of developing this condition.12 Increased levels primarily of alanine aminotransferase (ALT) and triglycerides, and secondarily of gamma-glutamyltransferase (GGT), appear to be the most sensitive biochemical indicators of the presence of hepatic steatosis.11, 12 Despite the growing body of evidence on the importance of visceral adiposity and the independent role of weight with respect to alcohol consumption in determining fat accumulation in the liver, to date very few studies, mostly related to restricted samples and often in clinical settings,13–15 have focused on the relationship between simply assessed anthropometric indices of abdominal visceral adipose tissue accumulation and the most commonly used biochemical liver tests, such as ALT, aspartate aminotransferase (AST), and GGT. Very recently, Ruhl and Everhart,16 using data from the Third National Health and Nutrition Examination Survey (NHANES III), found that waist-to-hip circumference ratio (WHR), as a measure of central adiposity, was more strongly associated with elevated ALT concentration than BMI. In epidemiological research, BMI has been used extensively to assess relative weight and obesity status and their relation to several health outcomes, including hepatic enzymes, whereas anthropometrical measures of central adiposity, such as waist circumference, WHR, and abdominal height, widely used to assess the relationship between body fat distribution and metabolic and cardiovascular diseases, have been scarcely considered in their association with liver enzyme activity.

The aim of this population-based study was to evaluate the relative role of abdominal fat accumulation compared to overall BMI on the relationship between body weight and liver enzymes as well as assessing the existence of gender-based differences in this association.

Abbreviations

ALT, alanine amino-transferase; AST, aspartate amino-transferase; GGT, gamma-glutamyltransferase; BMI, body mass index; WHR, waist-to-hip ratio.

Materials and Methods

Study Population.

The present study is based on data obtained from a randomly chosen sample of residents of Erie and Niagara Counties in New York State. The overall sample, identified as the “Western New York Health Study,” was collected between September 1995 and May 2001 as part of a series of studies focusing on a number of chronic conditions. The details of the overall study design, participant enrollment, and methodology have been described elsewhere.17 The study protocol was approved by the University at Buffalo Institutional Review Board.

Two sources were used to identify potential participants: 1) Department of Motor Vehicles of New York State for people age 35–64, and 2) Health Care Financing Administration for people age 65–80. Of the 6,837 potential participants identified, contacted, and assessed as eligible for our study, a total of 4,065 (participation rate = 59.5%) agreed to participate. The exclusion criteria applied to the present analyses were a self-reported history of chronic or acute hepatitis, cirrhosis or noncirrhotic liver disease (n = 173); a self-reported history of prevalent coronary heart disease (prior myocardial infarction, coronary artery bypass graft surgery, angioplasty, or diagnosed angina pectoris, n = 536); missing information on anthropometric measurements (n = 403); missing data for education, smoking, and drinking habits (n = 49); missing blood determination of liver enzymes (n = 199); or outliers for enzyme levels (n = 1, for GGT = 4,256). The remaining 2,704 participants (1,519 women, 1,185 men) were included in this analysis.

Excluded individuals for whom information was available were different from included counterparts with respect to the distribution of gender and race, with a higher prevalence of men and nonwhite participants among the former. Excluded individuals were also significantly older, less educated, and characterized by higher mean values for all anthropometric measurements (BMI, abdominal height, waist and hip circumferences, WHR) compared to included subjects. Excluded individuals were significantly more likely to be former smokers and reported significantly higher average levels of total pack-years than included individuals. With regard to drinking status, the excluded participants were significantly more likely to be both lifetime abstainers and former drinkers compared with included participants, whereas the two groups of individuals did not differ for total amount of alcohol consumed in the past 30 days. For liver enzymes, excluded individuals who had liver enzyme measures had significantly higher AST and GGT (but not ALT) mean values compared with included participants, as a result of the higher prevalence of liver disease among the former.

General Examination.

All agreeing participants were invited to the Center for Preventive Medicine at the University at Buffalo for an interview and physical examination that lasted ∼2.5 hours. As part of the examination, participants were characterized in detail with regard to their health conditions and lifestyle habits. In addition to the questions regarding drinking behavior, participants were asked about their current and lifetime smoking habits. For the purpose of this study, participants reporting current use of cigarettes were classified as current smokers; participants reporting not having smoked at least 100 cigarettes in their lifetime were classified as never smokers. Other participants were classified as former smokers. A self-reported medical history was obtained to determine the prevalence of physician-diagnosed diseases including acute or chronic hepatitis and cirrhosis or noncirrhotic liver disease. For women, menopausal status was determined using criteria taking into account age, surgical or natural cessation of menses, hormone use, and bilateral oopherectomy.

Anthropometry.

Anthropometrical measurements were made by trained interviewers on participants wearing light clothing and no shoes. Measurements included abdominal height, waist and hip circumferences, height, and weight. Abdominal height, measured using the Holtain-Kahn abdominal caliper18 (Fig. 1), has been shown to be highly correlated with the volume of visceral fat as determined by multiscan tomography.19–22 Three separate measurements were made to the nearest 0.1 cm of the sagittal (e.g., anteroposterior) abdominal diameter. If the three readings were not within 0.5 cm of each other, the three readings were repeated until they were all within 0.5 cm of each other. During the study we examined the intra- and interobserver variability of the abdominal height measurement. The intraobserver variability, considering 2 observers, 4 observed volunteers, and 2 observations for each observer, and evaluated by the intraclass correlation (ICC) coefficient was 0.99 (0.96 lower bound). The interobserver variability was 0.99 (0.97 lower bound). Waist circumference was determined with participants standing erect with the abdomen relaxed, arms at the side, and feet together. The tape was horizontally placed between the bottom of the rib cage and the top of the iliac crest (hip bones) around the smallest circumference between these two reference points. The measurement was taken at the end of a normal expiration, without the tape compressing the skin, to the nearest 0.1 cm. As for hip circumference, the measurement was made with participants standing erect with the abdomen relaxed, arms at the side, and feet together and the tape horizontally placed around the hips at the biggest circumference point between the iliac crest and either the crotch for women or the head of the femur for men. The measurement was taken as well at the end of a normal expiration, without the tape compressing the skin, to the nearest 0.1 cm.

Figure 1.

The Holtain-Kahn abdominal caliper.

Height was measured on a permanently mounted vertical board (Perspective Enterprises, Kalamazoo, MI), according to a standardized protocol. Weight was measured to the nearest tenth of a pound on a calibrated balance beam scale (Detecto, Webb City, MO). BMI was calculated as weight (kg) divided by height in meters2.

Laboratory Analyses.

A fasting blood sample was obtained for determination of routine chemistry between 7:30 and 9:30 AM after a fasting of 8–12 hours. Immediately following phlebotomy, tubes were wrapped in aluminum foil to protect them from light and kept at room temperature for 30 minutes and allowed to clot. Blood tubes were centrifuged at 3,000g for 10 minutes and 1.5 ml of serum was transferred to polypropylene screw cap vials and placed in a cooler with a cold pack. Samples were delivered by courier to Millard Fillmore Center for Laboratory Medicine (Amherst, NY) for analysis the same day. Hepatic enzymes ALT,23 AST,23 and GGT24 were measured by kinetic enzyme assays as part of a chemistry profile on a Paramax Automated Chemistry System.

Alcohol Intake.

A specific focus of the interview was dedicated to assess the past and current use of total and beverage specific alcohol consumption. Information about alcohol intake was obtained with a computer-assisted in-person interview.25 Interviewers underwent extensive training and standardization. In addition to their current consumption (past 30 days), participants were asked about their lifetime alcohol intake with the use of the Cognitive Lifetime Drinking History (CLDH).26, 27 The interview included questions about the size of the container (glass, bottle) used and the amount of alcoholic beverage usually consumed in that particular container. Participants were shown models and photos of different containers with lines indicating the potential level of beverage within each container, in order to improve the accuracy of their estimates. Responses to the above questions were used to compute the following variables for the present analyses: 1) lifetime abstainers – participants who reported consumption of less than 12 drinks during their lifetime or in any 1-year period; 2) noncurrent drinkers – participants who reported 12 or more drinks during their lifetime or in any 1-year period, but did not consume an alcoholic beverage at least once in the past 30 days; 3) current drinkers – did consume at least 1 alcoholic beverage in the past 30 days; 4) total ounces of ethanol consumed in the past 30 days.

Statistical Analysis.

All analyses were conducted using the Statistical Package for Social Sciences (SPSS-11.0, Chicago, IL).28 The distribution of liver enzymes was normalized by natural log transformation and log-transformed values were used in all analyses while the other continuous variables were not significantly skewed or their transformation did not significantly affect the results of the analyses. Descriptive statistics were examined and differences by gender and menopausal status were compared using Independent-Sample t tests for continuous variables and χ2 for categorical ones. Interaction was investigated by including interaction terms of the considered anthropometric indices multiplied by age, total alcohol intake, and other covariates. Among women, tests for interaction between anthropometric measures (abdominal height and BMI) and menopausal status (pre- and postmenopausal) were significant for all three hepatic enzymes (with abdominal height for ALT, with BMI for AST and GGT). Consequently, all analyses including women were performed with stratification by menopausal status. Pearson linear correlation and partial correlation analyses were used to test the bivariate associations between the various anthropometric measures as well as the associations between anthropometric measures and liver enzymes. To further investigate the association between anthropometric indices and hepatic enzymes, we used multiple linear regression analysis. Based on the results of bivariate analyses, ALT and GGT were chosen as dependent variables while the independent variables were abdominal height and BMI. We included in the baseline model the following covariates: age, race (nonwhite, white), education (years), smoking status (never, former, and current) and cumulative tobacco smoke exposure (pack-years of smoking), drinking status (lifetime abstainers, former, and current) and total ounces of ethanol in the past 30 days. Statistical significance was considered if P values were less than 0.05 (two-tailed) unless otherwise indicated.

Results

The descriptive characteristics of the study participants are shown in Table 1. On average, women were significantly younger and less educated than men. As for the considered anthropometric measures, women had significantly lower mean values compared to men for all parameters. For categorical variables, we did not detect a significant difference in the distribution of race categories between the two sexes while, regarding smoking status, women were more likely than men to be never smokers and less likely to be former smokers and reported lower mean values of lifetime pack-years. For drinking status, women were more likely to be lifetime abstainers and consumed less alcohol overall in the 30 days prior to the interview than men. With respect to liver enzymes, women showed significantly lower average levels compared to men. When women were analyzed according to menopausal status, we found that premenopausal participants were, on average, significantly younger (as expected) and more educated and reported significantly lower mean values for all anthropometric indices but height than postmenopausal women. They were also significantly more likely to be never smokers and showed lower mean values of lifetime pack years, as expected. For drinking status, premenopausal women were more likely to be current drinkers and less likely to be lifetime abstainers in comparison with postmenopausal participants, even though we did not detect a significant difference between the 2 groups in the total amount of alcohol consumed in the past 30 days. With regard to biochemical hepatic tests, female participants in premenopausal status showed significantly lower mean values than postmenopausal women for all the considered enzymes.

Table 1. Characteristics of Participants by Gender and Menopausal Status—The Western New York Health Study, 1995–2001
Variable (unit)Men (n = 1185) Mean (SD)All Women (n = 1519) Mean (SD)Premenopausal Women (n = 511) Mean (SD)Postmenopausal Women (n = 1008) Mean (SD)
  • a

    P ≤ 0.05

  • **, *

    P ≤ 0.01

  • ***

    P ≤ 0.001 (P value for comparison between men and women or between premenopausal women and postmenopausal women); significance testing for liver enzymes is based on log transformed data.

  • Abbreviations: SD, standard deviation; AST, Aspartate amino-transferase; ALT, alanine amino-transferase; GGT, gamma-glutamyltransferase.

Age (years)57.7 (12.1)***55.9 (11.4)43.9 (4.5)***62.0 (8.8)
Education (years)13.9 (2.6)**13.6 (2.3)14.3 (2.2)***13.2 (2.3)
Weight (kg)86.7 (14.7)***72.6 (16.1)72.7 (17.6)72.5 (15.3)
Height (m)1.76 (0.07)***1.62 (0.06)1.64 (0.06)***1.61 (0.06)
Body mass index (kg/m2)28.1 (4.4)*27.7 (5.9)27.1 (6.2)**28.0 (5.6)
Abdominal height (cm)22.0 (3.3)***20.4 (3.7)19.6 (3.8)***20.8 (3.6)
Waist circumference (cm)97.9 (11.7)***86.0 (14.3)83.6 (14.9)***87.2 (13.8)
Hip circumference (cm)102.9 (10.2)**104.1 (13.1)102.7 (13.9)**104.7 (12.6)
Waist/hip ratio0.95 (0.06)***0.83 (0.08)0.81 (0.08)***0.83 (0.08)
Pack years of smoking16.1 (21.3)***9.8 (16.6)6.0 (10.6)***11.7 (18.6)
Past 30 days total alcohol intake (ounces)12.8 (21.3)***5.0 (10.0)5.1 (8.7)4.9 (10.5)
ALT (U/L)22.7 (16.2)***16.2 (9.3)14.7 (9.2)***17.0 (9.3)
AST (U/L)26.2 (11.9)***22.9 (8.3)21.1 (9.3)***23.9 (7.6)
GGT (U/L)44.4 (60.8)***29.4 (41.2)25.4 (33.6)***31.5 (44.4)
 %%%%
Race    
 Nonwhite8.97.04.58.2
 White91.193.095.5**91.8
Smoking Status    
 Never smokers38.150.554.548.5
 Former smokers46.133.729.535.8
 Current smokers15.8***15.816.0*15.7
Drinking Status    
 Lifetime abstainers3.812.69.614.1
 Non-current drinkers21.924.623.525.2
 Current drinkers74.3***62.866.9*60.7

Pearson correlations between BMI, abdominal height, and other anthropometric indices of visceral fat distribution (waist circumference and WHR) are shown in Table 2 for women, stratified by menopausal status, and for men. The measure of abdominal height was very highly correlated with BMI. The correlations of the other anthropometric measures with WHR were consistently less strong. Premenopausal women consistently showed somewhat higher coefficients than either postmenopausal women or men. BMI and waist circumference were the strongest correlates of abdominal height in either pre- or postmenopausal female participants, while among men waist circumference was the strongest correlate of abdominal height.

Table 2. Simple Pearson Correlation Coefficients Between Anthropometric Variables—The Western New York Health Study, 1995–2001
VariableBMIAbdominal HeightWaist CircumferenceWaist/Hip Ratio
  • *P ≤ 0.05; **P ≤ 0.01

  • ***

    P ≤ 0.001.

Pre-menopausal women (n = 511)
Body mass index (BMI)1.000.910***0.906***0.415***
Abdominal height 1.0000.908***0.489***
Waist circumference  1.0000.632***
Waist/hip ratio   1.000
Post-menopausal women (n = 1008)
BMI1.0000.861***0.866***0.335***
Abdominal height 1.0000.863***0.432***
Waist  1.0000.635***
Waist/hip ratio   1.000
Men (n = 1185)
BMI1.0000.840***0.860***0.367***
Abdominal height 1.0000.880***0.436***
Waist  1.0000.547***
Waist/hip ratio   1.000

The results of partial correlation analyses between anthropometric measures and liver parameters after adjustment for age are summarized in Table 3. In premenopausal women, waist circumference and abdominal height showed higher correlations with ALT and GGT compared to both BMI and WHR. For AST, the correlations with anthropometric indices were consistently weaker than those observed between anthropometric indices and the other two liver enzymes. Among postmenopausal women and men, abdominal height was the strongest correlate of both ALT and GGT, while for AST we reported no significant correlations with anthropometric indices. In addition, the relationship between abdominal height and liver enzymes (ALT and GGT) was attenuated but still significant after the further adjustment for BMI in all the considered categories of participants. By contrast, the correlation between BMI and the two hepatic enzymes was reduced and no longer significant when controlling for abdominal height in any of the three groups of participants. The relationship between waist circumference and ALT and GGT was also diminished but still significant (except for ALT among postmenopausal women) after the further adjustment for BMI. However, the correlation coefficients between waist circumference and ALT and GGT were consistently lower than those observed between abdominal height and the two liver enzymes (except for ALT among premenopausal women). The results of partial correlation analyses between anthropometric measures and liver enzymes were substantially unchanged after adjustment for other covariates, such as race, education, smoking status, pack-years of smoking, drinking status, and total ounces of ethanol in the past 30 days (data not shown).

Table 3. Partial Correlation Coefficients Between Anthropometric Variables and Liver Enzymes—The Western New York Health Study, 1995–2001
VariableALTASTGGT
  • a

    P ≤ 0.05

  • **

    P ≤ 0.01

  • ***, *

    P ≤ 0.001.

  • Coefficients adjusted for age.

  • Coefficients adjusted for age and abdominal height.

  • §

    Coefficients adjusted for age and body mass index.

Pre-menopausal women (n = 511)
BMI0.340***0.204***0.352***
Abdominal height0.360***0.172***0.387***
Waist circumference0.371***0.204***0.370***
Waist/hip ratio0.250***0.136**0.273***
BMI0.0320.116**0.001
Abdominal height§0.131**−0.0330.172***
Waist circumference§0.158***0.0470.129**
Post-menopausal women (n = 1008)
BMI0.167***−0.0100.156***
Abdominal height0.181***−0.0020.211***
Waist circumference0.156***−0.0430.195***
Waist/hip ratio0.109***−0.0500.181***
BMI0.022−0.016−0.054
Abdominal height§0.074*0.0130.154***
Waist circumference§0.022−0.0690.122***
Men (n = 1185)
BMI0.193***−0.0160.154***
Abdominal height0.227***0.0260.205***
Waist circumference0.218***0.0240.172***
Waist/hip ratio0.151***0.0430.113***
BMI−0.002−0.071−0.037
Abdominal height§0.122***0.074**0.141***
Waist circumference0.104***0.077**0.079**

Multiple linear regression models were performed to further assess the relative influence of anthropometric variables on liver enzymes. Based on the results of partial correlation analyses, BMI as a measure of overall adiposity and abdominal height as a marker of visceral fat accumulation were chosen as independent variables, together with other covariates (age, race, education, smoking status, pack-years of smoking, drinking status, and total ounces of ethanol in the past 30 days). For the dependent variables, ALT and GGT were selected since they exhibited more consistent and stronger correlations with anthropometric indices than AST at bivariate analyses. The β-coefficients with their standard error (SE) and the standardized coefficients for all the covariates and the two anthropometric variables when they were included in the model individually as well as the fraction of variance (R2) explained by the two models (first model with BMI, second model with abdominal height) are shown in Tables 4 and 5 for multiple linear regression analyses of ALT and GGT, respectively. For ALT (Table 4), in premenopausal women the two anthropometric variables were significantly associated with ALT when they were included individually in the model, but abdominal height showed a higher standardized coefficient than BMI and the model including the former index was characterized by a higher R2 than the one with BMI. Among postmenopausal women, BMI and abdominal height were still significant correlates of ALT in the two separate models. The R2 of the model with abdominal height was slightly higher and the standardized coefficient of abdominal height was again higher in comparison with the one of BMI. In men, the two anthropometric indices showed significant associations with ALT when they were included individually in the model. However, abdominal height was characterized by a higher standardized coefficient than BMI, and the R2 was again higher in the model with abdominal height compared to the one including BMI. The results of multiple linear regression analysis of GGT are summarized in Table 5. In premenopausal women, the two anthropometric indices were significantly associated with GGT in the two separate models. Abdominal height still showed a more elevated standardized coefficient than BMI and the R2 as well was higher in the model including abdominal height compared to the one with BMI. Among either postmenopausal women or men, the two anthropometric variables were significant correlates of GGT in the two individual models. However, abdominal height was consistently characterized by higher standardized coefficients than BMI as well as the fraction of variance explained by the model including abdominal height was higher than the one explained by the model with BMI, in both sexes.

Table 4. Multiple Linear Regression Coefficients for ALT—The Western New York Health Study, 1995–2001
VariableModel With Body Mass IndexModel With Abdominal Height
βSEStandardized CoefficientsR2βSEStandardized CoefficientsR2
  • *

    P ≤ 0.05

  • b

    P ≤ 0.01

  • ***

    P ≤ 0.001.

  • All models evaluating anthropometric measures (BMI and abdominal height) include the variables of the baseline model: age, race (nonwhite, white), education (years), smoking status (never, former, current) and total pack-years of smoking, drinking status (lifetime abstainers, former, and current) and total ounces of ethanol consumed in the past 30 days.

  • Abbreviations: β, β-coefficient; SE, standard error; R2, R square.

Pre-menopausal women (n = 511)
Age−0.0010.005−0.010 −0.0020.005−0.016 
Race−0.0710.107−0.028 −0.0360.106−0.014 
Education0.023*0.0100.096 0.025*0.0100.104 
Smoking status−0.0060.042−0.009 0.0090.0420.013 
Pack-years of smoking0.0020.0030.0420.1320.00080.0030.0170.145
Drinking status0.00050.036−0.001 −0.0060.036−0.007 
Alcohol intake0.0050.0030.078 0.0040.0030.070 
BMI0.031***0.0040.369  
Abdominal height 0.053***0.0060.389 
Post-menopausal women (n = 1008)
Age0.00060.0020.010 −0.000080.002−0.001 
Race−0.0040.056−0.002 0.0120.0560.007 
Education−0.0020.007−0.008 −0.00090.007−0.005 
Smoking status−0.0460.029−0.069 −0.0440.029−0.067 
Pack-years of smoking−0.0010.001−0.0380.039−0.0010.001−0.0520.046
Drinking status0.0150.0230.022 0.0160.0230.024 
Alcohol intake0.0020.0020.041 0.0010.0020.032 
BMI0.015***0.0030.171  
Abdominal height 0.026***0.0040.190 
Men (n = 1185)
Age−0.011***0.001−0.273 −0.012***0.001−0.294 
Race−0.0380.051−0.021 −0.0310.050−0.017 
Education0.00070.0060.004 0.0020.0060.008 
Smoking status−0.0490.026−0.067 −0.050*0.026−0.068 
Pack-years of smoking−0.00080.001−0.0320.138−0.00090.001−0.0390.152
Drinking status−0.0170.028−0.018 −0.0170.028−0.018 
Alcohol intake0.003***0.0010.142 0.003***0.0010.137 
BMI0.023***0.0030.194  
Abdominal height 0.035***0.0040.226 
Table 5. Multiple Linear Regression Coefficients for GGT—The Western New York Health Study, 1995–2001
VariableModel With Body Mass IndexModel With Abdominal Height
βSEStandardized CoefficientsR2βSEStandardized CoefficientsR2
  • *

    P ≤ 0.05

  • **

    P ≤ 0.01

  • ***

    P ≤ 0.001.

  • All models evaluating anthropometric measures (BMI and abdominal height) include the variables of the baseline model: age, race (nonwhite, white), education (years), smoking status (never, former, current) and total pack-years of smoking, drinking status (lifetime abstainers, former, and current) and total ounces of ethanol consumed in the past 30 days.

  • Abbreviations: β, β-coefficient; SE, standard error; R2, R square.

Pre-menopausal women (n = 511)
Age0.0080.0060.060 0.0070.0060.053 
Race−0.454***0.123−0.150 −0.411***0.122−0.136 
Education−0.0050.012−0.017 −0.0020.012−0.008 
Smoking status0.0070.0490.008 0.0250.0480.030 
Pack-years of smoking0.007*0.0030.1130.2130.0050.0030.0880.229
Drinking status−0.0050.042−0.005 −0.0100.041−0.11 
Alcohol intake0.013***0.0030.182 0.013***0.0030.175 
BMI0.037***0.0040.364  
Abdominal height 0.064***0.0070.388 
Post-menopausal women (n = 1008)
Age0.0010.0020.020 0.00070.0020.009 
Race−0.336***0.074−0.143 −0.310***0.074−0.132 
Education−0.0140.009−0.051 −0.0130.009−0.046 
Smoking status0.0310.0390.035 0.0350.0380.040 
Pack-years of smoking−0.00020.001−0.0050.061−0.00080.001−0.0220.076
Drinking status−0.0040.030−0.004 −0.00040.0300.000 
Alcohol intake0.007***0.0020.120 0.007***0.0020.113 
BMI0.019***0.0040.162  
Abdominal height 0.037***0.0060.205 
Men (n = 1185)
Age−0.003*0.002−0.059 −0.004**0.002−0.078 
Race−0.252***0.064−0.114 −0.244***0.064−0.111 
Education−0.0080.007−0.034 −0.0070.007−0.029 
Smoking status0.0090.0340.010 0.0090.0330.010 
Pack-years of smoking0.00030.0010.0090.0870.000040.0010.0010.102
Drinking status−0.0010.036−0.001 −0.0020.035−0.001 
Alcohol intake0.006***0.0010.220 0.006***0.0010.216 
BMI0.025***0.0040.173  
Abdominal height 0.041***0.0050.212 

Discussion

In the present study, we evaluated the relative role of central fat accumulation, as assessed by abdominal height, and overall adiposity, as determined by BMI, on the relationship between body weight and the most widely used liver function tests in a randomized sample of the general population. We found that abdominal height was consistently more correlated with ALT and GGT levels than BMI in both sexes. In addition, in a multivariate model, abdominal height followed by BMI were the most powerful independent predictors of ALT in both sexes and of GGT among women, either pre- or postmenopausal. These findings confirm the importance of overweight and body fat distribution as predictors of liver enzyme levels; these factors may be even more important than alcohol consumption, especially among women.2–7, 12

These findings support a role of central adiposity independent from overall adiposity in predicting increased levels of hepatic enzymes and, consequently, potential liver damage. Moreover, the associations between anthropometric measures and liver enzymes tended to be stronger in the group of premenopausal women than in the postmenopausal women. This finding may indicate that the relative contribution of excess body weight and, in particular, visceral fat accumulation in predicting higher levels of hepatic enzymes is more important at a younger age, consistent with the notion that these risk factors tend to decrease with increasing age as predictive. Nevertheless, we cannot exclude the possibility that changes in hormonal metabolism may be responsible for the observed differences between pre- and postmenopausal women.

For AST, we detected significant associations with anthropometric indices only among premenopausal women in multivariate analyses (data not shown). The last results are not surprising, since other studies have already reported weak or inconsistent associations between body weight and AST,3, 5, 6 whereas a previous observational study of premenopausal obese women has found a direct correlation between WHR as a measure of central adiposity and both ALT and AST.14

Increased body weight and obesity are well-known risk factors for elevated serum liver enzyme values, especially for ALT, both in adults and children,2–7 and have been associated, more recently, with increased risk of cirrhosis-related death or hospitalization.29 Many epidemiological studies, both cross-sectional and prospective, have analyzed the relationship between body weight and liver enzyme activity using BMI to assess relative weight. Nonetheless, growing evidence suggests that the body fat distribution may be even more important than the grade of obesity as determined by the BMI in the relationship between body weight and potential liver damage.10, 11, 13–16 Most of the studies that have focused so far on the association of simply measured anthropometric indices of abdominal adiposity with liver enzyme activity have included very restricted samples and often clinical settings.13–15 Van Barneveld et al.13 found a strong linear association between GGT and WHR as an indicator of fat distribution in a randomly selected group (n = 69) of 38-year-old Dutch men from the city of Ede, Netherlands. In this study, GGT was more strongly correlated with WHR than with BMI, although without significant difference in correlations. In a study of premenopausal obese women with menstrual irregularity (n = 58), women characterized by upper body segment obesity (android type), as assessed by WHR >0.85, showed stronger correlations with metabolic abnormalities, including triglycerides, AST, and ALT compared to women with lower body fat localization (gynoid obesity).14 The relationship between anthropometric measures of body fat distribution and biochemical complications, such as elevation of ALT values, was detected as well among outpatient obese Japanese children.15 In a very recent analysis on data from NHANES III, WHR was found to be more strongly associated with elevated ALT activity, defined as an ALT >43 U/L for both sexes, than BMI.16 Our study is the first large population-based investigation that has evaluated the relationship between indices of both overall adiposity (BMI) and visceral fat distribution (abdominal height) and the three biochemical tests commonly used to assess potential liver damage (ALT, AST, and GGT). From our results, abdominal adiposity, simply measured by anthropometry, appears to be a slightly but consistently stronger predictor of hepatic enzymes (ALT and GGT) and, consequently, potential liver injury, than the grade of relative weight as determined by the BMI in both sexes. However, it is important to note that these two measures were very highly correlated. These findings are consistent with the current knowledge on the independent role of body fat distribution in predicting the risk of related diseases.8, 9 In particular, abdominal height has been shown to be highly correlated with the volume of visceral fat as determined by multiscan tomography19–22 and has been extensively studied in the association with cardiovascular risk factors.21, 30 It appears to be, together with the waist circumference, the best anthropometric measure of the extent of abdominal adiposity that is closely related to abdominal visceral adipose tissue accumulation, whereas the WHR seems to provide information on the regional distribution of adipose tissue that is independent of the degree of obesity and less closely linked to the amount of abdominal visceral adipose tissue. Moreover, it has been shown that central body fat distribution can correlate with the development of fatty liver,10, 11 which has been postulated as the most likely cause of abnormal biochemical liver tests in overweight and obese individuals2 and to be the most common cause of liver injury in the United States.31, 32 In addition, increased levels of ALT and GGT, together with triglycerides, seem to be the most sensitive biochemical indicators of hepatic steatosis in both sexes.11, 12 Our findings are further suggestive of this relationship between fatty liver and biochemical abnormalities assuming the abdominal height measure as a marker for hepatic steatosis.

Our study is not without limitations. First of all, the cross-sectional design does not allow us to make any conclusive statement about the temporality of the observed associations. Furthermore, the suboptimal participation rate (59.5%) may leave the possibility for selection bias and restrict the generalization of our findings to the general public. However, it should be pointed out that our participation rate reflects the overall participation rate of similar studies conducted in the USA that require a visit to a clinical center and a complex examination and interview. In addition, we cannot rule out the presence of additional unknown confounding variables that we were unable to control in our analyses. Moreover, we were not able to exclude other risk factors for liver disease, such as drug use, past blood transfusions, or high-risk sexual activity. The presence of individuals in our sample with these risk factors (who may experience elevation of liver enzymes for reasons other than overweight or central fat distribution) may have resulted in an underestimation of the association of central adiposity with liver enzymes.

A further methodological issue resides in the high collinearity between anthropometric indices, which has already been reported in several previous studies,21, 30 and is not surprising, considering the wide range of adiposity within our sample. However, from the findings of multiple linear regression analyses abdominal height consistently appears as a better predictor of elevated ALT and GGT levels than BMI, even though the overall variance of both enzymes was not largely increased in the models including abdominal height compared to the ones with BMI among either women or men. In addition, these results were consistent as well in both postmenopausal women and men who reported a somewhat lower collinearity between the considered anthropometric variables compared with premenopausal women. Moreover, the partial correlation analyses showed that the relationship between abdominal height and liver enzymes was independent of the BMI, whereas the latter was no longer significantly associated with ALT and GGT after adjustment for abdominal height.

The strength of this study consists in the very detailed information elicited on several covariates related to hepatic enzymes as well as the excellent intra- and interobserver variability of the abdominal height measurement (intraclass correlation was 0.99 for both parameters). A further strength is that we enrolled participants randomly selected from the general population.

In conclusion, we found that abdominal height may be a better predictor of increased hepatic enzymes (ALT and GGT) than BMI and that both explained more of the variance in enzyme levels than other factors, including alcohol consumption. These results are among the first from a population-based study, since to date previous evidence on the specific association between anthropometric indices of central adiposity and biochemical hepatic tests has been related to selected samples in clinical settings. Our findings support a role of visceral fat accumulation independent from overall BMI in predicting increased levels of hepatic enzymes and, consequently, potential liver damage. Longitudinal studies are necessary to address this issue and demonstrate a temporal relation of this association.

Ancillary