Relationships Between Estimates of Adiposity, Insulin Resistance, and Nonalcoholic Fatty Liver Disease in a Large Group of Nondiabetic Korean Adults

The initial study population consisted of 56,249 subjects (33,546 men and 22,703 women) who volunteered for a health status evaluation in between 1 January and 30 September 2005 at Kangbuk Samsung Hospital, College of Medicine, Sungkyunkwan University. Data were obtained from a retrospective medical record review, and the study protocol was approved by the Kangbuk Samsung Hospital Ethics Committee. In addition to a complete medical history, physical examination, and chemical screening battery, all 56,249 subjects underwent an abdominal ultrasound and had blood drawn for serum levels of viral markers. To avoid confounding factors, the results of these preliminary observations resulted in the exclusion of a significant number from this analysis, based on: 1) a past history of diabetes or a fasting plasma glucose concentration >125 mg/dl (n = 2,630), 2) a history of malignancy (n = 293), 3) consumption of alcohol in amounts in excess of 70 g/week for women (n = 986) and 140 g/week for men (n = 8,688), or 4) hepatitis C antibody (HCV Ab) positivity (n = 107), hepatitis B surface antigen (HBsAg) positivity (n = 2,408), or solitary hepatitis B core antibody (HBcAb) positivity (n = 2,679). Also excluded from this analysis were subjects who took a variety of drugs reported to affect liver function (n = 376) and individuals with abnormal abdominal sonography results (evidence of malignancy, cirrhosis, or gall bladder disease) (n = 3,060). Thus, as some individuals met more than one exclusion criteria, the final number of subjects available for study was 36,654 (19,618 men and 17,036 women).

All subjects were seen after an overnight fast. Height and weight were determined, and BMI was expressed as weight (in kilograms) divided by the square of the height (in meters). Waist circumference was measured at the midlevel between the lowest rib and the iliac crest with the subject standing and breathing normally. Blood samples were collected, plasma separated, and alanine aminotransferase (ALT) measured by ultraviolet without P5P method (Advia 1650 Autoanalyzer; Byer Diagnostics, Leverkusen, Germany). HBsAg and HBsAb were measured by chemiluminescent microparticle immunoassay (Architect i2000 SR; Abbott, Abbott Park, IL). HBcAb immunoglobulin G was measured by radioimmunoassay (Titertek, AL). HCV Ab was measured by polymerase chain reaction (Cobas Amplicor; Roche, Basel, Switzerland). Insulin concentrations were determined by immunoradiometric assay (Biosource, Nivelles, Belgium), with intra- and interassay CVs of 2.1–4.5 and 4.7–12.2%, respectively. An enzymatic colorimetric test was used to measure total cholesterol and triglyceride concentrations (Hitachi 912 analyzer; Roche Diagnostics). The selective inhibition method was used to measure the level of HDL cholesterol, and a homogeneous enzymatic colorimetric test was used to measure the level of LDL cholesterol (Advia 1650 Autoanalyzer; Byer Diagnostics).

Abdominal ultrasonography (Aspen, Acuson, PA) was performed to detect the presence of fatty infiltration in the liver by a core laboratory of experienced radiologists, all of whom used standard criteria in evaluating the images for hepatic fat (11). The severity of the fatty liver was classified into three groups, grades 1–3, according to the Mittelstaedt classification (11). Subjects with both fatty liver and an ALT of <30 units/l for men and <19 units/l for women, as described by Prati et al. (12), were classified as having fatty liver with a normal ALT (FL-NALT). Patients with fatty liver and a serum ALT >30 units/l for men and >19 units/l for women were classified as having fatty liver with a high ALT (FL-HALT). These ALT cut points were determined after a review of previous work both in Koreans (13,14) and non-Koreans (12,15) in an effort to use sex-specific cut points that accurately detected abnormal values. Finally, subjects without fatty liver on ultrasound but an ALT >30 units/l for men or >19 units/l for women (n = 5,414 subjects) were not included in this analysis, but the mean FPI and TG/HDL-C ratio concentrations for this group were compared with the other groups and are presented.

Since specific measures of insulin-mediated glucose uptake had not been performed, the two following surrogate estimates of insulin sensitivity were used to assess the relationship between insulin resistance/hyperinsulinemia and fatty liver. FPI concentrations are significantly correlated with direct measures of insulin resistance (10,16). The plasma concentration of TG/HDL-C ratio is as closely related to direct measures of insulin resistance as FPI concentration (10) but has been shown to be superior to both triglycerides and HDL cholesterol alone (17).

Statistical analysis of the data were performed using SPSS version 14.0 (SPSS, Point Richmond, CA), and the continuous variable data are presented as means ± SD. Student's unpaired t test was used to compare the means of grade 1 fatty liver versus grades 2 and 3 combined. ANOVA was used to evaluate the overall differences in means. ANCOVA was used to assess the effect of weight (BMI and waist circumference) on the relationship between markers of insulin resistance and the presence of fatty liver. For the comparison of nominal variables, the χ² method was used for cross-tabulation analysis. Multiple logistic regression analysis was used to analyze relationships between age, BMI, waist circumference, and insulin and the TG/HDL-C ratio and the presence of FL-NALT and FL-HALT. P values <0.05 were considered to be statistically significant.

Overall, the 36,654 subjects were relatively young (mean ± SD age 41 ± 9 years [men 41 ± 8, women 41 ± 9]) and nonobese (BMI 23.0 ± 3.0 kg/m² [men 24.3 ± 2.8, women 22.2 ± 2.9] and waist circumference 78 ± 9 cm [men 83 ± 8, women 72 ± 8]). Hepatic ultrasonography indicated that 27,632 subjects were without evidence of fatty liver (75.4%), while the remaining subjects had grade 1 (21.4%, 6,149 men, 1,683 women), grade 2 (3.2%, 977 men, 181 women), or grade 3 (0.09%, 29 men, 3 women) fatty liver. In Table 1, the subjects are separated by fatty liver grade and sex. Subjects with grades 2 and 3 were combined into one category due to the very small numbers in grade 3. For the remainder of the presented analyses the fatty liver grades are not separated, and the data are presented for two groups: FL-NALT and FL-HALT.

The relevant clinical characteristics of the 36,654 subjects, divided into the three experimental groups (control, FL-NALT, and FL-HALT) are shown in Table 1, with the two fatty liver groups also divided into fatty liver grades. In general, the fatty liver groups had a higher overall (BMI) and abdominal (waist circumference) obesity than the control group, and the FL-HALT group was significantly more obese by both measures than the FL-NALT group. These data also show a significant trend of worsening metabolic variables (FPI, triglyceride, and HDL cholesterol concentrations and TG/HDL-C ratio) across the groups from the control to the FL-NALT and FL-HALT groups. After division of the two fatty liver groups into grades of fatty liver, the same can be said of men with grade 2 fatty liver in comparison with those with grade 1 fatty liver, within both the FL-NALT and the FL-HALT groups. Thus, in men, the degree of steatosis was positively associated with worsening metabolic variables. For women, although the division into grades of fatty liver also demonstrated an association between worsening metabolic variables and a higher grade of fatty liver, women with grade 2/3 fatty liver in the two fatty liver/ALT groups had equivalent mean values for variables.

Table 2 presents the mean values for FPI concentration and the TG/HDL-C ratio of the three experimental groups this time adjusted for differences in age, BMI, and waist circumference. These results indicate that the estimates of insulin resistance increased in magnitude in parallel with presence of fatty liver and increased ALT level, with each group being significantly different from the other two groups.

To quantify the adverse impact of obesity on the degree of liver disease, odds ratios (ORs) were calculated between age, BMI, and waist circumference and the presence of FL-NALT or FL-HALT (Table 3). For men, in this relatively young population, being older did not seem to have an adverse effect on the presence of NAFLD. However, in women, the two fatty liver groups were significantly older than the control subjects. In terms of adiposity, the larger the BMI or waist circumference, the greater the OR of having either FL-NALT or FL-HALT. The increase in ORs ranged from 10 to 30%, with the values for BMI and waist circumference of seemingly comparable magnitude.

As shown in Table 4, the impact of insulin resistance was assessed with the calculation of ORs between FPI and TG/HDL-C ratio concentrations and the presence of FL-NALT or FL-HALT. For this purpose, the experimental population was divided into quartiles on the basis of their FPI concentration or their TG/HDL-C ratio. Age, BMI, and waist circumference were included in the regression model for both the FPI and TG/HDL-C ratio analyses, so that the increasing quartiles of the two markers of insulin resistance are independent of these three variables. In terms of FPI concentration, when comparing each of the three highest insulin quartiles with the lowest, ORs for the presence of FL-HALT increased significantly in both men and women, with the OR increasing five- to sixfold in the quartile with the highest FPI concentration. Similar results are seen for TG/HDL-C ratio. When examining the ORs for FPI and TG/HDL-C ratio concentrations in predicting the presence of FL-NALT, the results are similar for women, albeit of smaller magnitude. However, for men, the OR for FL-NALT decreased with the highest FPI quartile, indicating a decrease in risk with this degree of hyperinsulinemia. This is difficult to explain but could indicate that in men an increasing FPI concentration is less associated with steatosis alone and more associated with steatohepatitis. This is with the caveat that ALT is only a moderately accurate marker of inflammatory activity. Thus, apart from this one exception, the higher the FPI and TG/HDL-C ratio concentrations, the significantly greater the OR of having either FL-NALT or FL-HALT.

A total of 5,414 subjects (3,412 women, 2,002 men) with an isolated elevated ALT (>30 units/l for men and >19 units/l for women), i.e., no fatty liver on ultrasound, were excluded from the above analyses. We compared the mean FPI and TG/HDL-C ratio concentrations for this group with those of our other groups and found that for both of these measures the mean values for the isolated ALT group were significantly lower than those in either the FL-NALT or the FL-HALT groups after adjusting for age, BMI, and waist circumference (data not shown).

Both obesity and insulin resistance have been identified as factors associated with deposition of fat in the liver, although it is unclear whether they are a cause of hepatic fat (18–21) or a result of hepatic fat (22–25). Because obesity and insulin resistance are themselves significantly correlated (26,27), it is difficult to clarify which of these two variables is most closely related to increased hepatic fat content. The problem is further confounded by questions as to the relative adverse impact of overall obesity as compared with abdominal obesity. This study is, to the best of our knowledge, the largest that has been performed in analyzing associations between measures of overall and abdominal obesity, insulin resistance, hepatic transaminases, and fatty liver, using a comprehensive evaluation including ultrasound, hepatic enzymes, and serum viral markers, in a nondiabetic apparently healthy population.

Based on the results in Tables 1 and 2, it appears that both surrogate estimates of insulin resistance used in this study (FPI and TG/HDL-C ratio concentrations) are independently associated with severity of liver disease. Thus, although the results in Table 1 indicate that both estimates of obesity (BMI and waist circumference) and the two surrogate estimates of insulin resistance all increased significantly as a function of the presence of steatosis and a raised ALT, it can be seen from Table 2 that FPI and TG/HDL-C ratio concentrations continued to be statistically associated with these features of liver disease when adjusted for differences in age, BMI, and waist circumference. Furthermore, when the subjects were divided into quartiles based on magnitude of surrogate estimates of insulin resistance, the OR of having evidence of FL-HALT increased significantly with each successively higher quartile (Table 4). As in the results in Table 2, these findings were independent of differences in age, BMI, and waist circumference and, with the exception of relatively minor quantitative differences, were true of both sexes. Thus, in answer to the question we posed, the evidence strongly supports the notion that although both overall and abdominal obesity are associated with fatty liver, the specific relationship between insulin resistance and fatty liver is independent of either overall or abdominal obesity.

It is of interest that the relationships between either index of obesity and hepatic fat content were quite comparable, so that neither waist circumference nor BMI was more predictive of hepatic fat than the other. However, the population under study was a very specific population of East Asian ancestry, with a lower mean BMI and waist circumference than seen in Caucasians, and these findings may not be applicable to other ethnic groups. Further, it should be emphasized that the specific cut points that performed the best in this analysis may only be applicable to individuals of Korean ethnicity.

Although the results of this study seem relatively straight forward, certain weaknesses should be acknowledged. First, subjects with sonographic evidence of cirrhosis were excluded, and in doing so some NAFLD chronic liver disease or cirrhosis may have also been excluded, explaining why we found such small numbers with grade 3 fatty liver. Second, NAFLD cirrhosis, where the hepatic fat tends to be greatly diminished, can be mistaken sonographically for normal nonfatty liver (28). Third, and most importantly, our conclusions are based on the combined use of hepatic ultrasonography and ALT measurements as markers of NAFLD, without morphological examination of hepatic tissue. We used “FL-HALT” to attempt to measure prevalence of inflammatory activity in the liver, to distinguish this from benign fatty liver. We acknowledge that serum ALT levels have been shown to be at best a moderate predictor of hepatic necroinflammatory damage or steatohepatitis (29–31); however, we would argue that the large number of apparently healthy individuals studied precludes the possibility of obtaining histology.

In summary, we demonstrate that although both adiposity and surrogate estimates of insulin resistance (FPI and TG/HDL-C ratio concentrations) are associated with the estimates of NAFLD in this healthy nondiabetic Korean population, these measures of insulin resistance predict NAFLD independently of both overall and abdominal overweight/obesity.

We acknowledge the efforts of the health screening group at Kangbuk Samsung Hospital, Seoul, South Korea.