Globally, nonalcoholic fatty liver disease (NAFLD) is the most prevalent chronic liver disease, with a prevalence of up to 30%.¹ Individuals with NAFLD have experienced higher all-cause mortality than the general population.¹ NAFLD-related fibrosis is rising in the United States (US), which could increase all-cause mortality.² Therefore, identifying individuals at risk of developing NAFLD and NAFLD-related fibrosis is essential for reducing the public health burden. Recommending lifestyle modification and early interventions in these high-risk populations might prevent the development of NAFLD and NAFLD-related fibrosis and reduce the burden of NAFLD-related morbidities and mortalities.

Obesity, a well-known risk factor for NAFLD and NAFLD-related fibrosis, is heterogeneous in the outcome because of the regional body fat distribution. Irrespective of the general obesity assessed by body mass index (BMI), body fat distribution is a significant risk factor for cardiometabolic abnormalities.³ BMI can not reflect the level and distribution of body fat. Regarding NAFLD, the results were inconsistent; some of the studies showed that visceral and subcutaneous fat correlate with NAFLD, whereas other studies showed that only visceral fat is associated with NAFLD.³ NAFLD-related fibrosis had a higher risk for all-cause mortality, and mortality risk increased as the fibrosis stage advanced.⁴ A study showed the independent association of visceral fat with significant fibrosis in patients with NAFLD.⁵ Above mentioned studies regarding this topic were limited by small sample sizes, specific race/ethnicity, and highly selected populations, which might account for the discrepant results. Not all individuals with obesity develop NAFLD and fibrosis, indicating that the role of obesity in the pathogenesis of NAFLD is complicated.⁶ Importantly, it was well-known that there were racial/ethnic differences in body fat distribution.⁷ For a given BMI, non-Hispanic Blacks have less body fat than non-Hispanic Whites, and non-Hispanic Asians have more body fat than non-Hispanic Whites.⁸ Therefore, we hypothesize that body fat distribution may impact NAFLD and significant fibrosis differently according to race/ethnicity. Few studies determined the association between body fat distribution and NAFLD and significant fibrosis based on diverse races/ethnicities. We evaluate the relationship between body fat distribution and NAFLD/significant fibrosis according to race/ethnicity in the US general population.

Subjects and study design

We analyzed the recent National Health and Nutrition Examination Survey (NHANES) 2017 to 2018 data, employing a multi-stage, clustered, and stratified probability sampling design to retrieve a nationally representative population of the US non-institutionalized civilians.⁹ The National Center for Health Statistics’ Institutional Review Board approved the original NHANES survey (Protocol #2011-2017 and #2018-01), and all participants reviewed and signed informed consent. Because the data used in the study was fully de-identified, this analysis was waived by Stanford University’s Institutional Review Board (IRB-57117).

A total of 2,740 adults (18 to 59 years of age) were examined for laboratory tests and dual-energy X-ray absorptiometry (DXA) at a mobile examination center. Among these, we excluded 494 participants with hepatitis C virus (by hepatitis C antibody), hepatitis B virus (by hepatitis B surface antigen), significant alcohol use (>20 g/day in women and >30 g/day in men), steatogenic medication for more than 6 months (corticosteroid, amiodarone, tamoxifen, valproate, and methotrexate), other races including multiracials, and/or those for whom data on BMI and/or transient elastography were not available. The final cohort consisted of 2,246 participants with complete data.

Clinical and laboratory evaluations

We used a previously described method.¹⁰ We defined race/ethnicity as non-Hispanic Whites, non-Hispanic Blacks, Hispanics, or non-Hispanic Asians. We defined marital status as marriage or living with a partner versus others. Educational status was dichotomized as high school graduation versus no high school graduation. We defined current smokers as individuals who reported ongoing current smoking among those who had smoked at least 100 cigarettes in their lifetime. We calculated alcohol consumption based on the amount and frequency of alcohol consumption using a self-reported questionnaire.¹¹ We defined hypertension as systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, and/or current treatment with anti-hypertensive medication. We defined diabetes as fasting plasma glucose levels ≥126 mg/dL, glycosylated hemoglobin ≥6.5%, and/or the use of hypoglycemic agents or insulin. We defined leisure-time physical activity according to the ‘2018 Physical Activity Guidelines for Americans’ (adults engaged in ≥150 minutes/week of moderate-intensity physical activity, 75 minutes/week of vigorous-intensity physical activity, or an equivalent combination).¹²

Definition of NAFLD and significant fibrosis

We used previously described methods for these definitions.¹³ Individuals were examined for liver stiffness measurement (LSM) and the controlled attenuation parameter (CAP) score by the Fibroscan 502 V2 Touch (Echosens).⁹ Incompleteness results for transient elastography were reported if they had stiffness interquartile range/median ≥30%, <10 complete LSMs, or fasted <3 hours.⁹ We defined NAFLD as CAP scores of 263 dB/m or more¹⁴ and significant fibrosis as LSM value of 8 kPa or higher^15-17 without significant alcohol consumption, other causes of liver disease, and use of steatogenic medication.

Body fat measurements

The NHANES DXA whole-body scans provide nationally representative data on abdominal fat distribution for age, sex, and racial/ethnic groups.¹⁸ DXA whole-body scans were eligible for participants aged 8 to 59, excluding pregnancy, self-reported radiographic contrast (barium) use in the past 7 days, and measured weight over 204 kg or height over 1.98 m.¹⁸ The whole-body scans were examined using the Hologic Discovery model A densitometers software version APEX 3.2 (Hologic Inc.).¹⁸ Visceral and subcutaneous adipose tissue volumes were defined by the Hologic APEX version 4.0 software.¹⁸ The visceral fat volume inside the abdominal cavity and subcutaneous fat volume outside the abdominal cavity were measured at the approximate interspace of the L4 and L5 vertebra.¹⁸

Statistical analysis

We analyzed the data by applying appropriate sample weights, stratification, and clustering to retrieve representative population-level data for the entire US non-institutionalized civilians because of the complex survey design of the NHANES. The independent relationship between visceral or subcutaneous fat volumes and CAP score or LSM value was determined by multiple regression analysis and the determination of the standardized correlation coefficients. Based on the weighted sample distribution, we calculated the weighted mean and standard deviation (SD) of visceral fat volume and subcutaneous fat volume according to sex and race/ethnicity. The visceral and subcutaneous fat volumes were standardized to a mean of 0 and an SD of 1 based on sex and race/ethnicity. The tests for the odds ratios (OR) and significance of the differences among the visceral and subcutaneous fat volumes were performed to estimate the association between each type of fat and NAFLD/significant fibrosis. The OR per 1-SD was used to show the relative strength of the relationship across race/ethnicity. After adjusting for clinical and metabolic confounders, multivariable logistic regression was performed to investigate the independent association between body fat distribution and NAFLD and significant fibrosis. Using Taylor series linearization, we performed analyses using STATA version 17.0 (Stata Corp.).

Characteristics of the study population

As mentioned in the ‘Methods,’ this study enrolled 2,246 individuals, corresponding to 115.2 million US adults. The weighted prevalence of non-Hispanic Whites, non-Hispanic Blacks, Hispanics, and non-Hispanic Asians was 60.5% (95% confidence interval [CI], 54.5 to 66.2; 69.7 million), 12.6% (95% CI, 8.9 to 17.4; 14.5 million), 19.7% (95% CI, 14.4 to 26.4; 22.7 million), and 7.2% (95% CI, 5.1 to 10.1; 8 million), respectively. Supplementary Table 1 shows the study population’s baseline characteristics. Body fat distribution in men according to race/ethnicity was provided in Fig. 1. In men, visceral and subcutaneous fat volume among non-Hispanic Blacks was lowest compared to other races/ethnicities. Fig. 2 shows body fat distribution in women according to race/ethnicity. Visceral fat volume was lowest in non-Hispanic Asians and non-Hispanic Blacks in order compared with other races/ethnicities. However, subcutaneous fat volume was highest among non-Hispanic Blacks, who had the lowest visceral fat volume out of total fat volume and lowest among non-Hispanic Asians. Visceral fat volume was higher in men than women, while subcutaneous fat volume was higher in women than in men across the race/ethnicity.

Association between adipose tissue volume and NAFLD

Table 1 showed correlations adjusted for age and sex and additional variables between body fat distribution and CAP scores or LSM. Visceral and subcutaneous fat volumes were statistically correlated to CAP score in both models. Subcutaneous fat volume was more strongly correlated to CAP score in non-Hispanic Blacks than other races/ethnicities. In contrast, visceral and subcutaneous fat volumes showed weaker correlations to CAP scores in non-Hispanic Asians than in other racial/ethnic groups. As shown in Table 2, in the age and sex-adjusted models, the ORs for NAFLD per 1-SD of visceral fat volume and subcutaneous fat volume were 5.52 (95% CI, 3.69 to 8.26) and 4.05 (95% CI, 3.19 to 5.14), respectively. After further adjusting for education status, marital status, BMI, smoking status, diabetes, hypertension, leisure-time physical activity, total cholesterol, high-density lipoprotein-cholesterol, and total calorie intake, the OR for NAFLD per 1-SD in visceral fat volume and subcutaneous fat volume was 2.31 (95% CI, 1.50 to 3.39) and 1.93 (95% CI, 1.43 to 2.61), respectively. This association persisted after simultaneously adjusting for visceral and subcutaneous fat volume. Both visceral fat volume and subcutaneous fat volume were associated with NAFLD (OR per 1-SD of visceral fat volume, 2.05 [95% CI, 1.36 to 3.09]; and OR per 1-SD of subcutaneous fat volume, 1.48 [95% CI, 1.04 to 2.09]). When we performed similar analyses based on race/ethnicity, there were some differences in the impact of body fat distribution and NAFLD between race/ethnicity. In terms of the impact of visceral fat volume on NAFLD, non-Hispanic Blacks had the highest odds (OR, 2.86; 95% CI, 1.45 to 5.62), and non-Hispanic Whites (OR, 2.29; 95% CI, 1.19 to 4.40) and non-Hispanic Asians (OR, 1.61; 95% CI, 1.13 to 2.29) were in order. Although there was a significant association between subcutaneous fat volume and NAFLD in the total population, this association remained solely significant among non-Hispanic Blacks (OR, 1.55; 95% CI, 1.01 to 2.38).

Association between adipose tissue volume and significant fibrosis

Regarding LSM (Table 1), visceral and subcutaneous fat volumes were statistically correlated to LSM except for non-Hispanic Whites. Visceral fat volume presented the highest correlation coefficients for LSM in non-Hispanic Asians and non-Hispanic Blacks, whereas subcutaneous fat volume showed weaker correlations with LSM in Hispanic and non-Hispanic Asians.

As shown in Table 3, a significant association between visceral fat volume, not subcutaneous fat volume, and significant fibrosis was noted in the total population regardless of NAFLD status (OR per 1-SD of visceral fat volume, 1.37 [95% CI, 1.02 to 1.72; P=0.036]; and OR per 1-SD of subcutaneous fat volume, 1.53 [95% CI, 0.90 to 2.61; P=0.111]). When we performed analyses among individuals with NAFLD, we found significant associations between subcutaneous fat volume (OR, 2.10; 95% CI, 1.34 to 3.29; P=0.003) or visceral fat volume (OR, 1.35; 95% CI, 1.05 to 1.73; P=0.023) and significant fibrosis (Table 4). In terms of the impact of subcutaneous fat volume on significant fibrosis in NAFLD based on race/ethnicity, Hispanics had the highest odds (OR, 2.74; 95% CI, 1.32 to 5.70), and non-Hispanic Whites (OR, 2.35; 95% CI, 1.11 to 4.98) and non-Hispanic Asians (OR, 2.01; 95% CI, 1.01 to 4.01) were in order. Although there was a significant association between visceral fat volume and significant fibrosis in NAFLD, this estimate remained similar (24%–46% for each race/ethnicity vs. 35% for entire NAFLD) but non-significant across the race/ethnicity.

Using nationally representative data on abdominal fat distribution for age, sex, and racial/ethnic groups, we found that visceral adiposity was more strongly associated with NAFLD than subcutaneous adiposity across the various races/ethnicities. Evidence suggests that not all fat contributes to disease risk in NAFLD in the same way. Visceral abdominal fat is more harmful than subcutaneous fat because visceral fat cells release adipokines that contribute to cardiometabolic abnormalities, including NAFLD.¹⁹ Several mechanisms explain visceral adiposity’s metabolically adverse effects: (1) a more metabolically active endocrine organ and (2) increased amounts of free fatty acids reach the liver due to the anatomic location allowing direct access to the portal venous system.²⁰

Although visceral adiposity was associated with significant fibrosis in the total population and individuals with NAFLD, we found no significant association between visceral adiposity and significant fibrosis across race and ethnicity. Because we used only one cycle of the NHANES 2017 to 2018 dataset, statistical power due to the small sample across race and ethnicity may diminish. The NHANES 2017 to 2020 dataset had a unique opportunity to determine NAFLD and fibrosis using transient elastography. However, visceral adipose tissue and subcutaneous adipose tissue defined by the DXA scan analysis were included only in the NHANES 2017 to 2018. Therefore, we were unable to combine other cycles to increase the statistical power. In addition, studies have found that the relative distribution of body fat varies across racial and ethnic groups, suggesting that this might explain differences in association with significant fibrosis across racial and ethnic groups. For example, the impact of visceral and subcutaneous adiposity on cardiometabolic health has been investigated, showing racial/ethnic differences in the associations.^21-23 Subcutaneous abdominal fat in relation to cardiometabolic risk factors consistently follows similar patterns with visceral abdominal fat among some races/ethnicities, suggesting that subcutaneous abdominal fat is also associated with ethnic-specific cardiometabolic risk factors.²³ Interestingly, we found that subcutaneous fat volume was more strongly associated with significant fibrosis than visceral fat volume in some races/ethnicities. Although explaining the mechanistic pathway between subcutaneous fat volume and significant fibrosis is challenging, there may be several explanations. First, a recent in vivo and ex vivo study showed the potential impact of subcutaneous abdominal adipose tissue inflammation and exosomes on the pathogenesis of insulin resistance among individuals with NAFLD.²⁴ A statistically significant difference in subcutaneous abdominal adipose tissue immune cell populations and expression of most proinflammatory cytokines in people with NAFLD than those without NAFLD.²⁴ Second, subcutaneous adipose tissue is divided into two layers: superficial and deep subcutaneous adipose tissue.^25,26 Deep subcutaneous adipose tissue displayed an intermediate level of macrophage accumulation between superficial subcutaneous adipose tissue and visceral adipose tissue.^25,26 Lipolysis, lipogenesis, and inflammatory protein expression rates are higher in deep subcutaneous adipose tissue than in superficial subcutaneous adipose tissue.²⁶ Also, deep subcutaneous adipose tissue was associated with metabolic syndrome, increased inflammation, and oxidative stress, suggesting that deep subcutaneous adipose tissue may be an essential determinant of nonalcoholic steatohepatitis (NASH) and significant fibrosis. The abundance of macrophages in deep subcutaneous adipose tissue, but not in superficial subcutaneous adipose tissue, significantly increased among individuals with NASH and fibrosis with inflammation.²⁵ In individuals with NASH, longitudinal reductions in deep subcutaneous adipose tissue and potentially visceral adipose tissue volumes related to histologic improvement, independent of reduction in hepatic steatosis.²⁷ However, reduced superficial subcutaneous adipose tissue volume alone was not associated with histologic improvement in NASH.²⁷ These findings suggest that deep subcutaneous adipose tissue might be associated with the exacerbation of NASH, including liver inflammation, hepatocellular ballooning, and fibrosis.²⁷ Other studies have linked increased deep subcutaneous adipose tissue volume to altered insulin resistance,^26,28 which accelerates the development of NASH.²⁹ However, we were unable to determine deep or subcutaneous fat in our study, which was the limitation of our study. Third, subcutaneous fat is one of the main secretors of leptin, which was strongly correlated with NASH and significant fibrosis because leptin plays a role in regulating insulin levels.³⁰ Fourth, if the subcutaneous fat stores become saturated and limited storage capacity of subcutaneous fat, ectopic fat accumulation arises in NAFLD.³¹ We hypothesize that additional ectopic fat accumulation resulted in significant fibrosis in NAFLD, which may already be full of visceral adiposity. An animal study reported that steatosis and diabetes develop secondary to a plateau in adipose expansion, suggesting limited adipose capacity.³² Future studies are needed to determine the mechanistic pathway between subcutaneous fat and significant fibrosis in NAFLD across races/ethnicities.

Emerging evidence suggests racial/ethnic differences in body fat distribution. Although non-Hispanic Blacks had lower total visceral fat compared with non-Hispanic Whites at a similar BMI, non-Hispanic Blacks had paradoxically similar hepatic insulin sensitivity³³ and higher low-grade inflammatory markers, such as C-reactive protein and fibrinogen,^34,35 which may contribute to the increased impact of visceral fat on NAFLD. Especially, subcutaneous adiposity was more strongly associated with insulin resistance in non-Hispanic Blacks, suggesting that overall body fatness remains important in developing NAFLD in non-Hispanic Blacks.³⁵

In general, Asians have a higher percentage of visceral abdominal fat volume compared to Hispanics and Caucasians of the same age, sex, and BMI.³⁶ In this respect, there appears to be a more significant percentage of individuals with metabolic abnormality but normal weight phenotype among Asians compared with their European counterparts,³⁶ because BMI does not reflect regional body fat distribution. A recent study demonstrated a wide variation in the prevalence of metabolic abnormality but normal weight phenotype between major ethnic groups, with a rate of 21% in non-Hispanic Whites, 31% in non-Hispanic Blacks, 36% in Hispanics, 32% in Chinese Americans, and 44% in South Asians.³⁷ Asians may have the most deleterious body fat distributions of all, with lower subcutaneous fat and higher visceral fat despite lower absolute BMIs.^8,38 Impaired expansion of subcutaneous fat volume may predispose individuals to a metabolically unhealthy phenotype among Asians.³⁹ A Korean longitudinal study showed that higher visceral adiposity at baseline was longitudinally associated with a higher incidence of NAFLD.⁴⁰ Regarding the severity of NAFLD, a Korean study reported that visceral adiposity was independently associated with histology-confirmed NASH and significant fibrosis.⁵ Therefore, lifestyle modification to decrease visceral adiposity may help prevent the development of NAFLD and slow the progression of NAFLD-related fibrosis, especially in Asians.

Our study has several strengths. First, the NHANES provides nationally representative abdominal fat distribution data for age, sex, and racial/ethnic groups. Second, the NHANES’ clinical data were of high quality, i.e., measurements were taken by trained personnel using a standardized protocol. Third, we defined NAFLD and significant fibrosis by transient elastography, of which sufficient accuracy for detecting steatosis and fibrosis against biopsy has been reported.¹⁵ Therefore, our results could apply to clinical situations and be used to develop screening strategies in the US general population.

We acknowledge that this study has limitations. First, this cross-sectional analysis was unable to establish causality between body fat distribution and NAFLD. Second, we were unable to get liver histological samples, which is the gold standard for fibrosis. Third, no universal cut-off guideline for CAP score and LSM exists. However, we used the most validated cut-off point for CAP score and liver stiffness in several studies.^15-17

In conclusion, we showed that visceral adiposity was more strongly associated with NAFLD than subcutaneous adiposity; non-Hispanic Blacks had the highest odds, and non-Hispanic Whites and non-Hispanic Asians were in order. There is a stronger significant association between subcutaneous fat and significant fibrosis in individuals with NAFLD than visceral fat. Our study suggests that subcutaneous fat is not protective but a risk factor for significant fibrosis in individuals with NAFLD, especially in Hispanics, non-Hispanic Whites, and non-Hispanic Asians. These data suggest that a certain type of body fat may be a risk factor for NAFLD, whereas other types may be a risk factor for NAFLD-associated significant fibrosis. In addition, there were racial/ethnic differences in the association between body fat distribution on NAFLD and significant fibrosis.

Supplementary materials can be found online at https://doi.org/10.7570/jomes24005.

The authors declare no conflict of interest.

Study concept and design: DK and AA; acquisition of data: DK and AA; analysis and interpretation of data: DK, GC, and AA; drafting of the manuscript: DK; critical revision of the manuscript: DK, GC, and AA; statistical analysis: DK; administrative, technical, or material support: DK; and study supervision: DK and AA.