Waist and hip circumference are independently associated with the risk of liver disease in population‐based studies

While several anthropometric measures predict liver disease, the waist‐hip ratio (WHR) has shown superiority in previous studies. We analysed independent and joint associations of waist circumference (WC) and hip circumference (HC) with liver disease and liver‐related risk factors.


| INTRODUC TI ON
In the present era of the obesity epidemic, non-viral chronic liver disease is becoming an increasing health concern and economic burden. 1 Chronic liver disease is usually asymptomatic and may silently progress to cirrhosis and end-stage liver disease, constituting the 11th leading cause of death, worldwide. 2 Early identification of persons at risk for progressive liver disease requires that clinicians are aware of a cluster of risk factors, encompassed under the concept of non-alcoholic fatty liver disease (NAFLD) or metabolic dysfunctionassociated fatty liver disease (MAFLD). 3 Obesity, as measured by the body mass index (BMI), is a risk factor for future severe liver disease; however, the risk effect is small. [4][5][6] Studies comparing different anthropometric measures consistently report that liver-related outcomes are more strongly associated with measures of abdominal obesity than with BMI. [7][8][9] Specifically, the waist-hip ratio (WHR) reportedly predicts clinical liver-related outcomes better than waist circumference (WC) alone, BMI, or waist-height ratio. [7][8][9] However, it remains unclear which liver disease-relevant metabolic, lifestyle or genetic factors are better reflected by WHR than the other anthropometric measures.
WHR includes the measure of hip circumference (HC), which exhibits an independent association with metabolic factors. 10,11 HC can also independently predict all-cause mortality. 12,13 However, the association between HC and liver-related outcomes has been scarcely studied. HC may reflect both muscle mass and subcutaneous adiposity. Indeed, the loss of muscle mass (sarcopenia) has been implicated in the pathophysiology of NAFLD [14][15][16] and liver-related outcomes. 17,18 An improved understanding of the anthropometric measures predictive of clinical liver disease will be important for population risk stratification, which in turn is necessary for early identification of persons at risk. In the present population-based study, we aimed to compare the WHR to BMI with regards to their associations with several lifestyle, metabolic and genetic risk factors for liver disease, and to analyse the differential associations with WC and HC. In the second part of this study, we evaluated independent associations and interactions of WC and HC with liver disease.

| Cross-sectional study
Cross-sectional analyses were performed using data from the Health 2000 Survey, which was coordinated by the Finnish Institute for Health and Welfare (previously National Public Health Institute).
This survey originally included 8028 adults of ≥30 years of age, and the participation rate in the full examinations was 80%. 19 The cohort is generated using a regional two-stage stratified cluster sampling procedure and is considered representative of the entire Finnish population. After excluding individuals with missing WHR measurements (n = 1409), our final data-set comprised 6619 individuals.

| Longitudinal cohort
Longitudinal analyses were performed using Health 2000 Survey data along with data from FINRISK, which is a national population survey conducted in Finland every 5 years by the Finnish Institute for Health and Welfare, using random representative population samples. 20  To collect follow-up data, the included individuals were linked with several national registers using the unique personal identity code assigned to all Finnish residents. Data regarding hospitalisations were obtained from the Care Register for Health Care (HILMO), which registers all hospitalisations in Finland since 1969. Upon discharge, one or several ICD diagnoses are assigned to each hospitalisation, and these diagnosis codes are systematically recorded in the HILMO register. Data regarding malignancies were obtained from the Finnish Cancer Registry, which includes nationwide cancer records since 1953. Data regarding vital status and cause of death were obtained from Statistics Finland. In Finland, by law, each person who dies is assigned a cause of death (using ICD codes) on the official death certificate, which is issued by the treating physician based on medical or autopsy evidence, or forensic evidence when necessary.
Then the cause-of-death codes are verified by medical experts at the register and recorded according to systematic coding principles.
Data reporting to all of these registries is mandatory by law, and the general quality is consistent and virtually 100% complete. 21,22 The study end-points were fatal and non-fatal severe liver disease (requiring hospital admission or causing liver cancer or liver-related death). Table S1 lists the ICD codes used for defining the outcomes.

| Baseline variables
At baseline, data were collected through interviews (Health 2000), questionnaires, and health examinations by trained physicians and

PLAIN SUMMARY
High values of waist-hip ratio (WHR) are risk factors for severe liver disease.
Waist circumference (WC) and hip circumference (HC) have independent and joint associations with liver disease risk particularly among those with a high WHR.
Five liver disease-related risk genotypes are associated with either WC, HC or body mass index. Respondents were asked to report how often they consumed al-

| Bioimpedance data
An eight-terminal body composition analyser was used in the Health

| Statistical analyses
For comparisons between groups, we used the chi-square or Mann-Whitney test, as appropriate. We calculated pairwise partial correlations between anthropometric measures and various other variables -with adjustment for age only, or for age plus anthropometric confounders -using the Spearman method for two continuous variables, and the point-biserial method when comparing a continuous and a dichotomous variable. To study independent associations between anthropometric measures and 59 candidate genetic risk variants associated with NAFLD/liver disease in previous studies [32][33][34][35] (Table S2) To study independent associations of WC and HC with liver enzymes, we performed linear regression analyses. The covariant was age, the independent variables were WC, and HC, and the dependent variable was either ALT, AST or γ-GT. These analyses were performed separately for men and women. In separate regression models, we also assessed the interaction between WC and HC by additionally including the interaction term between WC and HC.
In the longitudinal dataset, we used Cox proportional hazards models -with time to first liver-related event as the outcome variable -to analyse the associations between anthropometric measures and liver outcomes. We used one Cox model with age, WC, and HC as covariates, with separate analyses for men and women.
Another model additionally included the following covariates: systolic and diastolic blood pressure (BP), BMI, daily cigarette consumption, weekly alcohol consumption, diabetes, exercise habits and serum levels of cholesterol, LDL-C, HDL-C, triglycerides and CRP. To assess the two-way interaction effect between WC and HC, we included the interaction term between WC and HC in the

| Baseline characteristics
At baseline, men and women significantly differed with regards to almost every variable, except for total cholesterol, triglycerides, percentage of diabetes, γ-GT, CRP and exercise, in the cross-sectional analysis ( Table 1).

| Correlations between anthropometric measures and lifestyle, metabolic and bioimpedance variables
Tables S5 and S6 show the age-or multivariable-adjusted correlations between anthropometric measures and various healthrelated variables in the cross-sectional study. Figure 1 shows the significant correlations (P < .0028 in A and B and P < .0036 in C and D). WHR and BMI were interrelated (r = +.656 for men and r = +.496 for women, P < .001 for both). When comparing WHR to BMI, all correlations with a coefficient stronger than 0.3 showed similar directions for both WHR and BMI (Figure 1). Bioimpedance measures tended to correlate more strongly with BMI than WHR ( Figure 1C,D). No single correlation was stronger for WHR than for BMI (Figure 1). When comparing WC to HC, correlations with health-related variables were generally weak (r < .25), and the directions were generally opposite for WC and HC ( Figure 1; Table   S5). Compared to WC, HC tended to correlate more strongly with muscle mass, but not with SMI or arm muscular circumference ( Figure 1C,D).

| Associations with liver-related genetic risk variants
Age-and sex-adjusted linear regression analyses were used to assess the independent associations of BMI, WC and HC with 59 different liver disease-related risk genotypes in the cross-sectional study. These analyses revealed that five different genetic variants showed significant associations with at least one of these three anthropometric measures: one for WC, two for HC and two for BMI ( Figure 2).

| Independent associations of fat mass and muscle mass with WHR and HC
Age-adjusted linear regression analysis in the cross-sectional study

| Associations of WC and HC with liver enzymes
Age-adjusted linear regression analyses in the cross-sectional study revealed that WC and HC were independently associated with +0.690 and −0.461 for women) (P < .001 for all). The interaction term between WC and HC was significant for AST among women (P interaction < 0.001) and for ALT among men (P interaction < 0.001). The interaction effect translated into a steeper rise, especially in AST, along with increasing WC, in the high-WHR group compared to the low-WHR group (Figure 3).  (Figure 4). Among

| D ISCUSS I ON
In the present population-based cross-sectional and longitudinal cohort study, we found that WC and HC each exhibited independent associations with both liver enzymes and risk of liver disease, among both men and women. An interaction effect between WC and HC was evident, as especially ALT values showed a steeper rise along with increasing WC, among individuals with a visceral-type body composition (defined as a WHR above the sex-specific median) compared to those with low WHR. Similarly, among individuals with high TA B L E 2 Independent associations of waist and hip circumference with incident liver outcomes by Cox Regression analyses Abbreviations: HC, hip circumference; P int., P for interaction; WC, waist circumference. a Adjusted for age.
b Adjusted for age, systolic and diastolic blood pressure, BMI, daily cigarette consumption, weekly alcohol consumption, diabetes, exercise habits, and serum levels of cholesterol, LDL-C, HDL-C, triglycerides, and CRP.

F I G U R E 4
Associations between waist circumference and risk of liver outcomes among men (A) and among women (B), and associations between hip circumference and risk of liver outcomes among men (C) and among women (D) in subsets of individuals with low and high waist-hip ratio. Analyses were performed by Cox regression adjusted for age. WHR, waisthip ratio WHR, the risk of incident liver disease increased more steeply along with increasing WC values. Our interpretation is that HC modifies the association between abdominal obesity and liver disease. Although this modification effect was more apparent among men than women, the lack of significance in women may be due to insufficient statistical power (type II error). In order to estimate the possible risk profile of liver fibrosis of our study population already at enrolment, we performed a subgroup analysis applying the FIB-4 score, the results of which are in line with earlier population-based studies. 36 In general, WC and the WHR both measure abdominal (visceral) fat, while HC reflects a combination of muscle mass, subcutaneous fat in the gluteofemoral region and skeletal structure. 10 On the other hand, BMI reflects overall obesity, but not body fat distribution. 37 Muscle mass and subcutaneous fat might have beneficial metabolic functions that partially counteract the adverse effects of visceral adiposity. 14,38,39 Several studies indicate that hepatic steatosis is associated with central fat mass, lower extremity fat mass and appendicular lean mass. 40 A large Korean population-based study reported that sarcopenic subjects with NAFLD had an approximately two-fold increased risk of significant liver fibrosis, which was independent of obesity, insulin resistance and liver enzyme levels. 18 Two meta-analyses indicate that sarcopenia is a risk factor for NAFLD onset and is also associated with NAFLD progression to steatohepatitis and liver fibrosis. 15,16 Sarcopenic obesity seems to portend a particularly high risk. 41 Moreover, low physical activity, loss of skeletal muscle mass and strength, and low HC contribute to MetS and cardiovascular risks, and negatively affect the liver via NAFLD development. 14,17,18 The fact that HC reflects muscle mass and subcutaneous adiposity may explain the independent association of HC with liver disease, as well as why HC modifies the WC-associated risk of liver disease.
Insulin resistance is a strong driver of both NAFLD and alcoholic liver disease. 42 It was recently shown that genetic loci associated with insulin resistance are also associated with lower HC, which has been interpreted as a reduced ability to expand peripheral fat compartments (low HC) when challenged by a positive energy balance. 43 The interaction effects between waist and hip circumference for the risk of liver outcomes are visualised separately for men (A) and women (B), with adjustment for age. The colour code represents the hazard ratio (z-axis), with a more reddish colour indicating a higher hazard ratio key anthropometric measures only weakly reflect alcohol consumption and binge drinking in women.
Distinct genetic mechanisms may be linked to gluteofemoral and abdominal fat distribution. 38,47 Among the 59 candidate genetic risk variants associated with liver disease in previous studies, our present study showed that two variants in the tumour necrosis factorα (TNFα) gene (rs1800629A and rs361525A) were strongly associated with HC, independently of WC and BMI. TNFα has previously been linked with both non-alcoholic and alcoholic steatohepatitis. 32 Um et al previously reported that the TNFα rs361525A variant is associated with WHR, but not with BMI, among obese Koreans. 48 We have confirmed this finding in the Finnish population, and further demonstrated that this association is specifically related to the HC. Circulating TNFα is an inflammatory cytokine with catabolic properties predisposing to muscle wasting, 49 and genetic variation in the TNFα locus is reportedly related to muscle mass. 50 This suggests that the association between TNFα and HC may be secondary to differences in muscle mass, but this possibility merits further study.
The strengths of our study include the large population-based cohorts suitable for cross-sectional and longitudinal analysis, as well as the linkage with long-term outcome data with reliable national healthcare registers. The health-examination survey data enabled the combined analysis of large amounts of phenotype data, including bioimpedance data, and genotype data, along with both liver enzymes and incident liver disease. The cohorts were representative of the Finnish population and were thus ethnically homogenous.
Baseline data were collected by trained healthcare professionals.
One limitation of the study was that the bioimpedance measures were performed 20 years ago with the techniques that were used at that time. Additionally, the baseline data related to alcohol, smoking and exercise relied on information provided by the individuals participating in the cohorts. Notably, this was a prospective study performed in a single country, and the analyses may need to be replicated in ethnically diverse populations.
In conclusion, the results of the present population-based study emphasise the importance of the WHR as a risk factor for severe liver disease. WC and HC have independent and joint associations with liver disease risk, which increased along with increasing WC and HC among individuals with a high WHR. Furthermore, fat mass and muscle mass were almost consistently independently associated with WHR and HC in both sexes. Five liver disease-related risk genotypes were associated with either WC, HC, or BMI. WC was associated with serum liver enzymes in a WHR-dependent manner.
Moreover, high HC was associated with several beneficial healthrelated characteristics.

CO N FLI C T S O F I NTE R E S T
The authors do not have any conflicts of interest.

AUTH O R CO NTR I B UTI O N S
OD: study design, interpretation of results and writing the manuscript; MJN: study design, interpretation of results and writing the manuscript; AJ, VS, SM, AL, MP: study design, data collection, interpretation of results and critical revision of the manuscript; FÅ: study design, statistical analyses, interpretation of results and writing the manuscript. All authors revised and approved the manuscript.

R EPRO D U CE M ATER I A L
Dr Fredrik Åberg has the permission to analyse FINRISK and Health 2000 data from the THL biobank.

DATA AVA I L A B I L I T Y S TAT E M E N T
FINRISK and Health 2000 data are available from the THL biobank based on a research application, as explained on the website of the THL biobank (https://thl.fi/en/web/thl-bioba nk/for-resea rchers).