Graphical Abstract
INTRODUCTION
Non-alcoholic fatty liver disease (NAFLD) is the most common chronic liver disease worldwide. Its global prevalence increased from 20% in 2010 to 25% in 2018 [1], and, in the Republic of Korea, it increased from 29.0% (2001–2007 average) to 31.0% (2008–2014 average) [2]. The incidence rate of NAFLD in Hong Kong is approximately 34/1,000 population per year [3]. Established major risk factors include diabetes mellitus (DM), obesity, dyslipidemia, and metabolic syndrome, all of which are closely related to insulin resistance [4-6]. Therefore, the European Association for the Study of the Liver recommends that patients with insulin resistance and/or metabolic risk factors should undergo diagnostic tests for NAFLD [7].
Current evidence suggests that both hepatic and peripheral insulin resistance are strongly associated with NAFLD [8-13]. Hepatic insulin resistance refers to impaired suppression of glucose production and lipid synthesis by insulin in the hepatocytes [10]. It restricts the receptor-mediated entry of insulin into the liver and, consequently, impairs insulin clearance. At the same time, there is an increase in insulin secretion (hyperinsulinemia), contributing to hepatic steatosis [11]. Peripheral insulin resistance is also associated with an increased concentration of intrahepatic triglycerides [12]. It may contribute to liver steatosis by impairing the ability of insulin to suppress lipolysis, resulting in an increased delivery of free fatty acids to the liver, as well as by hyperinsulinemia, which stimulates de novo lipogenesis via sterol regulatory element binding protein-1c [13].
The gold standard method for direct evaluation of systemic insulin resistance is the euglycemic-hyperinsulinemic clamp (glucose clamp) [14]. Because of its invasiveness and high cost, however, indirect methods are generally used in clinical practice. The homeostatic model assessment for insulin resistance (HOMA-IR) is the most widely used insulin resistance index because of its strong correlation with the glucose clamp and its reflection of hepatic insulin resistance [15]. As hepatic insulin resistance is closely related to NAFLD, the HOMA-IR is commonly used in research to verify the relationship between insulin resistance and NAFLD [16,17]; however, this fails to take account of the contribution of peripheral insulin resistance to hepatic steatosis [9,13]. Meanwhile, the novel metabolic score for insulin resistance (METS-IR) is predictive of visceral fat and diabetes, was validated against the glucose clamp, and reflects peripheral insulin resistance [18]. Although it was validated in a Latin American population, its usefulness for the prediction of NAFLD in East Asian individuals is unclear.
Therefore, this study aimed to compare the use of the HOMA-IR and METS-IR for the prediction of prevalent and incident NAFLD in a large, community-based, prospective Korean cohort.
MATERIALS AND METHODS
Study population
All data originated from the Korean Genome and Epidemiology Study (KoGES), a longitudinal, prospective cohort study conducted by the Korea Centers for Disease Control and Prevention to evaluate the risk factors for non-communicable diseases [19]. In total, 10,030 adults, aged 40 to 69 years and living in urban (Ansan) and rural (Ansung) areas, were recruited into the KoGES_Ansan_Ansung cohort and studied biennially, beginning with the baseline survey conducted in 2001– 2002 and ending with the eighth follow-up in 2017–2018. Among 10,030 participants at baseline, 633 cumulative deaths were recorded between baseline and the fifth follow-up [19]. The participation rates for the first to sixth follow-ups were 86.1%, 76.0%, 68.2%, 68.9%, 65.3%, and 62.8%, respectively [19]. In addition, the eighth follow-up was conducted in 6,157 participants [20].
From the 10,030 participants in the baseline survey, we excluded 1) those with a history of hepatitis (n=423); 2) men with alcohol consumption ≥30 g/day or women with alcohol consumption ≥20 g/day (n=964); 3) those with insufficient data for calculation of the NAFLD-liver fat score (n=276); and 4) those with insufficient data for calculation of the METS-IR and/or HOMA-IR (n=7) (Fig. 1). We analyzed the data from the remaining 8,360 participants (6,142 participants without NAFLD and 2,218 participants with NAFLD) to compare the predictive power for the prevalence of NAFLD of METS-IR and HOMA-IR.
Enlarge this image.Figure 1. Flow chart of the study population selection. KoGES, Korean Genome and Epidemiology Study; METS-IR, metabolic score for insulin resistance; HOMA-IR, homeostatic assessment for insulin resistance; NAFLD, non-alcoholic fatty liver disease; AUROC, area under the receiver operating characteristic curve.
Moreover, to compare the predictive power for the incidence of NAFLD of METS-IR and HOMA-IR, we also analyzed the data from a total of 5,438 participants without NAFLD at baseline after excluding 1) those with NAFLD at baseline (n=2,218) and 2) those without follow-up data (n=704) from 8,360 participants without NAFLD at baseline.
The KoGES_Ansan_Ansung cohort protocol was reviewed and approved by the Institutional Review Board (IRB) of the Korea Centers for Disease Control and Prevention. All the participants gave written informed consent. This study protocol conformed to the ethical guidelines of the 1964 Declaration of Helsinki and its later amendments and was approved by the IRB of Nowon Eulji Medical Center (IRB No. 2021-09-025).
Measurements
Weight and height were measured to the nearest 0.1 kg and 0.001 m, respectively, for calculation of the body mass index (BMI; kg/m2). Waist circumference (WC; cm) was measured in the horizontal plane, midway between the lowest rib and the iliac crest. After at least 5 minutes of rest in a sitting position, the patient’s systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured. SBP and DBP were defined as the average of the last two of three measured values, with at least a 1-minute interval between each. We calculated the mean blood pressure (MBP) as DBP + 1/3 × (SBP − DBP).
Participants were categorized into four groups according to smoking status: never smoker, former smoker, some day smoker, or every day smoker [21]. A current drinker was defined as a man or woman who respectively drank ≤30 g or ≤20 g alcohol per day. Physical activity was measured as metabolic equivalent of task (MET)-hours per week (MET-hr/wk), according to participants’ self-reported hours spent sleeping, being sedentary (0 MET), and being engaged in very light (1.5 MET), light (3 MET), moderate (5 MET), and heavy (7 MET) physical activities [22]. Participants were categorized into three groups according to hr/wk of physical activity: low (<7.5 MET-hr/wk), moderate (7.5–30 MET-hr/wk), and high (>30 MET-hr/wk) [23].
Blood samples of each participant were collected after at least 8 hours of fasting and analyzed with a Hitachi 700-110 Chemistry Analyzer (Hitachi, Ltd., Tokyo, Japan). Concentrations of fasting plasma glucose (FPG), serum insulin, high-density lipoprotein (HDL) cholesterol, triglycerides, aspartate aminotransferase (AST), alanine aminotransferase (ALT), and C-reactive protein (CRP) were measured. Low-density lipoprotein (LDL) cholesterol was calculated by using the Friedewald equation in the case of a triglyceride concentration <400 mg/dL [24].
Participants’ dietary intake was assessed by using a validated, semi-quantitative, 103-item food frequency questionnaire [25]. Daily total energy intake (kcal/day) was calculated.
DM was defined as having an FPG concentration ≥126 mg/dL, plasma glucose concentration 2 hours after a 75-g oral glucose tolerance test ≥200 mg/dL, glycosylated hemoglobin ≥6.5%, being treated with anti-diabetic medications, or being treated with insulin therapy [26]. Hypertension (HTN) was defined as having an SBP ≥140 mmHg and/or a DBP ≥90 mmHg or being treated with anti-hypertensive medications [27]. Dyslipidemia was defined as having concentrations of serum total cholesterol ≥240 mg/dL, LDL cholesterol ≥160 mg/dL, HDL cholesterol <40 mg/dL, or triglyceride ≥200 mg/dL, or being treated with lipid-lowering agents [28].
Assessment of insulin resistance
The METS-IR and HOMA-IR were calculated by using the following formulas [15,18]: (1) METS-IR = ln (2 × FPG [mg/dL] + fasting serum triglyceride [mg/dL]) × BMI (kg/m2) / ln (HDL cholesterol [mg/dL]); (2) HOMA-IR = (fasting serum insulin [μIU/mL] × FPG [mg/dL] / 405).
Assessment of NAFLD
NAFLD was defined as the NAFLD-liver fat score, calculated as follows: (3) NAFLD-liver fat score = −2.89 + 1.18 × metabolic syndrome (yes: 1, no: 0) + 0.45 × DM (yes: 2, no: 0) + 0.15 × insulin (µIU/mL) + 0.04 × AST (U/L) – 0.94 × AST/ALT. An NAFLD-liver fat score >−0.640 was considered indicative of NALFD [29].
Statistical analysis
Clinical characteristics of the 8,360 included participants were compared between those with and those without NAFLD at baseline. Baseline characteristics of the 5,438 participants without NAFLD at baseline were compared between those who developed NAFLD after the baseline survey and those who did not. Categorical variables including sex, smoking status, drinking status, physical activity, DM, HTN, and dyslipidemia were analyzed by using chi-square test and are presented as number (%). Some continuous variables including age, BMI, WC, MBP, FPG, HDL cholesterol, AST, ALT, total energy intake, NAFLD-liver fat score, METS-IR value, and HOMA-IR, which showed P-value for Kolmogorov-Smirnov test ≥0.05, were analyzed by using Student’s t-test and are presented means±standard deviations. The other continuous variables including insulin, triglyceride, and CRP, which showed P-value for Kolmogorov-Smirnov test <0.05, were analyzed by using Mann-Whitney test and are presented as medians (25th percentile, 75th percentile).
For the 8,360 participants at baseline, the predictive powers of the indices for prevalent NAFLD were compared by using area under the receiver operating characteristic (ROC) curves (AUCs). The cut-off points for such prediction were calculated by using the Youden index [30].
For the 5,438 participants without NAFLD at baseline, a Cox proportional-hazards model was fitted with spline curves to determine the dose-response relationship between each index and incident NAFLD. We calculated the hazard ratio (HR) with its 95% confidence interval (CI) for incident NAFLD for each 1-point increase in each index by using univariable and multivariable Cox proportional-hazards regression analysis. In model 1, we adjusted for age, sex, BMI, WC, physical activity, smoking status, and current drinking status. In model 2, we further adjusted for total caloric intake, serum CRP concentration, DM, HTN, dyslipidemia, and serum ALT concentration. The indices’ predictive powers and discriminatory capabilities for incident NAFLD were assessed by using Harrell’s concordance index and time-dependent ROC curve analysis [31-34]. Heagerty’s integrated AUC (iAUC), Heagerty’s AUC at 8 years, and Heagerty’s AUC at 16 years were used as time-dependent AUCs, with an unadjusted survival analysis framework approach [32-34]. We used a bootstrapping method to calculate the differences and 95% CIs of Heagerty’s iAUC and AUC between the indices. Contal and O’Quigley’s method was used to calculate cut-off points for the prediction of incident NAFLD with the indices [35]. Subgroup analysis by obesity status and DM status were performed and are presented as a forest plot.
All statistical analyses were performed with SAS statistical software (version 9.4; SAS Institute Inc., Cary, NC, USA) and R software (version 4.1.1; R Foundation for Statistical Computing, Vienna, Austria). The significance level was set at P<0.05.
RESULTS
Characteristics of the study population
The number of participants with NAFLD was 2,218. The mean age, BMI, WC, MBP, total caloric intake, NAFLD-liver fat score, METS-IR, and HOMA-IR; the mean concentration of FPG, serum AST, and ALT; as well as the median concentrations of serum insulin, triglycerides, and CRP were higher in participants with than in those without NAFLD at baseline (Table 1). There was a higher proportion of former smokers and more participants who were in engaged in low amounts of physical activity, had DM, had HTN, and had dyslipidemia among those with NAFLD at baseline. The mean value of HDL cholesterol and the proportion of current drinkers were lower in participants with NAFLD at baseline.
Table 1.
Baseline characteristics of the population with or without NAFLD at the baseline survey
| Variable | Without NAFLD (n=6,142) | With NAFLD (n=2,218) | Total (n=8,360) | P-value* |
|---|---|---|---|---|
| Men sex | 2,526 (41.1) | 958 (43.2) | 3,484 (41.7) | 0.096 |
| Age (years) | 51.7±8.9 | 54.3±8.7 | 52.4±8.9 | <0.001 |
| Body mass index (kg/m2) | 23.8±2.9 | 26.5±3.0 | 24.6±3.2 | <0.001 |
| Waist circumference (cm) | 79.9±8.1 | 88.9±7.7 | 82.3±8.9 | <0.001 |
| MBP (mmHg) | 94.0±12.8 | 102.8±12.4 | 96.3±13.3 | <0.001 |
| Smoking status | 0.004 | |||
| Never smoker | 4,039 (66.7) | 1,377 (63.0) | 5,416 (65.7) | |
| Former smoker | 760 (12.5) | 334 (15.3) | 1,094 (13.3) | |
| Some day smoker | 129 (2.1) | 52 (2.4) | 181 (2.2) | |
| Every day smoker | 1,129 (18.6) | 421 (19.3) | 1,550 (18.8) | |
| Current drinker | 2,591 (42.6) | 833 (38.0) | 3,424 (41.3) | <0.001 |
| Physical activity | 0.021 | |||
| Low, <7.5 METs-hr/wk | 472 (8.0) | 210 (9.9) | 682 (8.5) | |
| Moderate, 7.5–30 METs-hr/wk | 3,639 (61.6) | 1,295 (61.1) | 4,934 (61.5) | |
| High, >30 METs-hr/wk | 1,794 (30.4) | 615 (29.0) | 2,409 (30.0) | |
| FPG (mg/dL) | 82.8±13.4 | 96.5±30.0 | 86.4±20.2 | <0.001 |
| Insulin (µIU/mL) | 6.3 (4.8, 8.2) | 10.1 (7.5, 12.6) | 7.0 (5.2, 9.7) | <0.001 |
| Triglyceride (mg/dL) | 119.0 (92.0, 156.0) | 187.0 (144.0, 256.0) | 133.0 (98.0, 185.0) | <0.001 |
| HDL cholesterol (mg/dL) | 46.1±10.0 | 40.3±8.5 | 44.5±9.9 | <0.001 |
| AST (U/L) | 26.5±7.1 | 34.9±29.6 | 28.7±16.8 | <0.001 |
| ALT (U/L) | 22.1±8.9 | 39.6±45.0 | 26.8±25.6 | <0.001 |
| CRP (mg/dL) | 0.1 (0.1, 0.2) | 0.2 (0.1, 0.3) | 0.2 (0.1, 0.3) | <0.001 |
| Total energy intake (kcal/day) | 1,929.2±699.5 | 1,977.1±731.8 | 1,941.9±708.5 | 0.009 |
| Type 2 diabetes | 196 (3.2) | 699 (31.5) | 895 (10.7) | <0.001 |
| Hypertension | 1,800 (29.3) | 1,371 (61.8) | 3,171 (37.9) | <0.001 |
| Dyslipidemia | 2,375 (38.7) | 1,613 (72.7) | 3,988 (47.7) | <0.001 |
| NAFLD-liver fat score | -1.9±0.6 | 0.5±1.5 | -1.3±1.4 | <0.001 |
| METS-IR | 35.8±5.6 | 43.3±6.2 | 37.8±6.6 | <0.001 |
| HOMA-IR | 1.3±0.6 | 2.6±2.0 | 1.7±1.3 | <0.001 |
Values are presented as mean±standard deviation, median (25th percentile, 75th percentile), or number (%).
NAFLD, non-alcoholic fatty liver disease; MBP, mean blood pressure; MET, metabolic equivalent of task; FPG, fasting plasma glucose; HDL, high-density lipoprotein; AST, aspartate aminotransferase; ALT, alanine aminotransferase; CRP, C-reactive protein; METS-IR, metabolic score for insulin resistance; HOMA-IR, homeostatic model assessment for insulin resistance.
* P-value for the comparison of the baseline characteristics between participants with NAFLD and those without NAFLD at the baseline survey. Significance was set at P<0.05.
The mean age, BMI, WC, MBP, NAFLD-liver fat score, METS-IR, and HOMA-IR; the mean concentrations of FPG, serum AST, and ALT; as well as the median concentrations of serum insulin, triglycerides, and CRP were higher in participants who developed NAFLD after baseline than in those who did not (Table 2). The proportions of participants with DM, HTN, and dyslipidemia were also higher in participants who developed NAFLD, whereas the mean value of HDL cholesterol was lower in participants who developed NAFLD.
Table 2.
Baseline characteristics of participants without NAFLD at baseline who followed-up at least once after baseline survey
| Variable | Not developed NAFLD (n=3,060) | Newly developed NAFLD (n=2,378) | Total (n=5,438) | P-value* |
|---|---|---|---|---|
| Men sex | 1,280 (41.8) | 979 (41.2) | 2,259 (41.5) | 0.643 |
| Age (years) | 51.4±9.0 | 52.0±8.5 | 51.7±8.8 | 0.011 |
| Body mass index (kg/m2) | 23.0±2.7 | 24.9±2.7 | 23.8±2.9 | <0.001 |
| Waist circumference (cm) | 77.6±7.7 | 83.1±7.4 | 80.0±8.1 | <0.001 |
| MBP (mmHg) | 92.1±12.5 | 96.3±12.4 | 94.0±12.6 | <0.001 |
| Smoking status | 0.895 | |||
| Never smoker | 2,031 (67.2) | 1,554 (66.3) | 3,585 (66.8) | |
| Former smoker | 380 (12.6) | 306 (13.0) | 686 (12.8) | |
| Some day smoker | 65 (2.2) | 49 (2.1) | 114 (2.1) | |
| Every day smoker | 547 (18.1) | 436 (18.6) | 983 (18.3) | |
| Current drinker | 1,287 (42.4) | 1,034 (43.9) | 2,321 (43.1) | 0.299 |
| Physical activity | 0.312 | |||
| Low, <7.5 METs-hr/wk | 222 (7.5) | 171 (7.5) | 393 (7.5) | |
| Moderate, 7.5–30 METs-hr/wk | 1,813 (61.5) | 1,360 (59.6) | 3,173 (60.7) | |
| High, >30 METs-hr/wk | 911 (30.9) | 750 (32.9) | 1,661 (31.8) | |
| FPG (mg/dL) | 80.9±8.4 | 84.6±14.8 | 82.5±11.8 | <0.001 |
| Insulin (µIU/mL) | 6.2 (4.7, 7.9) | 6.6 (5.0, 8.7) | 6.3 (4.8, 8.2) | <0.001 |
| Triglyceride (mg/dL) | 109.0 (85.0, 141.0) | 133.0 (104.0, 181.0) | 119.0 (92.0, 156.0) | <0.001 |
| HDL cholesterol (mg/dL) | 47.7±10.3 | 43.9±9.0 | 46.0±9.9 | <0.001 |
| AST (U/L) | 26.1±6.7 | 26.8±7.1 | 26.4±6.9 | <0.001 |
| ALT (U/L) | 20.8±8.1 | 23.8±9.5 | 22.1±8.9 | <0.001 |
| CRP (mg/dL) | 0.1 (0.0, 0.2) | 0.2 (0.1, 0.3) | 0.1 (0.1, 0.2) | <0.001 |
| Total energy intake (kcal/day) | 1,930.9±690.7 | 1,949.9±728.3 | 1,939.2±707.3 | 0.336 |
| Type 2 diabetes | 49 (1.6) | 108 (4.5) | 157 (2.9) | <0.001 |
| Hypertension | 736 (24.1) | 848 (35.7) | 1,584 (29.1) | <0.001 |
| Dyslipidemia | 948 (31.0) | 1,143 (48.1) | 2,091 (38.5) | <0.001 |
| NAFLD-liver fat score | -2.1±0.6 | -1.7±0.6 | -1.9±0.6 | <0.001 |
| METS-IR | 33.9±4.9 | 38.3±5.3 | 35.8±5.5 | <0.001 |
| HOMA-IR | 1.3±0.6 | 1.4±0.6 | 1.3±0.6 | <0.001 |
Values are presented as mean±standard deviation, median (25th percentile, 75th percentile), or number (%).
NAFLD, non-alcoholic fatty liver disease; MBP, mean blood pressure; MET, metabolic equivalent of task; FPG, fasting plasma glucose; HDL, high-density lipoprotein; AST, aspartate aminotransferase; ALT, alanine aminotransferase; CRP, C-reactive protein; METS-IR, metabolic score for insulin resistance; HOMA-IR, homeostatic model assessment for insulin resistance.
* P-value for the comparison of the baseline characteristics between participants who developed NAFLD and those who did not develop NAFLD. Significance was set at P<0.05.
During a total of 55,887.0 person-years of follow-up, 2,378 participants (43.7%) newly developed NAFLD. The mean follow-up time was 13.6 years, the incidence rate per 1,000 person-years was 42.6, and the incidence rate per 2 years ranged from 4.1 to 11.0 (Table 3).
Table 3.
Incidence of non-alcoholic fatty liver disease during follow-up
| Year range | Follow-up | Total | Incidence cases | Incidence rate (cases/2 years) |
|---|---|---|---|---|
| 2001–2002 | Baseline | 5,438 | ||
| 2003–2004 | 2 years | 3,745 | 411 | 11.0 |
| 2005–2006 | 4 years | 4,443 | 405 | 9.1 |
| 2007–2008 | 6 years | 4,116 | 432 | 10.5 |
| 2009–2010 | 8 years | 4,163 | 384 | 9.2 |
| 2011–2012 | 10 years | 3,931 | 184 | 4.7 |
| 2013–2014 | 12 years | 3,745 | 180 | 4.8 |
| 2015–2016 | 14 years | 3,844 | 235 | 6.1 |
| 2017–2018 | 16 years | 3,582 | 147 | 4.1 |
Comparison of predictive power between the indices for prevalent NAFLD
The AUCs of the METS-IR and HOMA-IR were 0.824 (0.814–0.834) and 0.831 (0.821–0.842), respectively, with no significant difference in their predictive powers (P=0.276) (Fig. 2). The cut-off points for the prediction of prevalent NAFLD with the METS-IR and HOMA-IR were 39.3 and 1.8, respectively.
Enlarge this image.Figure 2. Comparison of predictive power for prevalent non-alcoholic fatty liver disease of metabolic score for insulin resistance and homeostasis model assessment for insulin resistance. For the 8,360 participants at baseline, the predictive powers for prevalent NAFLD of METS-IR and HOMA-IR were compared by using area under the receiver operating characteristic curves. The cut-off points for such prediction were calculated by using the Youden index. ROC, receiver operating characteristic; NAFLD, non-alcoholic fatty liver disease; METS-IR, metabolic score for insulin resistance; HOMA-IR, homeostatic assessment model for insulin resistance; AUC, area under the receiver operating characteristic curve.
Longitudinal relationship between the indices and incident NAFLD
There was a positive dose-response association between the METS-IR and incident NAFLD, whereas the HOMA-IR had a J-shaped association with incident NAFLD (Fig. 3). The HR for incident NAFLD for each 1-point increase in the METS-IR and the HOMA-IR were 1.12 (95% CI, 1.11–1.13) and 1.41 (95% CI, 1.32–1.51), respectively (Table 4). In model 2, the adjusted HR were 1.11 (95% CI, 1.09–1.13) and 1.30 (95% CI, 1.21–1.39), respectively.
Enlarge this image.Figure 3. Cox proportional spline curves showing dose-response association between metabolic score for insulin resistance/homeostatic model assessment for insulin resistance and the incidence of non-alcoholic fatty liver disease. (A) Metabolic score for insulin resistance. (B) Homeostatic model assessment for insulin resistance. For the 5,438 participants without NAFLD at baseline, a Cox proportional-hazards model was fitted with spline curves to determine the dose-response relationship between METS-IR/HOMA-IR and incident NAFLD. METS-IR, metabolic score for insulin resistance; NAFLD, non-alcoholic fatty liver disease; HOMA-IR, homeostatic assessment model for insulin resistance.
Table 4.
Cox proportional hazard regression model for incident non-alcoholic fatty liver disease of two different insulin resistant indices
| Incident NAFLD------------ | |||
|---|---|---|---|
| HR | 95% CI | P-vaule | |
| METS-IR (per 1 increment) | |||
| Unadjusted | 1.12 | 1.11–1.13 | <0.001 |
| Model 1 | 1.12 | 1.10–1.13 | <0.001 |
| Model 2 | 1.11 | 1.09–1.13 | <0.001 |
| HOMA-IR (per 1 increment) | |||
| Unadjusted | 1.41 | 1.32–1.51 | <0.001 |
| Model 1 | 1.28 | 1.19–1.37 | <0.001 |
| Model 2 | 1.30 | 1.21–1.39 | <0.001 |
Model 1: adjusted for age, sex, body mass index, waist circumference, physical activity, smoking status, and drinking status; model 2: adjusted for the variables used in model 1 plus daily total energy intake, serum C-reactive protein level, type 2 diabetes mellitus, hypertension, dyslipidemia, and serum alanine aminotransferase level.
NAFLD, non-alcoholic fatty liver disease; HR, hazard ratio; CI, confidence interval; METS-IR, metabolic score for insulin resistance; HOMA-IR, homeostatic model assessment for insulin resistance.
Comparison of predictive power of the two indices for incident NAFLD
Harrell’s concordance index of the METS-IR was 0.697 (95% CI, 0.689–0.705), which was significantly higher than that of the HOMA-IR (0.556; 95% CI, 0.546–0.566; P<0.001) (Table 5). Similarly, Heagerty’s iAUC of the METS-IR was 0.683 (95% CI, 0.671–0.695), higher than that of the HOMA-IR (0.551; 95% CI, 0.539–0.563; P<0.001). Heagerty’s incident/dynamic AUC for the METS-IR at 8 years (0.669; 95% CI, 0.663–0.675) and at 16 years (0.670; 95% CI, 0.662–0.678) were also higher than that of the HOMA-IR (0.557; 95% CI, 0.553–0.561 and 0.556; 95% CI, 0.548–0.564, respectively; both P<0.001). The cut-off points for prediction of incident NAFLD with the METS-IR and HOMA-IR were 35.7 and 1.4, respectively. The joint use of the METS-IR and HOMA-IR for the prediction of incident NAFLD did not significantly differ from that with the METS-IR alone, except at 8 years, for which the predictive value of the two indices combined was higher than that of the METS-IR alone. Figure 4 presents a forest plot showing the predictive power for incident NAFLD by subgroups according to obesity and DM status. The Heagerty’s iAUCs of the METS-IR was 0.674 (95% CI, 0.627–0.721) in the DM subgroup, 0.655 (95% CI, 0.639–0.671) in the non-DM subgroup, 0.610 (95% CI, 0.594–0.626) in the obese subgroup, and 0.605 (95% CI, 0.566–0.644) in the non-obese subgroup, respectively, which was significantly higher than that of the HOMA-IR in each subgroup (P<0.001 in all subgroups).
Enlarge this image.Figure 4. Forest plot showing the predictive power for incident non-alcoholic fatty liver disease by subgroups according to obesity and diabetes mellitus status. Heagerty’s integrated AUC was used as time-dependent AUC over the 16-year of follow-up period, with an unadjusted survival analysis framework approach. A bootstrapping method to calculate the differences and 95% CI of Heagerty’s integrated AUC between the METS-IR and HOMA-IR. DM, diabetes mellitus; METS-IR, metabolic score for insulin resistance; HOMA-IR, homeostatic model assessment for insulin resistance; AUC, area under the receiver operating characteristic curve; CI, confidence interval.
Table 5.
Comparison of predictive ability for incident non-alcoholic fatty liver disease between METS-IR and HOMA-IR using time-dependent receiver operating characteristics curves analysis
| Cut-off point | Harrell’s C index | Heagerty’s iAUC | Heagerty’s incident/dynamic AUC (8 years) | Heagerty’s incident/dynamic AUC (16 years) | |
|---|---|---|---|---|---|
| METS-IR, (1) | 35.7 | 0.697 (0.689 to 0.705) | 0.683 (0.671 to 0.695) | 0.669 (0.663 to 0.675) | 0.670 (0.662 to 0.678) |
| HOMA-IR, (2) | 1.4 | 0.556 (0.546 to 0.566) | 0.551 (0.539 to 0.563) | 0.557 (0.553 to 0.561) | 0.556 (0.548 to 0.564) |
| METS-IR+HOMA-IR, (3) | 0.700 (0.692 to 0.708) | 0.685 (0.673 to 0.697) | 0.673 (0.667 to 0.679) | 0.672 (0.664 to 0.680) | |
| Difference (1)-(2) | 0.131 (0.113 to 0.149) | 0.131 (0.113 to 0.149) | 0.112 (0.104 to 0.12) | 0.114 (0.104 to 0.124) | |
| Difference (1)-(3) | -0.003 (-0.011 to 0.005) | -0.002 (-0.016 to 0.012) | -0.005 (-0.009 to -0.001) | -0.002 (-0.008 to 0.004) | |
| Difference (2)-(3) | -0.144 (-0.150 to -0.138) | -0.133 (-0.145 to -0.121) | -0.116 (-0.120 to -0.112) | -0.116 (-0.122 to -0.110) | |
| P-value: (1) vs. (2) | <0.001 | <0.001 | <0.001 | <0.001 | |
| P-value: (1) vs. (3) | 0.453 | 0.775 | 0.012 | 0.505 | |
| P-value: (2) vs. (3) | <0.001 | <0.001 | <0.001 | <0.001 |
Significance was set at P<0.05.
METS-IR, metabolic score for insulin resistance; HOMA-IR, homeostatic model assessment for insulin resistance; iAUC, integrated area under the receiver operating characteristic curve; AUC, area under the receiver operating characteristic curve.
DISCUSSION
We compared the predictive power of two insulin resistance indices for prevalent and incident NAFLD by using data of a large-scale, community-based, prospective cohort. Both the METS-IR and the HOMA-IR were statistically significantly related to prevalent and incident NAFLD.
The optimal threshold of the HOMA-IR for prediction of prevalent NAFLD was 1.8, lower than the generally accepted cut-off value of 2.5 to define insulin resistance. Considering that both HOMA-IR and METS-IR had similar predictive powers for prevalent NAFLD in this study, it could be better for healthcare providers to apply HOMA-IR value of 1.8 or METS-IR value of 39.3 for the early detection of NAFLD.
Interestingly, the baseline METS-IR and HOMA-IR were both predictive of incident NAFLD over the study period. In addition, the predictive powers of the two indices for incident NAFLD had maintained at both 8 and 16 years after baseline. The optimal thresholds for the prediction of incident NAFLD (35.7 for the METS-IR and 1.4 for the HOMA-IR) were lower than those for prevalent NAFLD (39.3 and 1.8). The METS-IR may be preferable to the HOMA-IR for the prediction of incident NAFLD, as it had a higher predictive power (Heagerty’s iAUC, 0.683 vs. 0.551). Joint use of the METS-IR and HOMA-IR did not increase the predictive power for incident NAFLD over that of the METS-IR alone. This suggests that peripheral insulin resistance is more closely related to the development of NAFLD than hepatic insulin resistance.
There are several possible explanations for the superior predictive power of the METS-IR to that of the HOMA-IR for incident NAFLD in this study. First, as mentioned above, the development of NAFLD could be more closely related to peripheral than to hepatic insulin resistance. NAFLD is closely related to metabolic dysfunctions such as overweight/obesity, abdominal obesity, hypertriglyceridemia, low HDL cholesterolemia, chronic inflammation, impaired fasting glucose/DM, and insulin resistance [36]. Qureshi et al. [9] reported that, among 113 participants who were morbidly obese and non-diabetic, those with NAFLD (fatty liver and non-alcoholic steatohepatitis [NASH]) had higher whole-body insulin resistance than those without, whereas hepatic insulin resistance was higher only in those with NASH than in those without NAFLD, suggesting that fatty liver is a hepatic manifestation of peripheral insulin resistance. Our results support such a hypothesis. Hence, the METS-IR may be more representative of peripheral metabolic dysfunction than the HOMA-IR. As the METS-IR uses BMI and concentrations of FPG, serum triglycerides, and serum HDL cholesterol, it makes sense that it would be representative of metabolic dysfunctions related to NAFLD and be predictive for incident NAFLD. Second, use of the HOMA-IR may underestimate insulin resistance in Koreans because Asians have a smaller pancreas and lower insulin secretory capacity than Westerners with the same BMI [37,38]. Moreover, since the traditional Korean diet comprises a high intake of carbohydrate, translating to a high glycemic index [39], the HOMA-IR, which is calculated as the product of the concentrations of FPG and fasting serum insulin, may not accurately reflect the metabolic effects of post-prandial glucose and insulin on liver steatosis.
Several limitations of this study warrant discussion. First, NAFLD was defined according to a surrogate marker, namely, the NAFLD-liver fat score. In addition, there was a lack of information about the use of certain medications which can induce NAFLD. Second, insulin resistance was not classified into hepatic insulin resistance and peripheral insulin resistance, because there was a lack of data regarding plasma glucose and serum insulin concentrations 30 minutes after the 75-g oral glucose tolerance test, which are required to calculate hepatic insulin resistance [40]. Third, neither the METS-IR nor the HOMA-IR had an excellent predictive power for incident NAFLD. However, considering that the predictive power of the METS-IR for incident NAFLD at 8 and 16 years after baseline was close to 0.7, it may nonetheless prove a useful indicator for the early detection of individuals at risk of NAFLD, prompting early lifestyle intervention. In addition, since both the METS-IR and the HOMA-IR are time-varying variables, the effect of changes in these resistance indices on incident NAFLD should be verified in future studies. Finally, generalization of the results is limited to Korean population, as Koreans are relatively homogeneous in terms of genetic traits and lifestyles.
In conclusion, both the METS-IR and the HOMA-IR are highly predictive of prevalent NAFLD in middle aged and older Korean adults. Moreover, the METS-IR is superior to the HOMA-IR for the prediction of incident NAFLD. Our findings suggest that the METS-IR is more useful than the HOMA-IR for the early detection and prevention of NAFLD in Korean population. Further clinical trials are warranted to determine which index is most valuable with regard to NAFLD.
Authors’ contributions
Study concept and design: Jun-Hyuk Lee, Jee Hye Han, and Sang Bong Ahn; Data collection: Jun-Hyuk Lee, Kyongmin Park, Hye Sun Lee, and Hoon-Ki Park; Data analysis and interpretation: Jun-Hyuk Lee, Jee Hye Han, and Sang Bong Ahn; Manuscript writing: Jun-Hyuk Lee, Jee Hye Han, and Sang Bong Ahn; Final approval of the manuscript: All authors.
Acknowledgments
This paper was supported by Eulji University in 2021 (grant number: EJRG-21-13). Data in this study were from the Korean Genome and Epidemiology Study (KoGES; 4851-302), National Research Institute of Health, Centers for Disease Control and Prevention, Ministry for Health and Welfare, Republic of Korea.
Conflicts of Interest
The authors have no conflicts to disclose.
Abbreviations
ALT, alanine aminotransferase; AST, aspartate aminotransferase; AUC, area under the receiver operating characteristic curve; BMI, body mass index; CI, confidence interval; CRP, C-reactive protein; DBP, diastolic blood pressure; DM, diabetes mellitus; FPG, fasting plasma glucose; HDL, high-density lipoprotein; HOMA-IR, homeostatic model assessment for insulin resistance; HR, hazard ratio; HTN, hypertension; iAUC, integrated area under the receiver operating characteristic curve; IRB, Institutional Review Board; KoGES, Korean Genome and Epidemiology Study; LDL, low-density lipoprotein; MBP, mean blood pressure; MET, metabolic equivalent of task; METS-IR, metabolic score for insulin resistance; NAFLD, non-alcoholic fatty liver disease; NASH, non-alcoholic steatohepatitis; ROC, receiver operating characteristic; SBP, systolic blood pressure; WC, waist circumference
1. Younossi Z, Tacke F, Arrese M, Chander Sharma B, Mostafa I, Bugianesi E, et al. Global perspectives on nonalcoholic fatty liver disease and nonalcoholic steatohepatitis. Hepatology 2019;69:2672-2682.
2. Im HJ, Ahn YC, Wang JH, Lee MM, Son CG. Systematic review on the prevalence of nonalcoholic fatty liver disease in South Korea. Clin Res Hepatol Gastroenterol 2021;45:101526.
3. Marchesini G, Mazzotti A. NAFLD incidence and remission: only a matter of weight gain and weight loss? J Hepatol 2015;62:15-17.
4. Kang SH, Lee HW, Yoo JJ, Cho Y, Kim SU, Lee TH, et al. KASL clinical practice guidelines: management of nonalcoholic fatty liver disease. Clin Mol Hepatol 2021;27:363-401.
5. Duseja A, Chalasani N. Epidemiology and risk factors of nonalcoholic fatty liver disease (NAFLD). Hepatol Int 2013;7 Suppl 2:755-764.
6. Gruben N, Shiri-Sverdlov R, Koonen DP, Hofker MH. Nonalcoholic fatty liver disease: a main driver of insulin resistance or a dangerous liaison? Biochim Biophys Acta 2014;1842:2329-2343.
7. European Association for the Study of the Liver (EASL); European Association for the Study of Diabetes (EASD); European Association for the Study of Obesity (EASO). EASL-EASD-EASO clinical practice guidelines for the management of non-alcoholic fatty liver disease. J Hepatol 2016;64:1388-1402.
8. Birkenfeld AL, Shulman GI. Nonalcoholic fatty liver disease, hepatic insulin resistance, and type 2 diabetes. Hepatology 2014;59:713-723.
9. Qureshi K, Clements RH, Saeed F, Abrams GA. Comparative evaluation of whole body and hepatic insulin resistance using indices from oral glucose tolerance test in morbidly obese subjects with nonalcoholic fatty liver disease. J Obes 2010;2010:741521.
10. Santoleri D, Titchenell PM. Resolving the paradox of hepatic insulin resistance. Cell Mol Gastroenterol Hepatol 2019;7:447-456.
11. Najjar SM, Perdomo G. Hepatic insulin clearance: mechanism and physiology. Physiology (Bethesda) 2019;34:198-215.
12. Hwang JH, Stein DT, Barzilai N, Cui MH, Tonelli J, Kishore P, et al. Increased intrahepatic triglyceride is associated with peripheral insulin resistance: in vivo MR imaging and spectroscopy studies. Am J Physiol Endocrinol Metab 2007;293:E1663-E1669.
13. Utzschneider KM, Kahn SE. Review: the role of insulin resistance in nonalcoholic fatty liver disease. J Clin Endocrinol Metab 2006;91:4753-4761.
14. DeFronzo RA, Tobin JD, Andres R. Glucose clamp technique: a method for quantifying insulin secretion and resistance. Am J Physiol 1979;237:E214-E223.
15. Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC. Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia 1985;28:412-419.
16. Salgado AL, Carvalho Ld, Oliveira AC, Santos VN, Vieira JG, Parise ER. Insulin resistance index (HOMA-IR) in the differentiation of patients with non-alcoholic fatty liver disease and healthy individuals. Arq Gastroenterol 2010;47:165-169.
17. Isokuortti E, Zhou Y, Peltonen M, Bugianesi E, Clement K, Bonnefont-Rousselot D, et al. Use of HOMA-IR to diagnose non-alcoholic fatty liver disease: a population-based and interlaboratory study. Diabetologia 2017;60:1873-1882.
18. Bello-Chavolla OY, Almeda-Valdes P, Gomez-Velasco D, ViverosRuiz T, Cruz-Bautista I, Romo-Romo A, et al. METS-IR, a novel score to evaluate insulin sensitivity, is predictive of visceral adiposity and incident type 2 diabetes. Eur J Endocrinol 2018;178:533-544.
19. Kim Y, Han BG; KoGES group. Cohort profile: the Korean genome and epidemiology study (KoGES) consortium. Int J Epidemiol 2017;46:e20.
20. National Institute of Health. Follow-up status according to each KoGES cohort (Kor). National Institute of Health web site, <https://www.kdca.go.kr/contents.es?mid=a40504030900>. Accessed 6 Mar 2022.
21. Centers for Disease Control and Prevention (CDC). Adult Tobacco Use Information. CDC web site, <https://www.cdc.gov/nchs/nhis/tobacco/tobacco_glossary.htm>. Accessed 4 Jul 2021.
22. Ainsworth BE, Haskell WL, Whitt MC, Irwin ML, Swartz AM, Strath SJ, et al. Compendium of physical activities: an update of activity codes and MET intensities. Med Sci Sports Exerc 2000;32(9 Suppl):S498-S504.
23. Jang M, Won C, Choi H, Kim S, Park W, Kim D, et al. Effects of physical activity on fractures in adults: a community-based Korean cohort study. Korean J Sports Med 2017;35:97-102.
24. Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem 1972;18:499-502.
25. Ahn Y, Kwon E, Shim JE, Park MK, Joo Y, Kimm K, et al. Validation and reproducibility of food frequency questionnaire for Korean genome epidemiologic study. Eur J Clin Nutr 2007;61:1435-1441.
26. American Diabetes Association. 2. Classification and diagnosis of diabetes: standards of medical care in diabetes-2019. Diabetes Care 2019;42(Suppl 1):S13-S28.
27. Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL Jr, et al. The seventh report of the joint national committee on prevention, detection, evaluation, and treatment of high blood pressure: the JNC 7 report. JAMA 2003;289:2560-2572.
28. Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults. Executive summary of the third report of the national cholesterol education program (NCEP) Expert panel on detection, evaluation, and treatment of high blood cholesterol in adults (Adult Treatment Panel III). JAMA 2001;285:2486-2497.
29. Kotronen A, Peltonen M, Hakkarainen A, Sevastianova K, Bergholm R, Johansson LM, et al. Prediction of non-alcoholic fatty liver disease and liver fat using metabolic and genetic factors. Gastroenterology 2009;137:865-872.
30. Youden WJ. Index for rating diagnostic tests. Cancer 1950;3:32-35.
31. Harrell FE Jr, Lee KL, Matchar DB, Reichert TA. Regression models for prognostic prediction: advantages, problems, and suggested solutions. Cancer Treat Rep 1985;69:1071-1077.
32. Heagerty PJ, Lumley T, Pepe MS. Time-dependent ROC curves for censored survival data and a diagnostic marker. Biometrics 2000;56:337-344.
33. Heagerty PJ, Zheng Y. Survival model predictive accuracy and ROC curves. Biometrics 2005;61:92-105.
34. Kamarudin AN, Cox T, Kolamunnage-Dona R. Time-dependent ROC curve analysis in medical research: current methods and applications. BMC Med Res Methodol 2017;17:53.
35. Contal C, O’Quigley J. An application of changepoint methods instudying the effect of age on survival in breast cancer. Comput StatData Anal 1999;30:253-270.
36. Jarvis H, Craig D, Barker R, Spiers G, Stow D, Anstee QM, et al. Metabolic risk factors and incident advanced liver disease in non-alcoholic fatty liver disease (NAFLD): a systematic review and meta-analysis of population-based observational studies. PLoS Med 2020;17:e1003100.
37. Rhee SY, Kwon MK, Park BJ, Chon S, Jeong IK, Oh S, et al. Differences in insulin sensitivity and secretory capacity based on OGTT in subjects with impaired glucose regulation. Korean J Intern Med 2007;22:270-274.
38. Roh E, Kim KM, Park KS, Kim YJ, Chun EJ, Choi SH, et al. Comparison of pancreatic volume and fat amount linked with glucose homeostasis between healthy Caucasians and Koreans. Diabetes Obes Metab 2018;20:2642-2652.
39. Lee YJ, Song S, Song Y. High-carbohydrate diets and food patterns and their associations with metabolic disease in the Korean population. Yonsei Med J 2018;59:834-842.
40. Abdul-Ghani MA, Matsuda M, Balas B, DeFronzo RA. Muscle and liver insulin resistance indexes derived from the oral glucose tolerance test. Diabetes Care 2007;30:89-94.
Jun-Hyuk Lee
Department of Family Medicine, Nowon Eulji Medical Center, Eulji University School of Medicine, Seoul, Korea
Kyongmin Park
Department of Medicine, Graduate School of Hanyang University, Seoul, Korea
Hye Sun Lee
Biostatistics Collaboration Unit, Department of Research Affairs, Yonsei University College of Medicine, Seoul, Korea
Hoon-Ki Park
Department of Medicine, Graduate School of Hanyang University, Seoul, Korea
Jee Hye Han
Department of Family Medicine, Nowon Eulji Medical Center, Eulji University School of Medicine, Seoul, Korea
Sang Bong Ahn
Department of Internal Medicine, Nowon Eulji Medical Center, Eulji University School of Medicine, Seoul, Korea
© 2022. This work is published under http://creativecommons.org/licenses/by-nc/3.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
; Ahn, Sang Bong