Markers of insulin resistance associated with non-alcoholic fatty liver disease in non-diabetic population

Insulin resistance (IR) plays an important role in the development of non-alcoholic fatty liver disease (NAFLD). IR markers are divided into two types: (1) insulin-based IR marker, homeostatic model assessment of IR (HOMA-IR); and (2) non-insulin-based IR markers, such as triglyceride-glucose (TyG) index, TyG index with body mass index (TyG-BMI), triglyceride/high-density lipoprotein cholesterol ratio (TG/HDL-c), and metabolic score for IR (METS-IR). The non-insulin-based IR markers are often associated with lipids. The aim of this study was to analyse the association between IR markers and NAFLD in non-diabetic population. Baseline data of NAFLD and non-NAFLD groups were compared. Logistic regression was used to evaluate the relationship between five IR markers and NAFLD risk. The odds ratios (ORs) and 95% confidence intervals (CIs) of IR markers were calculated. Receiver operating characteristic (ROC) curves and area under the curve (AUC) were used to evaluate the ability of different IR markers to detect NAFLD. Subgroup analyses were performed in obese and non-obese subgroups. This study found a positive correlation between NAFLD risk and elevation in five IR markers (HOMA-IR, TyG, TyG-BMI, TG/HDL-c, and METS-IR). In non-obese subjects, the AUC of TyG-BMI was larger than that of the other four IR markers to detect NAFLD. The AUC of HOMA-IR was larger than that of the other four IR markers to detect NAFLD in obese subjects. In non-diabetic population, the five IR markers are associated with the risk of NAFLD, including non-obese and obese NAFLD. TyG-BMI and HOMA-IR can be used to detect non-obese and obese NAFLD, respectively, with better detection ability compared with the other IR markers.

The global prevalence of non-alcoholic fatty liver disease (NAFLD) is around 25% 1 , with an increasing trend observed in China 2 .It is often associated with metabolic comorbidities 3 .Age is also considered a risk factor 4 .Although NAFLD is typically accompanied by obesity, it can also occur in individuals who are not obese, which is known as non-obese NAFLD.Subjects with non-obese NAFLD appear to have clinical outcomes comparable to or worse than subjects with obese NAFLD 5,6 .Non-obese NAFLD may be more prevalent in Asia than in other regions 7 .Because the indication for screening for non-obese NAFLD is unclear, patients may be easily missed when they do not fit the classic phenotype of obesity.Liver biopsy is the gold standard for diagnosis and staging of NAFLD, but it is limited by its invasiveness.Therefore, there is an urgent need for techniques to detect NAFLD, especially non-obese NAFLD, in order to early identify specific populations at risk for the disease.
Currently, diagnostic methods, such as ultrasound (US), computed tomography (CT), and magnetic resonance imaging (MRI) are routinely used to screen NAFLD.However, the high cost of CT and MRI is a limiting factor.US is relatively inexpensive but poorly sensitive for mild steatosis 8 .Moreover, US is often reported independently by a single physician, and diagnostic differences may exist between physicians.
Insulin resistance (IR) plays an important role in the development of NAFLD 9,10 .Baseline insulin resistance is associated with the progression of liver fibrosis in non-diabetic patients with NAFLD 11 .Therefore, IR markers may be the first step in early detection of the disease in non-diabetic populations.
The hyper-insulinemic euglycemic clamp is the gold standard for assessing insulin sensitivity 12 , but the cost and complexity limit its clinical application.Several studies have reported methods to predict IR, Commonly used IR markers including homeostatic model assessment of IR (HOMA-IR) 13,14 , triglyceride-glucose (TyG) index, TyG index with body mass index (TyG-BMI), triglyceride/high-density lipoprotein cholesterol ratio (TG/

Statistical methods
Statistical analysis was performed using SPSS Statistics for Windows, version 26.0.Due to the skewed distribution of the data, continuous data were expressed as medians and interquartile ranges.Differences between dichotomous variables were analysed by chi-square test.Mann-Whitney U test was used to analyse the differences between the two groups of continuous variables.Multivariate logistic regression models with IR marker values as categorical variables were constructed based on quartiles of HOMA-IR, TyG, TyG-BMI, TG/HDL-c, and METS-IR in the overall population, non-obese, and obese subgroups, respectively.Regression model 1 was a rough model without adjustment.Model 2 was adjusted for age, sex, and history of hypertension to assess the relationship between the five IR markers and the risk of NAFLD.The odds ratio (OR) and 95% confidence interval (CI) for IR markers were calculated.Receiver operating characteristic (ROC) curves and area under the curve (AUC) were used to evaluate the predictive ability of different IR markers for NAFLD.A P value of < 0.05 defined statistical significance.

Characteristics of participants
Compared with non-NAFLD group, the NAFLD group had significantly higher proportion of males and those with hypertension; Values of BMI, FPG, UA, TG, AST, ALT, GGT, HOMA-IR, TyG, TG/HDL-c, METS-IR, and TyG-BMI were significantly elevated.The results revealed 928 obese and 1220 non-obese subjects.The BMI of NAFLD group was significantly higher than that of non-NAFLD group in both obese or non-obese subgroups (Table 1).
The risk of NAFLD increased significantly with increasing TyG, TyG-BMI, HOMA-IR, TG/HDL-c, and METS-IR quartiles.Based on the ORs, we found that the positive effect of the highest quartile of TyG-BMI was most pronounced in the adjusted Model 2 (Table 2).www.nature.com/scientificreports/ In the total population, the AUC of each IR marker was greater than 0.5 (P < 0.05), indicating specific predictive values for NAFLD.The AUC values decreased in the following order: TyG-BMI (0.788, 95% CI 0.766-0.811),METS-IR (0.783, 95% CI 0.760-0.806),HOMA-IR (0.782, 95% CI 0.756-0.808),TyG (0.738, 95% CI 0.711-0.765),TG/HDL-c (0.738, 95% CI 0.711-0.764).The P value greater than 0.05 for the difference in AUC between any two of the first three IR markers indicates lack of statistically significant differences between TyG-BMI, METS-IR and HOMA-IR.The AUCs of TyG-BMI, METS-IR and HOMA-IR were significantly higher than those of TyG and TG/HDL-c (P < 0.05), respectively.Thus, the ability of TyG-BMI, METS-IR and HOMA-IR to detect NAFLD was significantly higher than that of the other 2 IR makers in the overall non-diabetic population.The AUC of TyG-BMI was the largest (Table 3, Fig. 1).

The association of 5 IR markers with NAFLD in non-obese populations
The relationship between the 5 IR markers and NAFLD in the non-obese population was similar to that in the overall population with the risk of NAFLD increasing with the TyG, TyG-BMI, HOMA-IR, TG/HDL-c and METS-IR quartiles.Based on ORs (OR 58.288, 95% CI 13.729-247.468,and OR 33.071, 95% CI 10.110-108.177),the top two IR markers were METS-IR and TyG-BMI, which were most pronounced after adjustment for age, sex, and history of hypertension in Model 2 (Table 4).

The association of 5 IR markers with NAFLD in obese populations
The risk of NAFLD was increased with the increase of HOMA-IR, TyG, TG/HDL-c, TyG-BMI and METS-IR quartiles.The highest ORs were found in the highest quartile group of HOMA-IR in Models 1 and 2 (Table 6).

Discussion
The study findings suggest that IR is strongly associated with the risk of NAFLD in patients with obese and nonobese NAFLD.All IR markers were significantly higher in the NAFLD group than in the non-NAFLD group in the overall study population.After adjusting for confounding factors using logistic regression, the ORs for TyG, TyG-BMI, TG/HDL-c, METS-IR, and HOMA-IR tended to increase with increasing quartile levels, in the total population, as well as in obese and non-obese subgroups.In the non-obese subgroup, the AUC (0.817, 95% CI 0.778-0.856) of TyG-BMI was the highest among all 5 IR markers, suggesting that TyG-BMI has a better predictive value for NAFLD in non-diabetic and non-obese subjects.In the obese subgroup, the AUC of HOMA-IR was the largest (AUC 0.724, 95% CI 0.689-0.760)compared with the other four IR markers (P < 0.05).Thus, HOMA-IR is a better predictor for NAFLD in non-diabetic obese subjects compared with other IR markers.
Our study corroborates previous reports of IR markers associated with the risk of NAFLD 15,19,20 .However, in various studies, the association between IR markers and the risk of NAFLD has shown inconsistent results.Le et al. found that TyG-BMI and METS-IR demonstrated superior discriminatory ability for NAFLD compared to TyG and TG/HDL-c in subjects aged ≥ 60 years 15 .However, the study did not differentiate between obese and non-obese participants.A study found that elevated METS-IR was independently associated with new-onset NAFLD in non-obese populations 21 .Fedchuk et al. found that TyG had a significant diagnostic accuracy for the presence of steatosis with an AUC of 0.90 22 .Zhang et al. analysed 6809 Chinese subjects with BMI < 25.0 kg/m 2 and reported that TyG-BMI was more accurate in identify NAFLD than TyG, with an AUC of 0.835 23 , However, HOMA-IR and TG/HDL-c were not compared in that study.In a study of 826 individuals with type 2 diabetes, TyG-BMI (AUC 0.727) was found to be a better predictor of NAFLD than TyG, HOMA-IR, or TG/HDL-c, the predictive value of TyG-BMI for NAFLD in obese subjects was higher than in non-obese subjects 19 .Although the results of these studies differed, they consistently showed a strong correlation between IR markers and NAFLD.These differences may be attributed to the different populations, races, and diagnostic methods used.As commonly used diagnostic tools, a finding suggests variation in the prevalence of NAFLD diagnosed by CT and US 24 .
In our study, the association between IR markers and obese and non-obese NAFLD does not show a consistent pattern.The role of TyG-BMI in predicting non-obese NAFLD was better than the other 4 IR markers.HOMA-IR is more effective in predicting obese NAFLD.The reason for this distinction is not clear.TyG-BMI is calculated using FPG, TG and BMI.The diagnostic value of TyG-BMI in non-obese individuals may be superior as it is based on body fat distribution.Studies suggest that normal BMI in non-obese people still indicates an independent risk factor for NAFLD 23,25 , which may be due to the higher body fat content of individuals with non-obese NAFLD than in non-obese non-NAFLD subjects 25 .
The study has several advantages.First, NAFLD was investigated instead of metabolic-associated fatty liver disease (MAFLD) in this study.The diagnosis of NAFLD and MAFLD is based on pathological hepatic steatosis and radiographic diffuse fatty liver.However, the diagnostic criteria for MAFLD are also based on overweight and obesity, T2DM, or the presence of two or more metabolic risk abnormalities 26 .This may lead to the exclusion of certain lean or non-obese individuals 27 .Second, this study compared the predictive value of insulin-based IR marker (HOMA-IR) and non-insulin-based IR markers (TyG, TyG-BMI, TG/HDL-c, and METS-IR) for NAFLD in a large non-diabetic population and in obese and non-obese sub-groups.To the best of our knowledge, this study represents the first comparative analysis of five IR markers like this.Third, the CT results of this study were diagnosed by two experienced doctors, thus minimising the diagnostic deviation.
However, the study also has limitations.First, the study was conducted in a single medical centre.Second, CT was used in this study to diagnose NAFLD, which may miss diagnosis compared with liver biopsy.

Conclusion
The risk of NAFLD is positively associated with elevations in both insulin-based and non-insulin-based IR markers, including HOMA-IR, TyG, TyG-BMI, TG/HDL-c, and METS-IR in non-diabetic population.Results of sub-group analyses indicate that TyG-BMI (AUC 0.817, 95% CI 0.778-0.856)and HOMA-IR (AUC 0.724, 95% CI 0.689-0.760)are more effective than the other four IR markers for detection of non-obese and obese NAFLD, respectively.

Table 1 .
Comparison of baseline characteristics between groups with and without NAFLD.

Table 4 .
Logistic regression analysis of 5 IR markers and NAFLD in non-obese population.Model 1: unadjusted; Model 2: adjusted for age, sex, and history of hypertension.

Table 6 .
Logistic regression analysis of 5 IR Markers and NAFLD in obese population.Model 1: unadjusted; Model 2: adjusted for age, sex, and history of hypertension.