A prediction model of liver fat fraction and presence of non-alcoholic fatty liver disease (NAFLD) among patients with overweight or obesity

Abstract

The global prevalence of non-alcoholic fatty liver disease (NAFLD) has attained a level of 25.24%. The prevalence of NAFLD in China has exhibited an upward trajectory in parallel with the increasing incidence of obesity over the preceding decade. In order to comprehensively assess hepatic lipid deposition in individuals with overweight or obesity, we have devised a pioneering prognostic formula that capitalizes on clinical parameters. To this end, we have conducted a cross-sectional cohort study involving 149 overweight or obese subjects. Magnetic resonance imaging proton density fat fraction (MRI-PDFF) has been employed to evaluate the extent of liver fat accumulation. Through univariate analysis, we have identified potential factors, and the definitive elements in the prediction model were selected utilizing the forward stepwise regression algorithm. The Shang Hai Steatosis Index (SHSI) incorporates alanine aminotransferase (ALT), aspartate aminotransferase (AST), fasting insulin, and 1-h postload glycaemic levels, thereby furnishing the capability to predict NAFLD with an area under the receiver operator characteristic (AUROC) of 0.87. By establishing a threshold value of 10.96, determined through Youden’s index, we have achieved a sensitivity of 69.57% and a specificity of 88.24%. The Spearman correlation coefficient between liver fat fraction ascertained by MRI-PDFF and that predicted by the SHSI equation amounts to 0.74. Consequently, the SHSI equation affords a dependable means of predicting the presence of NAFLD and liver fat accumulation within the overweight and obese population.

WITH THE RAPID ADVANCEMENT OF THE ECONOMY, there has been a discernible upswing in the consumption of high-calorie diets and the adoption of sedentary lifestyles. The global prevalence of non-alcoholic fatty liver disease (NAFLD) currently stands at 25.24% (95% CI: 22.10–28.65) [1]. NAFLD is characterized by hepatic steatosis in the absence of secondary factors contributing to liver fat accumulation, and it can be diagnosed via imaging or histological examinations [2]. Notably, NAFLD exhibits a close correlation with metabolic comorbidities such as obesity, type 2 diabetes, hyperlipidemia, hypertension, and metabolic syndrome [1]. Within the confines of China, NAFLD has emerged as the primary cause of chronic liver disease, surpassing even hepatitis B, and accounting for 49.3% of all liver ailments [3]. Owing to its subtle onset and lack of overt symptoms, NAFLD is frequently discovered during routine medical checkups and commonly overlooked by patients themselves. However, if left unchecked, it can progress to fibrosis, cirrhosis, or hepatocellular carcinoma over time [4]. Subjects afflicted with NAFLD exhibit higher overall mortality rates as well as increased liver-related mortality, with adjusted hazard ratios of 1.038 and 9.32, respectively. Among patients with NAFLD, cardiovascular diseases and neoplasms constitute the primary causes of death [5].

Liver biopsy remains the gold standard for diagnosing NAFLD. However, its invasive nature restricts its widespread application. Among non-invasive imaging assessments, magnetic resonance imaging proton density fat fraction (MRI-PDFF) emerges as a noteworthy imaging biomarker for hepatic steatosis, offering qualitative estimations. Its diagnostic performance, as indicated by the area under the receiver operator characteristic curve (AUROC), is exceptionally high at 0.99 (95% confident interval (CI), 0.98–1.00) for the detection of any grade of liver steatosis [6]. In order to streamline the diagnostic process, various prediction models utilizing clinical parameters have been devised to evaluate hepatic steatosis. These include the fatty liver index (FLI) [7], the Hepatic Steatosis Index (HSI) [8], the NAFLD liver fat score (NLFS) [9], the SteatoTest [10], and the Framingham Steatosis Index (FSI) [11]. However, with the exception of a single study involving liver biopsy, none of these models have incorporated MRI-PDFF as an imaging-based diagnostic criterion. Notably, MRI-PDFF appears to be the most precise and noninvasive method available thus far for the diagnosis of fatty liver [12].

The prevalence of NAFLD in China has exhibited a concurrent rise with the escalating prevalence of obesity over the past decade [13]. Notably, approximately 75–92% of individuals with morbid obesity are afflicted with NAFLD [14]. Conversely, more than 80% of NAFLD patients fall within the overweight and obese categories [15], and they typically present compromised metabolic and histological profiles [16]. NAFLD is characterized by the suppression of hepatic glucose production (HGP), which in turn disrupts both fasting and postprandial glucose homeostasis [17]. Curiously, none of the existing models have accounted for postprandial glycemic and insulin levels. It has been documented that patients experiencing 1-h postprandial hyperglycemia faced a significantly heightened risk of hepatic steatosis when compared to those with normal 1-h postprandial blood glucose levels [18, 19]. Moreover, NAFLD is intimately associated with insulin resistance.

Consequently, our objective was to construct a prediction equation with the aim of enhancing the accuracy of NAFLD prediction and enabling a quantitative assessment of liver fat percentage among overweight and obese individuals.

Material and Methods

Subjects and study design

A total of 149 overweight or obese individuals, ranging in age from 18 to 70, were enrolled in the study between September 2020 and June 2021, specifically from the Obesity Outpatient Clinic of Endocrinology (ClinicalTrials.gov: NCT05779644). The inclusion criteria encompassed the following: 1. Participants aged 18 to 70 years old; 2. Body mass index (BMI) ≥24 kg/m² [15]. Conversely, the exclusion criteria consisted of the following: 1. Known factors contributing to liver dysfunction, including alcoholic liver disease, autoimmune liver disease, chronic viral hepatitis, and drug-induced liver injury. 2. Secondary causes of obesity, such as Cushing syndrome, eating disorders, prolonged use of glucocorticoids and immunosuppressants. 3. Usage of antidiabetic or anti-obesity medications within the past three months. 4. Pregnancy or lactation. 5. Magnetic resonance imaging (MRI) contraindications. The study adhered to the principles outlined in the Declaration of Helsinki and received approval from the Renji Hospital Affiliated to Medical College of Shanghai Jiaotong University Clinical Research Ethics Committee (ID 2019-035). All participants were provided with a consent form. Flow diagram showed the recruitment procedure (Fig. 1).

Fig. 1

Flow diagram of the recruitment procedure.

Somatometric characteristics and laboratory assessments

Body weight was measured and rounded to the nearest 0.10 kg with light clothing. Height was recorded with a precision of 0.10 cm. Waist circumference was measured with an accuracy of 0.10 cm, specifically at the midpoint between the lower rib margin and iliac crest. Hip circumference was determined as the largest measurement around the buttocks. BMI was calculated by dividing the body weight by the square of the height (kg/m²). Furthermore, the waist-hip ratio was computed by dividing the waist circumference by the hip circumference.

Laboratory parameters were assessed according to the prescribed guidelines provided by the manufacturers at a centralized laboratory. Enzymatic methods were employed to measure aspartate aminotransferase (AST), alanine aminotransferase (ALT), γ-glutamyl transpeptidase (GGT), triglycerides and total cholesterol. High-density lipoprotein (HDL) cholesterol and low-density lipoprotein (LDL) cholesterol were determined using homogeneous assays (Roche Diagnostics, Basel, Switzerland). Blood glucose concentrations were evaluated by the hexokinase method. Insulin and C peptide were measured by chemiluminescence assays (Roche Diagnostics, Basel, Switzerland). The 75 g oral glucose tolerance test (OGTT) were conducted at 8:00 a.m., following an overnight fasting period of no less than 12 hours. Venous blood samples were collected prior to ingestion of the glucose (t = 0), as well as at 30, 60, 120, and 180 minutes after the administration of 75 g of glucose.

MRI image analysis

MRI was utilized to assess the liver fat fraction. All abdominal MRI scans were conducted using a 3.0 T wide bore scanner (Ingenia, Philips Healthcare) within the Department of Radiology. The MRI mDIXON-Quant sequence was employed to acquire the images during a single breath-hold, with parameters set as previously described [20]. The liver fat fraction was calculated by identifying regions of interest (ROI) through the use of the ISP V9 workstation (Ingenia, Philips Healthcare) based on the raw imaging data [20]. Specifically, a mean liver fat fraction greater than 6.4% served as the diagnostic threshold for NAFLD [21].

Statistical analyses

The normal distribution of variables was assessed using the Kolmogorov-Smirnov test. Continuous variables are shown as mean ± standard deviation (SD) or median with interquartile range (IQR). All participants were included in the discovery group. To examine the correlations between anthropometric characteristics, blood biochemistry indices and liver fat content, Spearman’s coefficients were calculated. Subsequently, the significant factors identified in the Spearman correlation test were chosen for further analysis. Non-normally distributed data were log-transformed. Multivariable linear regression analyses were used to establish a prediction model for liver fat content. The final predictors of liver fat content were selected by stepwise forward selection method (ALT, AST, fasting insulin, and 1-h postload glycemic levels). The Shang Hai Steatosis Index (SHSI) equation was constructed based on the results of the multivariate logistic regression analyses. Bootstrap resampling method was used for internal validation. Values were estimated from 1,000 bootstrap replicates with SPSS 25 built-in bootstrap procedure.

The association between SHSI and MRI-PDFF was examined using the Spearman correlation test after bootstrap resampling. Participants with MRI-PDFF exceeding 6.4% were diagnosed as NAFLD. Logistic regression models were employed to evaluate the accuracy of NAFLD prediction based on the SHSI, and the AUROC was used to assess the performance of the final equation. Furthermore, a comparison was conducted among different prediction formulas. The statistical calculations were performed using SPSS 25 for Windows (SPSS, Chicago, IL), and the graphics were generated by R 4.0.1. (R Foundation for Statistical Computing, Vienna, Austria) and MedCalc (MedCalc Software, Mariakerke, Belgium).

Results

Subject characteristics

A total of 149 individuals who were either overweight or obese participated in this study. The participants had a median age of 34.00 years-old (with an IQR of 29.00–44.00). Among the participants, 74 (49.66%) were male, and 75 (50.34%) were female, as shown in Table 1. The average body mass index (BMI) was 32.26 kg/m² (with a median of 29.04–36.99) based on the IQR. All subjects underwent MRI-PDFF. Fig. 2 displays liver maps from different stages of NAFLD patients. It is important to note that all images were captured at a similar level above the Porta hepatic plane. On average, the liver fat fraction among all subjects was 15.70%. According to the MRI-PDFF results, 77.18% of the individuals were diagnosed with NAFLD using a threshold of 6.4%.

Table 1

Characteristics of the subjects

	All subjects (n = 149)
Somatometric characteristics
Age	34.00 (29.00–44.00)
Male/female	74/75
Weight (kg)	90.00 (79.25–115.00)
BMI (kg/m2)	32.26 (29.04–36.99)
Waist circumference (cm)	109.00 (97.00–120.00)
Hip circumference (cm)	112.00 (105.75–120.00)
Waist-hip ratio	0.96 ± 0.08
Biochemical parameters
ALT (U/L)	35.00 (21.00–63.00)
AST (U/L)	23.00 (17.00–37.00)
GGT (U/L)	37.00 (21.00–52.00)
Plasma lipids
Triglycerides (mmol/L)	1.77 (1.30–2.49)
Total cholesterol (mmol/L)	5.23 (4.52–5.85)
HDL cholesterol (mmol/L)	1.08 ± 0.23
LDL cholesterol (mmol/L)	3.20 ± 0.82
75 g-oral glucose tolerance test
Fasting glucose (mmol/L)	5.14 (4.72–5.70)
0.5-h glucose (mmol/L)	9.14 (8.01–10.21)
1-h glucose (mmol/L)	9.85 (8.06–11.61)
2-h glucose (mmol/L)	7.77 (6.46–9.08)
3-h glucose (mmol/L)	5.46 (4.01–6.67)
Glucose AUC	23.61 (21.19–26.88)
Fasting C-peptide (ng/mL)	3.51 (2.85–4.88)
0.5-h C-peptide (ng/mL)	9.46 (7.45–12.84)
1-h C-peptide (ng/mL)	12.09 (9.54–15.79)
2-h C-peptide (ng/mL)	12.16 (9.35–15.16)
3-h C-peptide (ng/mL)	7.76 (5.68–10.34)
C-peptide AUC	30.18 (24.79–39.31)
Fasting insulin (mIU/L)	13.96 (10.26–21.05)
0.5-h insulin (mIU/L)	88.96 (55.68–136.69)
1-h insulin (mIU/L)	108.16 (65.48–159.16)
2-h insulin (mIU/L)	84.61 (49.60–135.86)
3-h insulin (mIU/L)	30.01 (16.19–57.94)
insulin AUC	237.12 (156.10–346.97)
Liver fat
MRI-PDFF (%)	15.70 (6.76–24.85)
NAFLD (%)	115 (77.18%)

AST, aspartate aminotransferase; ALT, alanine aminotransferase; GGT, γ-glutamyl transpeptidase; HDL, high-density lipoprotein; LDL, low-density lipoprotein; AUC, area under curve; NAFLD, non-alcoholic fatty liver disease; MRI-PDF, magnetic resonance imaging proton density fat fraction

Continuous values with normal distribution are shown as mean ± standard deviation (SD), other continuous variables are shown as median with interquartile range (IQR).

Fig. 2

Distinct levels of liver MRI-PDFF are observed across nearly homogeneous regions located superior to the Porta hepatis. MRI-PDFF, magnetic resonance imaging proton density fat fraction.

Associations between parameters and MRI-PDFF

The Spearman correlation coefficients between MRI-PDFF and various parameters within the entire subject cohort were documented (Table 2). MRI-PDFF exhibited positive associations with male gender, body weight, BMI, waist circumference, hip circumference, waist-hip ratio, liver enzymes, triglyceride, HDL cholesterol, fasting glucose levels, glucose levels at 0.5, 1, and 2 hours, glucose area under the curve (AUC), fasting C-peptide levels, C-peptide levels at 3 hours, fasting insulin levels, insulin levels at 1, 2, and 3 hours, and insulin AUC (Table 2).

Table 2

Univariate analysis of Spearman correlation coefficients between MRI-PDFF and the other parameters

	Rho	p value
Somatometric characteristics
Age	–0.161	0.050
Male/female	0.213	0.009
Weight (kg)	0.351	0.000
BMI (kg/m2)	0.342	0.000
Waist circumference (cm)	0.303	0.000
Hip circumference (cm)	0.224	0.006
Waist-hip ratio	0.281	0.001
Biochemical parameters
ALT (U/L)	0.699	0.000
AST (U/L)	0.606	0.000
GGT (U/L)	0.512	0.000
Plasma lipids
Triglycerides (mmol/L)	0.334	0.000
Total cholesterol (mmol/L)	0.095	0.250
HDL cholesterol (mmol/L)	–0.190	0.020
LDL cholesterol (mmol/L)	0.022	0.791
75 g oral glucose tolerance test
Fasting glucose (mmol/L)	0.281	0.001
0.5-h glucose (mmol/L)	0.242	0.003
1-h glucose (mmol/L)	0.366	0.000
2-h glucose (mmol/L)	0.307	0.000
3-h glucose (mmol/L)	0.108	0.191
Glucose AUC	0.351	0.000
Fasting C-peptide (ng/mL)	0.410	0.000
0.5-h C-peptide (ng/mL)	–0.023	0.781
1-h C-peptide (ng/mL)	0.052	0.531
2-h C-peptide (ng/mL)	0.154	0.060
3-h C-peptide (ng/mL)	0.212	0.010
C-peptide AUC	0.134	0.103
Fasting insulin (mIU/L)	0.410	0.000
0.5-h insulin (mIU/L)	0.039	0.641
1-h insulin (mIU/L)	0.171	0.037
2-h insulin (mIU/L)	0.271	0.001
3-h insulin (mIU/L)	0.347	0.000
insulin AUC	0.239	0.003

AST, aspartate aminotransferase; ALT, alanine aminotransferase; GGT, γ-glutamyl transpeptidase; HDL, high-density lipoprotein; LDL, low-density lipoprotein; AUC, area under curve; NAFLD, non-alcoholic fatty liver disease; MRI-PDF, magnetic resonance imaging proton density fat fraction

The multivariable linear regression model incorporated variables that displayed significant correlations with MRI-PDFF. Non-normally distributed data were subjected to a logarithmic (base 10) transformation. Through a forward stepwise regression selection process and exclusion of collinear variables, ALT, fasting insulin, 1-h glucose, and AST were included in the final multivariable linear regression model (Table 3). The adjusted r² of the linear regression model in the discovery group was 0.52 (p < 0.0001). The beta coefficient represents the impact of the independent variable on the dependent variable. The following preliminary formula was derived.

Table 3

Multivariate linear regression model to predict liver fat

Liver fat (%- log)	Beta coefficient	Standard error	p value
ALT (U/L- log)	0.933	0.156	0.000
fasting insulin (mIU/L- log)	0.294	0.087	0.001
1-h glucose (mmol/L- log)	0.740	0.173	0.000
AST (U/L- log)	–0.570	0.206	0.006
Constant	–0.625	0.194	0.002

ALT, alanine aminotransferase; AST, aspartate aminotransferase

Log (Liver fat [%]) = –0.625 + 0.933*log (ALT [U/L]) + 0.294*log (fasting insulin [mIU/L]) + 0.74*log (1-h glucose [mmol/L]) – 0.57*log (AST [U/L])

Then, SHSI equation is derived: Liver fat (%) = 10 ^{(–0.625 + 0.933*log (ALT [U/L]) + 0.294*log (fasting insulin [mIU/L]) + 0.74*log (1-h glucose [mmol/L]) – 0.57*log (AST [U/L]))}

Prediction of NAFLD

The Spearman correlation coefficient between MRI-PDFF and SHSI yielded a value of 0.74 (95%CI, 0.71–0.77, p < 0.0001) based on 1,000 bootstrap replicates (Fig. 3). To predict NAFLD by SHSI, the AUROC was found to be 0.87 (95%CI, 0.81–0.92, p < 0.0001), accompanied by a sensitivity of 69.57% and specificity of 88.24% after bootstrap validation (Fig. 4). Employing Youden’s index on the receiver operating characteristic (ROC) curve for the SHSI equation, the cutoff value of 10.96 was identified as the threshold with the highest sum of sensitivity and specificity for the identification of individuals with NAFLD.

Fig. 3

The association between hepatic fat content assessed via MRI-PDFF and predicted by SHSI equation. r = 0.74 (95% CI 0.65–0.81), p < 0.0001. MRI-PDFF, magnetic resonance imaging proton density fat fraction.

Fig. 4

ROC curves of the comparison among different prediction formulas. * ROC curve of SHSI score was based on 1,000 bootstrap replicates. SHSI, Shang Hai Steatosis Index; FLI, fatty liver index; FSI, Framingham Steatosis Index; HSI, Hepatic Steatosis Index; NLFS, NAFLD liver fat score.

Discussion

In the present investigation, it was observed that 77.18% of the overall overweight/obese participants exhibited NAFLD within a sample size of 149 individuals. Univariate analysis was employed to identify potential risk factors, and subsequently, forward stepwise regression selection was applied to establish the predictive model. The SHSI equation was utilized to assess the accumulation of hepatic fat, considering parameters such as ALT, AST, fasting insulin, and 1-hour postload glycaemic levels. The threshold value of 10.96 was identified based on Youden’s index. The AUROC was calculated as 0.87, accompanied by a sensitivity of 69.57% and specificity of 88.24%.

Thus far, numerous research investigations have developed equations for the prediction of NAFLD based on anthropometric indices and laboratory indicators. Noteworthy examples include the FLI [7], HSI [8], NLFS [9], SteatoTest [10], and FSI [11]. In our current study, the SHSI stands out as the initial index to forecast the accumulation of liver fat and the presence of NAFLD using OGTT. In comparison to fasting glucose, OGTT emerges as a highly sensitive assessment and an early marker for the dysregulation of glucose [22].

The ROC curve of the SHSI, which closely resembled the NLFS, exhibited significant distinctions when compared to FLI, FSI, HSI (Fig. 4). The SHSI equation incorporates parameters such as ALT, AST, fasting insulin, and 1-hour glycaemic levels. Notably, mild elevation in liver enzyme levels is frequently observed in individuals with NAFLD [23]. A prospective study revealed that even individuals with higher ALT levels within the reference range are more susceptible to NAFLD [24]. In accordance with our findings, some prediction models have included ALT and AST in their final equations [8, 9, 11]. Insulin resistance serves as a pivotal factor in the pathogenesis of NAFLD [23, 25]. Conversely, excessive accumulation of liver fat impairs insulin clearance efficiency and induces insulin resistance through the secretion of proteins, lipids, other metabolites, and miRNAs [26, 27]. Fasting insulin levels, which play a significant role in insulin resistance, were included in our prediction model. The liver assumes a crucial role in maintaining blood glucose homeostasis. 1-hour post-load glycaemic levels demonstrate sensitivity in identifying individuals at risk of NAFLD [18, 19]. Moreover, 1-hour post-load glycaemic levels have been shown to be a stronger predictor of future type 2 diabetes compared to 2-hour post-load blood glucose levels [28]. These findings support the notion of the early detection of metabolic disorders through sensitive changes in 1-hour post-load blood glucose.

Most of the established models for diagnosing NAFLD have traditionally relied on methods such as ultrasound, biopsy and computed tomography (CT). However, these approaches possess certain limitations. Ultrasound exhibits relatively good accuracy in detecting moderate-to-severe hepatic steatosis, but it lacks the required specificity and sensitivity for detecting mild hepatic steatosis [29, 30]. Liver biopsy is considered the gold standard for diagnosis, but it is an invasive procedure and carries a risk of sampling error [31, 32]. Similarly, CT, like ultrasound, demonstrates specificity in detecting moderate-to-severe hepatic steatosis, but it is less sensitive in identifying mild steatosis [33]. Moreover, the use of CT is restricted due to concerns regarding radiation exposure.

In comparison to other studies, our research utilized MRI-PDFF as the reference method for measuring liver steatosis, which offers a comprehensive assessment of liver fat content. Unlike most existing models that only provide qualitative results, MRI-PDFF confers significant advantages in terms of quantitative detection of liver fat. It exhibits particular sensitivity in determining steatosis grade and quantifying changes [34]. Moreover, an informal consensus has been reached, acknowledging MRI-PDFF as the most appropriate standardized magnetic resonance (MR)-based biomarker for quantifying tissue fat content. It is characterized by accuracy, standardization, quantification, objectivity, and practicality [35]. Furthermore, a meta-analysis comprising 23 studies demonstrated the excellent linearity and precision of MRI-PDFF across various conditions [36]. Hence, prediction models based on the reference standard of MRI-PDFF are considered highly reliable.

There are several limitations. Firstly, the sample size of participants was relatively modest. We enrolled drug-naïve individuals with overweight or obesity, while excluding those who were using antidiabetic or anti-obesity medications to mitigate potential pharmacological influences. Secondly, we did not incorporate individuals of normal weight or lean stature into our study cohort. It is noteworthy that the prevalence of NAFLD is considerably greater among individuals with overweight or obesity compared to the general population. Thirdly, we regrettably did not execute external validation. These aspects should be taken into account in future investigations.

Conclusion

The SHSI generally offers an alternative approach to anticipate the existence of NAFLD and hepatic lipid accumulation through the amalgamation of ALT, AST, fasting insulin, and 1-hour postload glycemic levels among the overweight or obese populace. This method can be employed to expedite diagnosis and encourage timely intervention prior to the transition from uncomplicated NAFLD to irreversible nonalcoholic steatohepatitis.

Acknowledgments

The authors express their gratitude to all patients and investigators for their valuable cooperation and significant contributions.

Author Contributions

JM and QL were responsible for the study design. JC, JY and JJF conducted the statistical analyses and drafted the manuscript. SYH and WZ performed the MRI examination and analyzed the MRI image. JY, MLY and HX provided the clinical evaluation. QJL collected serum samples and clinical data. The final version of the manuscript was approved by all authors.

Funding

This study received funding from the Shanghai Municipal Education Commission–Gaofeng Clinical Medicine Grant Support (NO.20181807), Shanghai Pujiang Program (NO.2019PJD027), Ministry of Education, Science and Technology Development Center–New Generation of Information Technology Innovation Program (NO.2019ITA01004).

Data Availability

All datasets generated or analyzed during the present study are not publicly available, but can be obtained from the corresponding author upon reasonable request.

Declarations

Conflict of interest

All the authors declare no competing interests.

Ethics approval and consent to participate

The study was carried out in accordance with the principles outlined in the Declaration of Helsinki and received approval from the Clinical Research Ethics Committee of Renji Hospital, affiliated with the Medical College of Shanghai Jiaotong University (ID 2019-035). All participants provided informed consent by signing a consent form.

References

• 1   Younossi  ZM,  Koenig  AB,  Abdelatif  D,  Fazel  Y,  Henry  L, et al. (2016) Global epidemiology of nonalcoholic fatty liver disease-meta-analytic assessment of prevalence, incidence, and outcomes. Hepatology 64: 73–84.
• 2   Chalasani  N,  Younossi  Z,  Lavine  JE,  Charlton  M,  Cusi  K, et al. (2018) The diagnosis and management of nonalcoholic fatty liver disease: practice guidance from the American Association for the Study of Liver Diseases. Hepatology 67: 328–357.
• 3   Wang  FS,  Fan  JG,  Zhang  Z,  Gao  B,  Wang  HY (2014) The global burden of liver disease: the major impact of China. Hepatology 60: 2099–2108.
• 4   Fan  JG,  Kim  SU,  Wong  VW (2017) New trends on obesity and NAFLD in Asia. J Hepatol 67: 862–873.
• 5   Ong  JP,  Pitts  A,  Younossi  ZM (2008) Increased overall mortality and liver-related mortality in non-alcoholic fatty liver disease. J Hepatol 49: 608–612.
• 6   Park  CC,  Nguyen  P,  Hernandez  C,  Bettencourt  R,  Ramirez  K, et al. (2017) Magnetic resonance elastography vs. transient elastography in detection of fibrosis and noninvasive measurement of steatosis in patients with biopsy-proven nonalcoholic fatty liver disease. Gastroenterology 152: 598–607.e2.
• 7   Bedogni  G,  Bellentani  S,  Miglioli  L,  Masutti  F,  Passalacqua  M, et al. (2006) The fatty liver index: a simple and accurate predictor of hepatic steatosis in the general population. BMC Gastroenterol 6: 33.
• 8   Lee  JH,  Kim  D,  Kim  HJ,  Lee  CH,  Yang  JI, et al. (2010) Hepatic steatosis index: a simple screening tool reflecting nonalcoholic fatty liver disease. Dig Liver Dis 42: 503–508.
• 9   Kotronen  A,  Peltonen  M,  Hakkarainen  A,  Sevastianova  K,  Bergholm  R, et al. (2009) Prediction of non-alcoholic fatty liver disease and liver fat using metabolic and genetic factors. Gastroenterology 137: 865–872.
• 10   Poynard  T,  Ratziu  V,  Naveau  S,  Thabut  D,  Charlotte  F, et al. (2005) The diagnostic value of biomarkers (SteatoTest) for the prediction of liver steatosis. Comp Hepatol 4: 10.
• 11   Long  MT,  Pedley  A,  Colantonio  LD,  Massaro  JM,  Hoffmann  U, et al. (2016) Development and validation of the framingham steatosis index to identify persons with hepatic steatosis. Clin Gastroenterol Hepatol 14: 1172–1180.e2.
• 12   Piazzolla  VA,  Mangia  A (2020) Noninvasive diagnosis of NAFLD and NASH. Cells 9: 1005.
• 13   Zhou  J,  Zhou  F,  Wang  W,  Zhang  XJ,  Ji  YX, et al. (2020) Epidemiological features of NAFLD from 1999 to 2018 in China. Hepatology 71: 1851–1864.
• 14   Erickson  SK (2009) Nonalcoholic fatty liver disease. J Lipid Res 50 (Suppl): S412–S416.
• 15   Ye  Q,  Zou  B,  Yeo  YH,  Li  J,  Huang  DQ, et al. (2020) Global prevalence, incidence, and outcomes of non-obese or lean non-alcoholic fatty liver disease: a systematic review and meta-analysis. Lancet Gastroenterol Hepatol 5: 739–752.
• 16   Chen  F,  Esmaili  S,  Rogers  GB,  Bugianesi  E,  Petta  S, et al. (2020) Lean NAFLD: A distinct entity shaped by differential metabolic adaptation. Hepatology 71: 1213–1227.
• 17   Titchenell  PM,  Lazar  MA,  Birnbaum  MJ (2017) Unraveling the regulation of hepatic metabolism by insulin. Trends Endocrinol Metab 28: 497–505.
• 18   Fiorentino  TV,  Andreozzi  F,  Mannino  GC,  Pedace  E,  Perticone  M, et al. (2016) One-hour postload hyperglycemia confers higher risk of hepatic steatosis to HbA1c-defined prediabetic subjects. J Clin Endocrinol Metab 101: 4030–4038.
• 19   Sesti  G,  Hribal  ML,  Fiorentino  TV,  Sciacqua  A,  Perticone  F (2014) Elevated 1 h postload plasma glucose levels identify adults with normal glucose tolerance but increased risk of non-alcoholic fatty liver disease. BMJ Open Diabetes Res Care 2: e000016.
• 20   Zhang  Y,  Wang  C,  Duanmu  Y,  Zhang  C,  Zhao  W, et al. (2018) Comparison of CT and magnetic resonance mDIXON-Quant sequence in the diagnosis of mild hepatic steatosis. Br J Radiol 91: 20170587.
• 21   Tang  A,  Desai  A,  Hamilton  G,  Wolfson  T,  Gamst  A, et al. (2015) Accuracy of MR imaging-estimated proton density fat fraction for classification of dichotomized histologic steatosis grades in nonalcoholic fatty liver disease. Radiology 274: 416–425.
• 22   Inzucchi  SE (2012) Clinical practice. Diagnosis of diabetes. N Engl J Med 367: 542–550.
• 23   Brunt  EM,  Wong  VWS,  Nobili  V,  Day  CP,  Sookoian  S, et al. (2015) Nonalcoholic fatty liver disease. Nat Rev Dis Primers 1: 15080.
• 24   Chang  Y,  Ryu  S,  Sung  E,  Jang  Y (2007) Higher concentrations of alanine aminotransferase within the reference interval predict nonalcoholic fatty liver disease. Clin Chem 53: 686–692.
• 25   Lonardo  A,  Lombardini  S,  Ricchi  M,  Scaglioni  F,  Loria  P (2005) Review article: hepatic steatosis and insulin resistance. Aliment Pharmacol Ther 22 (Suppl 2): 64–70.
• 26   Bril  F,  Lomonaco  R,  Orsak  B,  Ortiz-Lopez  C,  Webb  A, et al. (2014) Relationship between disease severity, hyperinsulinemia, and impaired insulin clearance in patients with nonalcoholic steatohepatitis. Hepatology 59: 2178–2187.
• 27   Watt  MJ,  Miotto  PM,  De Nardo  W,  Montgomery  MK (2019) The liver as an endocrine organ-linking NAFLD and insulin resistance. Endocr Rev 40: 1367–1393.
• 28   Pareek  M,  Bhatt  DL,  Nielsen  ML,  Jagannathan  R,  Eriksson  KF, et al. (2018) Enhanced predictive capability of a 1-hour oral glucose tolerance test: a prospective population-based cohort study. Diabetes Care 41: 171–177.
• 29   Zhang  YN,  Fowler  KJ,  Hamilton  G,  Cui  JY,  Sy  EZ, et al. (2018) Liver fat imaging-a clinical overview of ultrasound, CT, and MR imaging. Br J Radiol 91: 20170959.
• 30   Ferraioli  G,  Soares Monteiro  LB (2019) Ultrasound-based techniques for the diagnosis of liver steatosis. World J Gastroenterol 25: 6053–6062.
• 31   Nalbantoglu  IL,  Brunt  EM (2014) Role of liver biopsy in nonalcoholic fatty liver disease. World J Gastroenterol 20: 9026–9037.
• 32   Tapper  EB,  Lok  AS (2017) Use of liver imaging and biopsy in clinical practice. N Engl J Med 377: 756–768.
• 33   Pickhardt  PJ,  Park  SH,  Hahn  L,  Lee  SG,  Bae  KT, et al. (2012) Specificity of unenhanced CT for non-invasive diagnosis of hepatic steatosis: implications for the investigation of the natural history of incidental steatosis. Eur Radiol 22: 1075–1082.
• 34   Noureddin  M,  Lam  J,  Peterson  MR,  Middleton  M,  Hamilton  G, et al. (2013) Utility of magnetic resonance imaging versus histology for quantifying changes in liver fat in nonalcoholic fatty liver disease trials. Hepatology 58: 1930–1940.
• 35   Reeder  SB,  Hu  HH,  Sirlin  CB (2012) Proton density fat-fraction: a standardized MR-based biomarker of tissue fat concentration. J Magn Reson Imaging 36: 1011–1014.
• 36   Yokoo  T,  Serai  SD,  Pirasteh  A,  Bashir  MR,  Hamilton  G, et al. (2018) Linearity, bias, and precision of hepatic proton density fat fraction measurements by using MR imaging: a meta-analysis. Radiology 286: 486–498.