ALT level has acceptable sensitivity for detecting NAFLD, although it is not correlated with the histologic grade of NAFLD [3,24,25]. Moreover, some children with NAFLD have normal ALT and ALT levels, which can be elevated in other liver diseases [3,26]. Meanwhile, aspartate aminotransferase (AST) level is correlated with hepatic fibrosis progression, although it has not been suggested as an independent screening tool for NAFLD in children [3,24,25].

The North American Society of Pediatric Gastroenterology, Hepatology, and Nutrition guidelines recommend ALT as a screening test due to its minimal invasiveness and adequate sensitivity [3]. The guidelines suggest NAFLD screening for obese children aged 9–11 years and overweight children with risk factors such as central obesity, insulin resistance, dyslipidemia, sleep apnea, and/or family history of NAFLD. Further evaluations for NAFLD are suggested in children with ALT levels higher than the upper limit value for longer than 3 months, suggested as 26 U/L for boys and 22 U/L for girls based on the National Health and Nutrition Examination Survey [27]. A Canadian study suggested the upper limit of ALT as 30 mg/dL in children 1–12 years old and 24 mg/dL in those aged 13–19 years [28]. In Korea, the 95th percentile value for ALT level was suggested as 24.1 U/L for boys and 17.7 U/L for girls in a population-based study [29]. Another Korean study suggested an upper limit of ALT of 39 U/L for boys and 31 U/L for girls and those of AST as 39 U/L for boys and 41 U/L for girls aged 2–7 years and 34 U/L for girls aged 8–14 years [30].

To overcome the limitations of liver biopsy and laboratory tests, various parameters such as the hepatic steatosis index (HSI) and fatty liver index, derived from AST, ALT, and/or platelets, have been suggested to predict hepatic steatosis and fibrosis in adults [31,32]. In addition, the triglyceride glucose (TyG) index, a parameter derived as the product of fasting triglyceride and glucose levels, has been suggested to be a robust predictive marker for NAFLD based on the relationship between NAFLD and dyslipidemia and IR [33]. However, investigations on the usefulness of these parameters in children and adolescents are limited.

A pediatric study showed a positive association between HSI and the presence of hepatic steatosis [34]. A population-based study reported that modified TyG indices, which combined the TyG index and obesity-related parameters, were superior to the TyG index for predicting pediatric NAFLD [20]. A cohort study reported that HSI was superior to the TyG index for NAFLD predictions in children [35]. Song et al. [21] suggested the AST-to-platelet ratio index (APRI), fibrosis-4 index, and HSI as predictive markers of pediatric NAFLD. In addition, this study reported that the modified APRI, which combined APRI and obesityrelated parameters, such as BMI and waist circumference, were superior to other parameters. For predicting hepatic fibrosis, the area under the curve of receiver operating characteristic curve values was 0.741 for APRI, 0.71 for NAFLD fibrosis score, and 0.546 for pediatric NAFLD fibrosis index in a retrospective study [36]. Formulas of biomarkers for pediatric NAFLD and fibrosis are shown in Table 1.

Ultrasonography is an easily accessible imaging modality without radiation exposure that can exclude other hepatic pathologies and assess hepatic fatty infiltration grade based on degree of echogenicity [3,23]. Hepatic steatosis shows high echogenicity because more echoes return to the ultrasound transducer due to disturbed propagation and attenuation of the sound wave by lipid droplets in hepatocytes [37]. The normal liver shows a homogenous echo texture similar to that of the renal cortex. A mildly fatty liver shows diffuse increased echogenicity versus the renal cortex, while a moderately fatty liver shows obscured vessel walls due to increased liver echogenicity. In a severely fatty liver, the diaphragm is not visible owing to acoustic attenuation. However, ultrasound is physician- and device-dependent and cannot be used for a quantitative analysis. Moreover, its accuracy is limited due to its low sensitivity and specificity [3].

Based on evidence of increased attenuation of the ultrasound beam in the fatty liver, the controlled attenuation parameter (CAP) is suggested as an imaging modality for the quantification of fat deposition in the liver [23,38]. The CAP estimates the decreased amplitude of ultrasound as it is amplified through the liver tissue [37]. In addition, the liver stiffness measurement (LSM) represents a measure of the shear wave velocity generated by a push pulse as it passes [39]. CAP and LSM can be obtained by a FibroScan (Echosens, Paris, France), an ultrasound-based vibration-controlled transient elastography device, to estimate liver stiffness [40]. In a pediatric study, 95th percentile values of CAP and LSM were 214.53 dB/m and 4.78 kPa, respectively [40]. However, the clinical application of CAP is limited due to significant measurement failure and an unavailability of anatomic imaging [23,38]. The LSM can be altered by several factors such as venous pressure, inflammation, amyloid deposition, and cholestasis [39].

To overcome the limitations of CAP, attenuation imaging (ATI) has recently been suggested to quantify liver fat using realtime ultrasound. ATI estimates the degree of ultrasound beam attenuation with depth using a rudimentary echo decay profile after removing the systemic effects of ultrasound [41]. Operators can easily measure ATI values during grayscale ultrasound by automatically removing vessels and artifacts. ATI is highly correlated with magnetic resonance imaging (MRI) proton density fat fraction (PDFF) and shows outstanding performance for the detection of fatty liver in adults [42,43]. In a pediatric study, ATI was useful for distinguishing fatty from normal liver and moderate to severe from mild fatty liver [23]. In this study, the median ATI values for normal, mild fatty, and moderate to severe fatty liver were 0.54, 0.63, and 0.73 dB/cm/MHz, respectively. The cutoff ATI value was 0.69 dB/cm/MHz for differentiating moderate to severe fatty liver from mild fatty liver.

MRI measures PDFF, the ratio of proton density in fatty liver and that in the whole liver, based on chemical shift imaging [41]. Magnetic resonance signals are attenuated or combined by the resonance frequency of fat and water molecules and different phases. Signals of fat and water are combined and the total signal intensity is higher on the in-phase echo time, while water and fat signals cancel each other out and the total signal intensity is reduced in the out-of-phase echo time. Thus, the normal liver does not show a difference in signal intensities between the inphase and out-of-phase images, whereas the signal intensity is reduced on the out-of-phase image in hepatic steatosis. Tang et al. [44] reported that MRI-PDFF was significantly correlated with histologic steatosis grade, while the area under the receiver operating characteristic curve was 0.989 for distinguishing participants with healthy liver from those with hepatic steatosis among patients with NAFLD, including children.

Magnetic resonance elastography (MRE) can be used to quantify hepatic fibrosis and cirrhosis. MRE estimates liver stiffness and fibrosis and distinguishes steatohepatitis from simple steatosis [37,38]. Yin et al. [45] reported that the sensitivity and specificity of MRE for the detection of hepatic fibrosis were 98% and 99%, respectively, using a cutoff value of 2.93 kPa. A prospective study reported that the mean liver stiffness was 2.45 kPa among children and adolescents, higher than that in adults (2.1 kPa) [46]. In this study, the 95th percentile of normal liver stiffness was 3.11–3.28 kPa in children and 2.70–3.35 kPa in adolescents. Another prospective study reported that the mean liver stiffness was 2.1 kPa among children with normal liver, while the 95th percentile of normal liver stiffness was 2.8 kPa [47].

Insulin resistance plays a major role in the pathogenesis of NAFLD, and ectopic fat accumulation plays a key role in insulin resistance [9,48]. Insulin downregulates gluconeogenesis by promoting glycogen synthesis in the liver and muscle and inhibiting lipolysis in adipose tissue [4,48,49]. Fat accumulation due to excess caloric intake and decreased physical activity induces adipocyte hypertrophy and dysfunction as well as insulin resistance, thereby increasing lipolysis and free fatty acid levels (Fig. 1) [48,50]. Free fatty acids taken up by the cell via fatty acid transport protein increases long-chain acyl-CoA and diacylglycerol, which induces insulin resistance by inhibiting insulin receptor function via activating protein kinase C in the liver and skeletal muscles (Fig. 2) [48]. In the skeletal muscle, decreased glycogen synthesis due to insulin resistance induces hyperglycemia and postprandial hyperinsulinemia, which induce hepatic de novo lipogenesis, hypertriglyceridemia, increased very low-density lipoprotein (VLDL) production, and decreased plasma high-density lipoprotein (HDL) cholesterol [51]. In the liver, hyperinsulinemia promotes the activation of sterol regulatory element-binding protein 1 (SREBP1) and hyperglycemia and activates carbohydrate response element-binding protein (ChREBP) [51,52]. SREBP1 and ChREBP promote de novo lipogenesis, lipid production from carbohydrate in the liver. Although de novo lipogenesis is not the main source of lipids, it can be a main factor for lipid supply in people with obesity and hyperinsulinemia [4]. Moreover, hyperinsulinemia promotes lipid accumulation by inactivating Forkhead box protein A2, a transcription factor promoting lipid oxidation in the liver [52]. The paradox in the pathogenesis of NAFLD wherein insulin fails to suppress glucose production while continuing lipogenesis in the liver can be explained by selective insulin resistance, which indicates different actions of insulin between glycogen synthesis and lipogenesis [51,53].

In summary, insulin resistance promotes lipolysis in adipose tissue, which increases the influx of free fatty acids into the skeletal muscle and liver [48,50,51]. In the muscle cells, insulin resistance inhibits glycogen synthesis, which increases glucose influx into the liver. Finally, hyperinsulinemia and increased de novo lipogenesis accelerate lipid accumulation in the liver [4].