Screening for insulin resistance in children with abdominal obesity: sex-specific performance of TyG index, FGIR, and VAI

Abstract

[Purpose]

This study aimed to compare the diagnostic performance of the triglyceride-glucose index (TyG), fasting glucose-to-insulin ratio (FGIR), and visceral adiposity index (VAI) for detecting insulin resistance (IR) and to establish sex-specific IR cutoffs in Korean schoolchildren with abdominal obesity.

[Methods]

This cross-sectional study included 728 fourth- to sixth-grade students (398 boys and 330 girls; mean age, 11.5 ± 1.1 years) from P City, Gyeonggi-do (June 2018-August 2019). Abdominal obesity was defined as waist circumference and waist-to-height ratio ≥ 90th percentile (high-risk subgroup, n = 192). IR was classified using sex-specific homeostasis model assessment of insulin resistance cutoffs (boys ≥ 3.54, girls ≥ 3.69). Receiver operating characteristic analysis with the Youden index was performed to compare TyG index, FGIR, and VAI in the high-risk subgroup.

[Results]

FGIR demonstrated superior performance in both sexes. The area under the curve (AUC) was 0.939 (95% confidence interval [CI], 0.903-0.976) for boys and 0.894 (95% CI, 0.835-0.954) for girls. Corresponding cutoff values were 8.435 and 7.473, with sensitivities of 94.9% and 91.2% and specificities of 79.6% and 81.2%, respectively. TyG and VAI exhibited lower AUCs (0.726‒0.768 and 0.679‒0.740, respectively). High-risk abdominal obesity was strongly associated with IR (adjusted odds ratio 6.175, 95% CI, 3.958-9.633).

[Conclusion]

Among Korean school-aged children with abdominal obesity, FGIR outperformed TyG and VAI as surrogate markers of IR. Accounting for pubertal insulin dynamics with sex-specific cutoff values supports the use of FGIR as a practical, low-cost tool for early screening and intervention in high-risk groups.

INTRODUCTION

Childhood and adolescent obesity have become a major public health challenge worldwide, with South Korea exhibiting a similar trend among school-aged children. According to the 2023 National Child Survey conducted by the Ministry of Health and Welfare, the prevalence of obesity among children aged 9-17 years reached 14.3%, representing an approximately 3.5-fold increase compared with that in 2018 [1]. Notably, overweight and obesity rates exceeded 20% across all age groups, with older school-aged children showing a marked increase. Importantly, a substantial proportion of children with obesity also have abdominal obesity, underscoring its clinical significance beyond cosmetic concerns. Childhood abdominal obesity is a critical precursor to adult metabolic disorders, including type 2 diabetes, dyslipidemia, and hypertension [2,3]. The primary pathophysiological association between obesity and these comorbidities is insulin resistance (IR). The presence of IR during childhood markedly increases the risk of chronic disease progression in adulthood [4,5]. Therefore, early identification and management of IR in high-risk pediatric populations are crucial for effective long-term disease prevention.

The euglycemic-hyperinsulinemic clamp is the gold standard for quantifying insulin sensitivity; however, its invasiveness, high cost, and complexity limit its clinical applicability [6]. Consequently, there is a growing demand for simple, noninvasive, and cost-effective surrogate markers for IR that can be used in large-scale clinical settings.

Representative indices include the triglyceride-glucose (TyG) index, fasting glucose-to-insulin ratio (FGIR), and visceral adiposity index (VAI) [7-10]. Although the predictive value of these markers has been extensively validated in adults [11,12], their application in children and adolescents requires further scrutiny owing to growth- and puberty-related changes in body composition and hormonal profiles [13-15]. Among these indices, the FGIR was originally developed to assess insulin sensitivity in adults [16] and later validated in children and adolescents with obesity [17]. Although FGIR shares fasting glucose and fasting insulin with the homeostasis model assessment of insulin resistance (HOMA-IR), the two indices differ in their arithmetic structure: HOMA-IR is the product of the two variables divided by a constant, whereas FGIR is their direct ratio, with lower values indicating greater IR. This simpler ratio formulation may facilitate its use in primary care and school-based screening.

In particular, the VAI, developed for adults, may be less accurate in children because it does not fully capture ageand sex-specific adiposity distribution patterns [18,19]. Furthermore, although several studies have assessed these indices individually, direct comparisons of the diagnostic accuracies of the TyG index, FGIR, and VAI in Korean upper-elementary-school-aged children (grades 4-6) remain limited. Given that IR levels and body fat distribution differ substantially by sex during this developmental period, establishing sex-specific diagnostic cutoff values is essential for accurate clinical application [20]. Accordingly, the objective of the present study was to compare the diagnostic performance of the TyG index, FGIR, and VAI as surrogate markers of IR in Korean schoolchildren. Specifically, we aimed to determine the optimal sex-specific cutoff values for each index to identify children at high risk of IR, particularly those with abdominal obesity, during late elementary school.

METHODS

Study Participants

This cross-sectional study was conducted from June 2018 to August 2019 among fourth- to sixth-grade school-aged children in P City, Gyeonggi-do, South Korea. A total of 745 healthy children were recruited. Participants were included if they had no chronic medical conditions, including dyslipidemia, diabetes, metabolic dysfunction, hematologic disorders, infectious diseases, or endocrine disorders, and were not taking regular medications. Prior to data collection, the study objectives and procedures were explained to the students and their parents. Written informed consent was obtained from all participants and their parents or legal guardians. Of the 745 students initially recruited, 17 were excluded owing to incomplete data (missing blood samples, n = 11; body composition, n = 2; and anthropometric measurements, n = 4). Ultimately, 728 students (398 boys and 330 girls) were included in the final analysis. The study protocol was approved by the Institutional Review Board of Sungkyunkwan University (approval no. SKKU 2018-06-005-003) and conducted in accordance with the Declaration of Helsinki (2013 revision).

Measurement

Anthropometrics and clinical assessments

Height and weight were measured using an automated anthropometer (DS-102; JEnix Co., Korea). Body mass index (BMI) was calculated as weight (kg) divided by height squared (m²). Waist circumference (WC) was measured at the midpoint between the lower costal margin and the superior iliac crest using a non-stretchable tape measure. Each measurement was performed twice, and the average value was calculated. If the first two measurements differed by more than 0.5 cm, a third measurement was taken, and the average of all three measurements was recorded. The waist-to-height ratio (WHtR) was calculated as WC divided by height. Abdominal obesity was classified using age- and sex-specific 90th-percentile cutoff values for WC and WHtR from the 2017 Korean National Growth Charts published by the Korean Pediatric Society and the Korea Disease Control and Prevention Agency. The at-risk group was defined as WC or WHtR at or above the 90th percentile, and the high-risk group as both WC and WHtR at or above the 90th percentile. The international WHtR cutoff of ≥ 0.5 [21] was additionally evaluated. Resting blood pressure was measured twice using an automated sphygmomanometer (Fj-500R; Jawon Medical, Seoul, Korea) after participants had rested quietly in a seated position for at least 10 min.

Biochemical analysis and definitions

Blood samples were collected between 8:00 and 9:00 a.m., after an overnight fast of at least 10 h. Fasting blood glucose, total cholesterol, triglycerides, and high-density lipoprotein cholesterol levels were measured using an automated biochemical analyzer (ADVIA 1650; Siemens, Washington, DC, USA). Serum insulin levels were measured using a gamma counter (1470 WIZARD; PerkinElmer, Turku, Finland). The following surrogate indices of IR and visceral adiposity were calculated:

・ HOMA-IR [22]:

$$HOMA-IR=\frac{Fasting\mathit{ }insulin (\mu U/mL)\times Fasting\mathit{ }glu\mathit{c}\mathit{o}\mathit{s}e (\mathit{mg}/\mathit{mL})}{405}$$

・ TyG index [23]:

$$TyG\mathit{ }index=\mathrm{In}\left[\frac{Fasting\mathit{ }TG (mg/dL)\times Fasting\mathit{ }glu\mathit{c}\mathit{o}\mathit{s}e (\mathit{mg}/dL)}{2}\right]$$

・ FGIR [16,17]:

$$FGIR=\frac{Fasting\mathit{ }glu\mathit{c}\mathit{o}\mathit{s}e (\mathit{mg}/dL)}{Fasting\mathit{ }insulin \left(\mu U/mL\right)}$$

・ VAI was calculated using sex-specific equations [11,12]:

$$\begin{array}{rl}& \text{Males : VAI}=\left(\frac{WC}{39.68+(1.88)\times BMI}\right)\times \left(\frac{TG}{1.03}\right)\times \left(\frac{1.31}{HDL-C}\right)\\ & \text{Females :}VAI=\left(\frac{WC}{36.58+(1.89)\times BMI}\right)\times \left(\frac{TG}{0.81}\right)\times \left(\frac{1.52}{HDL-C}\right)\end{array}$$

IR was defined using sex-specific HOMA-IR cutoffs established for Korean adolescents (≥ 3.54 for boys and ≥ 3.69 for girls) [24].

Covariates

To account for potential confounding effects on the associations between IR and metabolic markers, physical activity and sexual maturity were included as covariates.

Daily physical activity levels were objectively measured using a uniaxial accelerometer (Life Corder EX; Suzuken, Nagoya, Japan). Participants were instructed to wear the device on the right side of the waist for seven consecutive days, removing it only while sleeping, swimming, or bathing. Prior to data collection, individual characteristics, including age, sex, height, and weight, were programmed into the device. The device recorded activity intensity every 4 s and categorized it as total physical activity (TPA) or moderate-to-vigorous physical activity (MVPA) [25].

Pubertal development was assessed using a self-report questionnaire based on Tanner stages [26]. Participants rated their maturation level (stages 1-5) based on standardized illustrations and descriptions of secondary sexual characteristics. For males, this included genital and pubic hair development, whereas for females, it included breast and pubic hair development. Stage 1 indicates prepubertal status, and stage 5 represents complete adult maturity. Furthermore, age and experience of the first nocturnal emission in males and menarche in females were recorded. For the analysis, pubertal maturation was treated as a continuous variable (scored 1-5), with higher scores indicating more advanced pubertal development.

Statistical Analysis

All statistical analyses were performed using SPSS version 26.0 (IBM Corp., Armonk, NY, USA). Continuous variables are presented as mean ± standard deviation. Data normality was assessed using the Kolmogorov-Smirnov test. Statistical significance was defined as P < 0.05. First, independent t-tests were used to compare anthropometric and biochemical parameters between boys and girls. Second, a one-way analysis of variance (ANOVA) with polynomial contrasts was applied to examine the associations between obesity status and metabolic parameters, including tests for linear trends across groups. Third, logistic regression models were used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the risk of IR according to obesity status. These models were adjusted for age, physical activity level, and Tanner stage to control for potential confounding effects of growth and lifestyle factors. Finally, receiver operating characteristic (ROC) curve analysis was performed to evaluate the diagnostic performance of the TyG index, FGIR, and VAI in predicting IR in children with abdominal obesity (high-risk group: 94 boys and 98 girls). The area under the curve (AUC) was calculated to measure the overall diagnostic accuracy. The optimal sex-specific cutoff values were determined using the Youden index (maximum of [sensitivity + specificity - 1). The corresponding sensitivity, specificity, and related metrics were used to assess the clinical applicability.

RESULTS

Characteristics of Study Participants

A total of 728 Korean schoolchildren (398 boys and 330 girls, mean age 11.5 ± 1.1 years) were included in the final analysis. Boys had significantly higher weight, BMI, WC, and WHtR than girls, whereas girls exhibited higher insulin, HOMA-IR, and VAI values. Additionally, boys exhibited higher TPA and MVPA (all P < 0.05; see Table 1).

Table 1.

Baseline anthropometric and biochemical characteristics of the study participants by sex

Variables	Total (n = 728)	Boys (n = 398)	Girls (n = 330)	P-value
Age (years)	11.5 ± 1.1	11.5 ± 1.2	11.5 ± 1.1	0.582
Height (cm)	146.4 ± 8.0	146.4 ± 8.2	146.5 ± 7.8	0.812
Weight (kg)	42.2 ± 10.2	43.3 ± 10.8	41.0 ± 9.2	0.003
BMI (kg/m2)	19.5 ± 3.5	20.0 ± 3.6	18.9 ± 3.2	<0.001
WC (cm)	69.7 ± 9.9	70.9 ± 10.2	68.2 ± 9.3	<0.001
WHtR	0.48 ± 0.06	0.48 ± 0.06	0.47 ± 0.06	<0.001
Systolic BP (mmHg)	105.4 ± 16.3	104.9 ± 16.4	106.1 ± 16.2	0.304
Diastolic BP (mmHg)	65.6 ± 13.4	64.8 ± 12.4	66.6 ± 14.4	0.062
TC (mg/dL)	173.6 ± 33.3	172.4 ± 32.8	175.0 ± 33.8	0.301
TG (mg/dL)	91.0 ± 51.2	88.7 ± 53.4	93.9 ± 48.4	0.177
FBG (mg/dL)	92.2 ± 10.1	92.6 ± 10.0	91.7 ± 10.3	0.212
HDL-C (mg/dL)	55.1 ± 14.4	55.6 ± 15.0	54.5 ± 13.8	0.232
Insulin (μU/mL)	9.8 ± 6.3	9.1 ± 5.8	10.7 ± 6.8	<0.001
TC/HDL	3.34 ± 1.02	3.31 ± 1.07	3.37 ± 0.96	0.364
TG/HDL	1.83 ± 1.34	1.77 ± 1.35	1.90 ± 1.33	0.202
HOMA-IR	2.26 ± 1.54	2.09 ± 1.39	2.67 ± 1.68	0.001
TyG index	8.21 ± 0.51	8.17 ± 0.53	8.23 ± 0.48	0.038
FGIR	13.32 ± 8.57	14.19 ± 8.82	12.28 ± 8.16	0.003
VAI	2.63 ± 2.12	2.05 ± 1.64	3.32 ± 2.41	<0.001
TPA (min/day)	434.9 ± 268.5	473.9 ± 293.9	388.7 ± 226.7	<0.001
MVPA (min/day)	165.1 ± 112.4	183.9 ± 124.6	142.7 ± 91.2	<0.001

BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; BP, blood pressure; TC, total cholesterol; TG, triglycerides; FBG, fasting blood glucose; HDL-C, high-density lipoprotein cholesterol; HOMA-IR, homeostasis model assessment of insulin resistance; TyG, triglyceride-glucose index; FGIR, fasting glucose-toinsulin ratio; VAI, visceral adiposity index; TPA, total physical activity; MVPA, moderate-to-vigorous physical activity.

Parameters according to Abdominal Obesity Status

Table 2 presents participant characteristics stratified by their abdominal obesity risk status (non-obese, at-risk, and high-risk). In both sexes, most anthropometric parameters and IR-related markers showed significant increasing linear trends across the three risk groups (P < 0.001). Notably, insulin, HOMA-IR, TyG, and VAI increased progressively across abdominal obesity risk groups, from non-obese to high-risk, in both boys and girls (P < 0.001).

Table 2.

Comparison of anthropometric and biochemical parameters according to abdominal obesity risk status

Variables	Boys (n=398)				Girls (n=330)
	Non-obese (n = 251)	At-risk (n = 53)	High-risk (n = 94)	P for trend	Non-obese (n = 204)	At-risk (n = 28)	High-risk (n = 98)	P for trend
Age (years)	11.6 ± 1.2	11.8 ± 1.0	11.3 ± 1.0	0.058	11.6 ± 1.2	11.1 ± 1.0	11.3 ± 1.0	0.048
Height (cm)	145.4 ± 8.0	145.8 ± 8.2	149.4 ± 8.1	<0.001	146.2 ± 7.9	143.4 ± 10.3	148.0 ± 6.5	0.067
Weight (kg)	38.3 ± 7.6	46.8 ± 8.2	54.6 ± 10.1	<0.001	36.9 ± 7.0	41.7 ± 9.6	49.3 ± 7.3	<0.001
BMI (kg/m²)	18.0 ± 2.2	21.8 ± 2.1	24.3 ± 2.7	<0.001	17.1 ± 2.0	20.0 ± 2.6	22.4 ± 2.3	<0.001
WC (cm)	64.7 ± 6.1	76.3 ± 3.4	84.3 ± 6.1	<0.001	62.5 ± 5.3	69.9 ± 3.6	79.6 ± 5.1	<0.001
WHtR	0.45 ± 0.03	0.52 ± 0.02	0.56 ± 0.03	<0.001	0.43 ± 0.03	0.49 ± 0.02	0.54 ± 0.03	<0.001
Systolic BP (mmHg)	102.1 ± 15.3	105.6 ± 15.8	111.7 ± 17.5	<0.001	105.6 ± 17.1	104.5 ± 14.8	107.7 ± 14.5	0.288
Diastolic BP (mmHg)	62.3 ± 8.8	64.9 ± 11.7	71.3 ± 17.5	<0.001	66.6 ± 14.3	70.6 ± 25.5	65.5 ± 9.6	0.509
TC (mg/dL)	167.8 ± 31.0	185.6 ± 37.8	177.3 ± 32.2	0.015	173.8 ± 33.5	183.6 ± 34.3	174.9 ± 34.3	0.783
TG (mg/dL)	79.1 ± 48.2	103.3 ± 58.2	106.3 ± 58.0	<0.001	85.6 ± 37.9	119.8 ± 59.3	103.7 ± 59.3	0.002
FBG (mg/dL)	92.6 ± 10.3	91.9 ± 11.4	93.3 ± 8.2	0.536	91.4 ± 10.8	93.9 ± 10.4	91.7 ± 9.1	0.827
HDL-C (mg/dL)	57.3 ± 14.8	49.8 ± 14.7	54.1 ± 14.8	0.072	56.4 ± 13.7	54.9 ± 16.6	50.4 ± 12.2	<0.001
Insulin (μU/mL)	7.6 ± 4.2	10.7 ± 8.3	12.0 ± 6.4	<0.001	8.76 ± 5.19	12.29 ± 4.90	14.46 ± 8.35	<0.001
TC/HDL	3.10 ± 0.95	4.00 ± 1.36	3.47 ± 1.00	0.003	3.23 ± 0.90	3.57 ± 1.12	3.62 ± 0.98	0.001
TG/HDL	1.49 ± 0.99	2.37 ± 1.85	2.17 ± 1.62	<0.001	1.64 ± 0.93	2.35 ± 1.36	2.29 ± 1.82	<0.001
HOMA-IR	1.75 ± 1.01	2.46 ± 1.99	2.79 ± 1.56	<0.001	2.01 ± 1.29	2.93 ± 1.42	3.33 ± 2.06	<0.001
TyG index	8.07 ± 0.50	8.33 ± 0.51	8.36 ± 0.55	<0.001	8.17 ± 0.46	8.52 ± 0.49	8.34 ± 0.50	0.004
FGIR	16.17 ± 9.42	12.61 ± 8.24	9.77 ± 4.88	<0.001	14.46 ± 8.95	9.34 ± 5.30	8.56 ± 4.89	<0.001
VAI	1.68 ± 1.16	2.84 ± 2.21	2.60 ± 2.01	<0.001	2.82 ± 1.65	4.18 ± 2.46	4.11 ± 3.31	<0.001
TPA (min/day)	474.8 ± 294.5	385.9 ± 296.0	523.5 ± 282.1	0.181	387.0 ± 224.8	395.1 ± 212.7	390.5 ± 236.2	0.902
MVPA (min/day)	186.5 ± 126.7	148.7 ± 125.6	197.4 ± 115.3	0.481	140.1 ± 88.3	143.0 ± 80.5	148.0 ± 100.0	0.483

At-risk was defined as WC or WHtR ≥ 90th percentile for age and sex; High-risk was defined as both WC and WHtR ≥ 90th percentile. BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; BP, blood pressure; TC, total cholesterol; TG, triglycerides; FBG, fasting blood glucose; HDL-C, high-density lipoprotein cholesterol; HOMA-IR, homeostasis model assessment of insulin resistance; TyG, triglyceride-glucose index; FGIR, fasting glucose-to-insulin ratio; VAI, visceral adiposity index; TPA, total physical activity; MVPA, moderate-to-vigorous physical activity.

Odds Ratio for Insulin Resistance

Table 3 presents the ORs for IR stratified by abdominal obesity status. Across all models, high-risk abdominal obesity (both WC and WHtR ≥ 90th percentile) was a strong predictor of IR. After adjusting for age, physical activity level, and Tanner stage (Model 2), the high-risk group had significantly higher odds of IR in the overall cohort (OR, 6.175; 95% CI, 3.958-9.633; P < 0.001). Sex-specific analysis of the adjusted model confirmed this finding, with the highest ORs observed in high-risk boys (OR, 7.021; 95% CI, 3.546-13.900; P < 0.001) and girls (OR = 5.426; 95% CI, 2.987-9.856; P < 0.001). The at-risk group (WC or WHtR ≥ 90th percentile) was a significant predictor of IR only in girls (adjusted OR, 3.392; 95% CI, 1.326-8.681; P = 0.011), whereas the association was not statistically significant in boys.

Table 3.

Odds ratios for insulin resistance according to abdominal obesity status

	Model 1		Model 2
	OR (95%CI)	P value	OR (95%CI)	P value
Overall
Non-obese	Reference		Reference
At-risk	3.063 (1.693 ‒ 5.542)	<0.001	2.552 (1.368 ‒ 4.760)	0.003
High-risk	6.120 (4.002 ‒ 9.358)	<0.001	6.175 (3.958 ‒ 9.633)	<0.001
Boys
Non-obese	Reference		Reference
At-risk	2.498 (1.061 ‒ 5.879)	0.036	2.216 (0.922 ‒ 5.324)	0.075
High-risk	5.724 (3.025 ‒ 10.831)	<0.001	7.021 (3.546 ‒ 13.900)	<0.001
Girls
Non-obese	Reference		Reference
At-risk	4.633 (1.948 ‒ 11.017)	0.001	3.392 (1.326 ‒ 8.681)	0.011
High-risk	6.334 (3.558 ‒ 11.274)	<0.001	5.426 (2.987 ‒ 9.856)	<0.001

Model 1: Unadjusted; Model 2: Adjusted for age, physical activity levels, and Tanner stage. Insulin resistance was defined as HOMA-IR ≥ 3.54 for boys and ≥ 3.69 for girls. At-risk was defined as WC or WHtR ≥ 90th percentile for age and sex; High-risk was defined as both WC and WHtR ≥ 90th percentile. OR, odds ratio; CI, confidence interval

Predictive Performance and Cutoff Values of IR Indices

ROC curve analysis was performed to evaluate the diagnostic performance of the TyG index, FGIR, and VAI for predicting IR (Figures 1 and 2). Among boys, FGIR demonstrated the highest diagnostic performance, with an AUC of 0.939 (95% CI, 0.903-0.976; P < 0.001), followed by the TyG index (AUC, 0.768) and the VAI (AUC, 0.679). The optimal FGIR cutoff value for boys was 8.435, with 94.9% sensitivity and 79.6% specificity. In girls, the FGIR outperformed the other two indices (AUC, 0.894; 95% CI, 0.835-0.954; P < 0.001), with an optimal cutoff value of 7.473, 91.2% sensitivity, and 81.2% specificity. The optimal cutoff values for all three indices differed substantially by sex (Table 4), with the FGIR cutoff value being considerably higher in boys than in girls.

Figure 1. Comparison of ROC curves for IR prediction indices in boys with abdominal obesity.

Receiver operating characteristic (ROC) curves illustrate the diagnostic performance of the fasting glucose-to-insulin ratio (FGIR; red line), triglyceride-glucose (TyG) index (blue line), and visceral adiposity index (VAI; green line) for predicting insulin resistance (IR) in boys (n = 94). IR was defined as a HOMA-IR score ≥ 3.54. The reference line (diagonal black line) represents an area under the curve (AUC) of 0.50. Sensitivity, specificity, and optimal cutoff values are provided in Table 4.

Figure 2. Comparison of ROC curves for IR prediction indices in girls with abdominal obesity.

Receiver operating characteristic (ROC) curves illustrate the diagnostic performance of the fasting glucose-to-insulin ratio (FGIR; red line), triglyceride-glucose (TyG) index (blue line), and visceral adiposity index (VAI; green line) for predicting insulin resistance (IR) in girls (n = 98). IR was defined as a HOMA-IR score ≥ 3.69. The reference line (diagonal black line) represents an area under the curve (AUC) of 0.50. Sensitivity, specificity, and optimal cutoff values are provided in Table 4.

Table 4.

Optimal cutoff values of IR indices for predicting insulin resistance in children with abdominal obesity

	Indexes	Sensitivity	Specificity	Cutoff point	AUC	95% CI	P value
Boys	TyG	0.795	0.713	8.362	0.768	0.671‒0.864	<0.001
	VAI	0.744	0.731	2.367	0.740	0.646‒0.834	<0.001
	FGIR	0.949	0.796	8.435	0.939	0.903‒0.976	<0.001
Girls	TyG	0.702	0.754	8.338	0.726	0.635‒0.818	<0.001
	VAI	0.614	0.754	3.840	0.697	0.602‒0.791	<0.001
	FGIR	0.912	0.812	7.473	0.894	0.835‒0.954	<0.001

AUC, area under the curve; CI, confidence interval; TyG, triglyceride-glucose index; FGIR, fasting glucose-to-insulin ratio; VAI, visceral adiposity index.

DISCUSSION

In the present study, we evaluated the diagnostic accuracy of the TyG index, FGIR, and VAI in identifying IR in Korean schoolchildren with abdominal obesity and established sex-specific optimal cutoff values for each surrogate marker. FGIR demonstrated superior diagnostic performance to the TyG index and VAI in both boys and girls. These results indicate that indices incorporating fasting insulin levels may be more reliable than lipid-based indices for identifying IR in this age group.

HOMA-IR and FGIR are both derived from the same two variables: fasting glucose and fasting insulin. Therefore, the high AUC values observed for FGIR (0.939 in boys and 0.894 in girls) likely reflect its agreement with HOMA-IR as a screening measure of IR, rather than independent diagnostic validation against a gold standard measure of insulin sensitivity, such as the euglycemic-hyperinsulinemic clamp. Clamp-based studies are required to establish the diagnostic accuracy of FGIR for IR. Nevertheless, the FGIR directly incorporates fasting insulin, which is a key marker of compensatory hyperinsulinemia during this developmental period. In contrast, the TyG index depends on the interaction between lipids and glucose, and its diagnostic performance may be limited in school-aged children, whose fasting triglyceride concentrations are physiologically lower than those of adults [9]. Accordingly, FGIR represents a feasible screening tool for this population. However, FGIR should not be regarded as a biologically distinct alternative to HOMA-IR, as both indices are derived from the same fasting variables and reflect the same underlying physiology. The rationale for evaluating the FGIR relies on its arithmetic and operational properties: a direct ratio is computationally simpler than a product divided by a constant, and prior validation studies in pediatric populations have reported intuitive cutoff thresholds [16,17]. These features may facilitate its use for rapid bedside assessment and consistent application in school-based and primary care settings, where simpler indices are preferred. The findings of the present study should be interpreted as evidence of agreement between FGIR and HOMA-IR within a pediatric screening framework, rather than as evidence of independent biological validity.

Previous studies have shown that the TyG index is a useful screening tool for IR in children and adolescents [27-29]. However, these studies primarily included heterogeneous populations ranging from healthy-weight individuals to those with obesity, classified according to BMI, rather than focusing on children with abdominal obesity. The diagnostic accuracy of these indices appears to vary according to obesity phenotype, age, and sex. Specifically, children with abdominal obesity have a substantially higher risk of IR and cardiovascular complications than those with other obesity phenotypes [30]. Furthermore, because compensatory hyperinsulinemia often occurs during early puberty, insulin-based markers such as FGIR may reflect these metabolic shifts more sensitively than lipid-based indices during this developmental period [31,32]. Our data are consistent with those reported in previous pediatric studies, suggesting that insulin-based indices may be more sensitive to early metabolic changes than lipid-based markers. Additionally, Hammel et al. [33] noted that insulin-based measures could identify early metabolic warning signs in children with obesity more effectively than fasting glucose or HbA1c. Similarly, the original HOMA-IR model developed by Matthews et al. [22] underscored the central role of direct insulin measurement for accurately assessing IR.

The VAI was originally developed for adults and has not yet been fully validated for use in children. In our cohort, the modest AUC values support previous reports that adult-derived visceral adiposity indices do not adequately account for the dynamic changes in body composition and fat distribution that occur during childhood and early adolescence [18-20]. These findings suggest that VAI has limited utility for detecting IR in schoolchildren and highlight the need for pediatric measures of visceral adiposity.

The sex-specific patterns observed in our logistic regression analysis may reflect underlying physiological differences between boys and girls. In these children, abdominal obesity was a statistically significant predictor of IR only in girls. Sex differences in pubertal timing and fat distribution can partly explain this discrepancy. A substantial proportion of Korean girls in grades 4-6 have already entered early puberty, a stage during which estrogen-driven increases in subcutaneous fat and leptin may strengthen the association between adiposity and IR [34,35]. Furthermore, WC has previously been identified as an independent predictor of IR in girls, performing comparably to magnetic resonance imaging (MRI)-derived visceral fat measurements and surpassing BMI [36,37]. In boys of the same age, pubertal progression is typically less advanced, and the metabolic impact of abdominal fat may not yet be sufficiently manifested to reach statistical significance. Therefore, screening for abdominal obesity may need to be initiated earlier in girls than in boys.

The identification of sex-specific cutoff values provides a practical application for early IR screening. For FGIR, the optimal cutoff values were 8.435 for boys and 7.473 for girls. This difference may reflect sex-specific patterns of fat accumulation and insulin response during the transition to puberty. Park et al. [38] and Yoon et al. [15] reported that metabolic and body composition changes during this developmental stage differ between boys and girls, potentially affecting the sensitivity of IR indices. Furthermore, Choi et al. [39] suggested that sex-specific metabolic markers may facilitate the early identification of pediatric diabetes and metabolic diseases. Considering these developmental variations, the sex-specific thresholds established in this study may provide more accurate identification of IR in schoolchildren than a single universal cutoff. The identified cutoff values could be incorporated into pediatric screening strategies. In Korea, annual school health examinations and primary care visits commonly include fasting glucose measurements. Because FGIR requires only an additional fasting insulin measurement and does not require a lipid panel, it may offer practical advantages over the TyG index and VAI in routine clinical settings. Children with FGIR values below the proposed sex-specific thresholds may warrant further evaluation, monitoring, and lifestyle intervention before the development of overt metabolic disease. As the present cohort was recruited from a single region in South Korea, these thresholds should be regarded as preliminary, and external validation in independent Korean cohorts, other ethnic populations, and against clamp-derived measures is required before broader clinical adoption.

Notably, this study has several strengths compared with previous studies. We specifically recruited children with abdominal obesity, rather than using general BMI-based criteria, to focus on the subgroup at greatest metabolic risk. Furthermore, we objectively measured physical activity using accelerometers and assessed pubertal development using Tanner staging, two important factors that are often not adequately controlled for pediatric IR studies.

Nevertheless, certain limitations of this study should be considered when interpreting its findings. The shared variable structure between FGIR and HOMA-IR indicates that the diagnostic performance reported herein represents screening-level concordance rather than independent validation. The sex-specific FGIR cutoffs proposed in this study should be considered preliminary and require external validation in independent cohorts using the euglycemic-hyperinsulinemic clamp as the reference standard. Although total and moderate-to-vigorous physical activities were measured using accelerometry and included as covariates in the adjusted models, dietary intake was not assessed, and the cross-sectional design precluded inference of the temporal relationship between physical activity and adiposity. Consequently, causal relationships cannot be established, and longitudinal studies are needed to evaluate the evolution of these metabolic profiles over time. As noted earlier, the VAI was not originally developed for pediatric populations [12], which may partly explain its limited performance in our sample. The development of a pediatric-specific VAI remains an important area for future research [40]. Additionally, given that our participants were from a single region in South Korea, the identified cutoff values require validation before application to other geographic or ethnic populations. Finally, we could not account for dietary habits or genetic factors, and the use of self-reported Tanner stages may have introduced some subjectivity compared with clinician-led evaluations.