Introduction

According to the World Health Organization (WHO) criteria, a body mass index (BMI) of 30 kg/m2 or more defines obesity. Since 1990, its worldwide prevalence among adults has more than doubled, with the age-standardized prevalence rising from 6.6% to 15.8% in 2022; in Canada, the estimated prevalence even reached 26.2%1,2. Obesity, especially visceral obesity (VO), is a critical risk factor for metabolic disorders and cardiovascular diseases, as it is strongly associated with insulin resistance and systemic inflammation3,4,5. Unlike subcutaneous adipose tissue, visceral adipose tissue (VAT) plays a more significant role in the pathogenesis of metabolic syndrome and type 2 diabetes mellitus (T2DM)3,6,7. Accurate assessment of visceral adiposity is essential for evaluating metabolic health and guiding early interventions.

Dual-energy X-ray absorptiometry (DXA) has been recognized as a reliable tool for measuring VAT, which correlates well with computed tomography (CT) and magnetic resonance imaging (MRI) estimations of visceral fat mass and volume8,9. However, its routine clinical use is limited due to the cost and radiation exposure concerns. Consequently, there is a need for simple, cost-effective methods or markers to estimate visceral adiposity. Several indices have been proposed for this purpose, including the hypertriglyceridemic waist and the visceral adiposity index (VAI)10,11. Among them, the triglyceride-glucose (TyG) index—calculated from fasting triglycerides (TG) and glucose levels—was introduced by Guerrero-Romero et al. in 201012 and has since accumulated substantial supporting evidence. It has shown a high degree of sensitivity (96.5%) and specificity (85.0%) in testing insulin resistance defined by the euglycemic-hyperinsulinemic clamp and has thus been considered an accurate and an efficient surrogate indicator of insulin resistance12,13,14,15. Recent studies have shown associations between the TyG index and various metabolic disorders, including impaired glucose tolerance (IGT), diabetes, non-alcoholic fatty liver disease (NAFLD), metabolic syndrome, and cardiovascular diseases16,17,18,19,20,21,22,23. Moreover, a recent study has highlighted the association between the TyG index and CT-measured visceral adiposity in T2DM patients. Yang et al.24 reported that among patients with T2DM, those with VO showed significantly higher TyG index values than those without VO. They proved that the TyG index was an independent risk factor for VO in both males (odds ratio [OR] = 2.997) and females (OR = 2.233) among patients with T2DM24. Additionally, the TyG index demonstrated a strong predictive power for VO in patients with T2DM, and the combination of TyG index and BMI (TyG-BMI) showed an even higher predictive power than TyG index alone24. However, currently there is no established TyG index threshold for defining VO, and its relationship with visceral adiposity in generally healthy individuals remains unclear. Assessing this relationship in healthy adults is crucial for early identification of individuals at risk for developing metabolic disorders before the onset of overt disease.

This study aimed to investigate the association between the TyG index and DXA-measured visceral adiposity in a cohort of healthy adults without serious metabolic, cardiovascular, or endocrine diseases. In addition, we sought to determine whether the TyG index could serve as a valuable predictor of VO in this population. Meanwhile, we also evaluated whether the combination of TyG index with anthropometric measures can enhance the predictive accuracy for VO, which may provide a cost-effective surrogate marker for estimating visceral adiposity and have important implications for identifying people at early risk of metabolic disorders.

Methods

Subjects and study design

This was a secondary analysis utilizing our pre-existing dataset from the Complex Diseases in the Newfoundland population: Environment and Genetics (CODING) study, conducted between 2003 and 2017. Eligibility for the original CODING study was based on the following inclusion criteria: 1) ≥ 19 years of age; 2) at least a third generation Newfoundlander; 3) healthy, without any serious metabolic, cardiovascular, or endocrine diseases; and 4) women were not pregnant at the time of the study25,26,27. All subjects provided written informed consent, and the original CODING study was approved by the Health Research Ethics Authority (HREA) of Newfoundland, St. John’s, Canada, with project identification code #10.33. All methods were performed in accordance with the relevant guidelines and regulations.

For the present analysis, we applied the following specific exclusion criteria to the initial CODING cohort of 3211 participants to ensure data completeness for our variables of interest: (1) exclusion for missing primary outcome data (n = 148): including 51 individuals with missing TG, 1 individual with missing fasting plasma glucose (FPG), and 96 individuals with missing DXA results; (2) exclusion for missing covariate data (n = 142): including 21 individuals missing waist circumference (WC) and hip circumference (HC), 69 individuals missing caloric intake, and 52 individuals missing physical activity information. After these exclusions, a total of 2921 participants were included in the final analysis. The study screening procedure is illustrated in Fig. 1. The demographic and clinical characteristics of the included participants were presented in Table 1. Briefly, the sample consisted of 71.76% females with a mean (± standard error, SE) age of 43.99 ± 0.23 years and a mean BMI of 26.97 ± 0.09 kg/m2.

As this is a secondary analysis of a pre-existing cohort, a formal a priori power calculation was not conducted for this specific study. The sample size was determined by the total number of participants in the CODING dataset who met our inclusion and exclusion criteria. However, the large sample size of the CODING cohort provided substantial statistical power for the planned analyses. A post hoc power analysis (G*Power 3.1.9.7 software) indicated that the final sample of 2921 participants provided > 90% power (α = 0.05, two-tailed) to detect a small effect size (OR ≈ 1.3) in our logistic regression models, which is well below the effect sizes we identified (e.g., OR = 3.512 for TyG in females).

Fig. 1
Fig. 1
Full size image

A flowchart depicting the enrolled participants. A total of 3211 participants from CODING study (2003–2017) were initially included. After excluding those with incomplete or missing data on TG, FPG, DXA results, and relevant covariates, a final total of 2921 participants were included in the analyses. TG, triglycerides; FPG, fasting plasma glucose; DXA, dual-energy X-ray absorptiometry; WC, waist circumference; HC, hip circumference.

Anthropometric measurements and other information

The anthropometric data were measured and recorded by trained personnel from all subjects following a 12 h overnight fast, including sex, age, body weight, height, WC, and HC. The BMI was calculated as body weight / (height ×height) (kg/m2). The waist-to-height ratio (WHtR) was calculated as WC / height28. Additionally, physical activity levels were measured via ARIC Baecke Questionnaire, which consists of work, sport, and leisure time activity indices25,26. Moreover, all participants completed a self-administered screening questionnaire for collecting information about their personal health history.

Dietary caloric intake measurements

A 124-item semiquantitative Willett food frequency questionnaire (FFQ) was used to assess dietary intake for each participant, which is the most cost-effective dietary assessment method utilized in large-scale epidemiological studies29,30. The FFQ was obtained from subjects regarding the number of weekly servings consumed of common food items over the past 12 months. Then, the quantity of each food item was converted into a daily average intake value, which was used subsequently to calculate dietary calorie (kcal/d) based on the Nutribase Clinical Nutrition Manager (software version 9.0; Cybersoft Inc, Phoenix, AZ, USA) as we previously described25,26. Caloric intake was assessed and included as a covariate in statistical models to account for its potential confounding effect on body composition and metabolic outcomes.

Body composition measurements

The whole-body DXA scans (Lunar Prodigy; GE Medical Systems, Madison, WI, USA) were taken for measurement of body composition. Scans were performed following a 12 h fast, in accordance with International Society for Clinical Densitometry (ISCD) guidelines for best practice. Total body fat percentage (BF%), trunk fat percentage (TF%), android fat percentage (AF%), arm fat percentage (ArF%), leg fat percentage (LF%), and gynoid fat percentage (GF%) were measured while patients lied horizontally on the scanner following a 12 h fast, and details were shown in our previous study25,27. VAT mass and volume were precisely delineated using the Encore CoreScan software during the analysis of the scans27, which has been recognized to be an accurate method which correlates well with CT and MRI estimations of visceral fat mass and volume8,9. Daily quality assurance was performed on the DXA scanner, and the typical coefficient of variation during the study period was 1.3%25,26,27.

Laboratory measurements

Blood samples were collected from participants following 12 h of overnight fasting. Plasma and serum samples were isolated from whole blood for subsequent analyses. FPG, serum concentrations of triacylglycerols (TG), total cholesterol (TC), and high-density lipoprotein cholesterol (HDL-C) were measured by using Synchron reagents and performed on an Lx20 analyzer (Beckman Coulter, Brea, CA). Low-density lipoprotein cholesterol (LDL-C) was calculated using the following formula: [TC - HDL-C - (TG/2.2)] which is reliable in the absence of severe hyperlipidemia31,32. The triglyceride glucose (TyG) index was calculated as the natural logarithm (Ln) of [TG (mg/dl) × FPG (mg/dl)/2]12.

Statistical analyses

SPSS software version 26.0 was used for statistical analyses. All continuous variables were presented as mean ± SE, among which caloric intake, FPG, TG, VAT mass and VAT volume were log-transformed to normalize distributions in order to perform effective data analyses. Comparisons of continuous data between males and females and between the VO and non-VO subgroups were analyzed using the independent Student’s t-test on log-transformed data where necessary. Comparisons of continuous data among quartiles of TyG index were performed using one-way analysis of variance (ANOVA) test, followed by pairwise comparisons using least significant difference (LSD).

Sex-specific VO was defined as those with the VAT mass or VAT volume ≥ 85th percentile; the cut-off values are 1243.37 g and 1317.97 cm3 in female, and 2259.33 g and 2394.89 cm3 in male, respectively. The 85th percentile cut-off was selected as it is a well-established criterion for identifying individuals with high visceral adiposity burden in epidemiological studies and aligns with methodologies used in previous relevant research33,34. Clinical variables and TyG index were compared between the VO and non-VO groups. Pearson partial correlation analysis was used to identify significant correlations of TyG index with clinical measurements and body composition, in which the confounding variables including age, physical activity, and caloric intake were statistically controlled. Logistic regression analysis was used for identifying clinical risk factors associated with VO, with adjustment for potential confounders ascertained based on prior publications9,25,26,27. Receiver operating characteristic (ROC) curve analyses were applied to assess the diagnostic performance of the independent risk factors for VO according to the area under the curve (AUC), and the comparison of AUCs was further analyzed for the predicting power of VO. A two-tailed P value < 0.05 was considered statistically significant.

Results

Clinical and body composition characteristics by sex

This study encompassed a total of 2921 participants, comprising 825 males and 2,096 females, as detailed in Table 1. The disparity in sample sizes reflects a higher proportion of female volunteers in community-based health surveys, a common epidemiological pattern. Given the considerable difference in sample size, comparisons were made using appropriate statistical tests (independent t-test) on log-transformed non-normal data where necessary. Male participants were on average 3.76 years younger than females and exhibited significantly higher values in body weight, BMI, WC, physical activity, caloric intake, FPG, TG, and the TyG index (P<0.001), with no significant difference in WHtR. Conversely, males had lower levels of TC, and HDL-C than females, with no significant difference in LDL-C levels. Additionally, DXA results highlighted distinct fat distribution patterns between sexes, with females showed greater overall and peripheral adiposity compared to males. In terms of central fat distribution, males, despite having lower TF% and AF%, displayed significantly higher visceral adiposity, with VAT mass (1239.50 g vs. 686.16 g, P<0.001) and VAT volume (1313.87 cm3 vs. 727.33 cm3, P<0.001) being nearly twice as high in males as in female.

Table 1 Clinical and body composition characteristics by sex.
Full size table

Body composition variations by TyG index quartiles

Analysis of clinical obesity-related measurements and body composition, including visceral adiposity levels, across quartiles of the TyG index is presented in Tables 2 and 3. The findings indicated that body weight, BMI, WC, and WHtR increased with ascending quartiles of the TyG index in both sexes (P < 0.001 for all). DXA-derived body composition data revealed a direct, dose-dependent relationship between the TyG index and visceral adiposity for both sexes, with the most pronounced differences observed in visceral adiposity. Participants in the fourth quartile, as compared to those in the first, had markedly higher VAT mass and volume, increasing by 2.32-fold in males and 1.85-fold in females, respectively.

Table 2 Body composition variations by TyG index quartiles in Male.
Full size table
Table 3 Body composition variations by TyG index quartiles in Female.
Full size table

TyG index variations by VO status

Comparisons of clinical obesity-related measurements and the TyG index between non-VO and VO groups are displayed in Table 4. The VO group demonstrated higher body weight, BMI, WC, and WHR in both males and females. Additionally, individuals with VO exhibited higher TyG index levels than those without VO, with values of 8.95 vs. 8.39 in males and 8.81 vs. 8.20 in females (P<0.001 for both).

Table 4 Clinical characteristics and TyG index variations by VO status.
Full size table

Association of TyG index with clinical measurements and body composition

The association between TyG index and clinical measurements, as well as body composition, is illustrated in Fig. 2. After statistically controlling for confounding variables including age, physical activity, and caloric intake, the TyG index was found to be significantly correlated with all body composition indicators in both males and females, with moderate effect sizes (R2). Notably, the TyG index had a stronger association with central obesity indicators, including TF%, AF%, A/G, VAT mass, and VAT volume. Specifically, in males, the TyG index showed positive correlations with TF% (r = 0.37, R2 = 13.7%), AF% (r = 0.38, R2 = 14.4%), A/G (r = 0.33, R2 = 10.9%), VAT mass (r = 0.36, R2 = 13.0%), and VAT volume (r = 0.36, R2 = 13.0%). Females also displayed similar significant correlations with TF% (r = 0.32, R2 = 10.2%), AF% (r = 0.33, R2 = 10.9%), A/G (r = 0.37, R2 = 13.7%), VAT mass (r = 0.35, R2 = 12.3%), and VAT volume (r = 0.35, R2 = 12.3%).

Fig. 2
Fig. 2
Full size image

Correlations between clinical measurements and body composition in males (A) and females (B). In the partial correlation analysis, confounding variables including age, physical activity, and caloric intake were statistically controlled. ns P > 0.05 (non-significant). BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; TyG index, triglyceride- glucose index; BF, total body fat; TF, trunk fat; AF, android fat; A/G, android-to-gynoid fat ratio; VAT, visceral adipose tissue; ArF, arm fat; LF, leg fat; GF, gynoid fat.

Logistic regression analyses for identifying clinical risk factors associated with VO

After adjusting for confounding variables including age, physical activity, and caloric intake, logistic regression analysis showed that BMI (OR = 1.302 for males and 1.263 for females, respectively), WC (OR = 1.077 for males and 1.070 for females, respectively), and TyG index (OR = 1.575 for males and 3.512 for females, respectively) were independent risk factors associated with VO in both sexes (Fig. 3).

Fig. 3
Fig. 3
Full size image

Forest plot of independent risk factors for VO identified from logistic regression analyses. In the logistic regression analysis, confounding variables including age, physical activity, and caloric intake were statistically controlled. BMI, body mass index; WC, waist circumference; TyG index, triglyceride-glucose index; OR, odds ratio.

Predictive ability of TyG index for VO by ROC analyses

ROC analyses demonstrated robust predictive power of BMI, WC, WHtR, and TyG index for VO in males (AUC = 0.903, 0.914, 0.916, and 0.733, respectively) and females (AUC = 0.912, 0.912, 0.911, and 0.791, respectively) (Fig. 4). The TyG cutpoints for VO were 8.49 for males and 8.35 for females (Table 5). Subsequently, we further assessed the predictive power of TyG related indices. As shown in Fig. 5, among females, TyG-BMI, TyG-WC, and TyG-WHtR exhibited significantly higher predictive ability for VO, with AUC values of 0.930, 0.929, and 0.926, respectively, compared with BMI (AUC = 0.912, P<0.001), WC (AUC = 0.912, P<0.001), WHtR (AUC = 0.911, P<0.001), and the TyG index (AUC = 0.791, P<0.001). In males, TyG-BMI, TyG-WC, and TyG-WHtR also demonstrated higher predictive ability (AUC = 0.904, 0.902, and 0.903, respectively) than the TyG index (AUC = 0.733, P<0.001 for all), but showed no significance when compared with BMI (AUC = 0.903), WC (AUC = 0.914), and WHtR (AUC = 0.916).

Fig. 4
Fig. 4
Full size image

Predictive power of BMI, WC, WHtR, and TyG index for VO by ROC analyses in males (A) and females (B). ROC, receiver operating characteristic curve; AUC, area under the curve; BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; TyG index, triglyceride-glucose index.

Table 5 ROC analysis for the cut-off values of BMI, WC, and TyG-index.
Full size table
Fig. 5
Fig. 5
Full size image

Predictive power of TyG related indices for VO by ROC analyses in males (A–C) and females (D–F). ROC, receiver operating characteristic curve analyses; AUC, area under the curve; BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; TyG index, triglyceride-glucose index.

Discussion

This study investigated the association between TyG index and VO in a generally healthy Newfoundland population from the CODING study. As VO is a vital health issue related to metabolic disorders and cardiovascular diseases, investigating early screening and identification methods for populations at high risk of VO should be addressed. Our findings revealed that besides traditional anthropometric measures, including BMI, WC, and WHtR, biochemical markers like the TyG index, could also be regarded as a simple, low-cost surrogate marker for identifying individuals at risk of VO in the generally healthy population. According to our study, subjects with higher levels of BMI (>29.81 kg/m2 in males, >28.46 kg/m2 for females, respectively), WC (>101.95 cm in males, >97.20 cm in females, respectively), WHtR (>0.59 in males, >0.61 in females, respectively), and TyG index (>8.49 in males, >8.35 in females, respectively) were more prone to have VO, which indicates that further examination is warranted for those subjects.

Our study demonstrated significant sex differences in body fat distribution characteristics. Males exhibited nearly twice the VAT mass and volume compared to females, despite having lower BF%, and even lower TF% and AF%. This is in line with a number of previous studies showing that men are more prone to central fat accumulation among Canadian and American populations, especially visceral fat, which is metabolically more active and detrimental than subcutaneous fat3,4,35,36. Conversely, females displayed higher BF% and peripheral fat distribution, consistent with hormonal influences such as estrogen, which directly or through activation of their receptors on adipocytes and in adipose tissues, promotes lipid accumulation in subcutaneous depots on hips and thighs rather than in visceral depots37,38. Moreover, human genetics have also been proposed to explain the sex differences in adipose distribution39,40.

Emerging evidence suggests that central fat accumulation, especially visceral fat deposition, is a key factor underlying the pathogenesis of insulin resistance and thereby contributes to the development of metabolic disorders41,42. Free fatty acids (FFAs) induced lipotoxicity plays a crucial role in this association and has been recognized as the main cause of insulin resistance under obesity status43. Excessive FFAs released from dysfunctional adipose tissue due to obesity, lead to an augmented deposition of FFAs and synthesis of TG in different visceral organs, such as liver, which results in hypertriglyceridemia and subsequently enhances hepatic glucose output42,43. Thus, TyG index, being a product of TG and FPG, is not only considered a reliable marker of insulin resistance but also shows potential theoretically as an indicator of VAT levels. Indeed, in this study, a direct dose-dependent relationship between the TyG index and visceral adiposity was observed in a cohort of healthy Newfoundland population without serious metabolic, cardiovascular, or endocrine diseases. Participants in the highest TyG quartile exhibited VAT mass and volume increases of 2.32-fold in males and 1.85-fold in females compared to those in the lowest quartile. In addition, individuals with VO exhibited higher TyG index levels than those without VO, with values of 8.95 vs. 8.39 in males and 8.81 vs. 8.20 in females. Notably, the absolute TyG values associated with VO in this healthy population (8.95 in males and 8.81 in females) were markedly lower than those reported in studies of metabolically unhealthy cohorts from similar demographics. Yang et al.24 found that among the Chinese population with T2DM, patients with VO showed higher TyG index values than patients without VO in males (9.74 vs. 8.88) and females (9.59 vs. 9.01). This pattern is consistent with the findings of research on US adults with metabolic syndrome or T2DM, which reported that mean TyG indices exceeded 9.0 and that values above 9.5 were strongly correlated with high cardiovascular disease risk44,45. The lower values in our healthy cohort underscore the TyG index’s sensitivity to metabolic health status and reinforce its utility as a surrogate marker of visceral adiposity across the metabolic health spectrum.

Moreover, stronger correlations between the TyG index and central obesity indicators such as TF%, AF%, and A/G were found in this study, and this positive association has also been reported in children46, underscoring its utility as a marker for central fat accumulation. Furthermore, in this study, the TyG index showed positive correlations with VAT mass (r = 0.36 for males and 0.35 for females, respectively) and VAT volume (r = 0.36 for males and 0.35 for females, respectively), which also supports its role as a reliable predictor of visceral adiposity. These findings align with study reported by Yang et al.24, who revealed that the TyG index significantly correlated with CT-measured visceral adipose area (VAA) both in males (r = 0.380) and females (r = 0.365) with T2DM. Interestingly, in this study, although males had higher absolute levels of visceral adiposity, logistic regression analyses revealed that OR for VO associated with the TyG index was 3.512 in females compared to 1.575 in males. The stronger association in females may reflect sex-specific differences in adipose biology and metabolic responses, as evidenced by hormonal influences on lipid and glucose metabolism47,48.

BMI has been known as an anthropometric index to assess overall adiposity, whereas WC and WHtR are used to evaluate abdominal adiposity49. Our findings confirm the robust predictive power of these measures for VO, with WHtR exhibiting the highest predictive power for VO, which aligns with the study conducted by Swainson et al.28, who demonstrated that among common anthropometric indices (BMI, WC, WHR), WHtR was the strongest predictor of VAT mass in both sexes. However, the evidence related to TyG index, as a biochemical indicator, as well as its combination with anthropometric indices for predicting VO is limited. Two observational studies conducted in Japan among general population reported that TyG index had an equal predictive power with BMI and WC for NAFLD with the cut-off values of TyG index were 8.2150 and 8.2551 for the overall population, respectively. In addition, TyG-BMI and TyG-WC showed even higher predictive performance for NAFLD with their AUCs greater than 0.90 compared to TyG, BMI, or WC alone50,51. The above studies highlighted the predictive role of TyG index and its related indices on VO. Moreover, a recent study conducted among patients with T2DM in China revealed that TyG index ranked second to BMI for predicting VO, with the cut-off values of TyG index were 8.91 for males and 9.07 for females24. They further showed that TyG-BMI had a higher predictive power than BMI for VO in male patients with T2DM but no significance was found when compared with BMI in female patients with T2DM. Similarly, our study demonstrated that BMI and WC exhibited higher predictive power for VO than the TyG index with its cut-off values were 8.49 for males and 8.35 for females. After combining the TyG index with anthropometric measures, including TyG-BMI and TyG-WC, diagnostic performance for VO were significantly enhanced than BMI or WC in females. Whereas in males, compared with BMI or WC, combining the TyG index did not significantly improve their predictive power. This sex difference warrants further study. Therefore, the above studies indicated that TyG index had a good predictive performance for VO, with the consistency of TyG index cut-off values close to 8.5. These studies further suggest that while BMI and WC remain useful anthropometric measures, combining them with biochemical markers like the TyG index, provide a more accurate assessment method of visceral adiposity and may improve the early diagnosis of VO. Additionally, our study extends these observations to a generally healthy population, emphasizing the broader applicability of these combined indices.

Despite its strengths, this study has several limitations. First, our analysis focused specifically on the TyG index and its derived parameters. We acknowledge that other valuable indices for assessing visceral adiposity and metabolic risk exist, such as the hypertriglyceridemic waist and the VAI10,11, and their comparative predictive performance in our cohort represents an important area for future investigation. Furthermore, the cross-sectional design precludes causal inference about the relationship between the TyG index and VO, and longitudinal studies are warranted to assess whether interventions targeting reductions in the TyG index can effectively mitigate visceral adiposity and its associated health risks. The findings may not be generalizable beyond this Newfoundland cohort due to potential genetic or environmental differences. Future research should focus on validating these findings in larger, more diverse populations, directly comparing multiple indices, and exploring potential mechanisms underlying these associations.

Conclusion

In conclusion, this study demonstrates that the serum TyG index is a significant predictor of VO among healthy adults in Newfoundland. The combination of TyG index with anthropometric measures, including BMI, WC, and WHtR, further enhances their predictive utility in females. These findings highlight the potential utility of TyG index and its related indices in clinical practice as simple, low-cost screening tools for early identification of individuals at risk of VO, particularly valuable for screening at the population level in primary care.