2026 年 73 巻 2 号 p. 205-215
Diabetes mellitus (DM) is a key global public health issue with rising incidence. The triglyceride-glucose (TyG) index has been widely applied to assess insulin resistance and metabolic abnormalities in recent years. However, the relation of the TyG index with DM is elusive when combined with central obesity indicators. This research was conducted to probe into the relationship between TyG-derived indices and DM. A total of 10,729 participants from the NHANES database were enrolled, of whom 1,984 had DM. The linkage of five TyG-derived indices with DM was examined using a weighted logistic regression model. At the same time, stratified analysis was undertaken on different gender and age subgroups. To evaluate the predictive performance of each indicator, ROC curve analysis was undertaken to examine the predictive capability of different indicators. The findings indicated that all TyG-derived indices were greatly positively linked with the risk of DM. The AUC values of TyG-CI and TyG-WWI were considerably higher than those of the TyG index and other indicators, demonstrating a stronger capability to predict the risk of DM. In subgroup analyses, both TyG-CI and TyG-WWI exhibited high robustness across different populations regardless of gender or age. The indices with the TyG index combined with indicators related to central obesity, especially TyG-CI and TyG-WWI, are effective tools for predicting the risk of DM.
Diabetes mellitus (DM) is one of the most prevalent chronic diseases with high blood sugar (BS). According to the estimates of the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) in 2019, DM is the eighth major culprit of death and disability globally, affecting nearly 460 million people worldwide [1] and approximately 529 million people in 2021 [2]. The incidence of DM continues to rise. DM imposes a daunting economic burden on the public health system, with global health expenditures related to DM estimated at $966 billion in 2021, which is estimated to be up to $1.054 trillion by 2045 [3]. In the pathophysiology of DM, insulin resistance and β-cell dysfunction are considered core driving factors of type 2 DM (T2DM) [4]. The identification of high-risk populations in the early stages is instrumental for the prevention and control of DM.
Recently, the triglyceride-glucose (TyG) index has been regarded as an effective indicator for identifying individuals with insulin resistance and high risk of DM due to its convenience and effectiveness [5], having a bearing on the risk of cardiovascular disease (CVD) [6]. A study based on the Chinese population has uncovered a positive link between the TyG index and the risk of T2DM [7]. Additionally, a higher TyG index is connected with an elevated risk of all-cause and non-cardiovascular mortality in patients with T2DM under 65 years old [8]. Moreover, central obesity is an independent predictor of DM [9]. A meta-analysis exhibited that waist circumference (WC) >102 cm (men)/88 cm (women) is better at predicting the development of DM than body mass index (BMI) >30 [10]. After adjusting for confounding factors, based on propensity score matching analysis, the study found that compared to people without central obesity, those with central obesity had a 72% elevated risk of getting DM [11]. Therefore, combining the TyG index with various indicators reflecting central obesity, such as weight-adjusted waist index (WWI), conicity index (CI), body shape index (BSI), relative fat mass (RFM), and body roundness index (BRI) [12], may provide a more comprehensive and accurate assessment for predicting the risk of DM. However, current research on the TyG index mainly relies on blood glucose and lipid parameters and fails to integrate it with the key indicators of central obesity. In addition, although some studies have attempted to combine TyG with a single obesity indicator, there is a lack of systematic comparison and validation of multiple central obesity derived indicators.
We herein designed a study with a large sample of data from the National Health and Nutrition Examination Survey (NHANES) used to systematically evaluate the association between five TyG-central obesity combined indicators (TyG-BRI, TyG-BSI, TyG-WWI, TyG-CI, TyG-RFM) and DM for the first time, and compare their predictive effectiveness. By integrating central obesity indicators and breaking through the limitation of the traditional TyG index depending on blood glucose parameters, this study aimed to reveal whether the combination of “insulin resistance + central obesity” double targets can optimize the risk stratification of DM and provide more accurate tools for grassroot screening.
The NHANES systematically collected health and nutrition data from a nationally representative sample of non-institutionalized Americans using a stratified, multi-stage probability sampling method. We investigated the linkage of DM with TyG-RFM, TyG-BRI, TyG-CI, TyG-WWI, and TyG-BSI using data from six cycles from 2007 to 2018 (n = 59,842). After excluding missing and invalid data for DM (n = 2,486), missing and invalid data for TyG index (n = 39,231), missing data for height, WC, and weight (n = 785), and data of participants with missing data for covariate (n = 6,611), a total of 10,729 participants were left (Fig. 1).
The indices with the TyG index combined with indicators reflecting central obesity (BRI, BSI, WWI, CI, RFM) were the independent variables in this research. After at least 8 hours of fasting, fasting blood glucose (FBG) and fasting triglyceride (TG) were assessed using enzymatic colorimetric methods on blood samples. FBG was tested by applying the Roche Modular P or Cobas C501 and C311 systems. TG levels were examined by a Roche Modular P or Roche Cobas 6000 biochemical analyzer. The calculation formulas are as follows: TyG index = ln (TG (mg/dL) × FPG (mg/dL)/2) [13]; TyG-BRI index = TyG index × [364.2 – 365.5 × [1 – (WC/2π)2/(0.5 × Height)2]1/2] [14]; TyG-BSI index = TyG index × [WC/(BMI2/3 × Height1/2)] × 1,000 [15], where BMI = Weight/Height2. TyG-WWI index = TyG index × [WC/Weight1/2] × 100 [16]; TyG-RFM index = TyG index × [64 – (20 × Height/WC) + (12 × Sex)], where Sex = 1 for female and Sex = 0 for male [17]. TyG-CI index = TyG index × [WC/[0.109 × (Weight/Height)1/2]] [18]. Height and WC were measured in meters (m), while Weight was measured in kilograms (kg).
2.3 Dependent variablesPeople who had any of the following states met were classified into the DM group: (1) DM diagnosis informed by a health professional or doctor ; (2) fasting plasma glucose (FPG) ≥126 mg/dL; (3) glycated hemoglobin A1c (HbA1c) ≥6.5%; (4) taking antidiabetic medication [19].
2.4 VariablesThe covariates of this research included race, gender, BMI, smoking, alcohol consumption, age, education level, hypertension, CVD, poverty income ratio (PIR), physical activity (PA), insulin, high-density lipoprotein cholesterol (HDL-C), family history of DM, and low-density lipoprotein cholesterol (LDL-C). PIR is the ratio of family income to the poverty line. Three groups were generated: high-income (PIR ≥ 3.5), moderate-income (1.3 ≤ PIR < 3.5), and low-income (PIR < 1.3) groups based on PIR [20]. Individuals were classified into the following three BMI categories: obesity (>30 kg/m2), overweight (25–30 kg/m2), and underweight/healthy weight (<25 kg/m2) [21]. Smoking status was classified as current smoking (smoking a total of 100 or more cigarettes and currently smoking), former smoking (smoking a total of 100 or more cigarettes but currently not smoking), and never smoking (smoking a total of less than 100 cigarettes) [22]. The standard for alcohol consumption was drinking at least 12 drinks per year [23]. Participants with hypertension were defined as having any of the following: (1) diastolic blood pressure ≥80 mmHg or systolic blood pressure ≥130 mmHg; (2) hypertension diagnosis informed by a doctor or other health professional; (3) currently taking antihypertensive medication [24]. Standards for CVD were set as having a stroke, heart attack, coronary heart disease, angina, or congestive heart failure (CHF) [25]. Those who responded “yes” to the question of whether any of their close family members—including a mother, father, sister, or brother—had been informed by a professional having DM were considered to have a family history of the condition [26]. Participants were sorted into three groups based on PA: no-PA group, moderate-PA group, and vigorous-PA group (vigorous PA: activities that can result in a large amount of sweating or a considerable elevation in breathing or heart rate; moderate PA: activities that result in slight sweating or a slight to moderate elevation in breathing or heart rate) [27].
2.5 Statistical analysisAnalysis was finished by R software (V4.4.1). Participants were grouped according to total population characteristics to determine if they had DM. A baseline table was generated by utilizing the “tableone” package. Sample size and percentage (n (%)) were employed to show categorical variables. As for continuous variables, mean and standard deviation (mean (sd)) were employed. A weighted logistics regression model between 5 TyG-derived indicators and DM was constructed using the “survey” package. The restricted cubic spline (RCS) was set up by utilizing the “rms” package to figure out the nonlinear link of TyG-derived indicators with DM. Subgroup analyses of gender and age were done. We built two models to adjust covariates. Model 1 had adjustments in age, race, and gender, while Model 2 had adjustments for age, smoking, gender, BMI, drinking, race, PIR, CVD, PA, education level, hypertension, insulin, family history of DM, LDL-C, and HDL-C. To test the value of these indicators in predicting DM, the receiver operating characteristic (ROC) curve was graphed with the use of “pROC” package.
Since NHANES data does not distinguish the type of DM, this study cannot determine the specific proportion of Type 1 DM (T1DM)/Type 2 DM (T2DM) in DM subjects. TyG-derived indicators mainly reflect insulin resistance (the core pathological mechanism of T2DM), we therefore excluded subjects who were less than 30 years old at the time of diagnosis by referring to the early-onset epidemiological characteristics of T1DM [28, 29]. We conducted a sensitivity analysis to explore whether the association between TyG-derived indicators and DM is mainly driven by T2DM. In addition, in order to verify the predictive value of TyG-derived indicators in the early stage of diabetes, we limited the analysis objects to participants with FPG <126 mg/dL (including patients who were diagnosed but whose FPG did not reach the diabetes threshold), and conducted sensitivity analysis to exclude the dominant influence of hyperglycemia on the correlation of indicators. Using the same weighted logistic regression model (Model 2) and ROC curve analysis as the main analysis, we evaluated the effect and predictive power of the indicators in this subgroup. Statistics are deemed significant when p < 0.05.
This project enrolled a total of 10,729 participants, including 1,984 DM patients (13.7%). Two groups were identified according to whether the participants had DM or not. Participants in the DM group, compared with those in the non-DM group, were significantly older (59.00 ± 13.51 vs. 45.28 ± 16.56, p < 0.001), and predominantly obese people (62.6% vs. 32.3%, p < 0.001), male (52.6% vs. 48.5%, p = 0.010), with moderate and low PIR (64.6% vs. 56.2%, p < 0.001). The prevalence rates of hypertension (79.1% vs. 43.1%, p < 0.001) and CVD (22.8% vs. 6.1%, p < 0.001) in the DM group were considerably higher than those in the non-DM group. Additionally, the insulin levels and 5 TyG-derived indices (TyG-WWI, TyG-RFM, TyG-CI, TyG-BRI, TyG-BSI) in the DM group were remarkably higher than those in the non-DM group (Table 1).
| Characters | Total | Non-DM | DM | p |
|---|---|---|---|---|
| Overall | 10,729 | 8,745 (86.3) | 1,984 (13.7) | |
| Age | 47.16 (16.85) | 45.28 (16.56) | 59.00 (13.51) | <0.001 |
| Gender | 0.010 | |||
| Male | 5,268 (49.0) | 4,208 (48.5) | 1,060 (52.6) | |
| Female | 5,461 (51.0) | 4,537 (51.5) | 924 (47.4) | |
| Race | <0.001 | |||
| Mexican American | 1,611 (8.2) | 1,275 (8.0) | 336 (9.2) | |
| Other Hispanic | 1,109 (5.4) | 875 (5.3) | 234 (5.9) | |
| Non-Hispanic White | 4,732 (68.6) | 3,987 (69.4) | 745 (63.9) | |
| Non-Hispanic Black | 2,069 (10.6) | 1,599 (10.0) | 470 (13.9) | |
| Other race | 1,208 (7.3) | 1,009 (7.3) | 199 (7.1) | |
| BMI (kg/m2) | <0.001 | |||
| <25 | 3,104 (30.0) | 2,837 (33.0) | 267 (11.4) | |
| 25–30 | 3,627 (33.6) | 3,053 (34.8) | 574 (25.9) | |
| >30 | 3,998 (36.4) | 2,855 (32.3) | 1,143 (62.6) | |
| Alcohol | <0.001 | |||
| No | 2,596 (19.2) | 2,025 (18.2) | 571 (25.5) | |
| Yes | 8,133 (80.8) | 6,720 (81.8) | 1,413 (74.5) | |
| Smoke | <0.001 | |||
| Never | 5,951 (55.5) | 4,946 (56.3) | 1,005 (50.3) | |
| Past | 2,690 (25.7) | 2,024 (24.2) | 666 (34.8) | |
| Now | 2,088 (18.8) | 1,775 (19.5) | 313 (14.9) | |
| Education | <0.001 | |||
| Less than high school | 942 (4.5) | 666 (4.0) | 276 (7.6) | |
| High school or equivalent | 3,925 (33.1) | 3,098 (32.0) | 827 (39.7) | |
| College or above | 5,862 (62.4) | 4,981 (64.0) | 881 (52.6) | |
| PIR | <0.001 | |||
| Low | 32,650 (21.1) | 2,631 (20.8) | 634 (23.0) | |
| Medium | 4,122 (36.2) | 3,272 (35.4) | 850 (41.6) | |
| High | 3,342 (42.7) | 2,842 (43.8) | 500 (35.4) | |
| Hypertension | <0.001 | |||
| No | 5,085 (52.0) | 4,676 (56.9) | 409 (20.9) | |
| Yes | 5,644 (48.0) | 4,069 (43.1) | 1,575 (79.1) | |
| CVD events | <0.001 | |||
| No | 9,604 (91.6) | 8,088 (93.9) | 1,516 (77.2) | |
| Yes | 1,125 (8.4) | 657 (6.1) | 468 (22.8) | |
| Physical activity | <0.001 | |||
| None | 3,970 (41.9) | 3,539 (45.0) | 431 (22.7) | |
| Moderate | 3,430 (32.6) | 2,740 (31.8) | 690 (37.8) | |
| Intense | 3,329 (25.5) | 2,466 (23.2) | 863 (39.5) | |
| Family history of DM | <0.001 | |||
| No | 6,220 (60.4) | 5,529 (64.2) | 691 (36.1) | |
| Yes | 4,509 (39.6) | 3,216 (35.8) | 1,293 (63.9) | |
| Insulin (uU/mL) | 12.61 (13.79) | 11.29 (9.11) | 20.91 (28.07) | <0.001 |
| HDL (mg/dL) | 54.32 (16.05) | 55.24 (15.96) | 48.49 (15.34) | <0.001 |
| LDL (mg/dL) | 114.00 (35.17) | 115.67 (34.51) | 103.48 (37.47) | <0.001 |
| TyG-BRI | 46.41 (21.57) | 43.35 (19.39) | 65.71 (24.50) | <0.001 |
| TyG-BSI | 697.61 (73.12) | 686.52 (66.83) | 767.64 (72.30) | <0.001 |
| TyG-WWI | 93.89 (11.43) | 92.01 (10.32) | 105.82 (10.89) | <0.001 |
| TyG-RFM | 302.21 (80.18) | 293.74 (76.64) | 355.69 (81.43) | <0.001 |
| TyG-CI | 11.19 (1.33) | 10.97 (1.21) | 12.58 (1.26) | <0.001 |
n (%) represented the categorical variable and mean (sd) represented the continuous variable. n was unweighted. n (%), mean, and sd were weighted.
The link between TyG-derived indices and DM was analyzed in a weighted logistics regression model. The results manifested that five TyG-derived indicators were greatly positively linked with DM in a model that adjusted for all confounders. Among them, TyG-CI had the strongest link with DM (OR = 2.565, 95%CI: 2.332–2.822, p < 0.001) (Table 2). We further constructed RCS to dissect the non-linear linkage. There was a remarkable overall trend between TyG-derived indicators and DM (p-overall <0.05). As each indicator elevated, the risk of DM dramatically rose. Obvious non-linear linkages were discovered between TyG-BSI (p-non-linear = 0.0205) and DM as well as between TyG-WWI (p-non-linear = 0.0148) and DM. The non-linear links of the other 3 indicators with DM were not remarkable (Fig. 2).
| Characteristic | OR (95% CI) | ||
|---|---|---|---|
| Crude model | Model 1 | Model 2 | |
| TyG-BRI | 1.043 (1.040–1.046)*** | 1.045 (1.041–1.049)*** | 1.031 (1.024–1.037)*** |
| TyG-BSI | 1.018 (1.016–1.019)*** | 1.016 (1.015–1.018)*** | 1.015 (1.013–1.016)*** |
| TyG-RFM | 1.010 (1.009–1.011)*** | 1.024 (1.022–1.026)*** | 1.027 (1.024–1.030)*** |
| TyG-WWI | 1.133 (1.124–1.143)*** | 1.131 (1.120–1.142)*** | 1.113 (1.101–1.126)*** |
| TyG-CI | 2.965 (2.763–3.183)*** | 2.892 (2.671–3.130)*** | 2.565 (2.332–2.822)*** |
The crude model had no adjustments. Model 1 had adjustments in age, race, and gender, while Model 2 had adjustments for age, gender, BMI, drinking, race, smoking, PIR, CVD, PA, education level, hypertension, insulin, family history of DM, LDL-C, and HDL-C. *** p-value <0.001.
The RCS curves were adjusted for age, gender, race, BMI, smoking, drinking, PIR, education level, hypertension, CVD events, PA, family history of DM, insulin, LDL-C, and HDL-C.
Significant differences in WC among different genders and age groups have been explained in the former study [30]. Most of the TyG-derived indices in this research were closely linked to WC. In light of this, we launched subgroup analyses stratified by gender and age. In each subgroup, five TyG-derived indicators were considerably positively linked with DM risk, and this association was consistent across gender and age groups (Fig. 3).
Subsequent ROC curve analysis was undertaken to examine the predictive value of TyG-derived indices for DM. The TyG index and its derived indicators were capable of predicting DM well. The AUC values of TyG-CI and TyG-WWI were 0.801 (95%CI: 0.790–0.811) and 0.800 (95%CI: 0.789–0.810), respectively, higher than the TyG index and other indicators, indicating that these two indicators had better performance in predicting the risk of DM (Fig. 4A, Supplementary Table 1). The predictive effects of TyG-CI and TyG-WWI were better than other indicators (Fig. 4B–E, Supplementary Table 1) in different gender and age subgroups.
After excluding DM patients whose diagnosis age was less than 30 years, 9,987 participants were included in this study, including 1,242 DM patients. Further sensitivity analysis results showed that the association between TyG-derived indicators and DM was still highly significant (p < 0.001), and the effect amount was consistent with the main analysis (Supplementary Table 2).
The sensitivity analysis results of the subgroup without obvious hyperglycemia (FPG <126 mg/dL, n = 9,394, including 649 patients with diabetes) showed that after full adjustment, all TyG-derived indicators were still significantly related to the risk of diabetes (Supplementary Table 3). In addition, ROC analysis further demonstrated that TyG-CI and TyG-WWI had good risk prediction ability for diabetes (Supplementary Fig. 1).
Based on the NHANES database, this work systematically evaluated the linkage between five TyG-derived indicators (TyG-CI, TyG-BRI, TyG-WWI, TyG-RFM, TyG-BSI) and DM. All TyG-derived indices were remarkably positively linked with the risk of DM, with TyG-CI and TyG-WWI performing better than other indices in predicting the risk of DM.
The TyG index is a comprehensive biochemical indicator that manifests the combined effects of TG and glucose on human metabolism. It is widely considered an effective alternative indicator for identifying insulin resistance [5]. In a Brazilian population-based study, the TyG index outperforms the steady-state model assessment in predicting insulin resistance [31]. Postprandial clearance of triglyceride-rich lipoprotein (TRL) particles is reduced in diabetic patients, resulting in elevated TRL levels [32]. Higher fasting TG levels are considered predictors of impaired glucose tolerance in people without DM at baseline [33]. According to an observational study, higher TG levels have a bearing on the higher risk of DM diagnosis and higher DM mortality [34]. The results of a longitudinal queue study have revealed the level of TG as an independent predictor of T2DM in middle-aged and elderly people [35]. Building on the linkage between TG and DM, the TyG index further combines FBG to provide a comprehensive tool. A retrospective cohort study has manifested that a high TyG index is independently linked with an elevated DM risk in Chinese adults [36]. The TyG index had a favorable link with the gestational DM in pregnant women [37].
Central obesity has connections with elevated risks of DM and CVD [38, 39]. Compared to BMI, WC has a stronger correlation with DM [40]. The results of the Mendelian randomization study have further indicated a potential causal linkage between the waist-to-hip ratio (WHR), a biomarker of central obesity, and high BS, which may be related to insulin resistance [41]. The CI is an effective indicator reflecting central obesity, based on the assumption that the shape of fat accumulation around the abdomen resembles a double cone. CI has been confirmed to be linked with the risk of DM in Brazilian women [42]. WWI, as a new type of obesity index, can better distinguish the mass components of fat and muscle. There is a positive link between WWI and fat mass, as well as an adverse link between WWI and muscle mass [43]. This research further combined central obesity-related indices with the TyG index, and revealed that TyG-CI and TyG-WWI performed more accurately in predicting the risk of DM than the TyG index alone and other derived indicators. The composite indicator not only enhanced the predictive performance of DM but also functioned as a more valuable tool for early screening and personalized intervention of DM.
WC grows gradually with age [30], and correlates more strongly with DM risk in men than in women [40], which may be explained by the characteristics of gender differences in fat distribution and sensitivity to insulin. Men tend to have fat accumulated in the abdominal area, while women’s fat is more concentrated in the subcutaneous region [44]. Moreover, women have a greater proportion of type I fibers and capillaries in skeletal muscles, which is beneficial for enhancing the action of insulin [45]. Estrogen also exerts a protective effect against insulin resistance by activating the ERα pathway in insulin-sensitive tissues [46]. However, the subgroup analyses demonstrated that both TyG-CI and TyG-WWI could effectively predict the risk of DM in males, females, people ≥60 years old, and people <60 years old, indicating that these indicators had high robustness and applicability in a wide range of populations.
The specific mechanism by which TyG-derived indicators are implicated in DM risk is unilluminated, and the underlying mechanisms may involve insulin resistance, fat distribution, etc. Insulin resistance is a pivotal factor in T2DM. The decline in insulin sensitivity leads to restricted glucose utilization, ultimately resulting in high BS and DM [4]. Persistent visceral fat accumulation leads to increased adipocyte hypertrophy and macrophage infiltration. Adipocytes and macrophages secrete multiple adipokines, triggering chronic inflammatory responses that disrupt the normal function of insulin. Additionally, in the state of obesity, cells increase their uptake of free fatty acids. However, the insufficient β-oxidation capacity induces the accumulation of intermediate lipid metabolites that can interfere with the insulin signaling pathway, leading to insulin resistance [47]. Insulin resistance is the pathological basis of T2DM. The strong correlation between insulin resistance and TyG-derived indicators was verified in the sensitivity analysis of this study: after excluding DM patients under 30 years of age (a high incidence group of T1DM), the OR value of TyG-CI increased from 2.565 to 2.682. Combined with the epidemiological characteristics that T2DM accounts for 90–95% of adult DM in the United States [48, 49], it was confirmed that TyG-derived indicators mainly reflected the dual pathological process of “insulin resistance + visceral fat” of T2DM. This discovery highlights the biological rationality of TyG-CI/WWI as a warning tool for T2DM.
This study systematically verified the predictive value of the dual target combination of “insulin resistance (TyG-index) + central obesity” on the risk of DM based on a large sample of people for the first time. All TyG-derived indicators were significantly positively correlated with the risk of DM. Among them, TyG-CI and TyG-WWI had the best prediction efficiency, significantly better than the traditional TyG index and other combined indices. This discovery breaks through the limitations of the traditional TyG-index relying on blood glucose parameters, and reveals that integrating visceral fat distribution can improve the accuracy of risk stratification. Its advantage stems from the collaborative impact of the two on the core pathology of DM, combining visceral fat accumulation with insulin resistance. More specifically, TyG-CI is based on the double cone model of trunk fat distribution [50, 51], and the weight of abdominal fat is magnified through geometric calculation [WC/[0.109 × (Weight/Height) 1/2]]. TyG-WWI utilizes a standardized weight design [WC/Weight1/2] to eliminate muscle mass interference [16, 52] and accurately quantify visceral fat deposition. This combination overcomes the limitations of a single index and provides a more comprehensive metabolic anatomical assessment of the risk of DM. More importantly, through the sensitivity analysis of the FPG <126 mg/dL subgroup, we further confirmed the early warning value of TyG-CI and TyG-WWI. Even if the blood glucose does not reach the diagnostic threshold, TyG-CI and TyG-WWI are still significantly associated with the risk of diabetes, and the prediction efficiency is outstanding, indicating that they can catch the early metabolic dysfunction driven by insulin resistance + central obesity. This is beneficial in addressing the early insulin resistance of traditional screening and insufficient sensitivity to central obesity [53, 54]. TyG-CI/WWI only requires routine physical examination indicators (waist circumference, weight, blood lipids, blood glucose) and does not require complex testing. It is particularly suitable for the grassroot screening with limited resources and can early identify high-risk individuals with normal blood glucose but abnormal metabolism, serving as a unique tool for targeted prevention.
Although this research revealed a remarkable linkage between TyG-derived indices and the risk of DM, there are still certain shortcomings. Firstly, the project, as a cross-sectional study, fails to set up a causal relationship to figure out whether the elevation of TyG-derived indices directly leads to an elevated risk of DM. Meanwhile, we still need to be vigilant about reverse causality. The increase of TyG-derived indicators in DM patients may be partly due to the metabolic disorder caused by DM, such as the accumulation of TG caused by hyperglycemia or the influence of hypoglycemic drugs. Therefore, although this study revealed strong clinical associations, future prospective studies are still needed to clarify the direction of causal relationships. Secondly, although this study excluded DM patients diagnosed under the age of 30 (high-risk population for T1DM) for sensitivity analysis, the conclusion is mainly applicable for predicting the risk of T2DM. However, due to limitations in NHANES data that did not differentiate between T1DM and T2DM, this limitation may still affect the interpretation of the results and lead to potential bias. In the future, further verification is needed in a classified queue. Moreover, the samples in the NHANES database are mainly from the American population, necessitating further validation for the generalizability of the results in other regions or different racial groups. Lastly, although we adjusted a wide range of covariates, some factors that may affect TyG-derived indicators and DM risk were not included in the model, such as detailed dietary intake data, accurate quantification of PA intensity, and the use of specific drugs. These factors that were not fully measured may have residual confounding effects on the observed association. Future research needs to optimize model adjustments through more comprehensive exposure assessments.
The indices with the TyG index combined with central obesity-related indicators, especially TyG-CI and TyG-WWI, are effective predictive tools for DM risk. Through the simultaneous assessment of insulin resistance and visceral fat distribution, these indices provide a simple, dual-target warning strategy for identifying high-risk groups of DM (especially central obese people with normal BS), which is expected to optimize the early prevention and control practice of DM at the grassroot level (Graphical Abstract).
Not Applicable.
Conflict of interestThe authors declare that they have no conflicts of interest with the contents of this article.
Availability of data and materialsThe data and materials in the current study are available from the corresponding author on reasonable request.
Ethics approval and consent to participateNHANES is jointly managed by the National Center for Health Statistics (NCHS) and the Centers for Disease Control and Prevention (CDC) in the United States. All research proposals were approved by the NCHS Institutional Review Board and complied with its regulatory framework. All participants signed a written informed consent form before being included in the survey. This study, as a secondary analysis of publicly available anonymous data, strictly adhered to the ethical principles of medical research outlined in the Helsinki Declaration. During the research process, no identifiable personal information was used, which complied with privacy protection regulations.
Author contributionsConception and design: Xiaohua Yang
Collection and assembly of data: Zhuojing Cheng, Ting Sun
Data analysis and interpretation: Xiaohua Yang, Zhuojing Cheng, Ting Sun
Manuscript drafting: Xiaohua Yang, Zhuojing Cheng
Reviewing and editing: Ting Sun
Final approval of manuscript: All authors
