Using the triglyceride-glucose index and derived indexes to forecast progression from pre-diabetes to diabetes: A 3-year follow-up study

Abstract

BACKGROUND

Pre-diabetes is a transitional metabolic stage between health and diabetes, serving as a critical warning signal for disease progression. Early intervention targeting risk factors in prediabetic individuals prevents the progression to type 2 diabetes.

AIM

To investigate the predictive value of the triglyceride-glucose (TyG) index and its derived indicators for new-onset diabetes in patients with pre-diabetes.

METHODS

A prospective community-based cohort study was carried out based on subjects aged over 40 years with pre-diabetes in Dalian, Liaoning Province, China. A total of 1352 subjects with complete follow-up data attended the follow-up survey. Multivariable Cox regression models were performed to assess the association of the TyG index and its derived indicators with risk of diabetes in patients with pre-diabetes. The diagnostic values of the TyG index and derived indicators in predicting new-onset diabetes were analyzed, and suitable cutting points were determined using the receiver operating characteristic (ROC) curve.

RESULTS

During a 3-year follow-up period, 153 cases with incident diabetes were identified, with a cumulative incidence of diabetes of 11.3%; 12.6% (43/341) in males and 10.9% (110/1011) in females (χ² = 0.760, P = 0.375). After adjusting for confounding factors including age, gender, body mass index (BMI) and insulin levels, the risk of diabetes with higher TyG and derived indexes [TyG-BMI and TyG-waist circumference index (TyG-WC)] increased significantly. The TyG index [hazard ratio (HR) = 1.389, 95% confidence interval (CI): 1.011-1.908, P = 0.043], TyG-BMI (HR = 1.010, 95%CI: 1.005-1.015, P = 0.000) and TyG-WC (HR = 1.003, 95%CI: 1.001-1.005, P = 0.001) were all strongly positively correlated with the risk of future diabetes. The ROC curve analysis showed that the area under the curve (AUCs) of the TyG, TyG-BMI and TyG-WC for predicting new diabetes were 0.578 (95%CI: 0.533-0.624), 0.622 (95%CI: 0.574-0.670) and 0.609 (95%CI: 0.562-0.657), respectively. The difference in AUC between TyG-BMI and TyG was significant (P = 0.047), while the differences between TyG-BMI and TyG-WC (P = 0.464) and between TyG-WC and TyG (P = 0.175) were not. The TyG-BMI had a larger AUC than the TyG and TyG-WC, and its difference from TyG was significant. The best cut-off points for predicting new diabetes were TyG > 8.6, TyG-BMI > 247 and TyG-WC > 860. Although the AUC values were modest, these indices may serve as preliminary screening tools in resource-limited settings.

CONCLUSION

The TyG index and its derived indicators were risk factors for the pre-diabetes to diabetes outcome, and may be regarded as predictors of the outcome. The risk of conversion of pre-diabetes to diabetes increased with increases in the TyG index and its derived indicators. The TyG-BMI was better than TyG and TyG-WC in predicting the 3-year outcome for diabetes. Although these indices could aid in the initial risk stratification in primary care, their modest accuracy warrants cautious interpretation.

Key Words: Triglyceride-glucose index; Triglyceride-glucose-body mass index; Triglyceride-glucose-waist circumference index; Pre-diabetes; Diabetes; Prospective cohort study

Core Tip: Pre-diabetes is an abnormal state of glucose metabolism between normal glucose tolerance and diabetes. If early intervention is carried out for risk factors in a pre-diabetic population, the occurrence and development of diabetes can be prevented. The triglyceride-glucose (TyG) index is a new index calculated according to triglycerides and fasting plasma glucose test. The latest research suggests that the TyG index and its derivative indexes, such as the product of TyG index and body mass index and the product of TyG index and waist circumference are novel and surrogate markers for insulin resistance. We prospectively investigated the relationship between TyG index and derived indexes and incident type 2 diabetes using a large sample, community-based cohort, aim at to find a simple indicator predicting the progression of a pre-diabetic population to diabetes, to allow relevant early-stage intervention in a pre-diabetic population and so reduce diabetes incidence.

INTRODUCTION

Pre-diabetes, also known as impaired glucose regulation, is an abnormal state of glucose metabolism between normal glucose tolerance and diabetes, which is considered an inevitable stage of diabetes and an early warning signal of diabetes[1,2]. If early intervention is carried out for risk factors in a pre-diabetic population, the occurrence and development of diabetes can be prevented. At least one oral glucose tolerance test (OGTT) reexamination should be conducted annually according to intervention for adults with prediabetes: A Chinese expert consensus (2023 edition). The OGTT is considered the “gold standard” in the diagnosis of pre-diabetes and diabetes mellitus (DM) but is unsuitable for large-scale screening due to poor reproducibility and being time-consuming. Although glycated hemoglobin (HbA1c) testing and the continuous glucose monitoring system are established modalities for glycemic monitoring, their clinical implementation in resource-limited primary care settings is constrained by prohibitive costs and technical complexity, thereby limiting their utility in screening programs. It is important to find an optimal measure that is simple and accurate for glucose re-screening in individuals at high risk of developing DM.

Insulin resistance (IR) is a significant factor in the development from pre-DM to type 2 DM (T2DM)[1,3,4]. The hyperinsulinaemic-euglycaemic glucose clamp has been regarded as the reference standard for assessing the degree of IR; however, its clinical application is widely limited due to its invasive and cumbersome procedure. Emerging evidence suggests that disruptions in lipid homeostasis may also be intricately linked to the etiology and pathogenesis of T2DM. Triglyceridemia is an independent risk factor for diabetes, and also reflects the impact of lipotoxicity theory on insulin sensitivity and insulin secretion. The triglyceride-glucose (TyG) index is a new index calculated according to triglycerides (TGs) and the fasting plasma glucose test (FPG)[5,6], which are easily and efficiently implemented in primary health care centers, and represents a clinically accessible, economically viable for assessing IR[7,8]. Emerging evidence suggests that the TyG index and its derivative indexes, such as the product of TyG index and body mass index (BMI) (TyG-BMI) and the product of TyG index and waist circumference (WC) (TyG-WC), may offer superior predictive performance for IR and diabetes risk compared to the TyG index alone[9]. While some studies have explored the TyG index in mixed or high-risk populations, there is a scarcity of prospective data specifically focusing on the comparative predictive value of the TyG index, TyG-BMI and TyG-WC for the progression from pre-diabetes to diabetes in a well-defined, community-based pre-diabetic cohort over a multi-year follow-up. No prospective assessment of TyG-derived indices has been carried out in a cohort of Chinese pre-diabetic individuals this study aims to fill this gap and is the first to prospectively assess the predictive value of the TyG index and its derived indices (TyG-BMI and TyG-WC) for diabetes progression in a community-based Chinese pre-diabetic cohort, which was identified by OGTT.

MATERIALS AND METHODS

Study population

We used all data derived from the China Cardiometabolic Disease and Cancer Cohort study, which is a large prospective cohort study initiated by Ruijin Hospital Affiliated to the Medical School of Shanghai Jiao Tong University. A total of 10207 participants were contacted from two communities in the city of Dalian during screening. Subjects without diabetes history underwent a 75 g OGTT in 2011 and this cohort study was conducted from the baseline survey in 2011 up to 2014. Participants with incomplete follow-up data or missing key variables were excluded from the final analysis. A total of 1352 participants consisting of 341 male and 1011 female residents were eligible for the final analysis, and a flow chart of population exclusion is shown in Figure 1.

Figure 1 Flow chart of population exclusion for the final analysis. T2DM: Type 2 diabetes mellitus; NGT: Normal glucose tolerance.

All subjects participated in the study voluntarily and provided informed consent prior to enrollment. The study was approved by the Institutional Research Ethics Committee of Ruijin Hospital Affiliated to the Medical School of Shanghai Jiao Tong University.

Data collection

The following data regarding clinical history, anthropometric measurements and blood analyses were derived at baseline and the follow-up visit. Blood samples were obtained after at least 8 hours of overnight fasting. Levels of FPG, TG, total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C) and low-density lipoprotein cholesterol (LDL-C) were measured. Subjects without diabetes history underwent a OGTT and tested 2 hours post load plasma glucose (2hPG). Serum for lipid profiles was stored below -20 °C, and whole blood for HbA1c analysis was stored below 4 °C, with samples delivered to the Chemical Laboratory of Shanghai Ruijin Hospital by a professional cold-chain express company within 3 weeks. Fasting serum insulin (FINS) was detected by chemiluminescence method using an automatic detector.

BMI = weight (kg)/height² (m²); waist-hip ratio (WHR) = WC (cm)/hip circumference (HC) (cm); homeostasis model assessment (HOMA)-IR = FPG (mmol/L) × FINS (mU/L)/22.5; HOMA-β = 20 × FINS (mU/L)/(FPG - 3.5) (mmol/L); TyG index = ln[TG (mg/dL) × FPG (mg/dL)/2]; TyG-BMI = TyG index × BMI; TyG-WC = TyG index × WC (cm).

Definition

According to 2011 World Health Organization diagnostic criteria, diabetes was defined as FPG ≥ 7.0 mmol/L and/or 2hPG ≥ 11.1 mmol/L, or HbA1c ≥ 6.5%; normal glucose tolerance was defined as FPG < 6.1 mmol/ L and 2hPG < 7.8 mmol/L; impaired fasting glucose (IFG) was defined 6.1 mmol/L ≤ FPG < 7.0 mmol/L and 2hPG < 7.8 mmol/L; impaired glucose tolerance (IGT) was defined FPG < 7.0 mmol/L and 7.8 mmol/L ≤ 2hPG < 11.1 mmol/L.

Inclusive criteria

Subjects (aged ≥ 40 years) who were diagnosed with pre-diabetes (IFG and IGT) at baseline and completed a three-year follow up were selected as the research objects.

Excusive criteria

Subjects with diagnosed normal glucose tolerance, diabetes, cancer, chronic liver disease, chronic kidney disease and glucocorticoid treatment at the baseline were exclusive; if the subjects didn’t complete the follow up survey, subjects were also exclusive.

Statistical analysis

Continuous variables are expressed as mean ± SD and qualitative variables as numbers and percentages. Comparisons between the diabetic and non-diabetic groups were performed using two independent samples t-tests and, for categorical variables, the χ² test was used. Multivariable-adjusted Cox proportional-hazards regression analyses were performed to estimate the hazard ratios (HRs) for the TyG index and its derived indicators in relation to 3-year diabetes incidence. The area under the curve (AUC) of the receiver operating characteristics (ROC) curve was calculated to compare the predictive power of the TyG index and its derivative indexes. Pairwise comparisons of the ROC curves were performed using DeLong’s test. The optimal cut-off points for the TyG index, TyG-BMI and TyG-WC were determined using the Youden index from the ROC curve analysis.

RESULTS

General clinical characteristics of the subjects

Among the 1352 subjects, the average age was 58.44 ± 8.21 years. The incidence of diabetes was 11.3% (153/1352); 12.6% (43/341) in males and 10.9% (110/1011) in females (χ² = 0.760, P = 0.375). According to FPG, OGTT or HbA1c results and history of diabetes during follow-up, the subjects were divided into diabetes and non-diabetes groups. All the subjects were analyzed for age, weight, height, BMI, WC, HC, WHR, systolic blood pressure, diastolic blood pressure, FPG, 2hPG, HbA1c, TC, TG, HDL-C, LDL-C, serum uric acid, FINS, HOMA-IR, HOMA-β, TyG index, TyG-BMI, TyG-WC and other clinical characteristics (Table 1).

Table 1 Characteristics of participants, mean ± SD.

Variables	Total	Non-diabetes	Diabetes	P value	Cohen’s d value
Type proportion (IFG/IGT)	282/1070	252/947	30/123	0.752
Gender (male/female)	341/1011	289/901	43/110	0.375
Age (year)	58.44 ± 8.21	58.43 ± 8.35	58.58 ± 7.03	0.826	0.01943
Weight (kg)	67.52 ± 10.88	67.14 ± 10.75	70.52 ± 11.43	0.000	0.30464
Height (m)	160.73 ± 7.65	160.70 ± 7.63	160.99 ± 7.92	0.661	0.03729
BMI (kg/m2)	26.08 ± 3.38	25.94 ± 3.33	27.18 ± 3.60	0.000	0.35759
WC (cm)	90.73 ± 9.34	90.40 ± 9.25	93.29 ± 9.64	0.000	0.30592
HC (cm)	101.5 ± 7.23	101.26 ± 7.13	103.37 ± 7.73	0.001	0.28375
WHR	0.89 ± 0.06	0.89 ± 0.06	0.90 ± 0.06	0.061	0.16667
SBP (mmHg)	143.29 ± 20.46	143.04 ± 20.38	145.25 ± 21.00	0.069	0.1068
DBP (mmHg)	81.38 ± 11.21	81.21 ± 11.25	82.68 ± 10.87	0.126	0.13289
FPG (mmol/L)	5.91 ± 0.50	5.90 ± 0.50	6.07 ± 0.49	0.001	0.34342
2hPG (mmol/L)	8.40 ± 1.37	8.38 ± 1.36	8.62 ± 1.40	0.044	0.17389
HbA1c (%)	5.86 ± 0.36	5.83 ± 0.34	6.06 ± 0.47	0.000	0.56073
TC (mmol/L)	5.59 ± 1.07	5.58 ± 1.07	5.64 ± 1.05	0.517	0.0566
TG (mmol/L)	1.64 ± 0.97	1.63 ± 0.98	1.75 ± 0.87	0.139	0.1295
HDL-C (mmol/L)	1.39 ± 0.31	1.39 ± 0.32	1.33 ± 0.27	0.015	0.20266
LDL-C (mmol/L)	3.36 ± 0.89	3.35 ± 0.90	3.42 ± 0.85	0.347	0.07997
SUA (μmol/L)	321.8 ± 72.55	319.11 ± 71.28	342.82 ± 78.94	0.000	0.31526
FINS (mU/L)	9.54 ± 4.71	9.41 ± 4.58	10.55 ± 5.54	0.016	0.22429
HOMA-IR	2.52 ± 1.27	2.48 ± 1.26	2.84 ± 1.45	0.004	0.26503
HOMA-β	81.82 ± 43.3	81.26 ± 41.47	86.2 ± 55.57	0.289	0.10076
TyG index (2011)	8.82 ± 0.51	8.80 ± 0.51	8.93 ± 0.47	0.003	0.26509
TyG-BMI (2011)	230.3 ± 34.98	228.64 ± 34.25	243.27 ± 37.91	0.000	0.40497
TyG-WC (2011)	800.59 ± 100.3	796.31 ± 99.25	834.08 ± 102.50	0.000	0.37438
TyG index (2014)	8.83 ± 0.51	8.81 ± 0.50	9.00 ± 0.55	0.000	0.3615
TyG-BMI (2014)	233.40 ± 96.61	230.89 ± 96.30	251.49 ± 97.24	0.021	0.21287
TyG-WC (2014)	788.41 ± 103.09	782.12 ± 98.88	833.71 ± 120.50	0.000	0.46806

IFG: Impaired fasting glucose; IGT: Impaired glucose tolerance; TyG: Triglyceride-glucose; BMI: Body mass index; WC: Waist circumference; HC: Hip circumference; WHR: Waist-hip ratio; SBP: Systolic blood pressure; DBP: Diastolic blood pressure; FPG: Fasting plasma glucose; 2hPG: 2 hours post load plasma glucose; HbA1c: Glycated hemoglobin; TC: Total cholesterol; HDL-C: High-density lipoprotein cholesterol; LDL-C: Low-density lipoprotein cholesterol; SUA: Serum uric acid; FINS: Fasting serum insulin; HOMA-IR: Homeostasis model assessment-insulin resistance; HOMA-β: Homeostasis model assessment-beta.

Diabetes prevalence according to TyG index, TyG-BMI and TyG-WC quartiles

The grouping by quartile according to the baseline TyG index was group T1 (TyG < 8.46), group T2 (8.46 ≤ TyG < 8.80), group T3 (8.80 ≤ TyG < 9.12) and group T4 (TyG ≥ 9.12). For these quartiles, the corresponding incidences of type 2 diabetes were 7.3%, 11%, 13.4% and 13.5%, showing an increase with increasing quartiles (χ² = 8.448, P = 0.038; Table 2).

Table 2 Prevalence of diabetes in quartile of triglyceride-glucose index, n (%).

TyG index quartiles	Non-diabetes	Diabetes	Total	χ2 value	P value
T1 group	305 (92.7)	24 (7.3)	329 (100)	8.448	0.038
T2 group	309 (89.0)	38 (11)	347 (100)
T3 group	290 (86.6)	45 (13.4)	335 (100)
T4 group	295 (86.5)	46 (13.5)	341 (100)
Total	1199 (88.7)	153 (11.3)	1352 (100)
χ2 value for trend = 7.328, P value for trend = 0.007

TyG: Triglyceride-glucose.

Grouping by quartile according to the baseline TyG-BMI was group B1 (TyG-BMI < 205), group B2 (205 ≤ TyG-BMI < 227), group B3 (227 ≤ TyG-BMI < 252) and group B4 (TyG-BMI ≥ 252). After 3 years, the corresponding incidences of diabetes were 7.8%, 6.8%, 10.6% and 19.9% (χ² = 36.008, P = 0.000; Table 3).

Table 3 Prevalence of diabetes in quartile of triglyceride-glucose-body mass index, n (%).

TyG-BMI quartiles	Non-diabetes	Diabetes	Total	χ2 value	P value
B1 group	307 (92.2)	26 (7.8)	333 (100)	36.008	0.000
B2 group	314 (93.2)	23 (6.8)	337 (100)
B3 group	303 (89.4)	36 (10.6)	339 (100)
B4 group	274 (80.1)	68 (19.9)	342 (100)
Total	1198 (88.7)	153 (11.3)	1351 (100)
χ2 value for trend = 27.150, P value for trend = 0.000

TyG: Triglyceride-glucose; BMI: Body mass index.

The grouping according to the baseline TyG-WC quartiles was group W1 (TyG-WC < 728), group W2 (728 ≤ TyG-WC < 796), group W3 (796 ≤ TyG-WC < 868) and group W4 (TyG-WC ≥ 868). After 3 years, the corresponding incidences of diabetes were 7.4%, 8.7%, 10.5% and 18.5%, showing an increase with increasing quartiles (χ² = 25.197, P = 0.000; Table 4).

Table 4 Prevalence of diabetes in quartile of triglyceride-glucose-waist circumference, n (%).

TyG-WC quartiles	Non-diabetes	Diabetes	Total	χ2 value	P value
W1 group	312 (92.6)	25 (7.4)	337 (100)	25.197	0.000
W2 group	304 (91.3)	29 (8.7)	333 (100)
W3 group	306 (89.5)	36 (10.5)	342 (100)
W4 group	277 (81.5)	63 (18.5)	340 (100)
Total	1199 (88.7)	153 (11.3)	1352 (100)
χ2 value for trend = 20.821, P value for trend = 0.000

TyG: Triglyceride-glucose; WC: Waist circumference.

Relationships of TyG index, TyG-BMI and TyG-WC with incident diabetes

Cox regression models were established to evaluate the association of TyG index, TyG-BMI and TyG-WC with diabetes. After adjusting for age, gender, BMI and FINS, TyG index was strongly positively correlated with the risk of diabetes [HR = 1.389, 95% confidence interval (CI): 1.011-1.908, P = 0.043] in pre-diabetic patients. After adjusting for age, gender and FINS, TyG-BMI was strongly positively correlated with the risk of diabetes (HR = 1.010, 95%CI: 1.005-1.015, P = 0.000). After adjusting for age, gender and FINS, TyG-WC was strongly positively correlated with the risk of diabetes (HR = 1.003, 95%CI: 1.001-1.005, P = 0.001) (Table 5).

Table 5 Cox regression analyses for the risk of diabetes.

Variables	HR	95%CI	P value
TyG index	1.389	1.011-1.908	0.043
TyG-BMI	1.010	1.005-1.015	0.000
TyG-WC	1.003	1.001-1.005	0.001

TyG: Triglyceride-glucose; BMI: Body mass index; WC: Waist circumference; HR: Hazard ratio; CI: Confidence interval.

Predictive value of TyG index, TyG-BMI and TyG-WC for new-onset diabetes

The ROC curves of T2DM for the TyG index, TyG-BMI and TyG-WC are shown in Figure 2. The AUC of the ROC curve for the TyG index was 0.578 (95%CI: 0.533-0.624), for TyG-BMI was 0.622 (95%CI: 0.574-0.670) and for TyG-WC was 0.609 (95%CI: 0.562-0.657). The difference in AUC between TyG-BMI and TyG was significant (P = 0.047), while the differences between TyG-BMI and TyG-WC (P = 0.464) and between TyG-WC and TyG (P = 0.175) were not. Based on the ROC curves, the optimal predictive cut-off for the TyG index was 8.6, with a sensitivity of 79.7% and a specificity of 36.6%. Specifically, the cut-off value of TyG-BMI was 247 (sensitivity = 49.7%, specificity = 72.8%) and for TyG-WC was 860 (sensitivity = 45.8%, specificity = 73.6%).

Figure 2 The receiver operating characteristic curves for the incidence of diabetes. TyG: Triglyceride-glucose; BMI: Body mass index; WC: Waist circumference.

DISCUSSION

Pre-diabetes is a state between diabetes and normal blood glucose. The prevalence rate of pre-diabetes in China is experiencing rapid growth. As early as 2011, a flow adjustment result showed that the prevalence rate of pre-diabetes in people over 40 years old in Dalian was as high as 25.1%. The latest flow adjustment data in 2017 showed that the prevalence rate of pre-diabetes in China may currently be 35.2%[10]. Not only does a significantly higher proportion of the pre-diabetic population progress to diabetes compared to the general population, but there is also a significant increase in the risk of cardiac and cerebrovascular diseases. Therefore, diabetes health education and intervention for pre-diabetes patients is important to prevent and treat diabetes and its complications[11-13]. Individuals with pre-diabetes should undergo systematic glycemic surveillance, including periodic measurement of FPG, HbA1c or OGTT, which is critical for early detection of dysglycemia progression, diagnostic confirmation of diabetes onset and timely implementation of therapeutic interventions to prevent microvascular and macrovascular complications. The mean HbA1c in the incident diabetes group was below the diagnostic threshold of 6.5%, underscoring that a significant proportion of new cases were identified through OGTT before their HbA1c had elevated to the diagnostic level, highlighting the complementary role of OGTT in early detection.

While OGTT remains the reference standard for identifying pre-diabetes and DM, its clinical utility in population-based screening is constrained by limited reproducibility and labor-intensive procedures. This underscores the need to develop simplified yet precise glycemic assessment protocols for high-risk populations undergoing diabetes surveillance.

IR and β-cell dysfunction constitute the dual core pathogenesis of T2DM from pre-diabetes[8], and this metabolic derangement is exacerbated by lipid metabolism dysregulation. The pathological accumulation of TG within pancreatic β-cells has been shown to induce apoptotic cascades while simultaneously compromising their regenerative potential. The impaired oxidation of free fatty acids in non-adipose tissue causes cytosolic TG accumulation in non-adipose tissues, leading to IR and β-cell dysfunction[14]. The hyperinsulinemic-euglycemic clamp is universally accepted as the reference technique for assessing insulin-mediated glucose uptake and defining IR phenotypes, but is invasive and cumbersome, and only suitable for evaluation in a small number of studies and investigations.

The steady-state model evaluation HOMA-IR, an index for evaluating IR[15], is widely used in Asia. The results of this study showed that the baseline HOMA-IR value was higher in patients with T2DM (2.84 ± 1.45 vs 2.48 ± 1.26). According to the European Group for the Study of IR recommendation, HOMA-IR index ≥ 2.8 is tentatively defined as IR[16]. This suggests that IR is present at baseline, i.e. in pre-diabetic status, in subjects progressing to diabetes after 3 years. However, fasting insulin level should be measured when calculating HOMA-IR, which is difficult to achieve in most primary medical institutions[17].

There are many reasons for IR, with lipid metabolism disorder, oxidative stress, mitochondrial dysfunction and inflammatory reaction all closely related to IR[18]. Studies have shown that abnormal blood glucose and blood lipid (especially increased TG level), high BMI and WC are all risk factors for IR-related metabolic diseases. If multiple risk factors are present in the same individual, the incidence of IR will be greatly increased[19].

The TyG index combined with blood glucose and TG is likely related to IR-associated metabolic diseases[20]. The TyG index can be used as an index to diagnose IR in normal people and those with metabolic disorder[21,22]. This index only requires FPG and TG levels, which is more suitable for clinical practice. Simental-Mendía et al[23] found for the first time that the TyG index of healthy people is correlated with IR evaluated by HOMA-IR. Similar results were found in obese children and adolescents in South America[24]. Some scholars have compared the TyG index with the normal blood glucose hyperinsulinemia clamp test, and found that the TyG index had higher sensitivity and specificity in distinguishing insulin sensitivity reduction[25]. Vasques et al[26] in 2011 showed that the TyG index was better than HOMA-IR in evaluating IR of a Brazilian population. Many studies have confirmed that the TyG index is closely related to IR, and it is a simple and reliable index for diagnosing IR[24,27]. A follow-up study of the middle-aged general population in Korea shows that the increase of the TyG index is related to increased risk of diabetes[28,29]. According to the correlation between the TyG index and IR, its mechanism may be related to two main components of TyG index (FPG and TG), and these two elements play key roles in the occurrence and development of IR[30]. Excessive serum TG may lead to accumulation of fatty acids in non-adipose tissues such as liver, muscle and heart, leading to ectopic lipid deposition. “Lipotoxicity” is considered another mechanism of IR. When the body is in IR state, the decomposition of peripheral adipose tissue is enhanced, and the excessive free fatty acids produced enter the liver through the portal vein system, and accumulate abnormally in the liver, resulting in a corresponding increase in TG synthesis in the liver[31].

The TyG-BMI is the product of TyG and BMI, and was first reported in 2016[32]. The TyG-BMI is better at predicting IR than the TyG index alone[33]. Study showed that TyG-BMI is a simple, effective and clinical substitute marker for evaluating IR[32,34]. Lim et al[33] pointed out, in a cross-sectional study involving 11149 Korean subjects, that the ability of TyG-BMI to predict IR is significantly greater than that of the TyG index. A cohort study from China also shows that TyG-WC is also a novel index to evaluate IR, which can be used to early identify the pre-diabetes and diabetes risk of first-degree relatives of T2DM patients[35]. A study from India also confirmed that the TyG index and TyG-WC could predict incidence of pre-diabetes[36].

In this study, we focused on TyG-BMI and TyG-WC as they are among the most established and commonly reported TyG-derived indices for predicting diabetes and IR. The BMI is a general measure of overall adiposity, while WC more specifically reflects central (visceral) adiposity, which is strongly linked to IR and metabolic dysfunction. We chose these two to investigate whether combining the TyG index with a general (BMI) or a central (WC) adiposity measure would yield different predictive power for diabetes progression in pre-diabetes. While other indices like the TyG-waist-to-height ratio also exist and may have merit, our primary aim was to validate and compare the performance of these two foundational combined indices in our prospective cohort. Future studies could include a broader range of derived indices for a more comprehensive comparison.

The results of this study showed that baseline body weight, BMI, WC, HC, blood glucose indexes (FPG, 2hPG and HbA1c), blood lipid levels (TC, TG and LDL-C) and HOMA-IR were higher in the diabetic group. Moreover, the TyG index, TyG-BMI and TyG-WC were significantly higher in diabetic than non-diabetic subjects. This shows that although all the subjects in this study were in pre-diabetic state at baseline, there were some differences in their weight, WC, blood lipid, blood glucose and other indicators, which also indicate the development direction of blood glucose of these subjects. The incidence of diabetes increased across the quartiles of the TyG index, TyG-BMI and TyG-WC. Further research showed that the increase of the TyG index, TyG-BMI and TyG-WC were independent risk factors for the progression of a pre-diabetic population to diabetes. The focus of this study was whether the progression of blood glucose in a pre-diabetic population could be preliminarily determined by simple biochemical indicators and physical measurement indicators. Then the ROC curve was used to analyze and compare the efficacy of the TyG index, TyG-BMI and TyG-WC in predicting early onset of diabetes in a pre-diabetic population. The AUC of these three indexes to predict the prognosis of diabetes was 0.578-0.622, which has certain diagnostic value, but could not directly diagnose diabetes like blood glucose or HbA1c testing, but it can serve as a preliminary tool for risk-association indication. Its core clinical significance does not lie in accurately determining whether an individual will experience the outcome, but in identifying potentially high-risk populations and assisting in the initial screening for clinical decision-making. In our cohort, the AUC for HOMA-IR was 0.575 and for baseline HbA1c was 0.640. The performance of TyG-BMI (AUC = 0.622) was comparable to that of HbA1c. This indicates that TyG-BMI, a simple calculated index, performs similarly to a standard glycemic marker like HbA1c in this context. The TyG-BMI is a comprehensive statistic that includes TGs, fasting blood glucose, weight and height all easily obtainable indicators at primary service centers, straight forward to compute and not limited by time or cost. This makes TyG-BMI particularly suitable for large populations, and it could be used as a pre-judgment providing clues concerning high-risk groups. When TyG-BMI was 247, its sensitivity and specificity were 49.7% and 72.8%, respectively, suggesting that it could be a simple tool to predict diabetes progression in a pre-diabetic population.

The innovation of this project is the first analysis of the relationship between the TyG index and its derivative indexes of pre-diabetic patients and their 3-year risk of diabetes. Because of the tedious OGTT procedure, most large-scale screening at home and abroad rarely uses OGTT to evaluate blood glucose status, but only uses FPG to screen diabetes prevalence, which cannot evaluate the situation of some specific people, such as those who are pre-diabetic. In this project, OGTT was used to evaluate pre-diabetes, making the target population more representative. In addition, research on the follow-up data of community populations using a large sample size offers great value for promoting the whole-process health management of T2DM.

However, there are still some limitations in this study. The male to female ratio was unbalanced, and the population was mainly urban. Collecting data through community recruitment will inevitably miss some hidden diabetic patients. There was a certain rate of missing visits and, due to this bias, the research results may be overestimated or underestimated. Although we adjusted for key confounders, data on physical activity, dietary patterns and medication use were not available and may represent residual confounding. The use of these indices in screening may lead to false positives, potentially resulting in unnecessary anxiety or overtreatment. Therefore, they should be used as part of a comprehensive risk assessment rather than as standalone criteria. Considering these limitations, it is necessary to further explore the application value of these IR indexes in a strict prospective cohort study and future studies should incorporate advanced statistical measures such as net reclassification improvement or decision-curve analysis to better evaluate their clinical utility.

Taken together, IR indexes of TyG index, TyG-BMI and TyG-WC are all risk factors of the progression of diabetes in pre-diabetic subjects. The TyG-BMI was superior to the TyG index and TyG-WC in predicting the progression of diabetes in adult pre-diabetic subjects in Dalian.

CONCLUSION

The TyG index and its derived indicators were independent risk factors for diabetes progression in adults with pre-diabetes. The TyG-BMI demonstrated a modest advantage over TyG and TyG-WC in prediction, although its discriminative ability was limited. These indices may serve as simple, initial screening tools in primary care. Before clinical implementation, these findings should be validated in larger, multi-center cohorts with diverse populations and external validation.