Association between the cumulative average triglyceride glucose–body mass index and glycaemic state transitions in adults with pre-diabetes: evidence from the CHARLS cohort Authors: Ningling Dai, Haiya Wang Published: 2025 ABSTRACT Abstract AbstractObjectivesThe role of the triglyceride–glucose body mass index (TyG-BMI) in glycaemic state transitions among individuals with pre-diabetes remains unclear. This study explores the association between cumulative average TyG-BMI and the likelihood of regression to normoglycaemia or progression to diabetes.MethodsParticipants with pre-diabetes from the China Health and Retirement Longitudinal Study were included. Pre-diabetes was defined using American Diabetes Association criteria. Cumulative average TyG-BMI was calculated using data from 2011 and 2015. Ordinal logistic regression and restricted cubic spline (RCS) analyses were used to evaluate the association between cumulative average TyG-BMI and glycaemic transitions.ResultsA total of 556 participants (mean age 58.5 years; 45.9% male) were included; 350 (62.9%) remained pre-diabetic, 111 (19.96%) reverted to normoglycaemia and 95 (17.09%) progressed to diabetes. After adjustment for confounding factors, higher cumulative average TyG-BMI was significantly associated with lower odds of regression to normoglycaemia and higher odds of progression to diabetes (OR=0.50; 95% CI 0.33 to 0.62; p<0.001). The RCS analysis demonstrated a negative linear relationship between cumulative average TyG-BMI and glycaemic improvement (p for overall <0.001, p for nonlinear=0.149).ConclusionsHigher cumulative average TyG-BMI levels identify individuals with pre-diabetes who are less likely to revert to normoglycaemia and more likely to progress to diabetes, highlighting its value as a practical marker for early risk stratification and targeted clinical management of pre-diabetes.Keywords: Diabetes Mellitus, Type 2; Lipid disorders; Follow-Up Studies STRENGTHS AND LIMITATIONS OF THIS STUDY Participants were drawn from a nationally representative cohort (China Health and Retirement Longitudinal Study (CHARLS)). Participants were drawn from a nationally representative cohort (China Health and Retirement Longitudinal Study (CHARLS)). Cumulative average triglyceride–glucose body mass index (TyG-BMI) was calculated using two time points to reflect long-term metabolic exposure. Cumulative average triglyceride–glucose body mass index (TyG-BMI) was calculated using two time points to reflect long-term metabolic exposure. Ordinal logistic regression and restricted cubic spline models were applied to assess associations and test for non-linearity. Ordinal logistic regression and restricted cubic spline models were applied to assess associations and test for non-linearity. The observational design and age-restricted Chinese cohort limit causal inference and generalisability. The observational design and age-restricted Chinese cohort limit causal inference and generalisability. Introduction Diabetes is one of the most prevalent chronic diseases globally and is associated with severe complications, substantial healthcare costs and reduced life expectancy.1 In 2021, an estimated 529 million people were living with diabetes, the vast majority of whom had type 2 diabetes mellitus (T2DM).1 Pre-diabetes, defined by impaired fasting glucose or impaired glucose tolerance,2 represents an intermediate metabolic state that substantially increases the risk of progressing to T2DM.3 It is also associated with a higher likelihood of microvascular and cardiovascular complications.4,6 Regression to normoglycaemia, on the other hand, markedly reduces these risks.6 7 Therefore, identifying modifiable factors that influence glycaemic regression or progression is critical for targeted prevention strategies. Insulin resistance (IR) plays a central role in the development of T2DM and cardiometabolic diseases.8,13 Although the hyperinsulinaemic-euglycaemic clamp (HEC) remains the gold standard for assessing IR,14 its cost and complexity limit its clinical applicability. As a practical alternative, the triglyceride–glucose body mass index (TyG-BMI) has emerged as a convenient and reliable surrogate marker of IR. By integrating triglyceride (TG), fasting blood glucose (FBG) and body mass index (BMI), TyG-BMI provides a composite indicator of metabolic dysfunction and has shown strong predictive capability for IR compared with other surrogate markers.15 Numerous studies have demonstrated significant associations between TyG-BMI and multiple metabolic conditions, including diabetes, hypertension, cardiovascular diseases and non-alcoholic fatty liver disease (NAFLD).16,22 However, limited research has examined its role in glycaemic transitions among individuals with pre-diabetes. Existing studies have typically evaluated either regression or progression separately,23 24 and most relied on single-time point TyG-BMI measurements. To date, no study has comprehensively assessed the association between cumulative average TyG-BMI—which better reflects long-term metabolic exposure—and both glycaemic improvement and deterioration within the same cohort with pre-diabetes. Therefore, this study aims to investigate the association between cumulative average TyG-BMI and glycaemic state transitions, providing deeper insight into the long-term metabolic determinants of pre-diabetes trajectories. Methods Study population This prospective cohort study used data from the China Health and Retirement Longitudinal Study (CHARLS), a nationally representative cohort of Chinese adults aged ≥45 years from urban and rural areas. The baseline survey (2011–2012) included 10 257 households across 150 counties or districts. Participants were followed up biennially in 2013, 2015 and 2018. Detailed descriptions of the study design, sampling procedures, anthropometric measurements and laboratory assays have been published elsewhere.25 26 This analysis used data from wave 1 (2011) and wave 3 (2015). Of the 15 683 participants who provided blood samples in 2011, we excluded individuals who (1) had missing data on FBG, TG or BMI in 2011 or 2015; (2) had diabetes or normoglycaemia at baseline, defined according to the American Diabetes Association (ADA) criteria;2 (3) had no follow-up information in 2015; (4) were aged under 45 years and (5) had extreme TyG-BMI values, defined as below Q1 − 1.5 ×interquartile range or above Q3+1.5 × interquartile range. After applying these criteria, 556 adults with pre-diabetes were included in the final analytic sample (figure 1). pre-diabetes was defined as FBG between 100 and 125 mg/dL or glycated haemoglobin (HbA1c) between 5.7% and 6.4%, in accordance with ADA diagnostic criteria.2 The mean age of the included participants was 58.5 years (range: 45–82 years), and 45.9% were male. This study therefore focused on a subpopulation of middle-aged and older adults with pre-diabetes drawn from the CHARLS cohort. Flowchart showing the selection of study population. BMI, body mass index; FBG, fasting blood glucose; TG, triglyceride; TyGBMI, triglyceride–glucose body mass index. Assessment of the cumulative average TyG-BMI TyG-BMI was calculated as ln((TG [mg/dL] ×FBG[mg/dL]) / 2) × BMI [kg/m2].15 Cumulative average TyG-BMI was computed as the mean of 2011 and 2015 values to better capture long-term exposure. FBG[mg/dL]) / 2) × BMI [kg/m2].15 Cumulative average TyG-BMI was computed as the mean of 2011 and 2015 values to better capture long-term exposure. Ascertainment of pre-diabetes, diabetes and normoglycaemia Glycaemic status was defined according to the ADA criteria,2 classifying participants as having pre-diabetes (FBG 100–125 mg/dL or HbA1c 5.7–6.4%), diabetes (self-reported diagnosis; use of glucose-lowering medication; FBG ≥126 mg/dL, or HbA1c ≥6.5%) or normoglycaemia (FBG<100 mg/dL and HbA1c<5.7%). The study outcome was glycaemic status in 2015, classified into three mutually exclusive categories: progression to diabetes, remaining pre-diabetic or regression to normoglycaemia. Covariates Baseline covariates were selected based on prior evidence and included demographic characteristics (age, sex, marital status, residence and education), lifestyle factors (smoking and drinking status), clinical conditions (hypertension, heart disease, stroke, dyslipidaemia and kidney disease), anthropometric and physiological measurements (weight, height, waist circumference, and systolic blood pressure (SBP) and diastolic blood pressure) and biochemical parameters, including FBG, HbA1c, TG, total cholesterol (TC), high-density lipoprotein cholesterol (HDL-c), low-density lipoprotein cholesterol (LDL-c), creatinine, blood urea nitrogen (BUN), uric acid (UA) and C-reactive protein (CRP). Statistical analysis All statistical analyses were conducted using R software V.4.4.1. Statistical significance was defined as two-sided p<0.05. The baseline characteristics of the participants were summarised across three tertiles—low, middle and high—of the cumulative average TyG-BMI, which was categorised based on the sample distribution: low (≤188.07), middle (188.07–220.99) and high (≥221.00). Normally distributed continuous variables were reported as mean±SD, while skewed continuous variables were presented as median (25th–75th percentile). Differences between groups were assessed using analysis of variance (ANOVA), and pairwise comparisons were conducted using the Student-Newman-Keuls method when the ANOVA p value was <0.05. The Kruskal-Wallis test was used for variables that did not meet normality assumptions. Categorical variables were presented as frequencies and percentages, with comparisons between groups made using the χ2 test or Fisher’s exact test. For covariates with missing values, multiple imputation based on chained equations was applied under the assumption of missing at random. Multivariate ordinal logistic regression models were used to evaluate the association between cumulative average TyG-BMI tertiles (low, middle, high) and glycaemic state transitions among participants with pre-diabetes. Results were reported as ORs with 95% CIs. Glycaemic status was treated as an ordinal outcome with three categories: progression to diabetes, remaining in the pre-diabetic state and regression to normoglycaemia. To capture both linear and non-linear trends in the relationship between cumulative average TyG-BMI and glycaemic transitions, orthogonal polynomial coding was employed. Prior to modelling, univariate logistic regression and Least Absolute Shrinkage and Selection Operator (LASSO) regression were used to screen variables predictive of the outcomes. Three stepwise models were developed: model 1 adjusted for sex and age; model 2 further adjusted for marital status, education level, smoking status, drinking status, heart disease, stroke, dyslipidaemia and kidney disease; and model 3, built upon model 2, additionally adjusted for HbA1c, TC, TG, creatinine, UA and BUN. The variance inflation factor (VIF) for each variable was calculated to assess potential multicollinearity. Furthermore, a restricted cubic spline (RCS) with 3 knots in the multivariate ordinal regression model was employed to explore the dose–response relationship between the cumulative average TyG-BMI and glycaemic state transitions. Subgroup analyses based on covariates—including sex (male or female), residential area (rural or urban), education level (middle school and below or high school and above) and smoking and drinking status (yes or no)—were conducted to assess the robustness of our findings. Additionally, sensitivity analyses were performed, which excluded participants with heart disease at baseline. Patient and public involvement Patients or members of the public were not involved in the design, conduct, reporting or dissemination plans of this research. Result Baseline characteristics grouped by three tertiles of the cumulative average TyG-BMI A total of 556 participants were included in this study (figure 1). The mean age was 58.5 years, and 45.9% were male. Table 1 shows baseline characteristics by TyG-BMI tertiles. Participants in the middle and high tertiles were more likely to be female and urban residents and have higher education levels. They also had higher waist circumference, SBP, DBP, TC, TG and CRP, but lower HDL-c, and higher prevalence of hypertension, dyslipidaemia and stroke compared with the low tertile (all p<0.05). Table 1Baseline characteristics of participants stratified by tertiles of the cumulative average TyG-BMI*TotalLow (tertile 1)Middle (tertile 2)High (tertile 3)P valueSample size (n)556185185186Age (years, median, quartile)58.50 (54.00; 64.00)62.00 (56.00; 68.00)58.00 (53.00; 63.00)57.50 (51.00; 62.00)<0.001Male (n, %)255 (45.86)107 (57.84)81 (43.78)67 (36.02)<0.001Marital status (n, %)0.070 Married478 (85.97)149 (80.54)167 (90.27)162 (87.10) Other78 (14.03)36 (19.46)18 (9.73)24 (12.90)Education level (n, %)0.003 Middle school and below429 (77.16)154 (83.24)144 (77.84)131 (70.43) High school and above127 (22.84)31 (16.76)41 (22.16)55 (29.57)Residence (n, %)0.001 Urban185 (33.27)43 (23.24)68 (36.76)74 (39.78) Rural371 (66.73)142 (76.76)117 (63.24)112 (60.22)Smoking (n, %)138 (24.82)65 (35.14)47 (25.41)26 (13.98)<0.001Drinking (n, %)177 (31.83)72 (38.92)58 (31.35)47 (25.27)0.005Hypertension (n, %)164 (29.50)43 (23.24)39 (21.08)82 (44.09)<0.001Dyslipidaemia (n, %)50 (8.99)6 (3.24)13 (7.03)31 (16.67)<0.001Heart disease (n, %)56 (10.07)12 (6.49)17 (9.19)27 (14.52)0.010Stroke (n, %)15 (2.70)1 (0.54)4 (2.16)10 (5.38)0.004Kidney disease (n, %)38 (6.83)10 (5.41)15 (8.11)13 (6.99)0.547Waist circumference (cm, median, quartile)85.00 (78.20; 91.60)77.30 (73.20; 80.60)85.40 (81.40; 90.00)93.00 (89.00; 98.00)<0.001SBP (mm Hg, median, quartile)127.67 (115.67; 142.33)125.33 (113.67; 140.33)125.67 (115.33; 141.33)132.00 (119.83; 146.67)0.003DBP (mm Hg, median, quartile)76.00 (67.67; 84.00)73.00 (65.00; 79.67)75.00 (68.33; 82.67)79.67 (72.33; 88.25)<0.001TC (mg/dL, median, quartile)196.01 (170.78; 223.94)188.27 (164.30; 210.70)197.55 (175.90; 226.93)199.49 (173.49; 232.44)0.001TG (mg/dL, median, quartile)103.54 (77.00; 148.68)77.88 (61.95; 102.66)101.78 (81.42; 137.18)146.91 (107.31; 206.65)<0.001HDL-c (mg/dL, median, quartile)51.03 (41.27; 61.08)57.22 (48.33; 70.75)50.64 (42.91; 60.70)43.30 (36.05; 52.96)<0.001LDL-c (mg/dL, median, quartile)115.59 (94.62; 138.02)109.79 (87.37; 129.51)122.55 (104.77; 145.75)114.63 (94.91; 139.47)0.055UA (mg/dL, median, quartile)4.24 (3.60; 5.17)4.17 (3.56; 5.08)4.21 (3.59; 5.12)4.43 (3.80; 5.25)0.072BUN (mg/dL, median, quartile)15.20 (12.55; 18.52)15.35 (12.63; 19.02)15.13 (12.88; 18.57)15.11 (12.30; 17.45)0.127Creatinine (mg/dL, median, quartile)0.76 (0.64; 0.87)0.76 (0.64; 0.86)0.75 (0.64; 0.87)0.77 (0.66; 0.87)0.405HbA1c (%, median, quartile)5.30 (5.00; 5.50)5.30 (4.90; 5.60)5.20 (5.00; 5.40)5.30 (5.00; 5.60)0.199C-reactive protein (mg/dL, median, quartile)1.07 (0.58; 2.20)0.75 (0.47; 2.12)1.06 (0.60; 1.92)1.45 (0.74; 2.53)<0.001Glycaemic status in 2015 (n, %)<0.001 Diabetes95 (17.09)17 (9.19)30 (16.22)48 (25.81) Pre-diabetes350 (62.95)116 (62.70)121 (65.41)113 (60.75) Normoglycaemia111 (19.96)52 (28.11)34 (18.38)25 (13.44)*The TyG-BMI was calculated by the formula ln [Triglyceride (mg/dL) × Fasting blood glucose (mg/dL)/2] × BMI (kg/m2); the cumulative average TyG-BMI was determined by the expression (TyG-BMI2011+TyG-BMI2015)/2.BUN, blood urea nitrogen; DBP, diastolic blood pressure; HbA1c, glycated haemoglobin; HDL-c, high-density lipoprotein cholesterol; LDL-c, low-density lipoprotein cholesterol; SBP, systolic blood pressure; TC, total cholesterol; TG, triglyceride; TyG-BMI, triglyceride–glucose body mass index; UA, uric acid. Baseline characteristics of participants stratified by tertiles of the cumulative average TyG-BMI* The TyG-BMI was calculated by the formula ln [Triglyceride (mg/dL) × Fasting blood glucose (mg/dL)/2] × BMI (kg/m2); the cumulative average TyG-BMI was determined by the expression (TyG-BMI2011+TyG-BMI2015)/2. BUN, blood urea nitrogen; DBP, diastolic blood pressure; HbA1c, glycated haemoglobin; HDL-c, high-density lipoprotein cholesterol; LDL-c, low-density lipoprotein cholesterol; SBP, systolic blood pressure; TC, total cholesterol; TG, triglyceride; TyG-BMI, triglyceride–glucose body mass index; UA, uric acid. Baseline characteristics grouped by follow-up outcomes During the 4-year follow-up, 350 participants (62.9%) remained pre-diabetic, 111 (20.0%) reverted to normoglycaemia and 95 (17.1%) progressed to diabetes (online supplemental table S1). Participants who reverted to normoglycaemia had lower baseline waist circumference, DBP, HbA1c, TC, TG and cumulative average TyG-BMI (all p<0.05). Association between the cumulative average TyG-BMI and glycaemic state transitions At the fourth follow-up year, the incidence of progression from pre-diabetes to diabetes and reversion to normoglycaemia differed markedly across tertiles of cumulative average TyG-BMI (online supplemental table S1 and figure S1). Participants in the highest tertile showed the highest progression rate to diabetes (50.5%) and the lowest reversion to normoglycaemia (22.5%), whereas those in the lowest tertile had the lowest progression (17.9%) and the highest reversion (46.9%). Overall, higher cumulative average TyG-BMI was associated with an increased risk of progression and a decreased likelihood of glycaemic improvement. Covariates were first screened using univariate logistic regression and LASSO regression, and three adjusted models were developed, with all VIF values <5, indicating no multicollinearity (online supplemental file 1). In the fully adjusted ordinal logistic regression model (table 2), higher cumulative TyG-BMI was significantly associated with reduced odds of regression to normoglycaemia (OR=0.50; 95% CI 0.33 to 0.62; p<0.001). RCS analysis demonstrated a negative, linear relationship between cumulative TyG-BMI and glycaemic improvement (p overall<0.001, p nonlinearity=0.149), indicating that participants with higher TyG-BMI levels were less likely to revert to normoglycaemia (figure 2). Table 2Logistic regression analysis of the association between the cumulative average TyG-BMI and glycaemic state transitionsCumulative average TyG-BMI (tertiles)Model 1Model 2Model 3OR (95% CIs)P valueOR (95% CIs)P valueOR (95% CIs)P valueLow (tertile 1)0.44 (0.32 to 0.60)<0.0010.45 (0.33 to 0.62)<0.0010.50 (0.33 to 0.62)<0.001Middle (tertile 2)High (tertile 3)Model 1 adjusted for age and sex.Model 2 adjusted for age, sex, marital status, education level, smoking status, drinking status, heart disease, stroke, dyslipidaemia and kidney disease.Model 3 adjusted for age, sex, marital status, education level, smoking status, drinking status, heart disease, stroke, dyslipidaemia, kidney disease, HbA1c, TC, TG, creatinine, UA and BUN.BUN, blood urea nitrogen; HbA1c, glycated haemoglobin; TC, total cholesterol; TG, triglyceride; TyG-BMI, triglyceride–glucose body mass index; UA, uric acid. Logistic regression analysis of the association between the cumulative average TyG-BMI and glycaemic state transitions Model 1 adjusted for age and sex. Model 2 adjusted for age, sex, marital status, education level, smoking status, drinking status, heart disease, stroke, dyslipidaemia and kidney disease. Model 3 adjusted for age, sex, marital status, education level, smoking status, drinking status, heart disease, stroke, dyslipidaemia, kidney disease, HbA1c, TC, TG, creatinine, UA and BUN. BUN, blood urea nitrogen; HbA1c, glycated haemoglobin; TC, total cholesterol; TG, triglyceride; TyG-BMI, triglyceride–glucose body mass index; UA, uric acid. Restricted cubic spline curves displaying the association between cumulative average TyG-BMI and glycaemic improvement. The model was adjusted for age, sex, marital status, education level, smoking status, drinking status, heart disease, stroke, dyslipidaemia, kidney disease, HbA1c, TC, TG, creatinine, UA and BUN. BUN, blood urea nitrogen; HbA1c, glycated haemoglobin; TC, total cholesterol; TG, triglyceride; TyG-BMI, triglyceride–glucose body mass index; UA, uric acid. Subgroup analysis and sensitivity analysis Subgroup analyses were conducted to examine the association between cumulative average TyG-BMI and glycaemic state transitions across potential risk factors, as depicted in figure 3. The relationship was consistent across subgroups stratified by age, gender, residence, education level and smoking and drinking habits (all P for interaction >0.05). Sensitivity analyses further supported these findings, showing no significant variations after excluding participants with pre-existing heart disease at baseline (online supplemental table S3). Subgroup analyses of the association between the cumulative average TyG-BMI and glycaemic state transitions. The model was adjusted for age, sex, marital status, education level, smoking status, drinking status, heart disease, stroke, dyslipidaemia, kidney disease, HbA1c, TC, TG, creatinine, UA and BUN. Education level: low, middle school and below; high, high school and above. BUN, blood urea nitrogen; HbA1c, glycated haemoglobin; TC, total cholesterol; TG, triglyceride; TyG-BMI, triglyceride–glucose body mass index; UA, uric acid. Discussion This study presents the first investigation into the relationship between cumulative average TyG-BMI and glycaemic state transitions among middle-aged and elderly individuals with pre-diabetes in China. The findings suggest that higher cumulative average TyG-BMI levels are associated with a lower likelihood of reverting to normoglycaemia and a higher risk of progressing to diabetes. Furthermore, the dose-response relationship identified via RCS analysis reinforces the negative association between higher TyG-BMI and glycaemic improvement, underscoring TyG-BMI’s vital role in glycaemic regulation and the importance of its monitoring in pre-diabetes management. Pre-diabetes is an intermediary stage between normoglycaemia and diabetes, characterised by a high risk of advancing to diabetes, and a heightened risk of microvascular and macrovascular diseases, chronic kidney disease, neuropathy and even death.3 5 27 28 Fortunately, pre-diabetes can be reversed to normoglycaemia, while its progression to diabetes can be slowed or prevented through lifestyle changes or pharmacologic interventions.29 A systematic review and meta-analysis of randomised controlled trials provides robust evidence that lifestyle changes can effectively reverse pre-diabetes to a normoglycaemic state.30 Data from the Special Diabetes Program for Indians - Diabetes Prevention Program revealed that metformin use, weight loss and increased physical activity are associated with greater odds of reversion to normoglycaemia.31 An observational study from the USA demonstrated that intensive lifestyle changes and metformin treatment reduced the incidence of T2DM in individuals with pre-diabetes by 58% and 31%, respectively.32 Lifestyle changes have also been reported to lower the incidence of cardiovascular events, microvascular complications and all-cause mortality.33 A pre-diabetes diagnosis serves as a crucial opportunity to avert the progression to T2DM and its related complications. IR is considered the primary pathogenic mechanism leading to the onset and progression of diabetes and its complications, even before plasma glucose levels become elevated.4 13 In the pre-diabetic stage, IR triggers compensatory hyperinsulinaemia to maintain normoglycaemia, which eventually leads to β-cell failure and transition to T2DM.34 Therefore, identifying IR using simple and reliable surrogate markers is clinically important for individuals with dysglycaemia. Traditional methods for evaluating IR, such as the HEC, are invasive and impractical in population-based studies. In addition, several alternative metabolic markers—such as fasting insulin (FI), homeostatic model assessment of insulin resistance (HOMA-IR) or conventional lipid parameters—have limitations when applied in large epidemiological settings. FI and HOMA-IR require direct insulin measurements, which are affected by analytical variability and are not routinely collected in many cohort studies.35 Traditional lipid markers (eg, TC or LDL-c), although inexpensive, primarily reflect isolated lipid metabolism and may not adequately represent the broader metabolic disturbances—such as dyslipidaemia, excess adiposity and impaired glucose homeostasis—that are often observed alongside IR.11 These methodological constraints highlight the need for a practical composite indicator in research on early dysglycaemia. In this context, the TyG-based indices have been recognised as accessible alternatives, and TyG-BMI has been shown to be one of the strongest among them.15 36 37 Overweight and obesity are major risk factors for pre-diabetes and T2DM in adults,38 and BMI is widely used to assess adiposity. Similarly, findings from the Steno-2 cohort indicate that lipid control is as important as glycaemic control in diabetes management.39 By integrating TG, FBG and BMI, TyG-BMI captures key metabolic abnormalities involved in glucose dysregulation, making it a plausible indicator of diabetes risk. Evidence supporting this comes from several population-based studies. A cohort study from the China Rich Health Care Group reported evidence supporting a causal link between TyG-BMI and T2DM incidence.17 Another observational study identified a non-linear association between TyG-BMI and new-onset diabetes.16 In the National Health and Nutrition Examination Survey population, TyG-BMI showed a U-shaped association with all-cause mortality among individuals with diabetes.18 Additionally, a large longitudinal cohort demonstrated that higher TyG-BMI levels significantly increased the risk of developing pre-diabetes.40 Previous studies have also established a link between TyG-BMI and several conditions, including cardiovascular diseases, cardiovascular events, hypertension and NAFLD.19,2141 42 Consistent with previous studies linking TyG-BMI to IR and metabolic disease,20 26 27 our findings further demonstrate its predictive value for both glycaemic regression and progression in a population with pre-diabetes—a relationship not previously evaluated simultaneously within the same cohort. Using multivariate ordinal logistic regression models, the analysis assessed the impact of varying levels of cumulative average TyG-BMI on glycaemic state transitions. The results revealed that higher cumulative average TyG-BMI levels were associated with an increased likelihood of progressing from pre-diabetes to diabetes, and a decreased likelihood of transitioning to normoglycaemia. This suggests that participants with higher cumulative average TyG-BMI levels were significantly less likely to achieve glycaemic improvement. Furthermore, the analysis showed that none of the subgroups—including age, gender, residence, education level and smoking and drinking habits—significantly modified the relationship between cumulative average TyG-BMI and glycaemic state transitions, validating the stability of these findings. These results highlight the value of TyG-BMI as a predictive marker for glycaemic outcomes in patients with pre-diabetes, underscoring the need for early and targeted interventions to improve metabolic status. Monitoring TyG-BMI may offer a practical means of identifying individuals with pre-diabetes at high risk for disease progression, supporting targeted lifestyle or pharmacological interventions. Such interventions may include personalised dietary and exercise programmes, as well as targeted therapies aimed at reducing IR and improving overall metabolic health. There are several advantages worth highlighting in this study. First, the focus on a population with pre-diabetes allowed for a detailed analysis of glycaemic state transitions and re-evaluated its relationship with cumulative average TyG-BMI—an area less explored in previous research. Second, the use of cumulative average TyG-BMI improved the accuracy of representing long-term metabolic risk exposure compared with single-point measurements. Third, the study’s reliability is supported by credible medical data from the CHARLS database, which provides a prospective nationwide cohort with extended follow-up. Finally, by analysing both glycaemic progression and reversal within the same cohort, this study highlighted the role of TyG-BMI in pre-diabetes management and its potential to reduce diabetes incidence through early intervention. Building upon these strengths, future research could compare the predictive performance of TyG-BMI with established markers such as HOMA-IR and HbA1c. HOMA-IR requires direct insulin measurements, which may not be routinely available in large cohort studies, whereas HbA1c reflects longer-term glycaemic exposure but can be influenced by factors such as haemoglobin variants or anaemia. Such comparative investigations would help clarify the relative predictive utility of TyG-BMI and provide a broader perspective to guide risk stratification and early intervention in populations with pre-diabetes. Several limitations of this study should be acknowledged. First, as an observational study, it cannot establish a direct causal relationship between cumulative average TyG-BMI and glycaemic state transitions. Second, the study population was restricted to Chinese adults aged ≥45 years from the CHARLS cohort, which limits the generalisability of the findings to other age groups or ethnic populations. Third, TyG-BMI was calculated using data from only two time points (2011 and 2015), failing to comprehensively capture the dynamic changes in long-term metabolic exposure. Fourth, despite adjusting for potential confounders in the statistical models, the risk of residual confounding from unmeasured variables cannot be fully ruled out. These limitations highlight the need for future studies with longer follow-up, more frequent TyG-BMI measurements and diverse study populations to validate the current findings. Conclusions Higher cumulative average TyG-BMI levels identify individuals with pre-diabetes who are less likely to revert to normoglycaemia and more likely to progress to diabetes, highlighting its value as a practical marker for early risk stratification and targeted clinical management of pre-diabetes. Future studies should aim to validate these findings in larger, long-term cohorts and compare the predictive performance of TyG-BMI with established metabolic markers to further improve generalisability and clinical applicability. Supplementary material Acknowledgements We thank the CHARLS staff team for their efforts to provide data used in this study and all participants who contributed to the data. Footnotes Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Prepublication history and additional supplemental material for this paper are available online. To view these files, please visit the journal online (https://doi.org/10.1136/bmjopen-2025-099014). Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Consent obtained directly from patient(s). Ethics approval: This study involves human participants. The studies involving human participants were reviewed and approved by the Ethics Review Committee of Peking University (IRB00001052-11015). The participants provided written informed consent to participate in the study. Participants gave informed consent to participate in the study before taking part. Patient and public involvement: Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research. Data availability statement Data may be obtained from a third party and are not publicly available. References Supplementary Materials Data Availability Statement Data may be obtained from a third party and are not publicly available.