The modified cardiometabolic index versus triglyceride-glucose index in predicting type 2 diabetes incidence: a 12-year cohort study Authors: Wei Ge, Chenjie Sun, Zhijian Zhu, Bing Wang, Zhigang Lu, Yesheng Pan Published: 2025 ABSTRACT BackgroundType 2 diabetes (T2D) presents a growing global health burden, with early identification of high-risk individuals remaining a critical challenge. The modified cardiometabolic index (MCMI), which integrates visceral fat, lipid ratios, and glucose measures, has emerged as a promising alternative, offering a more comprehensive assessment of metabolic risk. However, no large-scale cohort study has directly assessed the long-term predictive performance of the MCMI for incident T2D, particularly in normoglycemic populations. This study investigates the 12-year predictive performance of the MCMI for incident T2D and compares its efficacy with that of the triglyceride-glucose (TyG) index.MethodsIn this longitudinal cohort study, 15,453 adults with normal baseline glucose were selected from the NAGALA study. The associations of the MCMI and the TyG index with T2D risk were examined using Cox regression models and restricted cubic spline (RCS) analysis. Predictive performance was compared through receiver operating characteristic (ROC) analysis, and subgroup analyses assessed consistency across different populations.ResultsAmong participants, 373 (2.41%) developed T2D during a mean 6.05-year follow-up. In unadjusted analyses, the TyG index showed an HR of 3.76 (95% CI 3.22–4.38, P < 0.001), while the MCMI demonstrated a stronger association (HR 6.25, 95% CI 5.28–7.40, P < 0.001). These relationships persisted in fully adjusted models. RCS analysis revealed the TyG index and the MCMI maintained a positive linear relationship with T2D risk. ROC analysis indicated superior predictive performance for the MCMI within 1 to 12 years compared to the TyG index and remained relatively consistent in various subgroups.ConclusionsThe MCMI shows a strong, linear association with incident T2D and offers better predictive performance than the TyG index. These findings support the potential clinical utility of the MCMI for T2D risk stratification in normoglycemic individuals. From a broader health system perspective, its application in community-based screening, particularly within underserved regions, may strengthen early detection and support more efficient and accessible preventive care, thereby helping to alleviate the future burden on health systems.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12944-025-02839-5. Introduction The worldwide prevalence of diabetes mellitus (DM) has escalated dramatically, with projections estimating 700 million affected individuals by 2045, and type 2 diabetes (T2D) represents nearly 90% of all cases, posing unprecedented challenges to healthcare systems and economies worldwide [1]. Timely identification of high-risk individuals is critical for implementing preventive strategies; however, conventional risk markers often lack sufficient sensitivity to detect metabolic dysregulation in the pre-hyperglycemic stage [2]. While insulin resistance (IR) is a primary determinant of T2D onset, emerging evidence suggests visceral adiposity may play an equally important role through mechanisms involving chronic inflammation and lipotoxicity [3]. These findings underscore the pressing need for integrated biomarkers that simultaneously capture both components of metabolic dysregulation. The hyperinsulinemic–euglycemic clamp is still regarded as the reference method for evaluating IR, while the homeostatic model assessment of IR (HOMA-IR) offers a simpler surrogate [4, 5]. However, the HIEG clamp’s invasive nature and HOMA-IR’s dependence on insulin measurements limit widespread clinical utility. The triglyceride-glucose (TyG) index has become a widely applied and accessible IR surrogate [6, 7], yet it has an inability to account for visceral adiposity. In contrast, the modified cardiometabolic index (MCMI) incorporates waist circumference (WC) (a robust proxy for visceral fat) along with fasting plasma glucose (FPG) in combination with the triglyceride-to-high-density lipoprotein cholesterol (HDL-C) ratio, providing a more comprehensive metabolic profile. Recent studies demonstrate that the MCMI outperforms the TyG index in predicting nonalcoholic fatty liver disease (NAFLD) and hepatic fibrosis [8], conditions that share core pathophysiological pathways with T2D, including IR, visceral adiposity, and dysregulated lipid metabolism [9–11]. Despite these insights, no large-scale prospective study has evaluated whether the MCMI can predict the long-term risk of T2D, particularly among individuals with normal baseline glucose levels. Given that the MCMI simultaneously captures insulin resistance-related dyslipidemia and visceral adiposity, it is hypothesized that MCMI would serve as a more powerful and clinically informative predictor of incident T2D than the TyG index. To test this hypothesis, 12-year longitudinal data from 20,944 participants with normal FPG at baseline was analyzed to address two key questions: (1) Could the MCMI predict the incidence of T2D? and (2) Is the MCMI superior to the TyG index in long-term T2D risk prediction? By answering these questions, this study aims to provide new evidence supporting the integration of visceral adiposity metrics with traditional glycemic and lipid parameters for improved long-term T2D risk stratification. Method Study population The present longitudinal study drew on data from the NAGALA cohort (1994–2016), which enrolls Japanese adults undergoing annual health check-ups at Murakami Memorial Hospital in Gifu. Exclusion Criteria (Total excluded: n = 5,498): (1) Missing baseline data (n = 863). (2) Baseline dysglycemia: diagnosed diabetes (n = 323) and impaired fasting glucose (FPG ≥ 6.1 mmol/L) (n = 808). (3) Baseline liver disease (excluding NAFLD) (n = 416). (4) Individuals reporting excessive alcohol intake (n = 739). (5) Those using medications at baseline (n = 2321). (6) Withdrawn with undocumented reasons (n = 10). Additionally, 11 individuals with incomplete information on WC, height, FPG, or lipid measurements were excluded, yielding 15,453 eligible participants for analysis (Fig. 1). The open-access dataset was obtained from Dryad (https://datadryad.org/). The present analysis specifically examined the MCMI in relation to T2D risk, aligning with NAGALA’s original aim of identifying modifiable chronic disease risk factors. Fig. 1The flowchart of this study The flowchart of this study The initial study protocol received approval from the Ethics Committee of Murakami Memorial Hospital, and all participants provided written informed consent prior to enrollment [12]. Because the present work is a secondary analysis of previously collected data, no additional ethics review was required. All methods complied with the Declaration of Helsinki. Data collection and measurements Trained personnel collected data through standardized questionnaires (demographics, smoking status [never/former/current], alcohol intake, and exercise habits [≥ 1 session/week]) and performed physical measurements (height, weight, WC, and systolic and diastolic blood pressure [SBP/DBP]). NAFLD was diagnosed by gastroenterologists via abdominal ultrasound. Fasting venous blood samples were analyzed using an automated biochemical analyzer for liver enzymes (γ-glutamyl transferase [GGT], aspartate aminotransferase [AST], alanine aminotransferase [ALT]), lipid profiles (HDL-C, TG, total cholesterol [TC]), and glycemic parameters (hemoglobin A1c [HbA1c], FPG). the TyG index (calculated as ln [TG (mg/dL) × FPG (mg/dL)/2] [13]), and the MCMI (calculated as ln [TG (mg/dL) × FPG (mg/dL)/HDL-C (mg/dL)] × WC (cm)/height (cm) [8]). Body mass index (BMI) was calculated as weight (kg)/height² (m²). Outcome The primary endpoint was incident T2DM, identified based on the American Diabetes Association criteria: (1) FPG ≥ 7.0 mmol/L, (2) HbA1c ≥ 6.5%, or (3) a physician-verified self-reported diagnosis during follow-up [14]. Statistical analysis R software (version 4.4.3) was used for all statistical procedures, adopting a two-sided significance threshold of P < 0.05. Normally distributed continuous variables were reported as mean ± standard deviation, nonnormal variables as median (interquartile ranges, IQR), and categorical data as counts (%). Between-group comparisons employed t-tests or Mann-Whitney U tests based on data distribution, with chi-square/Fisher’s exact tests for categorical analyses. Additionally, standardized differences were calculated to assess imbalance between T2D and non-T2D cohorts, and a standardized difference of > 0.1 was considered to be a significant difference [15]. Adhering to STROBE guidelines, fitted four incrementally adjusted Cox models to quantify the link between the TyG index, the MCMI, and incident T2D by calculating corresponding hazard ratios and 95% confidence intervals. Model 1 did not adjust the variables; Model 2 adjusted basic demographic confounders such as age and sex; Model 3 further controls traditional cardiovascular risk factors and lifestyle factors, including SBP, fatty liver, exercise habits, drinking behavior, and smoking behavior based on Model 2; and Model 4 adjusted ALT, AST, GGT, TC, and HbA1c based on Model 3. Previous studies have found that liver enzyme markers are associated with new-onset diabetes [16]. Kaplan-Meier survival curves were adopted to visually observe the time-related event relationship of the TyG index and the MCMI to T2D, and group differences were appraised using the log-rank test. At the same time, a restricted cubic spline (RCS) regression model with four nodes (5th, 35th, 65th, and 95th percentiles) was used to evaluate the dose-response relationship of the TyG index and the MCMI to T2D. To contrast the predictive performance of the TyG index and the MCMI for T2D development, time-dependent receiver operating characteristic (ROC) curves were generated. The comparison of predictive efficacy at different time points (1–12 years) was evaluated, and the area under the curve (AUC) values of the TyG index and the MCMI were compared using the DeLong test. At the same time, the incremental predictive value of the MCMI relative to the TyG index for T2D risk was quantified using IDI analysis at different time points (1–12 years). To deepen the assessment of the TyG index and MCMI in relation to T2D, subgroup analyses based on sex, blood pressure, BMI, WC, smoking status, alcohol intake, fatty liver, exercise habits, and age were carried out. Results This study included 15,453 participants, with 373 (2.41%) having confirmed T2D during a mean follow-up of 6.05 years. As shown in Table 1, compared to non-T2D participants, participants with T2D were significantly older and more likely to be male (all P < 0.05 and all standardized differences > 0.1). Patients with T2D exhibited worse clinical profiles, including higher SBP and DBP, increased BMI, larger WC, and greater fatty liver prevalence (all P < 0.05 and all standardized differences > 0.1). Biochemical analyses revealed significantly poorer metabolic status in T2D participants, with elevated liver enzymes (ALT, AST, and GGT), higher HbA1c, FPG, TC, and TG, but lower HDL-C (all P < 0.05 and all standardized differences > 0.1). Notably, T2D participants showed a higher TyG index (8.62 ± 0.64 vs. 8.02 ± 0.64, P < 0.001, standardized differences > 0.1) and the MCMI (2.83 ± 0.57 vs. 2.19 ± 0.53, P < 0.001, standardized differences > 0.1). Lifestyle patterns also differed, with T2D participants being irregular exercisers and having higher current smoking rates and greater alcohol consumption (> 140 g/week) (all P < 0.05 and all standardized differences > 0.1). Table 1Baseline characteristics of participants in this studyVariablesTotalNon-T2D groupT2D group P Standardized Difference(n = 15453)(n = 15080)(n = 373)Age, years43.71 (8.90)43.63 (8.89)47.14 (8.52)< 0.0010.403Male8419(54.48)8133(53.93)286(76.68)< 0.0010.492SBP, mmHg114.49 (14.97)114.31 (14.91)122.03 (15.59)< 0.0010.506DBP, mmHg71.58 (10.50)71.44 (10.47)77.18 (10.23)< 0.0010.555BMI, kg/m222.12 (3.13)22.04 (3.07)25.03 (3.82)< 0.0010.862WC, cm76.47 (9.11)76.25 (8.97)85.08 (10.20)< 0.0010.919Smoking status< 0.0010.449 None9027(58.42)8882(58.90)145(38.87) Past2949(19.08)2872(19.05)77(20.64) Current3477(22.50)3326(22.06)151(40.48)Drinking status0.0030.164 < 40, g/week10,772(69.71)10,531(69.83)241(64.61) 40–140, g/week2784(18.02)2719(18.03)65(17.43) > 140, g/week1897(12.28)1830(12.14)67(17.96)Fatty liver2737(17.71)2514(16.67)223(59.79)< 0.0010.99Exercise habits2706(17.51)2655(17.61)51(13.67)0.0480.108Blood markersALT, U/L17.00(13.00,23.00)17.00 (13.00, 23.00)24.00 (18.00, 39.00)< 0.0010.666AST, U/L17.00(14.00,21.00)17.00 (14.00, 21.00)20.00 (16.00, 26.00)< 0.0010.442GGT, U/L15.00(11.00,22.00)15.00 (11.00, 22.00)24.00 (17.00, 36.00)< 0.0010.474HDL-C, mg/dl56.54 ± 15.5756.81 (15.54)45.92 (12.72)< 0.0010.767TC, mg/dl198.22 ± 33.41197.93 (33.33)209.95 (34.68)< 0.0010.353TG, mg/dl65.00(44.00,99.00)64.00 (43.00, 98.00)107.00 (76.00, 171.00)< 0.0010.728HbA1c, %5.17 ± 0.325.16 (0.32)5.53 (0.37)< 0.0011.074FPG, mg/dl92.96 ± 7.4492.76 (7.34)101.14 (6.43)< 0.0011.213TyG index8.03 (0.65)8.02 (0.64)8.62 (0.64)< 0.0010.94MCMI2.21 (0.54)2.19 (0.53)2.83 (0.57)< 0.0011.158Values were expressed as mean (SD) or median (interquartile range) or n (%). SBP systolic blood pressure, DBP diastolic blood pressure, BMI body mass index, WC waist circumference, ALT alanine aminotransferase, AST aspartate aminotransferase, GGT gamma-glutamyl transferase, HDL-C high-density lipoprotein cholesterol, TC total cholesterol, TG triglyceride, HbA1c hemoglobin A1c, FPG fasting plasma glucose, TyG index triglyceride–glucose index, MCMI modified cardiometabolic index, T2D type 2 diabetes Table 1Baseline characteristics of participants in this studyVariablesTotalNon-T2D groupT2D group P Standardized Difference(n = 15453)(n = 15080)(n = 373)Age, years43.71 (8.90)43.63 (8.89)47.14 (8.52)< 0.0010.403Male8419(54.48)8133(53.93)286(76.68)< 0.0010.492SBP, mmHg114.49 (14.97)114.31 (14.91)122.03 (15.59)< 0.0010.506DBP, mmHg71.58 (10.50)71.44 (10.47)77.18 (10.23)< 0.0010.555BMI, kg/m222.12 (3.13)22.04 (3.07)25.03 (3.82)< 0.0010.862WC, cm76.47 (9.11)76.25 (8.97)85.08 (10.20)< 0.0010.919Smoking status< 0.0010.449 None9027(58.42)8882(58.90)145(38.87) Past2949(19.08)2872(19.05)77(20.64) Current3477(22.50)3326(22.06)151(40.48)Drinking status0.0030.164 < 40, g/week10,772(69.71)10,531(69.83)241(64.61) 40–140, g/week2784(18.02)2719(18.03)65(17.43) > 140, g/week1897(12.28)1830(12.14)67(17.96)Fatty liver2737(17.71)2514(16.67)223(59.79)< 0.0010.99Exercise habits2706(17.51)2655(17.61)51(13.67)0.0480.108Blood markersALT, U/L17.00(13.00,23.00)17.00 (13.00, 23.00)24.00 (18.00, 39.00)< 0.0010.666AST, U/L17.00(14.00,21.00)17.00 (14.00, 21.00)20.00 (16.00, 26.00)< 0.0010.442GGT, U/L15.00(11.00,22.00)15.00 (11.00, 22.00)24.00 (17.00, 36.00)< 0.0010.474HDL-C, mg/dl56.54 ± 15.5756.81 (15.54)45.92 (12.72)< 0.0010.767TC, mg/dl198.22 ± 33.41197.93 (33.33)209.95 (34.68)< 0.0010.353TG, mg/dl65.00(44.00,99.00)64.00 (43.00, 98.00)107.00 (76.00, 171.00)< 0.0010.728HbA1c, %5.17 ± 0.325.16 (0.32)5.53 (0.37)< 0.0011.074FPG, mg/dl92.96 ± 7.4492.76 (7.34)101.14 (6.43)< 0.0011.213TyG index8.03 (0.65)8.02 (0.64)8.62 (0.64)< 0.0010.94MCMI2.21 (0.54)2.19 (0.53)2.83 (0.57)< 0.0011.158Values were expressed as mean (SD) or median (interquartile range) or n (%). SBP systolic blood pressure, DBP diastolic blood pressure, BMI body mass index, WC waist circumference, ALT alanine aminotransferase, AST aspartate aminotransferase, GGT gamma-glutamyl transferase, HDL-C high-density lipoprotein cholesterol, TC total cholesterol, TG triglyceride, HbA1c hemoglobin A1c, FPG fasting plasma glucose, TyG index triglyceride–glucose index, MCMI modified cardiometabolic index, T2D type 2 diabetes Baseline characteristics of participants in this study Values were expressed as mean (SD) or median (interquartile range) or n (%). SBP systolic blood pressure, DBP diastolic blood pressure, BMI body mass index, WC waist circumference, ALT alanine aminotransferase, AST aspartate aminotransferase, GGT gamma-glutamyl transferase, HDL-C high-density lipoprotein cholesterol, TC total cholesterol, TG triglyceride, HbA1c hemoglobin A1c, FPG fasting plasma glucose, TyG index triglyceride–glucose index, MCMI modified cardiometabolic index, T2D type 2 diabetes MCMI and the TyG index in relation to incident T2D In this cohort analysis, collinearity analysis showed DBP (VIF > 5) was excluded from multivariable modeling due to high collinearity (Supplementary Tables 1 and Supplementary Table 2). Multiple COX regression analysis revealed both the TyG index and MCMI exhibited a strong and independent associations with incident T2D (all P < 0.001) (Table 2). Specifically, the MCMI consistently demonstrated higher hazard ratios than the TyG index across all models, with the greatest divergence in the unadjusted analysis (HR 6.25 [95% CI: 5.28–7.40] vs. 3.76 [3.22–4.38]). This predictive advantage persisted through sequential adjustments for demographic factors (Model 2: HR 5.92 [95% CI: 4.933–7.103] vs. 3.38 [95% CI: 2.848–4.011]), lifestyle variables (Model 3: HR 3.50 [95% CI: 2.790–4.378] vs. 2.12 [95% CI: 1.744–2.565]), and comprehensive metabolic markers including HbA1c (Model 4: HR 2.64 [95% CI: 2.09–3.34] vs. 1.75 [1.43–2.16]). Notably, the MCMI maintained nearly 50% greater effect sizes than the TyG index in the fully adjusted model, suggesting the MCMI’s potential superiority for T2D risk stratification independent of conventional cardiometabolic factors. These findings warrant consideration of the MCMI as a robust, composite biomarker for T2D prediction in clinical practice. Table 2Association of the MCMI and the TyG index with T2DModelMCMITyG indexHR (95% CI) P HR (95% CI) P Model 16.248(5.278, 7.395)< 0.0013.757(3.223, 4.379)< 0.001Model 25.919(4.933, 7.103)< 0.0013.380(2.848, 4.011)< 0.001Model 33.495(2.790, 4.378)< 0.0012.115(1.744, 2.565)< 0.001Model 42.640(2.086, 3.340)< 0.0011.753(1.426, 2.155)< 0.001Model 1 did not adjust the variables; Model 2 adjusted age and sex; Model 3 adjusted SBP, drinking status, smoking status, fatty liver, and exercise habits based on Model 2; Model 4 adjusted ALT, AST, GGT, TC, and HbA1c based on Model 3;SBP systolic blood pressure, DBP diastolic blood pressure, ALT alanine aminotransferase, AST aspartate aminotransferase, GGT gamma-glutamyl transferase; TC: total cholesterol; HbA1c: hemoglobin A1c; TyG index, triglyceride-glucose index; MCMI, modified cardiometabolic index; T2D, type 2 diabetes Table 2Association of the MCMI and the TyG index with T2DModelMCMITyG indexHR (95% CI) P HR (95% CI) P Model 16.248(5.278, 7.395)< 0.0013.757(3.223, 4.379)< 0.001Model 25.919(4.933, 7.103)< 0.0013.380(2.848, 4.011)< 0.001Model 33.495(2.790, 4.378)< 0.0012.115(1.744, 2.565)< 0.001Model 42.640(2.086, 3.340)< 0.0011.753(1.426, 2.155)< 0.001Model 1 did not adjust the variables; Model 2 adjusted age and sex; Model 3 adjusted SBP, drinking status, smoking status, fatty liver, and exercise habits based on Model 2; Model 4 adjusted ALT, AST, GGT, TC, and HbA1c based on Model 3;SBP systolic blood pressure, DBP diastolic blood pressure, ALT alanine aminotransferase, AST aspartate aminotransferase, GGT gamma-glutamyl transferase; TC: total cholesterol; HbA1c: hemoglobin A1c; TyG index, triglyceride-glucose index; MCMI, modified cardiometabolic index; T2D, type 2 diabetes Association of the MCMI and the TyG index with T2D Model 1 did not adjust the variables; Model 2 adjusted age and sex; Model 3 adjusted SBP, drinking status, smoking status, fatty liver, and exercise habits based on Model 2; Model 4 adjusted ALT, AST, GGT, TC, and HbA1c based on Model 3; SBP systolic blood pressure, DBP diastolic blood pressure, ALT alanine aminotransferase, AST aspartate aminotransferase, GGT gamma-glutamyl transferase; TC: total cholesterol; HbA1c: hemoglobin A1c; TyG index, triglyceride-glucose index; MCMI, modified cardiometabolic index; T2D, type 2 diabetes To further assess dose-response relationships, participants were categorized into four quartile groups (Q1-Q4) according to their MCMI and TyG index. Kaplan-Meier survival curves revealed a significantly elevated cumulative incidence of T2D in the highest MCMI quartile compared to the lowest quartile (Log-rank P < 0.001) (Fig. 2). A comparable dose-dependent association was observed for the TyG index (P < 0.001). RCS regression further confirmed a positive linear association between both metabolic markers (MCMI and TyG index) and T2D risk (P for overall for both < 0.001, P for nonlinear for both > 0.05) (Fig. 3). Fig. 2Kaplan-Meier survival curves for cumulative T2D risk by the quartiles of the TyG index and the MCMI Kaplan-Meier survival curves for cumulative T2D risk by the quartiles of the TyG index and the MCMI Fig. 3Restricted cubic spline models analyzed the association of the TyG index and the MCMI with T2D risk. Models adjusted for the same covariables as in model 4 Restricted cubic spline models analyzed the association of the TyG index and the MCMI with T2D risk. Models adjusted for the same covariables as in model 4 Comparison of predictive performance of the MCMI and the TyG index for T2D risk In time-dependent ROC analysis over a 12-year follow-up, the MCMI demonstrated superior and sustained predictive performance compared to the TyG index for T2D. Specifically, there was no statistical difference in the AUCs of the two indices in the first year (AUC 0.753 vs. 0.718, P = 0.063) and the second year (AUC 0.737 vs. 0.734, P = 0.832), but AUC values of the MCMI were higher than the TyG index. From the third year to the 12th year, the discriminative power of the MCMI was significantly superior to that of the TyG index (all P < 0.05) and reached its maximum advantage in the eighth year (AUC 0.812 vs. 0.747, P < 0.001), with a sensitivity and specificity of 70.1% vs. 63.3% and 78.8% vs. 75.0%, respectively (Table 3 and Fig. 4). Additionally, IDI was used to evaluate whether the MCMI improved T2D prediction compared with the TyG index. The MCMI demonstrated significant improvements in risk discrimination within 1–12 years, with IDI estimates ranging from 3.6% to 7.8% (all P < 0.05). The greatest improvement occurred at years 3 and 6, with an IDI of 7.8% (all P < 0.001). This indicated the MCMI was superior as a long-term T2D risk assessment tool to the TyG index (Fig. 5). Table 3Comparison of the predictive efficacy of the TyG index and the MCMI for T2D risk events at different time pointsTimeIndicatorsAUC95% CISensitivitySpecificity P 1st yearMCMI0.7530.655–0.8510.7650.6570.063TyG index0.7180.610–0.8260.8230.567Ref.2nd yearMCMI0.7370.669–0.8050.7540.6320.832TyG index0.7340.6624–0.8050.6600.769Ref.3rd yearMCMI0.7770.730–0.8230.6140.8150.003TyG index0.7380.688–0.7880.6140.767Ref.4th yearMCMI0.7600.716–0.8040.5970.8150.001TyG index0.7230.676–0.7700.5960.765Ref.5th yearMCMI0.7900.754–0.8260.6870.793< 0.001TyG index0.7330.693–0.7720.7160.657Ref.6th yearMCMI0.8000.769–0.8320.6930.791< 0.001TyG index0.7420.707–0.7780.7430.649Ref.7th yearMCMI0.8120.784–0.8400.7490.752< 0.001TyG index0.7540.721–0.7870.7440.648Ref.8th yearMCMI0.8120.785–0.8380.7010.788< 0.001TyG index0.7470.715–0.7780.6330.75Ref.9th yearMCMI0.7970.770–0.8240.7860.669< 0.001TyG index0.7390.709–0.7690.7980.586Ref.10th yearMCMI0.7860.758–0.8140.7850.658< 0.001TyG index0.7250.694–0.7560.7900.575Ref.11th yearMCMI0.7750.746–0.8040.6390.778< 0.001TyG index0.7200.688–0.7520.6740.669Ref.12th yearMCMI0.7840.752–0.8160.6570.792< 0.001TyG index0.7200.684–0.7550.6790.671Ref.TyG index triglyceride–glucose index, MCMI modified cardiometabolic index, T2D type 2 diabetes Table 3Comparison of the predictive efficacy of the TyG index and the MCMI for T2D risk events at different time pointsTimeIndicatorsAUC95% CISensitivitySpecificity P 1st yearMCMI0.7530.655–0.8510.7650.6570.063TyG index0.7180.610–0.8260.8230.567Ref.2nd yearMCMI0.7370.669–0.8050.7540.6320.832TyG index0.7340.6624–0.8050.6600.769Ref.3rd yearMCMI0.7770.730–0.8230.6140.8150.003TyG index0.7380.688–0.7880.6140.767Ref.4th yearMCMI0.7600.716–0.8040.5970.8150.001TyG index0.7230.676–0.7700.5960.765Ref.5th yearMCMI0.7900.754–0.8260.6870.793< 0.001TyG index0.7330.693–0.7720.7160.657Ref.6th yearMCMI0.8000.769–0.8320.6930.791< 0.001TyG index0.7420.707–0.7780.7430.649Ref.7th yearMCMI0.8120.784–0.8400.7490.752< 0.001TyG index0.7540.721–0.7870.7440.648Ref.8th yearMCMI0.8120.785–0.8380.7010.788< 0.001TyG index0.7470.715–0.7780.6330.75Ref.9th yearMCMI0.7970.770–0.8240.7860.669< 0.001TyG index0.7390.709–0.7690.7980.586Ref.10th yearMCMI0.7860.758–0.8140.7850.658< 0.001TyG index0.7250.694–0.7560.7900.575Ref.11th yearMCMI0.7750.746–0.8040.6390.778< 0.001TyG index0.7200.688–0.7520.6740.669Ref.12th yearMCMI0.7840.752–0.8160.6570.792< 0.001TyG index0.7200.684–0.7550.6790.671Ref.TyG index triglyceride–glucose index, MCMI modified cardiometabolic index, T2D type 2 diabetes Comparison of the predictive efficacy of the TyG index and the MCMI for T2D risk events at different time points TyG index triglyceride–glucose index, MCMI modified cardiometabolic index, T2D type 2 diabetes Fig. 4The comparison of recognition ability of the TyG index and the MCMI for T2D risk events at different time points The comparison of recognition ability of the TyG index and the MCMI for T2D risk events at different time points Fig. 5The comparison of predictive performance between the TyG index and the MCMI for T2D risk events using IDI The comparison of predictive performance between the TyG index and the MCMI for T2D risk events using IDI Furthermore, the consistent advantages were observed across all major demographic, anthropometric, and metabolic subgroups. Stratified analyses revealed robust discriminative ability for both sexes, with women exhibiting particularly strong predictive power (MCMI AUC = 0.88 vs. TyG index AUC = 0.81, P < 0.001), while men also showed significantly higher AUCs for the MCMI (0.73 vs. 0.67, P < 0.001). Similarly, age-stratified analyses demonstrated that the MCMI maintained superior performance in participants < 60 years (AUC = 0.78 vs. 0.72, P < 0.001), whereas differences were nonsignificant in the ≥ 60-year subgroup (AUC = 0.77 vs. 0.68, P = 0.148), which may be due to the reduced statistical power caused by the lower event rate in this subgroup. Both indices performed better in individuals with lower metabolic burden, including those with BMI < 25 kg/m² (AUC = 0.75 vs. 0.69, P < 0.001), normal WC (AUC = 0.78 vs. 0.70, P < 0.001), and normotensive individuals (AUC = 0.79 vs. 0.73, P < 0.001). However, the MCMI’s advantage was preserved even in those with obesity, central adiposity, or elevated blood pressure. Lifestyle factors further confirmed the MCMI’s stability, with superior performance in non-smokers (AUC = 0.81 vs. 0.74, P < 0.001), light drinkers (AUC = 0.81 vs. 0.74, P < 0.001), and low physically active individuals (AUC = 0.79 vs. 0.73, P < 0.001), demonstrating its generalizability across behavioral risk strata. In summary, the MCMI outperformed the TyG index in traditional clinically relevant subgroups, with particularly strong discrimination in women, younger individuals, and those with a relatively lower metabolic risk burden, suggesting the MCMI’s utility as a more reliable predictor of incident T2D across diverse populations compared to the TyG index (Fig. 6 and Supplementary Table 3). Fig. 6Receiver operating characteristic curves compared the predictive efficacy of the TyG index and the MCMI for T2D risk events at different subgroups Receiver operating characteristic curves compared the predictive efficacy of the TyG index and the MCMI for T2D risk events at different subgroups Subgroup analysis Stratified evaluations were conducted to explore how the TyG index and the MCMI relate to incident T2D across clinically relevant subgroups demonstrated that both markers exhibited robust associations in all examined subgroups, with consistent hazard ratios above 1.0. Although modest interactions were observed between the MCMI with age (P for interaction = 0.048) and sex (P for interaction = 0.035) and the TyG index with sex (P for interaction = 0.010), this did not substantially alter their strong, independent predictive value for T2D risk across all demographic strata. Notably, the MCMI consistently demonstrated higher hazard ratios than the TyG index across all subgroups (Supplementary Fig. 1). Discussion This 12-year prospective cohort study demonstrated that the MCMI exhibits superior discriminative ability for incident T2D compared with the widely used the TyG index, particularly in normoglycemic individuals. This study represents the first long-term validation of the MCMI’s predictive utility, highlighting its potential as a clinically actionable biomarker for long-term T2D risk assessment. Beyond traditional glycemic markers, the MCMI integrates visceral adiposity and dyslipidemia, providing a more holistic assessment of metabolic derangement. Collectively, these findings reinforce that the MCMI outperforms the TyG index in predictive capability, underscoring its clinical value as a more comprehensive metabolic risk indicator. These findings align with evidence that visceral adiposity and dyslipidemia contribute significantly to T2D pathogenesis [17, 18]. The MCMI’s stronger association with T2D likely reflects its dual capture of dysregulated lipid metabolism (via TG/HDL-C ratio) and central obesity (via waist-to-height ratio), both implicated in IR and β-cell dysfunction [10, 19, 20]. In contrast, the TyG index-though a robust IR surrogate-fails to account for visceral adiposity, a key driver of metabolic inflammation and lipotoxicity [21–23]. This distinction may explain why the MCMI maintained a nearly 50% greater effect size than the TyG index in fully adjusted models (HR 2.64 vs. 1.75), even after accounting for HbA1c and hepatic enzymes. This suggests that integrated metabolic indices accounting for both adiposity and IR may offer superior risk stratification. Prior studies have established the TyG index as a feasible IR surrogate [24, 25], but its inability to assess visceral adiposity limits its predictive power [26]. The MCMI, including fat indicators, could significantly improve the prediction performance of T2D compared to the TyG index. The recent work by Guo et al. [8] showing the MCMI’s superiority in predicting NAFLD lends biological plausibility to findings of this study [27], given the shared pathophysiological pathways between NAFLD and T2D. Unlike prior studies that evaluated short-term T2D risk, this study provides longitudinal evidence (12 years) of the MCMI’s discriminative advantage, with AUCs peaking at 0.812 (vs. TyG index 0.747) at 8 years. This time-dependent superiority highlights the MCMI’s potential for long-term risk prediction. Furthermore, IDI analyses confirmed significant improvements (3.6–7.8%) in predictive accuracy, reinforcing the MCMI as a clinically meaningful upgrade over the index. Therefore, individuals with high MCMI are recommended to promptly undergo lifestyle or pharmacological interventions before overt hyperglycemia develops. Subgroup analyses also reveal nuanced insights into the MCMI’s predictive advantages across clinically important populations. Specifically, the MCMI demonstrated particularly strong discriminatory performance in women (AUC 0.88 vs. the TyG index’s 0.81, P < 0.001), likely reflecting its superior sensitivity to sex-specific visceral adiposity patterns that drive diabetes risk independent of BMI [28]. Its enhanced discrimination may reflect greater sensitivity to estrogen-regulated lipid dysfunction, particularly visceral adiposity and hepatic IR in postmenopausal states [29, 30]. Among younger individuals (< 60 years; AUC 0.78 vs. 0.72), the MCMI’s performance may stem from its ability to detect early ectopic fat deposition-a key driver of prediabetes. Intriguingly, in normal-weight individuals (BMI < 25, AUC 0.75 vs. 0.69), the MCMI remained robust, aligning with the “metabolically obese, normal-weight” phenotype, where visceral fat and adipose-derived inflammation, such as elevated IL-6, contribute disproportionately to metabolic risk [31]. Meanwhile, these findings also highlighted its ability to uncover metabolically obese, normal-weight phenotypes often missed by conventional screening. While the predictive advantage persisted across most subgroups, the margin narrowed in elderly participants (≥ 60 years, P = 0.148) and obese individuals (BMI ≥ 30, P = 0.052), which may be due to the reduced statistical power caused by the lower event rate in these subgroups. In conclusion, these findings collectively suggest the MCMI’s particular value in identifying at-risk populations where traditional indicators like BMI or the TyG index may fail to capture early metabolic dysfunction, especially in women, younger adults, and normal-weight individuals with visceral adiposity. This consistent superiority across key stratified populations further confirms the MCMI as a more sensitive tool for early metabolic risk detection than the TyG index. In countries such as China and Japan, where the prevalence of T2D has surpassed 10–15% among adults and where delayed diagnosis frequently leads to late-stage presentation, early identification of individuals at high metabolic risk is of considerable importance. In this cohort, even among normoglycemic participants, 2.41% developed T2D, indicating a substantial pool of individuals who may remain undetected until advanced disease occurs. The MCMI demonstrated superior long-term predictive performance (AUC up to 0.812 at 8 years) compared with the TyG index (AUC 0.747), suggesting that its use could improve early stratification of high-risk individuals. Notably, the MCMI relies solely on WC, FPG, and lipid parameters, which are already routinely collected in community-based screening or occupational health check-up programs in many Asian countries. Thus, incorporating the MCMI into primary care screening algorithms would incur minimal additional healthcare cost, while enabling targeted lifestyle or pharmacologic interventions during the pre-diabetic stage. This strategy may help reduce progression to T2D and decrease long-term healthcare expenditure associated with diabetic complications. Strengths This study has several key strengths. It is the first long-term cohort study to directly compare the MCMI with the TyG index for predicting incident T2D. The inclusion of normoglycemic individuals allows accurate assessment of early risk detection. The use of multiple complementary analytical methods ensures robust results, and consistent findings across subgroups further support the stability and generalizability of the MCMI. Finally, the MCMI relies on simple, routinely collected clinical measures, making it practical and easily applicable in community and primary-care settings. Limitations Several limitations also warrant consideration when interpreting these findings. Single-ethnicity cohort (Japanese population), which may limit generalizability to other ethnic groups with different adiposity distribution patterns. Potential unmeasured confounders, such as dietary habits or genetic predispositions, which were not adjusted for. Lack of dynamic biomarker assessment-only baseline MCMI/TyG index values were analyzed, whereas repeated measurements might improve predictive accuracy. Lower event rate in subgroups, such as older adults and smokers, reducing statistical power for certain stratified analyses. Diagnostic reliance on non-OGTT criteria for T2D, which may miss early-stage dysglycemia or borderline cases detectable by 2-hour postprandial glucose [2]. Future research should validate the MCMI in multiethnic cohorts to assess cross-population applicability. Incorporate repeated measures of the MCMI to evaluate its dynamic predictive utility. Explore machine learning approaches combining the MCMI with other biomarkers, such as inflammatory markers, for enhanced precision. Conclusion In conclusion, this large prospective study establishes the MCMI as a robust and superior predictor of incident T2D compared to the TyG index, owing to its integration of central adiposity, dyslipidemia, and fasting glucose. The primary clinical relevance of the findings of this study lies in the MCMI’s potential to enhance early, community-based risk stratification. Given that the index is derived from simple, low-cost, and routinely collected clinical measures, it can be seamlessly incorporated into standard metabolic health assessments. This provides a practical opportunity to identify high-risk, normoglycemic individuals who are currently undetected by conventional screening and who may derive the greatest benefit from targeted preventive strategies, such as personalized lifestyle counseling and close monitoring. This approach is particularly viable for resource-limited settings and high-burden countries, where scalable and cost-effective tools are urgently needed to alleviate the growing burden of diabetes. Further research is needed to validate the MCMI across diverse ethnic groups to confirm its generalizability and establish population-specific cut-off points. Furthermore, the clinical utility of the MCMI must be tested in real-world intervention studies to determine whether its implementation in primary care protocols effectively translates into a reduction in diabetes incidence and improved long-term patient outcomes. Supplementary Information Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2. Supplementary Material 3. Supplementary Material 3. Supplementary Material 4. Supplementary Material 4.