Association of the triglyceride-glucose index and its combination with obesity indices with cardio-renal-metabolic multimorbidity: two decades of follow-up in the Tehran Lipid and Glucose Study

Abstract

Objective

To investigate the associations between the triglyceride-glucose (TyG) index and its obesity-related derivatives with the risk of incident cardio-renal-metabolic multimorbidity (CRMM) in a Middle Eastern adult population initially free of cardiovascular disease (CVD), type 2 diabetes mellitus (T2DM), and chronic kidney disease (CKD).

Methods

In this prospective cohort analysis of 5845 Iranian adults from the Tehran Lipid and Glucose Study, we evaluated the associations of the TyG index and its combinations with body mass index (BMI), waist circumference (WC), waist-height ratio (WHtR), and waist-hip ratio (WHR) with the incidence of CRMM. Multivariable Cox proportional hazards models were used to estimate the associations between TyG indices and CRMM risk. The Wald test was used to formally compare the effect sizes of each TyG-related index with that of the TyG index in multivariable models. The predictive performance of these indices was evaluated using Harrell’s C-index and the integrated discrimination improvement (IDI).

Results

Over a median follow-up of 15.3 years (IQR: 11.9–16.4), 344 individuals (5.9%) developed CRMM. Restricted cubic spline models demonstrated significant linear associations between TyG indices and CRMM risk. The corresponding HRs (95% CI) per 1-SD increase were 1.41 (1.24–1.60) for the TyG index, 1.52 (1.36–1.71) for TyG-BMI, 1.57 (1.38–1.78) for TyG-WC, 1.57 (1.38–1.78) for TyG-WHtR, and 1.42 (1.24–1.63) for TyG-WHR (all P < 0.001). The inclusion of anthropometric measures alongside the TyG index did not substantially enhance its association with CRMM risk (all P for differences ≥ 0.05). Incremental predictive performance analyses showed modest but statistically significant improvements when adding TyG and TyG-obesity indices to conventional risk factors (all P < 0.05), whereas incorporating anthropometric-based indices to a model already containing TyG did not yield additional predictive improvement (all P > 0.05). The majority of associations remained robust after adjustment for homeostatic model assessment of insulin resistance and in sensitivity analyses. The association between TyG-WHR and CRMM was more pronounced among non-obese than obese individuals (P for interaction < 0.001).

Conclusions

Higher levels of TyG and TyG-obesity indices were independently associated with an increased risk of CRMM; however, incorporating obesity indices did not confer substantial improvement over the TyG index alone.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-025-02944-w.

Introduction

Multimorbidity, the coexistence of two or more chronic conditions, affects about 37.2% of adults and is a major global health concern [1]. This burden is particularly pronounced in low- and middle-income countries (LMICs) with an overall multimorbidity prevalence of 36.4% [2]. Among the most frequently observed patterns in these settings are cardiometabolic disorders, including type 2 diabetes mellitus (T2DM) and cardiovascular disease (CVD) [2]. In the Middle East and North Africa (MENA), multimorbidity affects 33.1% of people, and cardiometabolic disorders are the most common patterns [2]. The American Heart Association (AHA) recently emphasized that cardiometabolic and cardiorenal diseases should not be considered distinct, as they share pathophysiological mechanisms, the interconnection of which has led to the concept of cardio-renal-metabolic multimorbidity [3–7].

The TyG index has emerged as a reliable and cost-effective surrogate marker of insulin resistance (IR) [8, 9], with prognostic utility in cardiac, renal, and metabolic diseases and multimorbidity [10–13]. According to the Iran STEPs 2021 national survey, hypertriglyceridemia affects 39.7% of Iranian adults, with particularly high rates among individuals with obesity or diabetes [14]. Moreover, based on the Iran STEPs surveys (2004–2021), the prevalence of abdominal obesity increased from 27.5 to 40.4% [15], underscoring the need for integrated strategies to address hypertriglyceridemia and abdominal obesity as key cardiometabolic risks in Iran.

Previous studies have indicated that combining the TyG index with obesity-related measures, such as body mass index (BMI), waist circumference (WC), and waist-height ratio (WHtR), provides a more comprehensive evaluation of metabolic risk [12, 16–18]. Previous studies among European individuals have reported that combining the TyG index with obesity-related indicators such as BMI significantly predicts incident and progression of CRMM [13]. TyG-WC and TyG-WHtR indices have also been shown to enhance the effectiveness of IR assessment and improve the ability to predict incident cardiometabolic multimorbidity and its progression beyond TyG alone [19]. However, in an East Asian population, TyG, TyG-BMI, TyG-WC, and TyG-WHtR showed comparable predictive ability for the risk of developing multimorbidity across 14 common chronic diseases [20]. Another recent study among Chinese individuals found that elevated TyG, TyG-BMI, TyG-WC, and TyG-WHtR were associated with increased CRMM risk, while differences in their predictive values were not evaluated [12]. Although the TyG index reflects IR, visceral adiposity can further aggravate IR through excess free fatty acids and proinflammatory cytokines, leading to ectopic fat deposition, mitochondrial dysfunction, oxidative stress, and systemic inflammation, all of which are mechanisms that synergistically increase cardiac and metabolic risks beyond those captured by the TyG index alone [21–23]. Thus, combining these measures can provide a more comprehensive assessment of the interplay between metabolic dysfunction and obesity.

In the present study, conducted in a MENA population, we aimed to examine the association between the TyG index and its combinations with obesity measures, including TyG-BMI, TyG-WC, TyG-WHtR, and TyG-waist-hip ratio (TyG-WHR), with incident CRMM over approximately two decades of follow-up in the Tehran Lipid and Glucose Study (TLGS) cohort.

Materials and methods

Study design

The Tehran Lipid and Glucose Study (TLGS) is a community-based, prospective cohort study initiated to investigate the prevalence and incidence of non-communicable diseases (NCDs) and their risk factors. Individuals from District 13 of Tehran were recruited using multistage random sampling in two phases: Phase 1 from 1999 to 2001 (n = 15,005) and Phase 2 from 2002 to 2005 (n = 3550). Follow-up assessments have been carried out approximately every three years, including Phase 3 (2005–2008), Phase 4 (2008–2011), Phase 5 (2011–2014), and Phase 6 (2015–2018). The study’s design, participant selection process, data collection methods, and survey instruments have been comprehensively described in earlier publications [24]. The current study received approval from the institutional review board (IRB) of the Research Institute for Endocrine Sciences (RIES) at Shahid Beheshti University of Medical Sciences, with the ethics code number IR.SBMU.ENDOCRINE.REC.1404.041. Written informed consent was obtained from all participants.

Study population

In the current study, 9747 participants aged 30 years and older (8071 in Phase 1 and 1676 in Phase 2) were considered for inclusion. Participants were screened for pre-existing NCDs, including T2DM, CKD, and CVD, and those with any of these conditions were excluded (n = 2202), leaving 7545 eligible individuals. Additional exclusions were applied to those with missing data on triglycerides (TG) or fasting plasma glucose (FPG) (n = 275), and those with missing anthropometric data (weight, height, WC, or hip circumference [HC]) (n = 187). Individuals with incomplete covariate information (n = 88) and those lacking any follow-up data to determine whether CRMM developed (n = 1150) were also excluded. After applying these criteria, the final study population comprised 5845 individuals, of whom 3158 were women (Fig. 1).

Fig. 1.

Flowchart of participant selection from the TLGS for inclusion in the current analysis. TLGS, Tehran Lipid and Glucose Study; T2DM, type 2 diabetes mellitus; CKD, chronic kidney disease; CVD, cardiovascular disease; TG, triglycerides; FPG, fasting plasma glucose; WC, waist circumference; HC, hip circumference

Clinical and laboratory measurements

Demographic information, marital status, education level, smoking habits, family history of diabetes and CVD, past medical history, and medication use were gathered using standardized TLGS questionnaires. Anthropometric assessments included height, weight, HC, and WC. Blood pressure (including systolic blood pressure [SBP] and diastolic blood pressure [DBP]) was measured twice on the right arm after a 15-min seated rest, with 5-min intervals between the readings, and the average of the two readings was used for analysis. For biochemical analyses, blood samples were collected from all participants between 7:00 and 9:00 AM after an overnight fasting period of 12–14 h. Parameters measured included FPG, creatinine, and lipid profile components such as total cholesterol (TC), TG, and high-density lipoprotein cholesterol (HDL-C). Two-hour post-challenge glucose (2 h-PG) was measured after intake of 82.5 g glucose monohydrate. Regarding insulin, serum samples were stored at − 70 °C and then transferred to the hormone laboratory at the Research Institute for Endocrine Sciences for analysis. Fasting insulin levels were measured using the electrochemiluminescence immunoassay technique, employing Roche Diagnostics kits and analyzed on the Roche/Hitachi Cobas e-411 analyzer [25]. The homeostatic model assessment of insulin resistance (HOMA-IR) was calculated as [FPG (mmol/L) × fasting serum insulin (mU/L)/22.5] [26]. Serum creatinine concentrations were measured using the kinetic colorimetric Jaffe method. The estimated glomerular filtration rate (eGFR) was calculated using the 2021 Chronic Kidney Disease Epidemiology Collaboration equation (CKD-EPI) formula [27]. Lipid profile measurements, including TC, TG, and HDL-C, were performed using enzymatic colorimetric methods. Low-density lipoprotein cholesterol (LDL-C) was determined using a modified version of the Friedewald formula [28]. All biochemical analyses, except insulin, were performed at the TLGS research laboratory on the day of blood collection using commercial kits (Pars Azmoon Inc., Tehran, Iran) and a Selectra 2 chemistry auto-analyzer (Vital Scientific, Spankeren, The Netherlands). Further details are provided in the Supplementary Material and elsewhere [24].

Definition of terms

BMI was calculated by dividing weight in kilograms by the square of height in meters (kg/m²), WHtR was calculated by dividing WC by height, and WHR was calculated by dividing WC by HC. The TyG index was computed as ln[fasting TG (mg/dL) × FPG (mg/dL)/2], and TyG-related indices were defined as follows [29]:

A positive family history of premature CVD was defined as coronary heart disease (CHD) and/or stroke in a first-degree relative occurring before age of 55 years for male relatives and before age of 65 years for female relatives. A positive family history of diabetes was defined by the presence of diabetes in a first-degree relative. CKD at baseline was defined as an eGFR of less than 60 mL/min/1.73 m², while CKD during follow-up was defined as an eGFR < 60 mL/min/1.73 m² accompanied by a ≥ 30% decline from the baseline value [30–35], to balance a clinically meaningful eGFR decline with a sufficient number of events for analysis. Hypertension was defined as SBP ≥ 140 mm Hg, DBP ≥ 90 mm Hg, or the use of antihypertensive drugs. Prediabetes was identified in subjects with FPG levels between 5.6 and 6.9 mmol/L or 2 h-PG levels between 7.8 and 11.0 mmol/L. T2DM was diagnosed based on FPG levels ≥ 7.0 mmol/L, 2 h-PG levels ≥ 11.1 mmol/L, or the use of anti-diabetic medications []. Dyslipidemia was defined as TC ≥ 240 mg/dL, and/or LDL-C ≥ 160 mg/dL, and/or HDL-C < 40 mg/dL, and/or TG ≥ 200 mg/dL, and/or the use of lipid-lowering medications [37]. Obesity was defined as general obesity (BMI ≥ 30 kg/m²) or central obesity (WC ≥ 95 cm), based on proposed WC cut-offs for the Iranian population [38].

Details on CVD definitions and outcome analysis are available elsewhere [24, 39]. In summary, each year, a trained nurse conducted follow-up interviews by phone to gather information on any medical events experienced by participants in the previous year. If an event was reported, a trained physician collected additional data from medical records, or from death certificates in cases of mortality, obtained during home or hospital visits. All gathered documents were reviewed by an outcome committee within the TLGS. The final diagnosis was determined based on the committee’s majority consensus. In this study, CVD events included angiographically proven coronary heart disease (CHD), stroke, unstable angina pectoris, probable myocardial infarction (MI), definite MI, and sudden cardiac death (SCD).

While no standard criteria exist regarding the number or type of conditions required to define multimorbidity [40], CRMM was defined as the presence of two or more of CVD, CKD, and T2DM [4, 6, 40–42].

Statistical analysis

Baseline characteristics were reported as mean (SD) for normally distributed continuous variables, median (IQR) for non-normal continuous variables, and count (%) for categorical variables. Student’s t-test, Kruskal–Wallis, and Chi-Squared tests compared characteristics across TyG tertiles and between respondents and non-respondents (missing data or no CRMM follow-up data).

A multivariable Cox regression estimated hazard ratios (HRs) with 95% confidence intervals for CRMM risk, calculated across tertiles of TyG and TyG-obesity indices (first tertile as reference) and per SD increase. P values for the trend across tertiles were calculated by treating TyG and TyG-obesity indices as continuous variables. The proportional hazard assumptions in the Cox regression models were checked using Schoenfeld’s global test of residuals. Model 1 adjusted for age and sex. Model 2 (primary model) further adjusted for marital status, educational level, smoking status, SBP, antihypertensive medication use, HDL-C, LDL-C, lipid-lowering medication use, eGFR, pulse rate, family history of diabetes, family history of premature CVD, and prediabetes. Covariate selection was informed by former studies involving TyG-related indices and cardiometabolic risk [13, 43, 44], as well as by components of the pooled cohort equations (PCE) risk score [45]. To assess the robustness of our findings, we conducted several sensitivity analyses. Model 3 extended Model 2 by additionally adjusting for HOMA-IR in a subsample with available serum insulin data (n = 3197), to examine whether TyG indices are associated with outcomes independent of direct measures of insulin resistance, as TyG has outperformed HOMA-IR regarding risk of all-cause and cardiovascular mortality [46]. Model 4 accounted for the competing risk of death using the Fine–Gray subdistribution hazards [47], with death from any cause prior to CRMM onset specified as the competing event. CVD deaths were treated as competing only if T2DM or CKD had not occurred beforehand; otherwise, they were classified as CRMM events. Deaths attributed to T2DM or CKD were not modeled separately, as these conditions were defined using biomarkers and medication data rather than mortality records [48]. Finally, Model 5 excluded CRMM events that occurred within the first 5 years of follow-up to account for potential reverse causality bias, consistent with previous cohort studies [44, 49].Additionally, to evaluate the non-linear relationship between TyG and TyG-obesity indices and CRMM, we performed a multivariable-adjusted restricted cubic spline analysis with four knots positioned at the 5th, 35th, 65th, and 95th percentiles of the distribution of these indices.

In a secondary analysis, recognizing that different combinations of cardiac, renal, and metabolic conditions may have distinct prognostic implications, we conducted stratified analyses by CRMM subtypes: CVD and T2DM, CKD and T2DM, CVD and CKD, and the triple-outcome subtype comprising CVD, CKD and T2DM.

To directly compare the associations of the TyG index with those of TyG-related obesity indices regarding the risk of incident CRMM, we conducted paired homogeneity tests using the Wald test for linear hypotheses. Prior to model entry, we assessed potential multicollinearity among TyG indices using the variance inflation factor (VIF), ensuring no collinearity concerns that could bias the comparisons (All VIF < 1.5). Furthermore, we constructed a base model comprising the covariates included in Model 2 of the main analysis. A novel model was then developed by adding the TyG index or each of the TyG-obesity indices (TyG-BMI, TyG-WC, TyG-WHtR, and TyG-WHR) to these covariates. To assess predictive performance, we estimated Harrell’s C-index with 95% bootstrap confidence intervals for each index in relation to incident CRMM. In addition, we evaluated incremental discrimination by calculating the integrated discrimination improvement (IDI), comparing the base and novel models to determine whether the inclusion of TyG-related indices enhanced prediction beyond conventional risk factors. Finally, we compared Harrell’s C-index and IDI between the model containing TyG plus Model 2 covariates and the models further incorporating TyG-BMI, TyG-WC, TyG-WHtR, or TyG-WHR.

Subgroup analyses were performed by age (30–44, 45–59, and ≥ 60 years), sex, current smoking, dyslipidemia, prediabetes, hypertension, and obesity, to assess the risk of CRMM based on the TyG and TyG-obesity indices, with the inclusion of multiplicative interaction terms to test for statistical significance of the interactions. We adjusted the P interaction values using a Bonferroni correction, setting the significance threshold at approximately 0.0014 (0.05/35 tests).

For the CVD outcome, the event date was defined as the exact date when the condition was first diagnosed. For T2DM and CKD cases, the event date was determined as the midpoint between the date of the follow-up visit that confirmed the diagnosis and the previous follow-up visit. The event date for CRMM was recorded as the date when the second NCD was documented. Censoring occurred if a participant was lost to follow-up, died, or reached the end of the study (i.e., year 2018).

To address potential selection and survival bias due to missing data, we conducted multiple imputation as a sensitivity analysis under the assumption of missing at random to impute for missing exposure variables and covariates using multiple imputations with chained equations (MICE), generating 20 datasets [50, 51]. The imputation model included all covariates and exposures considered in the analysis. Accordingly, we performed analyses using (1) complete-case data, excluding participants with missing values for variables in the adjusted model (n = 5845), and (2) multiply imputed data (n = 6073). Because the missing-at-random assumption underlying multiple imputation is inherently untestable and may not hold for outcomes, we conducted extreme-case sensitivity analyses using four alternative models [52, 53]. Two assumed that all participants lost to follow-up experienced the event: at Day 1 (early-failure, worst case) or at the last follow-up day (late-failure). The other two assumed no events among missing participants, censoring them either at Day 1 (minimal follow-up) or at the last day (maximal follow-up). This range of best-worst case assumptions is recommended for testing robustness under extreme missing-data conditions. Associations of TyG- obesity indices with the outcome were then re-analyzed in the multiply imputed dataset (n = 7545) under each scenario.

All the analyses were performed using STATA version 14 SE (Stata Corp LP, TX, USA) and R version 4.4.0. A two-tailed P value of less than 0.05 was considered statistically significant.

Results

Baseline characteristics of respondents and non-respondents are shown in Table S1. The age and proportion of women were comparable between the two groups. Non-respondents, however, included fewer married individuals, a higher proportion of current smokers, and more participants with general obesity, but fewer participants with dyslipidemia. Mean values of anthropometric indices, FPG, 2 h-PG, SBP, DBP, lipid parameters, and TyG and TyG-obesity indices did not differ significantly between the two groups.

The study included 5845 participants (3158 women) with a mean age of 44.5 ± 11.1 years (Table 1). Participants in higher TyG tertiles were older and had higher BMI, WC, WHtR, WHR, FPG, SBP, DBP, TC, TG, and LDL-C, while HDL-C and eGFR levels were lower. Additionally, the proportion of current smokers and individuals with a family history of diabetes was higher in the upper tertiles. Compared to the lowest tertile, the higher tertiles included a greater proportion of married individuals and those with less than 6 years of education. Participants in higher TyG tertiles also had a greater prevalence of general obesity and central obesity, with increasing percentages across the tertiles (all P < 0.05).

Table 1.

Characteristics of the study participants: Tehran lipid and glucose study 1999–2018

Variables	Overall n = 5845	Tertiles of TyG			P value
		Tertile 1 n = 1949	Tertile 2 n = 1948	Tertile 3 n = 1948
Continuous variables, mean (SD)
Age, mean (SD), year	44.53 (11.13)	42.46 (11.28)	45.29 (11.20)	45.83 (10.59)	< 0.001
BMI, mean (SD), kg/m2	27.21 (4.43)	25.45 (4.23)	27.55 (4.29)	28.63 (4.16)	< 0.001
WC, mean (SD), cm	89.45 (11.26)	83.85 (10.71)	90.45 (10.28)	94.04 (10.33)	< 0.001
WHtR, mean (SD)	0.55 (0.07)	0.51 (0.07)	0.55 (0.07)	0.57 (0.06)	< 0.001
WHR, mean (SD)	0.88 (0.08)	0.84 (0.08)	0.88 (0.07)	0.91 (0.07)	< 0.001
FPG, mean (SD), mg/dL	90.55 (9.59)	86.73 (8.38)	90.43 (8.90)	94.50 (9.82)	< 0.001
2h-PG, mean (SD), mg/dL	108.48 (29.30)	99.52 (25.99)	108.46 (28.65)	117.46 (30.31)	< 0.001
SBP, mean (SD), mmHg	117.97 (17.26)	112.79 (15.84)	118.57 (16.87)	122.55 (17.61)	< 0.001
DBP, mean (SD), mmHg	77.68 (10.49)	74.27 (9.88)	77.98 (10.34)	80.78 (10.22)	< 0.001
Pulse rate, mean (SD), beats/min	78.63 (11.29)	78.57 (11.47)	78.18 (11.14)	79.15 (11.23)	0.03
eGFR, mean (SD), mL/min/1.73 m2	83.13 (12.42)	84.63 (12.36)	82.98 (12.54)	81.77 (12.18)	< 0.001
TC, mean (SD), mg/dL	208.61 (43.14)	187.99 (35.31)	208.31 (37.59)	229.54 (45.41)	< 0.001
HDL-C, mean (SD), mg/dL	41.55 (10.90)	46.76 (11.24)	41.06 (9.99)	36.83 (8.97)	< 0.001
LDL-C, mean (SD), mg/dL	133.30 (34.80)	118.47 (29.43)	135.76 (31.58)	145.68 (37.30)	< 0.001
TG, median [IQR], mg/dL	147 (101–207)	87 (71–102)	147 (129–164)	243 (206–309)	< 0.001
TyG, mean [SD]	8.79 (0.56)	8.18 (0.26)	8.78 (0.13)	9.41 (0.33)	< 0.001
TyG-BMI, mean [SD]	240.14 (45.89)	208.63 (36.45)	242.22 (38.35)	269.59 (40.67)	< 0.001
TyG-WC, mean [SD]	789.34 (126.97)	687.30 (95.20)	795.09 (92.48)	885.69 (104.98)	< 0.001
TyG-WHtR, mean [SD]	4.87 (0.81)	4.25 (0.62)	4.91 (0.63)	5.44 (0.68)	< 0.001
TyG-WHR, mean [SD]	7.78 (1.02)	6.92 (0.74)	7.82 (0.69)	8.62 (0.81)	< 0.001
Categorical variables, number (%)
Women	3158 (54.03)	1207 (61.93)	1032 (52.98)	919 (47.18)	< 0.001
Married	5275 (90.25)	1709 (87.69)	1783 (91.53)	1783 (91.53)	< 0.001
Educational level (y)					< 0.001
< 6	1989 (34.03)	546 (28.01)	694 (35.63)	749 (38.45)
6–12	3035 (51.92)	1098 (56.34)	996 (51.13)	941 (48.31)
>12	821 (14.05)	305 (15.65)	258 (13.24)	258 (13.24)
Family history of CVD	935 (16.00)	300 (15.39)	307 (15.76)	328 (16.84)	0.44
Family history of diabetes	1550 (26.52)	476 (24.42)	514 (26.39)	560 (28.75)	0.01
Lipid-lowering drug use	124 (2.12)	8 (0.41)	35 (1.80)	81 (4.16)	< 0.001
Anti-hypertension drug use	272 (4.65)	42 (2.15)	99 (5.08)	131 (6.72)	< 0.001
Smoking status					0.001
Never	4410 (75.45)	1537 (78.86)	1452 (74.54)	1421 (72.95)
Former	440 (7.53)	125 (6.41)	152 (7.80)	163 (8.37)
Current	995 (17.02)	287 (14.73)	344 (17.66)	364 (18.69)
General obesity	1408 (24.09)	273 (14.01)	496 (25.46)	639 (32.80)	< 0.001
Central obesity	1918 (32.81)	338 (17.34)	662 (33.98)	918 (47.13)	< 0.001
Prediabetes	1445 (24.72)	246 (12.62)	441 (22.64)	758 (38.91)	< 0.001
Dyslipidemia	3921 (67.08)	743 (38.12)	1310 (67.25)	1868 (95.89)	< 0.001
Hypertension	2711 (46.38)	636 (32.63)	920 (47.23)	1155 (59.29)	< 0.001

Bold values indicate P < 0.05

Data were given as mean ± SD or median [IQR] or number (%), except for skewed variables (i.e., TG)

TyG, Triglyceride-glucose index; BMI, body mass index; WC, waist circumference; WHtR, waist-height ratio; WHR, waist-hip ratio; FPG, fasting plasma glucose; 2 h-PG, 2-hour post-challenge glucose; SBP, systolic blood pressure; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; TG, triglycerides; and CVD, cardiovascular disease

During a median follow-up of 15.3 years (IQR: 11.9–16.4), 344 (5.9%) participants developed CRMM. The distribution of CRMM events is shown in Fig. 2, illustrating the different combinations of CKD, T2DM, and CVD among individuals who developed multimorbidity. Among those with incident CRMM, the combinations of coexisting CRM morbidities, ordered by frequency, were: (1) CVD and T2DM (48.8%), (2) CKD and T2DM (19.5%), (3) CVD and CKD (19.2%), and (4) CVD, CKD and T2DM (12.5%).

Fig. 2.

Incidence of cardio-renal-metabolic multimorbidity. CKD, Chronic kidney disease; T2DM, type 2 diabetes mellitus; CVD, cardiovascular disease; CRM, cardio-renal-metabolic

Figure 3 shows the proportion of CRMM events across TyG tertiles in age (< 45, 45–59, and ≥ 60 years) and sex subgroups. In all subgroups, the highest TyG tertile had a higher incidence of CRMM than the first tertile. Figure 4 presents Kaplan–Meier curves for the cumulative incidence of CRMM across tertiles (T1–T3) of TyG and TyG-obesity indices. For all indices, incidence was highest in T3 and lowest in T1, with the curves progressively separating over time, indicating greater cumulative incidence in higher tertiles throughout the follow-up period. Survival curves differed significantly over the follow-up (log-rank test, P < 0.001 for all comparisons).

Fig. 3.

CRMM event rates stratified by tertiles (T1–T3) of TyG, TyG-BMI, TyG-WC, TyG-WHtR, and TyG-WHR across age groups (30–44, 45–59, and 60–82 years), shown separately for men and women. CRMM, Cardio-renal-metabolic multimorbidity; TyG, triglyceride-glucose index; BMI, body mass index; WC, waist circumference; WHtR, waist-height ratio; WHR, waist-hip ratio

Fig. 4.

Cumulative incidence of CRMM stratified by tertiles (T1–T3) of TyG, TyG-BMI, TyG-WC, TyG-WHtR, and TyG-WHR, based on Kaplan-Meier survival analysis. CRMM, Cardio-renal-metabolic multimorbidity; TyG, triglyceride-glucose index; BMI, body mass index; WC, waist circumference; WHtR, waist-height ratio; WHR, waist-hip ratio

As shown in Table 2, after adjustment for age, sex, marital status, educational level, smoking status, SBP, antihypertensive medication use, HDL-C, LDL-C, lipid-lowering medication use, eGFR, pulse rate, family history of diabetes, family history of premature CVD, and prediabetes (Model 2), each SD increase in TyG, TyG-BMI, TyG-WC, TyG-WHtR, and TyG-WHR was associated with a higher risk of incident CRMM, with HRs (95% CI) of 1.41 (1.24–1.60), 1.52 (1.36–1.71), 1.57 (1.38–1.78), 1.57 (1.38–1.78), and 1.42 (1.24–1.63), respectively. Compared to the first tertile, HRs (95% CI) for tertile 2 and tertile 3 were 1.41 (1.00–2.00) and 1.82 (1.27–2.59) for TyG, 1.88 (1.32–2.66) and 2.57 (1.81–3.65) for TyG-BMI, 2.14 (1.44–3.20) and 3.07 (2.07–4.57) for TyG-WC, 1.76 (1.21–2.57) and 2.77 (1.91–4.03) for TyG-WHtR, and 2.05 (1.35–3.13) and 2.93 (1.91–4.49) for TyG-WHR, respectively. Significant positive trends were observed for all TyG and TyG-obesity indices (all P for trend < 0.01). After adjusting for HOMA-IR in a subsample of TLGS participants with available serum insulin data (Model 3), the significant trend remained persistent for all TyG-obesity indices, but not for TyG. Sensitivity analyses accounting for the competing risk of death (Model 4) and the exclusion of events within the first 5 years of follow-up (Model 5) did not substantially alter the results.

Table 2.

Associations of TyG and TyG-related indices with incident cardio-renal-metabolic Multimorbidity

	Categorical							Continuous
	Tertile 1		Tertile 2		Tertile 3		P trend	Per 1 SD	P value
	HR (95% CI)	P value	HR (95% CI)	P value	HR (95% CI)	P value
TyG
No of cases/total	51/1949	…	103/1948	…	190/1948	…	…	344/5845	…
Model 1	Ref (1.00)	…	1.85 (1.32–2.59)	< 0.001	3.30 (2.42–4.49)	< 0.001	< 0.001	1.68 (1.51–1.86)	< 0.001
Model 2	Ref (1.00)	…	1.41 (1.00–2.00)	0.049	1.82 (1.27–2.59)	0.001	0.001	1.41 (1.24–1.60)	< 0.001
Model 3	Ref (1.00)	…	1.24 (0.77–1.98)	0.363	1.34 (0.81–2.22)	0.252	0.274	1.23 (1.02–1.50)	0.034
Model 4	Ref (1.00)	…	1.36 (0.96–1.93)	0.078	1.85 (1.30–2.65)	0.001	< 0.001	1.43 (1.25–1.64)	< 0.001
Model 5	Ref (1.00)	…	1.36 (0.96–1.94)	0.080	1.69 (1.18–2.43)	0.004	0.004	1.35 (1.18–1.54)	< 0.001
TyG-BMI
No of cases/total	47/1949	…	116/1948	…	181/1948	…	…	344/5845	…
Model 1	Ref (1.00)	…	2.51 (1.79–3.53)	< 0.001	4.47 (3.22–6.20)	< 0.001	< 0.001	1.84 (1.67–2.04)	< 0.001
Model 2	Ref (1.00)	…	1.88 (1.32–2.66)	< 0.001	2.57 (1.81–3.65)	< 0.001	< 0.001	1.52 (1.36–1.71)	< 0.001
Model 3	Ref (1.00)	…	1.89 (1.16–3.08)	0.010	2.28 (1.37–3.79)	0.001	0.002	1.41 (1.18–1.69)	< 0.001
Model 4	Ref (1.00)	…	1.92 (1.35–2.72)	< 0.001	2.60 (1.83–3.69)	< 0.001	< 0.001	1.49 (1.33–1.68)	< 0.001
Model 5	Ref (1.00)	…	1.94 (1.35–2.77)	< 0.001	2.61 (1.81–3.76)	< 0.001	< 0.001	1.50 (1.33–1.69)	< 0.001
TyG-WC
No of cases/total	33/1949	…	104/1948	…	207/1948	…	…	344/5845	…
Model 1	Ref (1.00)	…	2.64 (1.78–3.92)	< 0.001	5.18 (3.58–7.49)	< 0.001	< 0.001	1.92 (1.72–2.14)	< 0.001
Model 2	Ref (1.00)	…	2.14 (1.44–3.20)	< 0.001	3.07 (2.07–4.57)	< 0.001	< 0.001	1.57 (1.38–1.78)	< 0.001
Model 3	Ref (1.00)	…	2.02 (1.19–3.41)	0.008	2.20 (1.28–3.79)	0.004	0.009	1.39 (1.13–1.69)	0.001
Model 4	Ref (1.00)	…	2.25 (1.51–3.35)	< 0.001	3.13 (2.11–4.67)	< 0.001	< 0.001	1.54 (1.36–1.76)	< 0.001
Model 5	Ref (1.00)	…	2.03 (1.36–3.05)	0.001	2.99 (2.00-4.47)	< 0.001	< 0.001	1.53 (1.34–1.74)	< 0.001
TyG-WHtR
No of cases/total	40/1949	…	95/1948	…	209/1948	…	…	344/5845	…
Model 1	Ref (1.00)	…	2.16 (1.49–3.13)	< 0.001	4.71 (3.33–6.65)	< 0.001	< 0.001	1.93 (1.73–2.15)	< 0.001
Model 2	Ref (1.00)	…	1.76 (1.21–2.57)	0.003	2.77 (1.91–4.03)	< 0.001	< 0.001	1.57 (1.38–1.78)	< 0.001
Model 3	Ref (1.00)	…	1.45 (0.88–2.39)	0.141	1.96 (1.18–3.26)	0.009	0.008	1.35 (1.10–1.64)	0.003
Model 4	Ref (1.00)	…	1.77 (1.21–2.59)	0.003	2.83 (1.95–4.11)	< 0.001	< 0.001	1.55 (1.36–1.77)	< 0.001
Model 5	Ref (1.00)	…	1.73 (1.17–2.54)	0.005	2.74 (1.87–4.01)	< 0.001	< 0.001	1.54 (1.35–1.76)	< 0.001
TyG-WHR
No of cases/total	30/1949	…	104/1948	…	210/1948	…	…	344/5845	…
Model 1	Ref (1.00)	…	2.82 (1.87–4.25)	< 0.001	5.11 (3.43–7.62)	< 0.001	< 0.001	1.78 (1.58–1.99)	< 0.001
Model 2	Ref (1.00)	…	2.05 (1.35–3.13)	0.001	2.93 (1.91–4.49)	< 0.001	< 0.001	1.42 (1.24–1.63)	< 0.001
Model 3	Ref (1.00)	…	1.60 (0.94–2.73)	0.082	1.89 (1.08–3.31)	0.025	0.032	1.11 (0.90–1.37)	0.289
Model 4	Ref (1.00)	…	2.09 (1.36–3.21)	0.001	2.91 (1.87–4.51)	< 0.001	< 0.001	1.44 (1.25–1.64)	< 0.001
Model 5	Ref (1.00)	…	2.07 (1.36–3.16)	0.001	2.83 (1.83–4.36)	< 0.001	< 0.001	1.38 (1.20–1.59)	< 0.001

Bold values indicate P < 0.05

Model 1 was adjusted for age and sex.

Model 2 was additionally adjusted for marital status, educational level, smoking status, systolic blood pressure, antihypertensive medication use, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, lipid-lowering medication use, estimated glomerular filtration rate, pulse rate, family history of diabetes, family history of premature CVD, and prediabetes.

Model 3 incorporated the Model 2 covariates plus HOMA-IR in a subsample (n = 3197) with available serum insulin data.

Model 4 used the Model 2 covariates but accounted for competing risk of death.

Model 5 used the Model 2 covariates but excluded events within the first five years of follow-up.

TyG, Triglyceride-glucose index; BMI, body mass index; WC, waist circumference; WHtR, waist-height ratio; WHR, waist-hip ratio; HR, hazard ratio; CI, confidence interval; CVD, cardiovascular disease; HOMA-IR, homeostasis model assessment of insulin resistance

As presented in Table S2, subtype-specific analyses indicated that higher levels of TyG and TyG-obesity indices were all associated with a 35–75% increased risk of CRMM subtypes of CVD & T2DM and CKD & T2DM (all P < 0.05). For CVD and CKD, TyG-BMI (HR, 1.24; 95% CI 1.01–1.53) and TyG-WC (HR, 1.25; 95% CI 1.00–1.56) reached statistical significance. For the triple-outcome subtype, significant associations were observed for TyG (HR, 1.51; 95% CI 1.05–2.16) and TyG-BMI (HR, 1.40; 95% CI 1.02–1.93), whereas the other TyG-obesity indices were associated with a 21–39% higher risk, although not statistically significant.

A significant linear association was observed between the TyG index and TyG-obesity indices and the risk of CRMM, with all P values for overall association < 0.001 and no evidence of nonlinearity (all P for nonlinearity > 0.05), as shown in Fig. 5.

Fig. 5.

Associations of TyG, TyG-BMI, TyG-WC, TyG-WHtR, and TyG-WHR with the risk of CRMM, estimated using restricted cubic spline models. Shaded areas represent 95% confidence intervals. Models are adjusted for age, sex, marital status, educational level, smoking status, systolic blood pressure, antihypertensive medication use, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, lipid-lowering medication use, estimated glomerular filtration rate, pulse rate, family history of diabetes, family history of premature CVD, and prediabetes. For all parameters, P_overall < 0.05 and P_nonlinear > 0.05. Abbreviations: CRMM, cardio-renal-metabolic multimorbidity; TyG, triglyceride-glucose index; BMI, body mass index; WC, waist circumference; WHtR, waist-height ratio; WHR, waist-hip ratio

Table S3 compares the HRs for TyG and its combinations with obesity measures in relation to CRMM. After adjusting for TyG in multivariable models, the associations of TyG-BMI, TyG-WC, TyG-WHtR, and TyG-WHR with incident CRMM remained statistically significant. When directly compared, the effect sizes of TyG-obesity parameters were not significantly larger than that of TyG alone (all P differences ≥ 0.05).

We also assessed the incremental predictive value of TyG and TyG-related indices beyond the conventional risk factors (Table S4). In Analysis 1, the Harrell’s C-index (95% CI) improved from 0.824 (0.803–0.845) in the base model (Model 2), to 0.829 (0.808–0.850) with TyG, 0.833 (0.812–0.854) with TyG-BMI, 0.834 (0.814–0.855) with TyG-WC, 0.832 (0.812–0.853) with TyG-WHtR, and 0.831 (0.811–0.852) with TyG-WHR. Notably, the increments in Harrell’s C-index (ΔC) were statistically significant for TyG (ΔC = 0.005, P = 0.040), TyG-BMI (ΔC = 0.008, P = 0.013), TyG-WC (ΔC = 0.010, P = 0.009), TyG-WHtR (ΔC = 0.008, P = 0.022), and TyG-WHR (ΔC = 0.007, P = 0.019). In terms of discrimination improvement, TyG-BMI, TyG-WC, and TyG-WHtR showed significant IDI gains, whereas TyG and TyG-WHR did not reach statistical significance. In Analysis 2, where TyG was included in the base model, further addition of anthropometric-based indices yielded no significant improvements in predictive performance (all P > 0.05).

The results of the subgroup analyses are presented in Fig. 6 and Tables S5–S9. A significant interaction between obesity and TyG-WHR was observed (P for interaction < 0.001), with a stronger association in non-obese individuals, whereas the association among those with obesity was positive but not statistically significant. In addition, a marginally significant effect modification by obesity status was noted for TyG, with a more pronounced effect in non-obese individuals (P for interaction = 0.005).

Fig. 6.

Subgroup analysis of the association between TyG, TyG-BMI, TyG-WC, TyG-WHtR, and TyG-WHR and incident CRMM. Cox regression, adjusted for age, sex, marital status, educational level, smoking status, systolic blood pressure, antihypertensive medication use, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, lipid-lowering medication use, estimated glomerular filtration rate, pulse rate, family history of diabetes, family history of premature CVD, and prediabetes, was performed in subgroups defined by age, sex, smoking status (current vs. former/never), dyslipidemia, prediabetes, and obesity (general or central obesity vs. non-obese). Abbreviations: CRMM, cardio-renal-metabolic multimorbidity; TyG, triglyceride-glucose index; BMI, body mass index; WC, waist circumference; WHtR, waist-height ratio; WHR, waist-hip ratio, CI, confidence interval

Results of the sensitivity analysis with imputed values for the main exposures and covariates are presented in Table S10. Overall, the findings were largely consistent with those from the complete-case analysis in both the fully adjusted model and the model that accounted for HOMA-IR. Notably, the association between TyG-WHR and incident CRMM reached statistical significance in the model that included HOMA-IR. Worst-best case sensitivity analyses indicated that the associations between the TyG index and TyG-related obesity indices with incident CRMM were broadly consistent with the main results. Across the four scenarios, HRs remained directionally and significantly in line with the complete-case analysis, with only the early-failure scenario showing attenuation, while estimates consistently remained above 1 (Table S11).

Discussion

In a community-based cohort study involving Iranian adults initially free from CVD, T2DM, and CKD, followed for nearly two decades, we identified the following key findings:

1. A significant linear relationship was observed between all TyG-obesity indices and CRMM, with no evidence of nonlinearity. Each 1-SD increase in TyG, TyG-BMI, TyG-WC, TyG-WHtR, and TyG-WHR was associated with 41%, 52%, 57%, 57%, and 42% higher risk of CRMM, respectively, after adjusting for a comprehensive set of confounders. These associations remained robust across multiple sensitivity analyses, including additional adjustment for HOMA-IR and consideration of the competing risk of death.

2. The association of TyG-obesity indices with incident CRMM was comparable to that of the TyG index alone.

3. The association between TyG-WHR and CRMM was more pronounced among non-obese than obese individuals.

The TyG index, a surrogate marker of insulin resistance, has been strongly associated with cardiometabolic diseases such as T2DM and CVD, as well as renal disorders including CKD [8–10]. Prior research has highlighted both the progression of CVD in patients with CKD [54], and the development of CKD in individuals with pre-existing CVD [55]. The TyG index has also been associated with CVD mortality in individuals with T2DM [56]. Zhang et al., using data from the National Health and Nutrition Examination Survey (NHANES) 2009–2018, found that TyG-related indices were significantly associated with increased risks of all-cause and CVD mortality [57]. TyG-BMI was shown to be associated with the risk of CKD onset [58]. Chen et al. also found a U-shaped association between TyG-BMI and both all-cause and CVD mortality in individuals with CKD [59]. Furthermore, Li et al., using data from the CHARLS study, found that among individuals at cardiovascular-kidney-metabolic (CKM) syndrome stages 0–3, as defined by the American Heart Association (AHA) to include those with or without metabolic risk factors and CKD, each 10-unit increase in TyG-BMI was associated with a 6.5% higher risk of cardiovascular disease [60].

Few studies have evaluated the role of TyG indices in the context of CRM multimorbidity. Liu et al., using data from the CHARLS study (2011–2020), followed 7848 adults aged 45 years or older for a median of 9 years to examine the association between TyG indices and CRMM. They found that higher levels of TyG, TyG-WC, TyG-WHtR, and TyG-BMI were significantly associated with an increased risk of CRMM, with corresponding HRs for the highest versus lowest quartile of 1.53, 1.42, 1.51, and 1.40, respectively [12]. This study found a significant linear dose-response relationship for TyG, while TyG-WC, TyG-WHtR, and TyG-BMI exhibited non-linear associations with CRMM risk [12]. Tang et al., using data from the UK Biobank (2006–2020), followed 349,974 adults free of cardiovascular, renal, or metabolic diseases at baseline for a median of approximately 14 years to examine the association between the TyG-BMI and the progression of CRMM [13]. Each SD increase in TyG-BMI was independently associated with a 32% higher risk of the first CRM disease, a 24% higher risk of progressing to double CRM disease, and a 23% higher risk of developing triple CRM disease. Restricted cubic spline models showed a clear dose-response relationship [13]. In the current study, we found that the TyG index and TyG-related indices were all significantly associated with CRMM, and these associations were linear. Similar to Tian et al., [19] we observed long-term divergence in CRMM incidence across TyG tertiles, with persistently higher risk in those with elevated values, underscoring the TyG index as a simple marker for sustained CRMM risk. Subtype-specific analyses indicated that all TyG indices were associated with outcomes involving diabetes combined with either CVD or CKD. However, for the CVD and CKD subtype, only TyG-BMI and TyG-WC reached statistical significance. For the triple-outcome combination, TyG conferred the highest risk, followed by TyG-BMI.

Central obesity is strongly associated with the development of IR, systemic inflammation, and oxidative stress, all of which play a pivotal role in the progression of cardio-renal-metabolic diseases [7, 61]. WHtR and WC, as useful parameters for evaluating central obesity, have shown superior discriminatory power in identifying cardiometabolic risk compared to BMI [62, 63]. It has been shown that the association of overweight and obesity with CVD is substantially mediated by the TyG index, accounting for nearly half of the excess risk for overweight (47.8%) and about one-third for general (37.9%) and central obesity (31.1–35.0%) [64]. Several studies suggest that combining the TyG index with obesity-related measures may provide a more precise assessment of insulin resistance than the TyG index alone [65]. However, findings for TyG-related indices (TyG-BMI, TyG-WC, and TyG-WHtR) have been inconsistent across disease outcomes and study designs. Comparison between studies is challenging due to differences in study populations, variations in follow-up duration, outcomes assessed, and approaches to confounder adjustment, as well as differences in inclusion criteria and baseline participant characteristics.

In the UK Biobank cohort of 374,274 individuals, Tian et al. reported that TyG-WC and TyG-WHtR showed superior predictive ability compared with TyG alone for cardiometabolic multimorbidity [19]. Park et al. reported that TyG-WC and TyG-WHtR showed superior predictive performance for incident CVD compared to TyG, but only among participants without diabetes [66]. Moreover, in a cohort of 9432 U.S. adults with hypertension, TyG-WC and TyG-WHtR showed predictive ability for cardiovascular mortality comparable to the TyG index [67]. However, in a retrospective cohort of Chinese adults, while TyG-WHtR was independently associated with incident CVD, the TyG index consistently outperformed TyG-obesity indices in predicting CVD, including CHD and stroke [68]. Furthermore, another study of 24,215 normal-weight Chinese elderly individuals showed that the TyG index had the highest area under the curve (AUC) for predicting T2DM risk, compared with TyG-BMI, TyG-WC, and TyG-WHtR [69]. Feng et al. evaluated 17 obesity and IR-related indicators in relation to hypertension, dyslipidemia, diabetes, and multimorbidity, and identified the TyG index as one of the strongest predictors [70].

In our study, although TyG-BMI, TyG-WC, and TyG-WHtR yielded statistically significant gains in discrimination compared to the conventional model, these improvements were minimal, and the addition of anthropometric indices to a model already containing TyG did not significantly improve predictive performance. This pattern suggests that TyG alone captures much of the relevant metabolic risk information, and that its obesity derivatives may not provide substantial added value in predicting CRMM risk.

We also found that the effect of TyG-WHR, and to a lesser extent TyG, was more pronounced among non-obese individuals. Meisinger et al. showed that higher risk of T2DM was associated with increasing WHR, and further adjustment for BMI considerably attenuated this relation in both men and women [71]. Using data from the National Health and Nutrition Examination Survey (NHANES) 2017–2018, Xie et al. examined the association between WHR and non-alcoholic fatty liver disease (NAFLD) using multivariate logistic regression and two-sample Mendelian randomization analyses, showing that this association was more pronounced at lower BMI values and strongest in non-obese individuals [72]. In non-obese individuals, the TyG index effectively detect subtle dysmetabolic states, such as impaired insulin signaling caused by the accumulation of free fatty acids, unlike in obese individuals, where excess body fat can obscure these findings and insulin resistance may not provide additional predictive value beyond obesity measures [73, 74]. Prior studies have similarly reported stronger TyG associations with incidence of diabetes in non-obese phenotypes [75, 76]. It has been shown that increasing HOMA-IR trajectory predicts incident CVD and mortality in non-obese individuals but not in those with obesity [77]. Overall, because obesity itself is a strong risk factor for developing CVD, T2DM, and CKD [78, 79], the incremental impact of TyG and TyG-WHR may be attenuated in obese individuals.

The TyG index is a non-insulin-based index that is less costly than insulin-based markers. Because it can be derived from a single sample, it is well-suited to clinical and epidemiological studies and shows promise as a screening marker across cardiac, renal, and metabolic disorders [9]. Although meta-analyses evaluating its diagnostic performance have increased recently, substantial heterogeneity and controversy persist among these studies. Notably, its diagnostic accuracy is not as high as that of the hyperinsulinemic–euglycemic clamp (HIEC) or HOMA-IR, and the overall quality of evidence supporting its use as a surrogate biochemical marker of insulin resistance is low to moderate [80]. A bivariate diagnostic test-accuracy (DTA) meta-analysis reported pooled sensitivity and specificity of 80% and 81% for screening metabolic syndrome, respectively [81]. More recently, a comprehensive meta-analysis underscored its value primarily as a prognostic marker of adverse cardiometabolic outcomes rather than as a precise diagnostic tool [9].

Clinical implications

The TyG index, derived from routine clinical measurements, may serve as a practical tool for early identification of individuals at high risk for multiple chronic diseases, particularly in resource-limited settings where advanced diagnostic tools may be unavailable [8, 13]. This is particularly relevant in the MENA region, where the burden of NCDs is rising and cost-effective screening tools are essential. Moreover, early detection and prevention of CRM diseases are crucial and could guide targeted interventions, such as lifestyle modifications (e.g., diet and regular physical activity), to prevent progression of CRM diseases [42]. This is especially important given the rising global burden of multimorbidity, which poses significant challenges to healthcare systems, especially in aging populations and low- to middle-income countries [1, 2]. Future studies should replicate these findings in more diverse ethnic and demographic groups to improve generalizability and further investigate inflammatory, metabolic, and adiposity-related mechanisms that may underlie the observed associations. Given that our cohort was free of T2DM and CVD and had preserved kidney function at baseline, additional research is warranted to evaluate the predictive value of the TyG index and its derivatives in higher-risk populations. Finally, evaluating the cost-effectiveness of using TyG-based screening in preventive healthcare, especially in resource-limited settings, may help inform clinical and policy decisions.

Highlights and limitations

Our study holds important public health relevance. We examined the role of the TyG index and its obesity-related derivatives in the development of CRMM within a prospective population-based cohort of Iranian adults followed for over two decades. In addition to evaluating the individual and comparative association of these indices, we performed sensitivity analyses by further adjusting for HOMA-IR in the subsample with available insulin data, serving as a robustness check for insulin resistance-related confounding. Moreover, consistent results were obtained in competing-risk models, in which 312 mortality events (approximately 5% of the cohort) were treated as competing events, likely reflecting the relatively young baseline age, the low competing-event rate, and the strength of the observed associations between TyG-related and CRMM. However, several limitations should be considered. First, the observational design of the study introduces the potential for residual confounding, and thus, causal inferences should be made with caution. Second, as the cohort was conducted in Tehran, the capital city of Iran, and comprised exclusively Iranian participants, the generalizability of our findings may be limited to other ethnic groups and to rural or non-urban settings. Third, a key limitation is the lack of validation in external cohorts, which may restrict the generalizability of our findings beyond our population. Nevertheless, TLGS is among the largest, long-standing, community-based prospective cohorts in the MENA region, with rigorous outcome assessment and repeated measurements over two decades [39, 82]. Future studies, particularly in other MENA populations, with serial follow-up assessments and a rigorous adjudication process, are warranted to replicate and extend the findings of the current study. However, previous studies have shown consistent results across subgroup analyses by race regarding the outcomes of cardiovascular disease [83], stroke [84], hypertension [85], and advanced cardiovascular-kidney-metabolic syndrome [86]. Fourth, although baseline characteristics were comparable between respondents and non-respondents, non-random attrition cannot be entirely ruled out and may have led to over- or under-representation of individuals at higher cardiometabolic risk, potentially affecting our findings. However, we performed multiple imputation at baseline, together with worst-best case sensitivity analyses for loss to follow-up, to assess the robustness of our results. Fifth, we were unable to perform formal mediation analyses to address the direct effect of TyG-related indices in the development of CRMM, because key potential biomarkers were not available, including inflammatory markers [87] such as high-sensitivity C-reactive protein (hsCRP) and tumor necrosis factor-α (TNF-α) [19], oxidative stress markers such as gamma-glutamyltransferase (GGT) and serum urate [19, 88], biomarkers of liver function [89], and renal and vascular function measures such as cystatin C [9, 90], serum urate [19, 88], and microalbuminuria [91, 92], which limited our ability to disentangle the biological pathways underlying the observed associations. Previous evidence highlights the potential mediating roles of systemic inflammation, oxidative stress, and hepatic dysfunction in linking higher TyG-related indices to adverse cardiometabolic outcomes [19, 87]. In the UK Biobank study, biomarkers of liver function, renal function, and systemic inflammation collectively explained about one-third of the associations of TyG-WC and TyG-WHtR with cardiometabolic multimorbidity risk [19]. Further studies are required to establish the mediating role of potential contributors in the association between TyG-obesity indices and incident CRMM. Lastly, the TyG index was calculated from single baseline measurements of triglycerides and fasting plasma glucose, which may not fully reflect intra-individual variability over time. Temporal fluctuations in these biomarkers, influenced by factors such as diet and stress, could introduce random error and attenuate the index’s precision as a surrogate marker of insulin resistance [93].

Conclusions

In this two-decade community-based cohort, higher levels of the TyG index and TyG-related indices were independently associated with an increased risk of CRMM. However, adding general or central anthropometric measures to the TyG index did not significantly improve its ability to predict CRMM risk beyond that of the TyG index alone.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

S.M. conceived and designed the study. S.M., N.E., S.S., and F.H. acquired, analyzed, and interpreted the data. N.E., S.S., M.T., and F.H. drafted and revised the manuscript. All authors reviewed and approved the final manuscript.

Data availability

The datasets used during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Ethical approval of this study has been certified by the central institutional review board of the Research Institute of Endocrine Science, Shahid Beheshti University of Medical Science, Tehran, Iran. The outlined principles in the Declaration of Helsinki have been respected in this study. All participants have signed written consent.

Competing interests

The authors declare no competing interests.