Longitudinal triglyceride-glucose index trajectories and kidney outcomes in patients with metabolic dysfunction-associated fatty liver disease

Article information

Korean J Nephrol. 2026;.j.krcp.25.246

Publication date (electronic) : 2026 March 6

doi : https://doi.org/10.23876/j.krcp.25.246

Jong Hyun Jhee¹, Wooram Bae¹, Hyo Jeong Kim¹, Hye Sun Lee², Shinae Kang³, Jung Il Lee⁴, Hyeong Cheon Park^1,5, Hoon Young Choi^,1,5

¹Division of Nephrology, Department of Internal Medicine, Gangnam Severance Hospital, Yonsei University College of Medicine, Seoul, Republic of Korea

²Biostatistics Collaboration Unit, Department of Research Affairs, Yonsei University College of Medicine, Seoul, Republic of Korea

³Division of Endocrinology, Department of Internal Medicine, Yonsei University College of Medicine, Seoul, Republic of Korea

⁴Division of Gastroenterology, Department of Internal Medicine, Yonsei University College of Medicine, Seoul, Republic of Korea

⁵Severance Institute for Vascular and Metabolic Research, Yonsei University College of Medicine, Seoul, Republic of Korea

Correspondence: Hoon Young Choi Department of Internal Medicine, Gangnam Severance Hospital, Yonsei University College of Medicine, 211 Eonju-ro, Gangnam-gu, Seoul 06273, Republic of Korea. E-mail: hychoidr@yuhs.ac

Received 2025 July 24; Revised 2025 November 17; Accepted 2025 December 8.

Abstract

Introduction

Metabolic dysfunction-associated fatty liver disease (MAFLD) is a recently proposed diagnostic entity that has attracted considerable attention due to its superior ability to identify individuals at elevated risk for chronic kidney disease (CKD) compared to the traditional concept of nonalcoholic fatty liver disease (NAFLD) [1,2]. Unlike NAFLD, which is defined by the exclusion of alcohol and other liver diseases, MAFLD is diagnosed based on the presence of hepatic steatosis together with metabolic dysfunction, such as overweight/obesity, type 2 diabetes mellitus (DM), or metabolic abnormalities [3–5]. This positive diagnostic framework identifies a broader and more metabolically active population at higher risk for cardiovascular and kidney complications. In addition, it reinforces the concept that metabolic dysfunction represents a central mechanism underlying extrahepatic diseases such as CKD [5,6]. Given that metabolic abnormalities are key drivers of CKD onset and progression, understanding the metabolic pathways linking MAFLD and CKD has become increasingly important. CKD affects more than 10% of the global population and remains a progressive, irreversible condition that substantially increases cardiovascular morbidity and premature mortality, underscoring the need for early identification and prevention [7–10].

In this context, the triglyceride-glucose (TyG) index has emerged as a reliable surrogate marker for insulin resistance, a central feature of metabolic syndrome and a key driver of atherosclerosis-related cardiovascular disease [11–15]. The TyG index is particularly relevant to CKD because it reflects underlying metabolic disturbances implicated in kidney damage. Elevated TyG levels are associated with various pathophysiological processes that aggravate kidney injury, such as increased glomerular pressure, hyperfiltration, and activation of pro-inflammatory and pro-fibrotic pathways in renal tissues, thereby contributing to CKD progression [16,17]. Despite its potential, the predictive utility of the TyG index in CKD, especially among individuals with MAFLD, remains unclear.

Therefore, this study aimed to comprehensively investigate the association between the longitudinal trajectory of the TyG index over time and adverse kidney outcomes in individuals with MAFLD.

Methods

Study subjects

This study used data from the Gangnam Severance Medical Cohort (GSMC, 2006–2021). The GSMC is a retrospective cohort study of patients who visited the Gangnam Severance Hospital (a tertiary university hospital in Seoul, Republic of Korea) with at least one metabolic derangement, including DM, fatty liver, and hypertension (HTN). The aim of establishing the GSMC was to identify risk factors and assess the outcomes of metabolically unhealthy individuals in the urban areas of South Korea. The GSMC comprised the subcohorts DM (n = 2,021), fatty liver (n = 4,287), and HTN (n = 2,709). The subcohorts were determined by the disease category based on the diagnosis of each specialist (endocrinologist, gastroenterologist, and cardiologist) or by the prescription of medication. Finally, 9,017 participants who regularly visited the Gangnam Severance Hospital every 3 to 6 months between 2006 and 2021 were enrolled in the GSMC. For this study, participants aged ≤18 years, with missing data, with a baseline estimated glomerular filtration rate (eGFR) of less than 15 mL/min/1.73 m², on chronic dialysis or having undergone kidney transplantation, or with a follow-up duration of less than 5 years or event monitoring period shorter than 90 days were excluded (n = 3,757). Among the remaining participants, 868 individuals diagnosed with MAFLD were included in the final analysis (Fig. 1).

Figure 1.

Flowchart of the study population.

eGFR, estimated glomerular filtration rate; MAFLD, metabolic dysfunction-associated fatty liver disease.

^aDiagnostic criteria of MAFLD: hepatic steatosis confirmed by histology, imaging, or biomarkers + at least one criterion of 1) overweight/obesity (body mass index >23 kg/m²), 2) type 2 diabetes mellitus, and 3) two or more metabolic dysregulation.

The study protocol was approved by the Institutional Review Board of the Gangnam Severance Hospital, Yonsei University College of Medicine (No. 3-2021-0254). The requirement for written informed consent was waived due to the study’s retrospective nature.

Exposure

Participants were classified into two TyG index trajectory groups. The longitudinal trajectory of TyG index levels was assessed by latent variable mixture modeling based on changes in the TyG index levels during the exposure period (from baseline to 4th year) (Fig. 2). Latent variable mixture modeling was used to identify homogeneous patterns in time-serial measurements and to classify individuals into groups with similar trajectories, known as latent classes. For a given variable, the latent variable mixture model estimates the probability that each individual belongs to a particular latent class and assigns them accordingly [20]. For this study, we used a growth mixture model that included more than one latent class and a random effect. Time was considered as a fixed effect and the individual as a random effect [21,22]. Assuming the existence of multiple trajectory groups for which probabilities can be estimated, we created models with different numbers of trajectory groups (R version 4.1.2, LCMM packages; R Foundation for Statistical Computing). The spline function was used as a link function to fit the data to the trajectories. Model fit was evaluated based on the following criteria. The Bayesian information criterion (BIC) was used to assess the model performance. Each group was required to comprise at least 5% of the total participants. We tested models with two to four trajectory groups and compared their fit (Supplementary Table 1, available online). Although the model with three trajectory groups produced the lowest BIC values, two of the groups contained fewer than 5% of participants. Therefore, a two-group model was selected, identifying two distinct TyG index trajectories (decreasing or increasing) that met the criteria.

Figure 2.

Triglyceride-glucose (TyG) index trajectories by latent class linear mixed model.

Results

Baseline characteristics

The baseline characteristics of the study subjects are shown in Table 1. The mean age of study subjects was 52.3 ± 10.4 years and 504 (58.1%) were male. The mean TyG index was 9.1 ± 0.6, and the mean eGFR was 96.0 ± 15.1 mL/min/1.73 m². The TyG index trajectory during the 4-year exposure period yielded two groups: decreasing (426, 49.1%) vs. increasing (442, 50.9%). Between the two groups, there were no significant differences in age, sex, BMI, SBP, DBP, smoking or alcohol status, previous histories of HTN, DM, dyslipidemia, cardiovascular disease, or cerebrovascular accidents. In laboratory data, the levels of TyG index, total cholesterol, triglyceride, fasting glucose, and hemoglobin A1c were lower in the increasing TyG trajectory group than in the decreasing group at baseline. However, the levels of hemoglobin, eGFR, HDL-C, LDL-C, total bilirubin, aspartate transferase, alanine transaminase, gamma-glutamyl transferase, and CRP were similar between the two groups.

Table 1.

Baseline characteristics

Association between triglyceride-glucose index trajectories and kidney outcomes

During a median follow-up of 6.9 years (4.0–10.0 years) and 5,495 per 1,000 person-years (PYs), a total of 36 kidney outcome events were identified, comprising 26 cases of incident eGFR <60 mL/min/1.73 m² and 10 cases of ≥30% decline in eGFR. No participants progressed to end-stage kidney disease requiring dialysis or transplantation during the observation period. The incidence rates per 1,000 PYs were 4.02 and 9.06 in the decreasing and increasing TyG index trajectory groups, respectively (p = 0.02) (Table 2). Kaplan-Meier analysis showed a significantly higher risk of kidney outcomes in the increasing trajectory group (Fig. 3). To assess the independent predictive value of the TyG index trajectory group for kidney outcomes, multivariable cause-specific Cox analyses were performed (Table 3). First, in unadjusted model 1, the increasing TyG index trajectory group showed 2.27-fold greater risk for kidney outcomes than the decreasing group (95% confidence interval [CI], 1.12–4.62; p = 0.02). In model 2, after adjusting for age and sex, the increasing trajectory group revealed a 2.31-fold higher risk for kidney outcomes (95% CI, 1.13–4.70; p = 0.02). In model 3, after adjusting for confounders including age, sex, BMI, SBP, histories of HTN, DM, and dyslipidemia, smoking and alcohol statuses, CRP, and baseline eGFR levels, the increasing trajectory group showed significantly higher risk for kidney outcomes (adjusted hazard ratio [HR], 3.68; 95% CI, 1.68–8.05; p = 0.001). Finally, in model 4, after fully adjusting for the baseline TyG index, the increasing trajectory group consistently showed a higher risk for kidney outcomes (adjusted HR, 4.49; 95% CI, 2.05–9.83; p < 0.001).

Table 2.

Incidence rates of kidney outcomes according to triglyceride-glucose index trajectory group

Figure 3.

Kaplan-Meier curve.

Table 3.

Association between TyG index trajectory groups and kidney outcomes

Subgroup analyses

To further explore whether longitudinal TyG index trajectories modify the effect on kidney outcomes across different subgroups, we conducted subgroup analyses (Fig. 4). The subgroups were stratified by age (≥60 years vs. <60 years), sex (male vs. female), BMI (≥25 kg/m² vs. <25 kg/m²), and the presence of HTN or DM. In subgroup analyses, none of the p for interaction terms between TyG index trajectories for kidney outcomes and the subgroups reached statistical significance. These findings suggest that the association between the increasing TyG index trajectory and the higher risk of kidney outcomes remained consistent across all subgroups.

Figure 4.

Subgroup analysis.

BMI, body mass index; CI, confidence interval; DM, diabetes mellitus; HR, hazard ratio; HTN, hypertension.

Discussion

In this retrospective cohort study of 868 patients with MAFLD, we investigated the association between longitudinal changes in the TyG index and adverse kidney outcomes. Using latent class trajectory modeling, we identified two distinct TyG index trajectories (decreasing and increasing). The increasing TyG index trajectory group exhibited a significantly higher risk of composite kidney outcomes than the decreasing group, even after adjusting for potential confounders, including demographic variables, metabolic comorbidities, and baseline eGFR or TyG index levels. This association remained consistent across subgroups defined by age, sex, BMI, and the presence of HTN or DM.

Our findings contribute to a growing body of evidence highlighting the role of metabolic dysfunction in CKD progression among patients with MAFLD. While the TyG index has been widely recognized as a surrogate marker for insulin resistance and metabolic dysregulation, its utility as a dynamic prognostic indicator for kidney health has been relatively underexplored, particularly in populations with MAFLD [23,24]. To the best of our knowledge, this is the first study to analyze longitudinal TyG trajectories in a MAFLD population and demonstrate a significant association with kidney outcomes. Although previous cross-sectional and short-term longitudinal studies have reported associations between elevated TyG index and reduced eGFR or increased risk of incident CKD in the general population, these studies have not adequately considered the dynamic nature of TyG index over time [25]. For instance, a large Japanese cohort study by Okamura et al. [26] found that a higher baseline TyG index was associated with a greater risk of incident CKD over a median follow-up of 4 years. In a cohort study with a median follow-up of 10 years, Wei et al. [27] reported that high baseline TyG index levels were associated with an increased risk of CKD, yet they did not account for the longitudinal changes in TyG index levels. More recently, Hou et al. [28] examined the relationship between longitudinal TyG index trajectories and incident CKD, showing that long-term exposure to high TyG index levels was associated with an increased risk of CKD. Nevertheless, this study did not stratify participants based on distinct metabolic phenotypes such as MAFLD [28]. Interestingly, in our study, the increasing TyG index trajectory group exhibited lower baseline TyG index, glucose, and HbA1c levels compared with the decreasing group. This finding suggests that individuals with initially favorable metabolic profiles may still experience metabolic deterioration over time, leading to an increased risk of kidney outcomes. In contrast, those in the decreasing trajectory group may have shown improvement from a higher baseline metabolic burden. Importantly, even after adjusting for baseline TyG index and histories of DM and dyslipidemia in the multivariable Cox model, the increasing trajectory group remained significantly associated with adverse kidney outcomes. These results underscore that the direction and progression of metabolic changes over time may be more informative than a single baseline value for predicting long-term kidney risk. Building upon this concept, our study fills an important gap by employing latent class trajectory modeling to capture dynamic changes in the TyG index and demonstrating that an increasing trend in TyG index over time is significantly associated with a higher risk of adverse kidney outcomes among individuals with MAFLD.

Several pathophysiological mechanisms may explain the observed association between the increasing TyG trajectory and kidney outcomes. First, the TyG index reflects both dyslipidemia and hyperglycemia, two key drivers of glomerular injury [29,30]. Persistent elevation in the TyG index over time may indicate sustained metabolic stress, which in turn can lead to endothelial dysfunction, increased oxidative stress, and activation of inflammatory pathways, all of which contribute to kidney fibrosis and eGFR decline [31,32]. Second, in patients with MAFLD, insulin resistance is a core feature that not only promotes hepatic steatosis but also exerts systemic effects, including increased renal sodium reabsorption and altered glomerular hemodynamics [33–35]. These effects may amplify renal vulnerability, particularly when the TyG index remains elevated or worsens over time.

The clinical implications of our findings are noteworthy. The burden of CKD among individuals with MAFLD has been increasingly recognized. MAFLD encompasses a spectrum of liver abnormalities occurring in the context of metabolic dysfunction and is frequently associated with insulin resistance, obesity, and DM, factors known to accelerate CKD progression. Recent meta-analyses have shown that MAFLD is associated with higher odds of prevalent and incident CKD, independent of traditional risk factors [36]. However, risk stratification within the MAFLD population remains challenging due to its heterogeneity. In this regard, our study offers a novel approach to phenotyping patients based on the longitudinal behavior of a simple, non-invasive biomarker. The TyG index, calculated from fasting triglyceride and glucose levels, is inexpensive and easily obtained in routine clinical settings. By identifying patients with an increasing TyG trajectory, clinicians may be able to target interventions more effectively toward those at greatest risk of kidney function decline. Additionally, our findings support the conceptual overlap between MAFLD and cardio-kidney-metabolic syndrome [37]. The kidney and liver share common pathophysiological pathways in the context of metabolic stress, and biomarkers like the TyG index may reflect this shared vulnerability. As such, monitoring the TyG index could provide a unifying metric to assess multi-organ risk in patients with metabolic liver disease. Finally, our subgroup analyses revealed consistent associations across different demographic and clinical strata, including age, sex, BMI status, and the presence of HTN or DM. This suggests that the predictive value of the TyG trajectory is robust across various patient profiles. Interestingly, even in the non-diabetic and non-obese subgroups, an increasing TyG index was associated with worse kidney outcomes. This underscores the importance of early metabolic risk detection, even among individuals who may not meet conventional criteria for high-risk status. Furthermore, kidney outcomes were defined irrespective of the underlying renal etiology, as information on primary renal disease was not consistently available. Although MAFLD is most strongly linked to diabetic nephropathy, the consistent association between an increasing TyG trajectory and kidney outcomes across DM subgroups suggests a broader metabolic impact. These findings imply that metabolic dysfunction associated with MAFLD contributes to kidney function decline through multiple mechanisms beyond DM alone.

Several limitations warrant consideration. First, although only 36 primary events accrued during follow-up, the corresponding incidence rate (6.55 per 1,000 PYs) aligns with previously reported CKD incidence among adults aged 40 years or older in the Korean population-based cohort [38], suggesting that the limited event count reflects the low background incidence of CKD rather than incomplete event capture. Nonetheless, the small number of events reduces the precision of HR estimates, particularly in subgroup analyses, and these findings should therefore be interpreted with caution. Furthermore, the limited number of kidney outcome events raises the possibility of model overfitting. Covariates were selected a priori based on clinical relevance and previous studies, and the consistent results across sequential models indicate that overfitting is unlikely to have substantially affected the findings. Moreover, CKD incidence may vary across different ethnic groups. However, because our study population consisted exclusively of Korean individuals, ethnicity-specific differences could not be assessed. Therefore, caution is warranted when generalizing these findings to other populations. Second, the trajectory model was simplified into two classes to ensure model stability and interpretability. This approach may have obscured finer metabolic heterogeneity that could exist within the population. Future studies with larger cohorts and denser longitudinal measurements may better characterize additional metabolic trajectory subtypes and their differential effects on kidney outcomes. Third, the duration used to define TyG index trajectories may not represent the optimal period for trajectory analysis, as there is no universally accepted standard. However, previous studies have typically used an exposure window of 4 to 6 years for CKD to capture meaningful longitudinal trends while maintaining model stability and interpretability [39,40]. In our study, the 4-year exposure period was chosen for methodological consistency. However, the key clinical implication is that kidney risk assessment should not rely solely on a single TyG index measurement. Rather, clinicians should pay close attention to longitudinal trends and directional changes in metabolic markers when evaluating patients with MAFLD.

In conclusion, an increasing TyG index over time is independently associated with a higher risk of adverse kidney outcomes in patients with MAFLD. Monitoring the trajectory of the TyG index may serve as a simple yet powerful tool to identify individuals at elevated renal risk and guide personalized management strategies.

Supplementary Materials

Article information Continued