Association of triglyceride-glucose (TyG)-Derived inflammatory indices with the risk of all-cause and cardiovascular mortality in patients with chronic kidney disease: A retrospective cohort study based on NHANES 1999–2018

Abstract

Objective

The triglyceride-glucose (TyG) index has been explored as a prognostic marker in chronic kidney disease (CKD), but it does not capture the complex interplay of metabolic dysfunction and chronic inflammation inherent to CKD pathogenesis. This study aimed to explore the associations between TyG-derived inflammatory indices and the risks of all-cause and cardiovascular disease (CVD) mortality in patients with CKD.

Methods

This retrospective cohort study included adults with CKD from the National Health and Nutrition Examination Survey (NHANES) 1999–2018, with mortality follow-up through December 2019. The associations of the TyG index and five TyG-derived inflammatory indices (TyG-NLR, TyG-MLR, TyG-lgPLR, TyG-lgSII, and TyG-SIRI) with mortality risk were evaluated using receiver operating characteristic (ROC) curves and survey-weighted Cox proportional hazards regression models.

Results

A total of 3,534 patients with CKD were included. TyG-MLR showed the highest area under the ROC curve (AUC) among the evaluated indices, with an AUC of 0.66 for all-cause mortality and 0.68 for CVD mortality. After stratification by TyG-MLR quartiles (Q1–Q4), patients in the Q4 group had an estimated 3.3-year reduction in life expectancy at the age of 45 compared with the Q1 group.

Conclusions

This study suggests that TyG-derived inflammatory indices may hold clinical value in predicting mortality risk among patients with CKD. In the first systematic comparison of multiple such indices, TyG-MLR emerged as the best-performing indicator, although its discriminative performance was moderate. Prospective validation in independent cohorts is warranted.

Introduction

Chronic kidney disease (CKD) is a major global public health issue with a continuously rising prevalence and an increasingly heavy burden of mortality and disability. ¹ According to statistics from the World Health Organization, CKD and its complications cause more than 1 million deaths each year, and this number is still climbing due to population aging and the increasing prevalence of chronic diseases such as diabetes mellitus (DM) and hypertension. ² CKD is not only a precursor to end-stage renal disease but also an independent risk factor for major adverse events such as cardiovascular disease (CVD). ³ Therefore, it is of great clinical significance to find simple and effective risk assessment tools for the early identification of the risks of all-cause mortality and CVD mortality in CKD patients.

In recent years, insulin resistance (IR) has been recognized as a key pathological mechanism in the development and progression of CKD. The triglyceride-glucose index (TyG), as a surrogate marker for IR, has been increasingly widely used in risk assessment studies of metabolic diseases and CKD due to its simplicity in calculation and strong stability. ⁴ Multiple studies have shown that an elevated TyG index is closely associated with the development and progression of CKD and with an increased risk of end-stage renal disease. It also has a significant positive correlation with all-cause mortality and CVD mortality in CKD patients.^5–7 However, the pathophysiological basis of CKD extends far beyond metabolic disorders; chronic low-grade inflammation also plays a crucial role in its progression and poor prognosis. Decreased renal function can lead to impaired clearance of inflammatory factors, and chronic inflammation in turn exacerbates insulin resistance, forming a positive feedback loop of “metabolism-inflammation-organ damage”.^8–10 Therefore, a single TyG index may be insufficient to fully reflect the complex risk status of CKD patients.

To more comprehensively assess the adverse prognosis risk from combined metabolic and inflammatory effects, researchers have proposed composite indicators by combining the TyG index with peripheral blood inflammatory markers, including TyG-NLR (TyG × NLR, NLR = Neutrophil Count/Lymphocyte Count), TyG-MLR (TyG × MLR, MLR = Monocyte Count/Lymphocyte Count), TyG-lgPLR (TyG × lgPLR, lgPLR = lg(Platelet Count/Lymphocyte Count)), TyG-lgSII (TyG × lgSII, lgSII = lg (Neutrophil Count × Platelet Count/Lymphocyte Count)), and TyG-SIRI (TyG × SIRI, SIRI = Neutrophil Count × Monocyte Count/Lymphocyte Count). ¹¹ Preliminary studies have revealed that these TyG-derived inflammatory indicators exhibit certain predictive value in CVD events.^10,12,13 However, there is currently a lack of systematic research on their ability to predict mortality risk in the CKD population, especially a shortage of horizontal comparisons among multiple indicators.

Based on this, this study aims to systematically explore the relationships between the TyG index, its inflammation-derived indicators, and all-cause mortality as well as CVD mortality in CKD patients using nationally representative population data. It intends to identify the most predictive indicators, thereby providing a theoretical basis and practical tools for precise risk stratification and early intervention in CKD patients.

Methods

Population

This study utilized data from the National Health and Nutrition Examination Survey (NHANES) conducted across ten consecutive 2-year cycles from 1999 to 2018 (1999–2000, 2001–2002, 2003–2004, 2005–2006, 2007–2008, 2009–2010, 2011–2012, 2013–2014, 2015–2016, and 2017–2018). A total of 101,316 participants were enrolled across these cycles (detailed in Supplementary Table S1). Among them, 55,081 were adults. After excluding 1,547 pregnant individuals and 5,140 participants with a history of cancer, 7,803 participants were identified as having CKD. Of these, 3,537 had available data on TyG and inflammatory markers. After removing 3 participants lost to follow-up, a total of 3,534 participants were included in the final analysis. All participants with CKD who met the eligibility criteria were consecutively included.

This was a retrospective cohort study based on NHANES 1999–2018 with prospective mortality follow-up through December 2019. The reporting of this study conforms to the STROBE guidelines. ¹⁴ This study was conducted in accordance with the Declaration of Helsinki of 1975, as revised in 2024; the NHANES protocol was approved by the National Center for Health Statistics Research Ethics Review Board, with written informed consent obtained from all participants. All NHANES data are de-identified and publicly available; no identifiable patient information was accessed, and no additional institutional review board approval was required for this secondary analysis.

A diagnosis of CKD was made based on two criteria: either a participant answered “yes” to the query “Has a doctor or other medical professional ever told you that you have CKD?”; or they had an estimated glomerular filtration rate (eGFR) of less than 60 mL/min/1.73 m² and/or a random urinary albumin-to-creatinine ratio (ACR) of 30 mg/g or higher. ¹⁵ The eGFR was calculated using the 2009 Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula. ¹⁶ CKD is staged into five categories according to eGFR levels: Stage 1 (eGFR ≥90 mL/min/1.73 m²), Stage 2 (eGFR 60–89 mL/min/1.73 m²), Stage 3 (which is further divided into 3a: 45–59 and 3b: 30–44 mL/min/1.73 m²), Stage 4 (eGFR 15–29 mL/min/1.73 m²), and Stage 5 (eGFR <15 mL/min/1.73 m²).

Calculation of TyG and derived indices

The calculation of TyG and its derived indices was as follows: TyG = ln [TG (mg/dL) × Glu (mg/dL)/2];

$$NLR=Neutrophil\hspace{0.17em}Count/Lymphocyte\hspace{0.17em}Count;$$

$$MLR=Monocyte\hspace{0.17em}Count/Lymphocyte\hspace{0.17em}Count;$$

$$lgPLR=\mathit{lg}\left(Platelet\hspace{0.17em}Count/Lymphocyte\hspace{0.17em}Count\right);$$

$$lgSII=\mathit{lg}\left(Neutrophil\hspace{0.17em}Count\times Platelet\hspace{0.17em}Count/Lymphocyte\hspace{0.17em}Count\right);$$

$$SIRI=Neutrophil\hspace{0.17em}Count\times Monocyte\hspace{0.17em}Count/Lymphocyte\hspace{0.17em}Count.$$

$$TyG-NLR=TyG\times NLR;$$

$$TyG-MLR=TyG\times MLR;$$

$$TyG-lgPLR=TyG\times lgPLR;$$

$$TyG-lgSII=TyG\times lgSII;$$

$$TyG-SIRI=TyG\times SIRI.$$

Note: ln denotes the natural logarithm; lg denotes the base-10 logarithm.

The index demonstrating the highest AUC in the ROC analysis (TyG-MLR) was subsequently used to stratify patients into quartile groups (Q1–Q4) for further analyses.

Outcome

The primary outcome of interest was all-cause mortality, with CVD mortality as the secondary outcome. The International Classification of Diseases, 10th Revision (ICD-10) was utilized to categorize the causes of death. Participants were considered to have died from CVD if their death was coded under ICD-10 categories I00-I09, I11, I13, I20-I51, or I60-I69. The most recent follow-up data for NHANES was updated to December 2019.

Confounding variable

Participant-reported factors included age, gender, race/ethnicity, smoking status, and alcohol consumption. Smoking status was categorized as never, former, or current smoker based on self-reports; current smokers were defined as those who reported having smoked more than 100 cigarettes in their lifetime and were smoking at the time of the survey. Drinking status was categorized as never, former, or current drinker; current drinkers referred to those who reported any alcohol intake in the past 12 months.

During the survey, body mass index (BMI) of participants was measured by NHANES staff. Blood test indicators, such as neutrophils, lymphocytes, monocytes, creatinine, glucose, triglycerides (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and hemoglobin A1c (HbA1c), were analyzed by laboratories affiliated with NHANES in accordance with standardized protocols detailed on the NHANES official website. The Healthy Eating Index 2015 (HEI-2015) was calculated using 24-hour dietary recall data, following the method described in previous studies. ¹⁷

In the NHANES, comorbidities were assessed through a combination of self-reports and objective diagnostic criteria. Hypertension was diagnosed if participants reported a previous diagnosis of hypertension by a healthcare professional, had a systolic blood pressure (SBP) ≥130 mmHg or diastolic blood pressure (DBP) ≥80 mmHg, or were currently taking antihypertensive medications. DM was defined by self-reported diagnosis of DM, HbA1c ≥6.5%, fasting plasma glucose ≥7.0 mmol/L, random plasma glucose ≥11.1 mmol/L, 2-hour oral glucose tolerance test result ≥11.1 mmol/L, or current use of antidiabetic medications or insulin therapy. CVD history was identified based on self-reported history of professional diagnosis. Hyperlipidemia was determined as follows: TG ≥150 mg/dL, TC ≥200 mg/dL, LDL-C ≥130 mg/dL, HDL-C ≤40 mg/dL (for men) or ≤50 mg/dL (for women), or current use of lipid-lowering medications.

Statistical analysis

All analyses incorporated NHANES survey weights to account for the complex multi-stage sampling design. For the combined 1999–2018 cycles, 4-year MEC weights (WTMEC4YR, for 1999–2002) were multiplied by 2/10, and 2-year MEC weights (WTMEC2YR, for 2003–2018) were divided by 10, consistent with NHANES analytic guidelines for multi-cycle analyses. For the primary exposure variables (TyG and its derived inflammatory indices), complete-case analysis was used; participants with missing data on any of the six indices were excluded. For covariates with missing data (such as BMI, HEI-2015 score, alcohol use), multiple imputation was performed using the mice package, with 5 imputed datasets generated and results combined using Rubin’s rules. To assess potential selection bias from excluding participants without available TyG or inflammatory marker data, we compared baseline characteristics of included (n = 3,537) and excluded (n = 4,266) participants with CKD who met all other eligibility criteria (age ≥ 20 years, non-pregnant, no cancer history, and available CKD status). Baseline characteristics were broadly comparable (Supplementary Table S2).

In this study, the receiver operating characteristic (ROC) curve was used to evaluate the predictive efficacy of TyG and TyG-derived inflammatory indices in predicting the risk of death in the overall cohort. To compare the discriminative performance of TyG-MLR with standard clinical variables, we also calculated the C-index for each individual predictor (age, gender, race, smoking status, alcohol use, CVD, hypertension, diabetes, hyperlipidemia, HEI-2015, BMI) using survey-weighted Cox proportional hazards models. To assess the robustness of the index selection, a random split-sample internal validation was performed by dividing the cohort into a training set (70%) and a validation set (30%); the ranking of all six indices was compared between the two partitions. In addition, three Cox proportional hazards models were used to investigate the relationship between TyG and TyG-derived inflammatory indices and all-cause mortality and CVD mortality in CKD patients. Model 1 was an unadjusted model. Model 2 was adjusted for age, gender, and race. Model 3 was additionally adjusted for smoking status, alcohol use, hypertension, diabetes, hyperlipidemia, CVD, HEI-2015, BMI, and CKD stage on the basis of Model 2. The flexsurvreg function was used to fit the Gompertz survival model. After adjusting the covariates in Model 3, the loss of life expectancy of patients was evaluated based on the conditional survival probability.

To evaluate the incremental predictive value of the best-performing index identified in the ROC analysis beyond standard clinical variables, we calculated the change in C-statistic (ΔC-index) with bootstrap-derived 95% confidence intervals (B = 1000 resamples), the integrated discrimination improvement (IDI), and the continuous net reclassification improvement (NRI) with 1000 perturbations. The base model included age, gender, race, smoking status, alcohol use, hypertension, diabetes, hyperlipidemia, CVD, HEI-2015, BMI, and CKD stage. Analyses were performed for both all-cause and CVD mortality.

All statistical analyses were performed using R software (version 4.3.2, R Foundation for Statistical Computing, Vienna, Austria). A two-sided P value < 0.05 was considered statistically significant.

Results

Baseline characteristics

The study population (N=3,534) had a weighted mean age of 59.05 years (SE=0.45), with 56.82% females. The majority were Non-Hispanic White (65.83%), followed by Non-Hispanic Black (13.31%) and Mexican American (8.38%). Weighted mean BMI was 30.18 kg/m², eGFR 75.96 mL/(min·1.73m²), and UACR 209.10 mg/g. Metabolic parameters included TG 152.08 mg/dL, GLU 6.47 mmol/L, TC 195.02 mg/dL, LDL 112.02 mg/dL, and HDL 53.33 mg/dL. Blood counts showed neutrophil 4.42×10³/μL, monocyte 0.57×10³/μL, lymphocyte 1.94×10³/μL, and platelet 247.40×10³/μL. Inflammatory ratios were NLR 2.53, MLR 0.32, and PLR 141.80; TyG index was 8.88. Comorbidities included hypertension (65.92%), DM (36.84%), CVD (23.41%), and hyperlipidemia (81.65%). All-cause and CVD mortality rates were 32.08% and 12.06%, respectively (Table 1).

Table 1.

Baseline characteristics.

Variable	Total (N=3,534)	Q1 group (N=884)	Q2 group (N=883)	Q3 group (N=883)	Q4 group (N=884)	P-value
Age, years	59.05(0.45)	51.80(0.85)	56.70(0.77)	61.58(0.81)	65.78(0.59)	< 0.001
Gender, n (%)						< 0.001
Female	1867(56.82)	603(74.55)	511(59.23)	407(50.13)	346(44.44)	​
Male	1667(43.18)	281(25.45)	372(40.77)	476(49.87)	538(55.56)	​
Race, n (%)						< 0.001
Mexican American	620(8.38)	192(13.04)	179(8.82)	146(7.40)	103(4.57)	​
Non-Hispanic Black	803(13.31)	267(20.29)	214(13.28)	189(11.82)	133(8.32)	​
Non-Hispanic White	1572(65.83)	241(47.86)	359(66.90)	418(67.95)	554(79.28)	​
Other race	539(12.48)	184(18.81)	131(11.01)	130(12.83)	94(7.82)	​
BMI, kg/m2	30.18(0.19)	30.25(0.37)	30.37(0.34)	29.61(0.31)	30.49(0.43)	0.199
eGFR, mL/min/1.73 m2	75.96(0.80)	88.54(1.41)	79.27(1.27)	72.07(1.38)	64.68(1.06)	< 0.001
UACR, mg/g	209.10(14.49)	166.10(17.14)	204.06(29.80)	201.02(28.26)	263.71(37.88)	0.117
TG, mg/dl	152.08(2.78)	136.18(5.09)	153.85(6.43)	161.98(5.61)	155.11(5.38)	0.003
GLU, mmol/l	6.47(0.05)	6.24(0.10)	6.45(0.11)	6.54(0.11)	6.64(0.11)	0.032
TC, mg/dl	195.02(1.08)	196.20(1.99)	198.30(2.21)	195.31(2.15)	190.14(2.00)	0.047
LDL, mg/dl	112.02(0.85)	114.70(1.79)	115.36(1.88)	112.00(1.85)	105.77(1.66)	< 0.001
HDL, mg/dl	53.33(0.43)	55.07(0.70)	53.25(0.92)	51.70(0.62)	53.41(1.20)	0.007
Neu, 1000/ul	4.42(0.04)	3.95(0.07)	4.25(0.08)	4.54(0.07)	4.91(0.08)	< 0.001
Monocyte, 1000/ul	0.57(0.01)	0.44(0.01)	0.53(0.01)	0.61(0.01)	0.71(0.01)	< 0.001
Lymp, 1000/ul	1.94(0.02)	2.45(0.04)	2.09(0.03)	1.84(0.02)	1.43(0.02)	< 0.001
Plt, 1000/ul	247.40(1.70)	260.90(3.12)	244.05(2.79)	247.39(2.96)	238.45(3.97)	< 0.001
NLR	2.53(0.03)	1.71(0.03)	2.12(0.04)	2.56(0.04)	3.71(0.08)	< 0.001
MLR	0.32(0.00)	0.18(0.00)	0.26(0.00)	0.33(0.00)	0.52(0.01)	< 0.001
PLR	141.80(1.51)	115.95(1.99)	125.34(2.06)	143.09(2.00)	182.16(4.14)	< 0.001
SII	629.56(10.18)	457.03(12.60)	524.14(14.16)	640.25(12.30)	892.11(28.88)	< 0.001
SIRI	1.48(0.03)	0.72(0.01)	1.09(0.02)	1.51(0.03)	2.57(0.06)	< 0.001
TyG	8.88(0.02)	8.74(0.04)	8.89(0.03)	8.95(0.03)	8.93(0.04)	< 0.001
TyG_NLR	22.21(0.25)	14.95(0.27)	18.69(0.33)	22.82(0.33)	32.11(0.55)	< 0.001
TyG_MLR	2.85(0.03)	1.59(0.01)	2.27(0.01)	2.97(0.01)	4.53(0.04)	< 0.001
TyG_lgPLR	18.76(0.05)	17.77(0.09)	18.40(0.08)	19.04(0.07)	19.80(0.11)	< 0.001
TyG_lgSII	24.23(0.07)	22.73(0.14)	23.72(0.12)	24.68(0.11)	25.71(0.15)	< 0.001
TyG_SIRI	12.96(0.21)	6.33(0.13)	9.66(0.18)	13.46(0.22)	22.15(0.40)	< 0.001
HEI-2015 score	50.75(0.38)	49.79(0.61)	51.04(0.65)	50.31(0.74)	51.76(0.62)	0.108
Smoking status, n (%)						< 0.001
Never	1,798(50.89)	513(59.39)	475(54.44)	424(47.20)	386(43.09)	​
Former	1,100(30.26)	189(20.52)	240(26.23)	302(32.21)	369(41.78)	​
Now	633(18.75)	181(20.09)	166(19.33)	157(20.60)	129(15.13)	​
Alcohol use, n (%)						0.431
Never	573(14.83)	168(18.74)	139(15.61)	131(14.99)	135(16.35)	​
Former	840(21.05)	186(20.41)	213(22.44)	210(24.55)	231(25.47)	​
Now	1,732(54.67)	433(60.86)	435(61.94)	433(60.46)	431(58.18)	​
Hypertension, n (%)	2,526(65.92)	562(55.28)	610(61.43)	664(70.00)	690(76.73)	< 0.001
DM, n (%)	1,495(36.84)	320(30.14)	376(34.24)	400(40.48)	399(42.22)	< 0.001
CVD, n (%)	939(23.41)	138(12.50)	188(17.30)	262(27.68)	351(35.81)	< 0.001
Hyperlipidemia, n (%)	2,910(81.65)	694(75.86)	743(83.91)	742(83.46)	731(82.83)	0.008
All-cause mortality, n (%)	1,343(32.08)	218(18.63)	261(24.51)	354(34.62)	510(50.13)	< 0.001
CVD mortality, n (%)	518(12.06)	78(7.56)	83(9.29)	145(16.82)	212(29.19)	< 0.001

Abbreviation: BMI, Body Mass Index; UACR, Urinary Albumin-to-Creatinine Ratio; TG, Triglycerides; GLU, Glucose; TC, Total Cholesterol; LDL, Low-Density Lipoprotein Cholesterol; HDL, High-Density Lipoprotein Cholesterol; Neu, Neutrophil Count; Monocyte, Monocyte Count; Lymp, Lymphocyte Count; Plt, Platelet Count; NLR, Neutrophil-to-Lymphocyte Ratio; MLR, Monocyte-to-Lymphocyte Ratio; PLR, Platelet-to-Lymphocyte Ratio; SII, Systemic Inflammatory Index; SIRI, Systemic Inflammatory Response Index; TyG, Triglyceride-Glucose Index; HEI-2015, Healthy Eating Index 2015; DM, Diabetes Mellitus; CVD, Cardiovascular Disease.

ROC curve analysis of the relationship between TyG, TyG-Inflammatory indices and mortality risk

As shown in Figures 1A and B. For all-cause mortality, the TyG index yielded an AUC of 0.55 (95% CI: 0.53–0.56); TyG-NLR achieved an AUC of 0.62 (95% CI: 0.61–0.64); TyG-MLR showed an AUC of 0.66 (95% CI: 0.64–0.68); TyG-lgPLR presented an AUC of 0.58 (95% CI: 0.56–0.60); TyG-lgSII had an AUC of 0.58 (95% CI: 0.56–0.60); TyG-SIRI displayed an AUC of 0.63 (95% CI: 0.61–0.65). Among these indices for all-cause mortality, TyG-MLR exhibited the highest AUC, indicating superior discriminative ability in predicting all-cause mortality in CKD patients.

Figure 1.

ROC curves of the TyG index and five TyG-derived inflammatory indices for predicting. (A) all-cause mortality and (B) cardiovascular mortality in patients with CKD, NHANES 1999–2018.

For CVD mortality, the TyG index generated an AUC of 0.54 (95% CI: 0.51–0.57) with a cutoff of 8.88, sensitivity 0.49, and specificity 0.55; TyG-NLR achieved an AUC of 0.62 (95% CI: 0.60–0.65); TyG-MLR showed an AUC of 0.68 (95% CI: 0.65–0.70); TyG-lgPLR presented an AUC of 0.57 (95% CI: 0.55–0.60); TyG-lgSII had an AUC of 0.57 (95% CI: 0.54–0.60); TyG-SIRI displayed an AUC of 0.64 (95% CI: 0.61–0.66). For CVD mortality, TyG-MLR again showed the highest AUC, suggesting it outperforms other TyG-related indices in predicting CVD-specific mortality in CKD patients. Collectively, TyG-MLR demonstrates relatively superior predictive performance across the analyzed outcomes, contributing to comprehensive mortality risk stratification in CKD populations.

In the split-sample internal validation, TyG-MLR ranked first among all six indices in both the training and validation sets for all-cause mortality (AUC = 0.661 and 0.666, respectively) and CVD mortality (AUC = 0.678 and 0.674, respectively), and the overall ranking of the six indices was consistent between the two partitions for both outcomes, supporting the robustness of the index selection (Supplementary Table S3). In addition, the C-index of TyG-MLR alone exceeded those of all other individual clinical predictors except age for both all-cause and CVD mortality (Supplementary Table S4).

Adding TyG-MLR to the base model (including CKD stage and other standard clinical variables) significantly improved risk prediction (Supplementary Table S5). For all-cause mortality, the ΔC-index was 0.005 (bootstrap 95% CI: 0.002–0.010), the IDI was 0.010 (95% CI: 0.004–0.019, p = 0.004), and the continuous NRI was 0.111 (95% CI: 0.026–0.172, p = 0.012). For CVD mortality, the ΔC-index was 0.010 (bootstrap 95% CI: 0.004–0.016), the IDI was 0.025 (95% CI: 0.010–0.049, p < 0.001), and the continuous NRI was 0.190 (95% CI: 0.061–0.271, p = 0.008).

The association between TyG, TyG-inflammatory indices and mortality risk

Table 2 illustrates the weighted associations between TyG and its derived indices with all-cause and CVD mortality across three models. For all-cause mortality, Model 1 (unadjusted) shows that TyG is significantly associated with mortality, yielding a hazard ratio (HR) of 1.17 (95% confidence interval [CI]: 1.06–1.30, P = 0.002). Concurrently, all TyG-derived indices exhibit robust significance: TyG-NLR demonstrates an HR of 1.04 (95%CI: 1.03–1.04, P < 0.001), TyG-MLR an HR of 1.40 (95%CI: 1.32–1.48, P < 0.001), TyG-lgPLR an HR of 1.07 (95%CI: 1.03–1.11, P = 0.001), TyG-lgSII an HR of 1.07 (95%CI: 1.04–1.10, P < 0.001), and TyG-SIRI an HR of 1.04 (95%CI: 1.04–1.05, P < 0.001). In Model 3, the association of TyG with all-cause mortality becomes non-significant, with an HR of 1.06 (95%CI: 0.93–1.20, P=0.377). By contrast, all derived indices maintain statistical significance: TyG-NLR shows an HR of 1.02 (1.02–1.03, P < 0.001), TyG-MLR an HR of 1.18 (95%CI: 1.12–1.25, P < 0.001), TyG-lgPLR an HR of 1.05 (1.02–1.09, P = 0.006), TyG-lgSII an HR of 1.06 (1.03–1.09, P < 0.001), and TyG-SIRI an HR of 1.03 (1.02–1.04, P < 0.001).

Table 2.

The association between the TyG/TyG-derived inflammatory indices and mortality (weighted).

Variable	Model 1		Model 2		Model 3
	HR* (95%CI # )	P-value	HR (95%CI)	P-value	HR (95%CI)	P-value
All-cause mortality
TyG	1.17(1.06,1.30)	0.002	1.20(1.07,1.33)	0.002	1.06(0.93,1.20)	0.377
TyG-NLR	1.04(1.03,1.04)	<0.001	1.03(1.02,1.03)	<0.001	1.02(1.02,1.03)	<0.001
TyG-MLR	1.40(1.32,1.48)	<0.001	1.19(1.13,1.25)	<0.001	1.18(1.12,1.25)	<0.001
TyG-lgPLR	1.07(1.03,1.11)	0.001	1.07(1.03,1.10)	<0.001	1.05(1.02,1.09)	0.006
TyG-lgSII	1.07(1.04,1.10)	<0.001	1.08(1.05,1.11)	<0.001	1.06(1.03,1.09)	<0.001
TyG-SIRI	1.04(1.04,1.05)	<0.001	1.03(1.02,1.04)	<0.001	1.03(1.02,1.04)	<0.001
CVD mortality
TyG	1.23(1.05,1.43)	0.034	1.30(1.09,1.55)	0.030	1.10(0.85,1.42)	0.474
TyG-NLR	1.04(1.03,1.05)	<0.001	1.03(1.02,1.04)	<0.001	1.03(1.02,1.04)	<0.001
TyG-MLR	1.56(1.43,1.70)	<0.001	1.28(1.18,1.39)	<0.001	1.29(1.18,1.41)	<0.001
TyG-lgPLR	1.09(1.02,1.16)	0.010	1.12(1.05,1.20)	<0.001	1.12(1.03,1.21)	0.009
TyG-lgSII	1.08(1.04,1.12)	<0.001	1.11(1.07,1.16)	<0.001	1.10(1.04,1.16)	<0.001
TyG-SIRI	1.05(1.04,1.07)	<0.001	1.04(1.03,1.05)	<0.001	1.04(1.02,1.05)	<0.001

Model 1: Not adjusted.

Model 2: Adjusted by age, gender, race/ethnicity.

Model 3: Adjusted by age, gender, race/ethnicity, cigarette smoking status, alcohol consumption status, BMI, CKD Stages, Hypertension, DM, CVD, Hyperlipidemia, HEI-2015.

*Hazard ratio.

^#Confidence interval.

For CVD mortality, Model 1 reveals a significant association for TyG (HR = 1.23, 95% CI: 1.05–1.43, P = 0.034), while Model 3 shows attenuation of this association (HR = 1.10, 95% CI: 0.85–1.42, P = 0.474). In contrast, TyG-derived indices remain strongly associated in Model 3: TyG-NLR yields an HR of 1.03 (95%CI: 1.02–1.04, P < 0.001), TyG-MLR an HR of 1.29 (95%CI: 1.18–1.41, P < 0.001), TyG-lgPLR an HR of 1.12 (95%CI: 1.03–1.21, P = 0.009), TyG-lgSII an HR of 1.10 (95%CI: 1.04–1.16, P < 0.001), and TyG-SIRI an HR of 1.04 (95%CI: 1.02–1.05, P < 0.001). These findings highlight that while the standalone TyG index loses significance in the fully adjusted model, the TyG-derived inflammatory indices consistently demonstrate robust predictive value for both all-cause and CVD mortality.

Association between TyG, TyG-inflammatory indices and CKD stages

Using a generalized linear model, we analyzed the associations between the TyG index and its inflammatory-derived indices (TyG-NLR, TyG-MLR, TyG-lgPLR, TyG-lgSII, TyG-SIRI) with CKD stages classified by eGFR. In Model 1 (unadjusted), the TyG index showed no significant association with CKD stages (β = 0.02, 95%CI: -0.04 to 0.08, P = 0.503), while TyG-NLR exhibited a significant positive association (β = 0.01, 95%CI: 0.01 to 0.02, P < 0.001), TyG-MLR demonstrated a strong positive link (β = 0.21, 95%CI: 0.18 to 0.23, P < 0.001), and TyG-SIRI showed a significant association (β = 0.02, 95%CI: 0.01 to 0.02, P < 0.001); TyG-lgPLR (β = 0.02, 95%CI: 0.00 to 0.04, P = 0.052) and TyG-lgSII (β = 0.01, 95%CI: 0.00 to 0.03, P = 0.105) showed no significant associations. In the fully adjusted Model 3, the TyG index still had no significant association (β = -0.03, 95%CI: -0.10 to 0.04, P = 0.471), TyG-NLR retained a significant but minimal association (β = 0.00, 95%CI: 0.00 to 0.01, P = 0.010), TyG-MLR remained significantly associated with the largest effect size (β = 0.05, 95%CI: 0.02 to 0.09, P = 0.002), TyG-SIRI showed a significant small association (β = 0.01, 95%CI: 0.00 to 0.01, P = 0.032), while TyG-lgPLR and TyG-lgSII showed no significant associations (Table 3).

Table 3.

The association between the TyG/TyG-derived inflammatory indices and CKD stage (weighted).

Variable	Model 1		Model 2		Model 3
	β* (95%CI # )	P-value	β* (95%CI # )	P-value	β* (95%CI # )	P-value
TyG	0.02(-0.04,0.08)	0.503	-0.01(-0.06,0.04)	0.606	-0.03(-0.10, 0.04)	0.471
TyG-NLR	0.01(0.01,0.02)	<0.001	0.00 (0.00,0.01)	0.004	0.00 (0.00, 0.01)	0.010
TyG-MLR	0.21(0.18,0.23)	<0.001	0.06(0.03,0.09)	<0.001	0.05(0.02, 0.09)	0.002
TyG-lgPLR	0.02(0.00,0.04)	0.052	0.00 (-0.02,0.02)	0.861	0.00 (-0.02, 0.02)	0.944
TyG-lgSII	0.01(0.00,0.03)	0.105	0.00 (-0.01,0.01)	0.638	0.00 (-0.01, 0.02)	0.641
TyG-SIRI	0.02(0.01,0.02)	<0.001	0.01(0.00,0.01)	0.017	0.01(0.00, 0.01)	0.032

Model 1: Not adjusted.

Model 2: Adjusted by age, gender, race/ethnicity.

Model 3: Adjusted by age, gender, race/ethnicity, cigarette smoking status, alcohol consumption status, BMI, Hypertension, DM, CVD, Hyperlipidemia, HEI-2015.

^#Confidence interval.

Baseline characteristics after grouping by TyG-MLR

Notably, TyG-MLR demonstrated the highest AUC in ROC curve analyses for both all-cause and CVD mortality, as well as the largest effect size in the Cox regression analyses. Additionally, TyG-MLR exhibited the largest effect size in the association with CKD stages among all evaluated indices. Consequently, this study stratified participants into groups based on the quartiles of TyG-MLR. The baseline characteristics of these groups were shown in Table 1. The Q4 group (highest TyG-MLR) had a significantly higher mean age (65.78 (0.59)) compared to the Q1 group (51.80 (0.85)), and the proportion of females in the Q4 group (44.44%) was notably lower than that in the Q1 group (74.55%). The prevalence of Non-Hispanic White in the Q4 group (79.28%) was significantly higher than that in the Q1 group (47.86%), while the proportions of Mexican American and Non-Hispanic Black showed a decreasing trend from Q1 to Q4. In terms of renal function, the mean eGFR in the Q4 group (64.68 (1.06) mL/min/1.73m²) was significantly lower than that in the Q1 group (88.54 (1.41) mL/min/1.73m²). The levels of TG and Glu showed an increasing trend across quartiles, with the TG in Q3 reaching 161.98 (5.61) mg/dL and the Glu in Q4 being 6.64 (0.11) mmol/L. The prevalence of comorbidities including hypertension (76.73% vs. 55.28%), DM (42.22% vs. 30.14%), and CVD (35.81% vs. 12.50%) was significantly higher in the Q4 group compared to the Q1 group. In terms of mortality, the all-cause mortality (50.13% vs. 18.63%) and CVD mortality (29.19% vs. 7.56%) in the Q4 group were strikingly higher than those in the Q1 group.

Association between TyG-MLR and mortality risk

COX regression analysis based on quartiles of TyG-MLR (with Q1 as the reference group) revealed a graded association with mortality outcomes as presented in Table 4. For all-cause mortality, unadjusted Model 1 showed that Q2, Q3, and Q4 had HRs of 1.35 (95% confidence interval [CI]: 1.10–1.66, P=0.004), 2.12 (1.72–2.61, P<0.001), and 3.48 (2.78–4.35, P<0.001), respectively, with a significant trend (P<0.001). In Model 2, Q2 became non-significant (HR=1.00, 0.82–1.22, P=0.989), while Q3 (HR=1.22, 1.01–1.50, P=0.043) and Q4 (HR=1.66, 1.33–2.08, P<0.001) remained significant, with a significant trend (P<0.001).

Table 4.

The association between the TyG-MLR and mortality (weighted).

Variable	Model 1		Model 2		Model 3
	HR* (95%CI # )	P-value	HR (95%CI)	P-value	HR (95%CI)	P-value
All-cause mortality
Q1 group	Ref	​	Ref	​	Ref	​
Q2 group	1.35(1.10,1.66)	0.004	1.00(0.82,1.22)	0.989	1.00(0.79,1.26)	0.995
Q3 group	2.12(1.72,2.61)	<0.001	1.22(1.01,1.50)	0.043	1.22(0.98,1.53)	0.072
Q4 group	3.48(2.78,4.35)	<0.001	1.66(1.33,2.08)	<0.001	1.63(1.30,2.06)	<0.001
P for trend	​	<0.001	​	<0.001	​	<0.001
CVD mortality
Q1 group	Ref	​	Ref	​	Ref	​
Q2 group	1.26(0.85,1.86)	0.247	0.87(0.62,1.23)	0.443	0.93(0.62,1.39)	0.726
Q3 group	2.50(1.81,3.46)	<0.001	1.28(0.93,1.76)	0.126	1.39(0.98,1.97)	0.069
Q4 group	4.79(3.41,6.72)	<0.001	1.94(1.38,2.73)	<0.001	2.13(1.47,3.09)	<0.001
P for trend	​	<0.001	​	<0.001	​	<0.001

Model 1: Not adjusted.

Model 2: Adjusted by age, gender, race/ethnicity.

Model 3: Adjusted by age, gender, race/ethnicity, cigarette smoking status, alcohol consumption status, BMI, CKD Stages, Hypertension, DM, CVD, Hyperlipidemia, HEI-2015.

*Hazard ratio.

^#Confidence interval.

In the fully adjusted Model 3, Q4 remained significantly associated with all-cause mortality (HR=1.63, 1.30–2.06, P<0.001), whereas the association for Q3 was attenuated (HR=1.22, 0.98–1.53, P=0.072) and Q2 was non-significant (HR=1.00, 0.79–1.26, P=0.995); the linear trend across quartiles remained significant (P<0.001). For CVD mortality, Model 1 showed that Q3 and Q4 had HRs of 2.50 (1.81–3.46, P<0.001) and 4.79 (3.41–6.72, P<0.001), respectively, with Q2 non-significant (HR=1.26, 0.85–1.86, P=0.247) and a significant trend (P<0.001). In Model 2, only Q4 remained significant (HR=1.94, 1.38–2.73, P<0.001), while Q3 (HR=1.28, 0.93–1.76, P=0.126) and Q2 (HR=0.87, 0.62–1.23, P=0.443) were non-significant. In Model 3, Q4 remained significantly associated with CVD mortality (HR=2.13, 1.47–3.09, P<0.001), with Q3 attenuated (HR=1.39, 0.98–1.97, P=0.069) and Q2 non-significant (HR=0.93, 0.62–1.39, P=0.726); the trend across quartiles was significant (P<0.001).

Given the consistent graded associations across quartiles, the reduction in life expectancy for participants in Q4 relative to Q1 was evaluated. At the age of 45, patients in Q4 had an estimated 3.3-year reduction in life expectancy compared to Q1 (Figure 2).

Figure 2.

Loss of life expectancy. Estimated loss of life expectancy in the highest TyG-MLR quartile (Q4) versus the lowest quartile (Q1), based on Gompertz survival models adjusted for all Model 3 covariates.

Finally, as shown in Figure 3, it was observed that in the group with a higher TyG-MLR, the proportion of patients in CKD stages 3, 4, and 5 increased significantly.

Figure 3.

Prevalence of each CKD stage in different TyG-MLR quartile groups. Weighted prevalence of CKD stages 1–5 across TyG-MLR quartile groups (Q1–Q4).

Discussion

In this study based on a nationally representative sample, we systematically compared, for the first time, the performance of the TyG index and its five inflammation-derived indices (TyG-NLR, TyG-MLR, TyG-lgPLR, TyG-lgSII, TyG-SIRI) in assessing the mortality risk of CKD patients. The results showed that, except for TyG, all other indices were positively correlated with all-cause mortality and CVD mortality, indicating that the metabolic-inflammatory state is of great significance in the long-term outcomes of CKD. TyG-MLR demonstrated the best, though moderate, discriminative performance among the evaluated indices, and patients with a higher TyG-MLR had a significantly reduced life expectancy. These findings position TyG-MLR primarily as a prognostic marker that may aid in identifying high-risk CKD subgroups warranting closer clinical surveillance and optimization of evidence-based cardiovascular prevention strategies. However, given the observational design and moderate discriminative performance, these findings should be interpreted as hypothesis-generating rather than practice-changing.

Existing studies have confirmed that the TyG index, as a simple surrogate marker for insulin resistance, exhibits strong predictive ability in metabolic diseases, CVD events, and kidney diseases.^18,19 An 8-year longitudinal cohort study observed that a high TyG level was highly associated with the development of CKD ²⁰ ; Qin et al. found that a high TyG level was significantly correlated with the risks of all-cause mortality and CVD mortality in CKD patients. ²¹ However, no significant association between the TyG index and mortality risk in CKD patients was observed in this study, which may be due to the U-shaped relationship between TyG and mortality risk, as evidenced in our supplementary files (Supplementary Figure S1). Such a U-shaped association has also been observed in many other studies.^22,23

It should be noted that the traditional TyG index mainly reflects metabolic abnormalities and cannot fully capture the chronic low-grade inflammatory state that is prevalent in CKD patients. In recent years, researchers have attempted to construct composite scores that integrate TyG with peripheral blood inflammatory indicators. For example, Chen et al. observed in the general population that TyG-MLR, TyG-NLR, and TyG-SIRI were all significantly associated with the risks of all-cause mortality and CVD mortality, and their predictive ability was superior to that of TyG used alone. ¹¹ Wei et al. found in the middle-aged and elderly population that TyG-inflammatory indices were closely related to the prevalence of cataracts. ²⁴ However, in the field of CKD, relevant evidence remains relatively scarce. Although Chen et al. explored the application of TyG-inflammatory indices in the general population, they did not analyze the CKD population, nor did they adjust for eGFR or UACR in the regression analysis model. Due to the effects of uremic toxins and various other factors, insulin resistance and low-grade inflammation are widespread in CKD patients,^25,26 so the impact of CKD on insulin resistance and inflammation cannot be ignored. Therefore, this study focused on the CKD population.

The results of this study are consistent with previous literature to a certain extent. All five TyG-derived inflammatory indicators were associated with mortality risk in CKD patients, suggesting that these composite indices have robust predictive ability in this population. Moreover, ROC curve analysis indicated that the predictive performance of composite indices integrating inflammatory status was significantly enhanced. In particular, TyG-MLR demonstrated the best, though moderate, discriminative performance across multiple analyses, with the highest AUC and the largest hazard ratios in both continuous and grouped analyses. This finding has a pathophysiological basis for explanation: monocytes play a key role in CKD-related chronic inflammation and atherosclerosis. Monocytes from CKD patients are more likely to respond to inflammatory cytokines and enhance the secretion of pro-inflammatory cytokines,^27–29 while low lymphocyte levels reflect an immunosuppressive state. An elevated MLR indicates that inflammatory activation is dominant, and when combined with insulin resistance, it may form a stronger pro-pathological effect, thereby more significantly affecting the prognosis of CKD patients. Compared with other combinations such as NLR, PLR, or SII, TyG-MLR may more accurately capture the synergistic risk of “immune imbalance + metabolic disorder,” which explains its superior predictive performance over other indicators in this study.

Beyond risk stratification, our findings have potential implications for the management of high-risk CKD patients. The metabolic-inflammatory state captured by TyG-MLR reflects insulin resistance, chronic inflammation, and cardiometabolic dysregulation—pathways that are mechanistically linked to cardiovascular mortality in CKD. In this context, optimization of evidence-based cardiometabolic therapies may be particularly relevant for patients with elevated TyG-MLR. Sodium-glucose cotransporter 2 (SGLT2) inhibitors have emerged as a cornerstone of cardiovascular and renal protection in CKD and diabetes, with recent adjudicated randomized evidence demonstrating that empagliflozin and dapagliflozin reduce sudden cardiac death. ³⁰ Given the biological interconnectedness of CKD, insulin resistance, diabetes, inflammation, and cardiovascular mortality, elevated TyG-MLR could help identify a subgroup of CKD patients in whom intensification of SGLT2 inhibitor therapy, when clinically indicated, may yield particular benefit. Similarly, aggressive LDL-C reduction remains a cornerstone of cardiovascular prevention, especially in high-risk patients with CKD and metabolic-inflammatory dysregulation. ³¹ The atherogenic lipid burden interacts with insulin resistance and inflammation to accelerate cardiovascular mortality in CKD, and TyG-MLR may serve as an adjunctive tool to prioritize patients for comprehensive lipid-lowering strategies. Current guidelines support achieving very low LDL-C levels in high-risk populations, and the metabolic-inflammatory state captured by TyG-MLR may help identify patients most likely to benefit from such intensive lipid management. However, these therapeutic implications remain speculative and require validation in prospective interventional studies.

In conclusion, this study suggests the clinical significance of TyG-derived inflammatory indices in predicting mortality risk in the CKD population. It is the first to systematically compare the predictive efficacy of multiple indicators and finds that TyG-MLR consistently ranked first across the overall cohort and the split-sample internal validation. Meanwhile, CKD patients with higher TyG-MLR have a reduced life expectancy, which provides a basis for TyG-MLR to serve as a tool for health risk stratification. The parameters required for calculating TyG-MLR (triglycerides, fasting blood glucose, monocyte count, and lymphocyte count) are all derived from routine biochemical and blood routine tests, offering high accessibility and cost-effectiveness. Thus, it is expected to be applied in the health stratification management of the CKD population.

Limitations

First, NHANES is a cross-sectional survey linked to prospective mortality follow-up, and the analysis itself is observational. Therefore, the associations observed do not imply causality. In particular, reverse causality cannot be excluded: TyG-MLR may reflect underlying disease burden, subclinical comorbidities, or malnutrition-inflammation complex rather than representing a modifiable causal pathway. Therefore, our findings should be interpreted as hypothesis-generating, and external validation in independent CKD cohorts is warranted before clinical application. Second, the measurement of TyG-derived indices was based on a single assessment, which may underestimate the impact of fluctuations over time. Third, the discriminative performance of TyG-MLR was moderate (AUC 0.66–0.68), and its incremental contribution beyond standard clinical variables, while statistically significant, was modest. Finally, despite adjustment for multiple confounding factors, residual confounding from unmeasured variables cannot be excluded.

Conclusions

This study suggests that TyG-derived inflammatory indices may hold clinical value in predicting mortality risk among patients with CKD. In the first systematic comparison of multiple such indices, TyG-MLR emerged as the best-performing indicator, although its discriminative performance was moderate. Higher TyG-MLR was associated with reduced life expectancy, suggesting that TyG-MLR may serve as a tool for health risk stratification in this population. Prospective validation in independent cohorts is warranted.

ORCID iDs

Rongwen Lin https://orcid.org/0009-0000-5232-4099

Huijuan Zhao https://orcid.org/0009-0009-5378-915X

Bo Wu https://orcid.org/0000-0002-4004-1881

Ethical considerations

The ethical approval for NHANES was formally granted by the US National Center for Health Statistics Research (NCHS) Ethics Review Board—a key Institutional Review Board (IRB) responsible for overseeing research ethics. The specific protocol numbers include 98-12, 2011-17, Continuation of Protocol No. 2011-17, and 2018-01, with detailed approval records accessible at: https://www.cdc.gov/nchs/nhanes/about/erb.html?CDC_AAref_Val=https://www.cdc.gov/nchs/nhanes/irba98.htm. NHANES was conducted strictly in accordance with the Helsinki Declaration, and its research protocols obtained explicit approval from the above-mentioned National Center for Health Statistics Research Ethics Review Board—an authoritative Institutional Review Board (IRB). As NHANES is a publicly available dataset, the secondary analysis in this study is exempt from additional ethical approval under the US Health and Human Services (HHS) regulations specified in 45 CFR 46.104 (for reference: https://www.hhs.gov/ohrp/regulations-and-policy/regulations/45-cfr-46/common-rule-subpart-a-46104/index.html).

Consent to participate

Notably, written informed consent was obtained from all participants prior to their household interviews and health examinations. Participants were clearly assured that the collected data would be used solely for the stated research purposes and would not be disclosed or released to any third parties without their consent (further details available at: https://www.cdc.gov/nchs/nhanes/genetics/genetic_participants.htm).All human participants provided written informed consent prior to their involvement in NHANES.

Consent for publication

All authors agreed to publish the manuscript.

Data Availability Statement

All data are available as publicly accessible datasets through NHANES. It is open and publicly accessible through the following link: https://wwwn.cdc.gov/nchs/nhanes/Default.aspx.