Association between changes in estimated glucose disposal rate and composite kidney outcome in middle-aged and elderly individuals: a national cohort study

Abstract

Aim

To investigate the association between changes in estimated Glucose Disposal Rate (eGDR) and Composite Kidney Outcome (CKO) among middle-aged and older adults.

Methods

This study examined data from 4,007 individuals in the China Health and Retirement Longitudinal Study (CHARLS). The eGDR was estimated based on waist circumference, HbA1c, and hypertension status. Participants were stratified by cumulative eGDR quartiles and change patterns. The primary outcome was CKO, which included Chronic Kidney Disease (CKD)​ and rapid kidney function decline (RKFD). Logistic regression and restricted cubic spline (RCS) models were employed to assess the relationship between eGDR changes and outcomes, while receiver operating characteristic (ROC) curves evaluated the predictive performance of eGDR. Additionally, the Weighted Quantile Sum (WQS) regression model was used to quantify the contributions of eGDR components.

Results

Cumulative eGDR exhibited a linear and inverse relationship with the risk of CKO (OR = 0.96 (0.94 ~ 0.99), P for nonlinear = 0.214), and the greatest cumulative eGDR group (Q4) had a 53% lower risk of CKO than the reference group (OR = 0.47 (0.28 ~ 0.79), P = 0.004). The predictive performance of cumulative eGDR surpassed both baseline eGDR and traditional insulin resistance indices. HbA1c and hypertension were the most influential components in the model.

Conclusion

Long-term declines in eGDR were significantly associated with an increased risk of CKO. Cumulative eGDR showed enhanced predictive value, underscoring its promise as a biomarker for early CKD risk assessment.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13098-025-02028-5.

Introduction

Chronic kidney disease (CKD) affects over 10% of adults globally and is a major risk factor for end-stage renal disease (ESRD) and cardiovascular mortality, imposing substantial healthcare burdens worldwide [1–4]. Despite a decline in CKD prevalence in China over the past decade [5], the ongoing growth of the senior population and the rapid increase in metabolic disease incidence indicate that CKD remains a pressing public health challenge [6].

The development and progression of CKD are influenced by numerous factors, including conventional risk factors like hypertension, diabetes, and aging [7, 8], as well as novel risk factors such as obesity and metabolic disorders [9, 10]. Rapid Kidney Function Decline (RKFD), characterized by a swift reduction in glomerular filtration rate (eGFR), acts as a significant early indicator of CKD progression and is typically correlated with ESRD [11]. Consequently, identifying risk factors for RKFD and adopting early interventions are crucial for postponing CKD progression.

Insulin resistance plays a key role in CKD progression through inflammatory and metabolic pathways, forming a vicious cycle with renal dysfunction [12–14]. Although the hyperinsulinemic-euglycemic clamp (HEC) remains the gold standard for assessing insulin resistance [15], it is impractical for large-scale studies. The estimated glucose disposal rate (eGDR), derived from routine clinical measures, provides a validated and practical alternative [16].

The eGDR has been found to be significantly correlated with various disorders. For instance, multiple studies have shown that eGDR can independently predict cardiovascular events, with its predictive performance surpassing other insulin resistance indices [17–19]. Nevertheless, evidence regarding the association between eGDR and CKD remains limited. Vladu et al. discovered in a cross-sectional investigation of type 1 diabetes mellitus (T1DM) patients that eGDR has a protective impact against CKD, exceeding glycated hemoglobin and hypertension in terms of predictive performance [20]. Similarly, Peng et al. found essentially identical results in a cohort analysis of people with type 2 diabetes mellitus (T2DM), and the connection held true across gender-specific subgroups [21]. Furthermore, a randomized controlled trial (RCT) conducted in Italy demonstrated that eGDR is independently associated with microalbuminuria and macroalbuminuria in T2DM patients [22]. This finding is crucial as it extends the association between eGDR and kidney damage from mere functional decline (eGFR) to include structural damage indicated by albuminuria, thereby strengthening the biological plausibility of prior findings. However, most existing studies on eGDR have focused primarily on diabetic populations or specific demographic groups, limiting their generalizability to the broader population. While these studies have established the clinical relevance of eGDR, they have largely relied on single-timepoint measurements [23], which cannot capture the dynamic nature of insulin resistance. This is a critical limitation, as emerging evidence demonstrates that insulin resistance is a modifiable metabolic process - dietary interventions like the DASH diet and ADA guidelines have been shown to improve lipid profiles and reduce free fatty acids in diabetes [24], while green coffee extract supplementation can lower fasting glucose and triglycerides [25]. Given that longitudinal changes in insulin resistance may provide a more accurate reflection of metabolic status and disease risk than single measurements, investigating the long-term association between eGDR trajectories and CKD risk carries important scientific and clinical implications.

This study aims to investigate the association between changes in eGDR and the risk of incident CKD in a middle-aged and elderly population—a demographic particularly vulnerable to metabolic dysfunction and kidney disease - utilizing nationwide cohort data. By analyzing the dynamic changes in eGDR, we seek to offer deeper insights into the contribution of insulin resistance to the onset and advancement of CKD, as well as determine if changes in eGDR offer superior predictive performance compared to baseline measurements.

Materials and methods

Study design and population

The China Health and Retirement Longitudinal Study (CHARLS) is a national longitudinal cohort study administered by the National Development Research Institute of Peking University, with surveys conducted biennially. To date, the study has completed the baseline survey in 2011 and four follow-up surveys in 2013, 2015, 2018, and 2020 [26]. The main goal of CHARLS is to gather detailed micro-level data on households and individuals aged 45 and older, with the aim of conducting a thorough examination of population aging in China and generating evidence-based insights for policy formulation.

This research encompassed 11,847 participants who provided blood samples during the baseline wave of the survey [27]. The criteria for exclusion were as follows: (1) missing serum creatinine (SCr) or cystatin C (Cys C) measurements in blood samples; (2) baseline eGFR < 60 mL/min/1.73 m²; (3) missing WC values; (4) missing HbA1c values; (5) missing records of hypertension history; and (6) outlier values in WC. The specific criteria for inclusion and exclusion are depicted in Fig. S1. The final analytical cohort comprised 4,007 eligible participants.

Calculations of exposures

The eGDR was calculated based on WC, HbA1c, and hypertension diagnosis data collected during the baseline survey (Wave 1) and the third wave of follow-up (Wave 3), using the following formula: [28]

To assess the average exposure level of eGDR during the follow-up period, cumulative eGDR was calculated using the following formula: [29]

  

This formula for calculating cumulative exposure, adapted from prior epidemiological research on cumulative biomarkers, effectively integrates both the intensity and duration of insulin resistance, providing a more robust measure than single assessments and aligning with the study’s aim to evaluate long-term metabolic burden. The participants were then stratified into four quartiles based on cumulative eGDR.

Additionally, a K-means clustering model was employed to classify eGDR change patterns from 2011 to 2015. The optimal number of clusters was determined by evaluating solutions ranging from 1 to 10 clusters using the elbow method based on within-cluster sum of squares (SSE), where a distinct elbow was observed at k = 3. The robustness of this 3-cluster solution was further validated by a silhouette coefficient of 0.51, indicating a reasonable cluster structure. This method has been widely employed in prior research to detect different patterns of change in a variety of health markers. After determining the best model, individuals were allocated to the most likely class.

Definition of outcomes

The primary outcome of this study was defined as Composite Kidney Outcome (CKO), encompassing the incidence of CKD and RKFD. The secondary outcomes were independent events of CKD and RKFD, respectively. CKD was defined as having a sustained eGFR of less than 60 mL/min/1.73 m² at the third wave of follow-up [30]. RKFD was defined as an annualized eGFR decline rate of ≥ 5 mL/min/1.73 m² [31], equivalent to a cumulative eGFR decline of ≥ 20 mL/min/1.73 m² over the entire follow-up period. The eGFR was calculated using SCr and Cys C levels measured during the baseline survey (Wave 1) and the third follow-up wave (Wave 3), based on the combined formula developed by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) [32].

Covariates

The covariates were obtained from Wave 1 and Wave 3 surveys and encompassed the following categories of variables: (1) Social and demographic characteristics: including age, gender, marital status, smoking, and drinking statuses; (2) Physiological parameters: including blood pressure, body mass index (BMI), and others; (3) Laboratory tests: including three blood cell lines, blood glucose, lipid profile, and renal function indicators; (4) Comorbidities: including hypertension, diabetes, and cardiovascular diseases; (5) Disease treatment: including antihypertensive treatment, diabetes treatment, and lipid-lowering treatment. Smoking and drinking statuses were categorized as never, ever, and current, respectively. Information on comorbidities and disease treatments was self-reported through questionnaires. BMI was determined based on the equation: BMI (kg/m²) = weight (kg)/height² (m²). Additionally, the Metabolic Score for Insulin Resistance (METS-IR) was used as a comparative indicator, using the following formula: [33]

The number and proportion of missing covariate data are detailed in Fig. S2. Missing data were handled using multiple imputation, implemented with the “mice” package in R software.

Statistical analyses

The Shapiro-Wilk test was employed to evaluate the normal distribution of continuous variables (Table S1). Continuous variables exhibiting normal distribution were expressed as mean ± standard deviation (Mean ± SD), while those not conforming to normal distribution were expressed as median (interquartile range, IQR). Categorical variables were represented as frequencies (percentages). Descriptive analyses were conducted utilizing ANOVA or χ² tests, contingent upon the variable type. Multicollinearity between variables was assessed through the generalized variance inflation factor (GVIF), and those with GVIF > 5 were removed from the models. (Table S2).

To evaluate the role of eGDR as an independent predictor, three logistic regression models were developed: model 1 was unadjusted; model 2 was adjusted for age and gender; and model 3 included adjustments for age, gender, marital status, smoking status, drinking status, SBP, DBP, BMI, WBC, Hb, PLT, FPG, BUN, UA, TC, TG, HDL-C, hs-CRP, diabetes mellitus, heart disease, stroke, hyperlipidemia, antihypertensive treatment, diabetes treatment, and lipid-lowering treatment. The results were expressed as Odds Ratios (ORs) accompanied by 95% confidence intervals (95% CIs). Cumulative eGDR was incorporated into the models as both a continuous and a categorical variable, with its change classes also included as categorical variables. Furthermore, trend analysis was performed to assess the linear relationship between cumulative eGDR and the outcomes.

To investigate the dose-response relationship between eGDR and outcomes, restricted cubic spline (RCS) models were developed, with knots positioned at the 10th, 50th, and 90th percentiles of the independent variable. The predictive performance of baseline eGDR, METS-IR, and cumulative eGDR was evaluated using receiver operating characteristic (ROC) curves and the area under the curve (AUC). Net Reclassification Improvement (NRI) and Integrated Discrimination Improvement (IDI) were employed to assess the incremental value of cumulative eGDR in predicting outcomes.

Subgroup analyses were conducted by stratifying participants based on age (> 60 years vs. ≤ 60 years), gender (male vs. female), and BMI (> 24 kg/m² vs. ≤ 24 kg/m²), with interaction effects between subgroups and eGDR evaluated using ANOVA. To evaluate the contributions of WC, hypertension, and HbA1c to the overall effect of eGDR, a weighted quantile sum (WQS) regression model was applied, with 1,000 iterations conducted using the bootstrap method. The WQS model assigned weights to WC, hypertension, and HbA1c, constrained to a range of 0 to 1 with a sum of 1, where higher weights indicated greater importance of the corresponding variable in predicting outcomes.

All statistical analyses were performed using R software (version 4.4.2), and a two-sided P-value < 0.05 was considered statistically significant.

Results

Baseline characteristics of the study population

This study included 4,007 participants with a median age of 58 years, 45.1% of whom were male. Table 1 illustrates the baseline characteristics of participants stratified by the K-means clustering method. In 2011, the mean eGDR of the participants was 10.36mg/kg/min. As people aged, the eGDR declined to 9.59mg/kg/min by 2015. Additionally, the cumulative eGDR for the entire cohort was 39.67mg/kg/min. Table S3 further provides detailed baseline characteristics stratified by quartiles of cumulative eGDR.

Table 1.

Baseline characteristics of participants stratified by K-means clustering analysis of eGDR

Group	Overall n = 4007	Class 1 n = 974	Class 2 n = 2083	Class 3 n = 950	P value
Demographic characteristics
Age (years)	58.00 (52.00, 65.00)	60.00 (54.00, 66.00)	57.00 (50.00, 64.00)	58.00 (52.00, 65.00)	< 0.001
Gender, n (%)
Female	2200 (54.9)	590 (60.6)	1078 (51.8)	532 (56.0)	< 0.001
Male	1807 (45.1)	384 (39.4)	1005 (48.2)	418 (44.0)
Marital status, n (%)
Married	3394 (84.7)	813 (83.5)	1786 (85.7)	795 (83.7)	0.162
Others	613 (15.3)	161 (16.5)	297 (14.3)	155 (16.3)
Smoking, n (%)
Never	2504 (62.5)	652 (66.9)	1250 (60.0)	602 (63.4)	< 0.001
Current	1181 (29.5)	210 (21.6)	687 (33.0)	284 (29.9)
Ever	322 (8.0)	112 (11.5)	146 (7.0)	64 (6.7)
Drinking, n (%)
Never	2432 (60.7)	636 (65.3)	1223 (58.7)	573 (60.3)	< 0.001
Current	1258 (31.4)	235 (24.1)	725 (34.8)	298 (31.4)
Ever	317 (7.9)	103 (10.6)	135 (6.5)	79 (8.3)
Physiological parameters
SBP (mmHg)	126.00 (113.50, 141.00)	141.50 (127.50, 156.00)	118.50 (108.50, 130.25)	129.50 (117.50, 143.00)	< 0.001
DBP (mmHg)	74.00 (66.50, 82.50)	81.50 (73.50, 89.00)	70.50 (63.50, 78.00)	75.50 (68.50, 84.00)	< 0.001
WC2011 (cm)	85.00 (78.40, 92.00)	91.20 (86.00, 98.00)	80.40 (75.80, 86.00)	90.80 (81.00, 96.40)	< 0.001
WC2015 (cm)	86.00 (79.35, 93.00)	92.30 (86.60, 99.00)	82.00 (76.40, 87.05)	92.00 (82.20, 98.00)	< 0.001
BMI (kg/m2)	23.20 (20.97, 25.80)	25.61 (23.35, 28.15)	21.86 (20.13, 23.67)	24.81 (21.93, 27.43)	< 0.001
Laboratory tests
WBC (109/L)	5.90 (4.90, 7.20)	6.10 (5.10, 7.50)	5.85 (4.90, 7.10)	5.90 (4.90, 7.10)	< 0.001
Hb (g/dL)	14.30 (13.10, 15.60)	14.50 (13.30, 15.80)	14.10 (13.00, 15.40)	14.40 (13.20, 15.60)	< 0.001
PLT (109/L)	207.00 (162.00, 254.00)	208.00 (165.00, 259.75)	205.00 (161.00, 252.00)	207.50 (162.00, 256.75)	0.202
HbA1c2011 (%)	5.10 (4.90, 5.40)	5.20 (5.00, 5.60)	5.10 (4.80, 5.30)	5.20 (4.90, 5.50)	< 0.001
HbA1c2015 (%)	5.80 (5.50, 6.20)	6.00 (5.70, 6.50)	5.70 (5.50, 6.00)	5.90 (5.60, 6.40)	< 0.001
FPG (mg/dL)	102.42 (94.32, 112.68)	105.84 (98.64, 120.24)	99.54 (92.16, 108.18)	104.76 (96.12, 116.28)	< 0.001
BUN (mg/dL)	14.96 (12.49, 17.84)	14.82 (12.52, 17.84)	15.24 (12.69, 17.94)	14.71 (12.25, 17.48)	0.01
eGFRcr-cys2011 (ml/min/1.73m2)	91.72 (80.54, 103.13)	88.88 (78.35, 100.70)	93.07 (81.92, 104.00)	91.39 (81.32, 103.79)	< 0.001
eGFRcr-cys2015 (ml/min/1.73m2)	98.58 (85.75, 109.18)	93.74 (81.15, 106.09)	101.05 (88.27, 110.53)	97.82 (85.65, 108.16)	< 0.001
Cys C (mg/L)	0.96 (0.85, 1.08)	0.97 (0.86, 1.10)	0.95 (0.84, 1.08)	0.96 (0.84, 1.07)	0.052
UA (mg/dL)	4.22 (3.50, 5.05)	4.43 (3.67, 5.36)	4.09 (3.42, 4.91)	4.28 (3.53, 5.04)	< 0.001
TC (mg/dL)	190.59 (166.62, 214.95)	196.39 (173.20, 222.58)	185.95 (163.73, 210.70)	192.53 (167.78, 215.34)	< 0.001
TG (mg/dL)	106.20 (75.22, 155.76)	129.65 (92.04, 187.40)	95.58 (69.03, 133.63)	114.17 (79.65, 180.32)	< 0.001
HDL-C (mg/dL)	49.10 (40.21, 59.54)	45.23 (37.50, 53.35)	52.19 (43.30, 62.24)	46.39 (37.89, 56.83)	< 0.001
LDL-C (mg/dL)	114.05 (92.78, 136.86)	119.85 (96.26, 143.04)	110.95 (91.62, 132.60)	117.33 (92.78, 139.47)	< 0.001
hs-CRP (mg/L)	0.98 (0.53, 1.98)	1.40 (0.73, 2.73)	0.80 (0.46, 1.63)	1.08 (0.57, 1.99)	< 0.001
eGDR2011 (mg/kg/min)	10.36 (8.10, 11.19)	6.66 (5.98, 7.26)	11.08 (10.63, 11.57)	9.70 (9.08, 10.31)	< 0.001
eGDR2015 (mg/kg/min)	9.59 (6.89, 10.62)	6.07 (5.29, 6.72)	10.58 (10.11, 11.13)	8.14 (7.06, 9.00)	< 0.001
Cumulative eGDR (mg/kg/min)	39.67 (30.50, 43.48)	25.63 (22.75, 27.83)	43.31 (41.53, 45.21)	35.89 (32.80, 37.80)	< 0.001
Comorbid diseases
Hypertension2011, n (%)	1070 (26.7)	931 (95.6)	0 (0.0)	139 (14.6)	< 0.001
Hypertension2015, n (%)	1431 (35.7)	961 (98.7)	10 (0.5)	460 (48.4)	< 0.001
Diabetes, n (%)	241 (6.0)	146 (15.0)	33 (1.6)	62 (6.5)	< 0.001
Heart disease, n (%)	482 (12.0)	234 (24.0)	144 (6.9)	104 (10.9)	< 0.001
Stroke, n (%)	81 (2.0)	40 (4.1)	26 (1.2)	15 (1.6)	< 0.001
Hyperlipidemia, n (%)	406 (10.1)	222 (22.8)	95 (4.6)	89 (9.4)	< 0.001
Therapy for diseases
Antihypertensive treatment, n (%)	774 (19.3)	655 (67.2)	9 (0.4)	110 (11.6)	< 0.001
Diabetes treatment, n (%)	153 (3.8)	96 (9.9)	18 (0.9)	39 (4.1)	< 0.001
Lipid lowering treatment, n (%)	208 (5.2)	131 (13.4)	36 (1.7)	41 (4.3)	< 0.001
Outcomes
CKO, n (%)	255 (6.4)	96 (9.9)	90 (4.3)	69 (7.3)	< 0.001
CKD, n (%)	90 (2.2)	44 (4.5)	23 (1.1)	23 (2.4)	< 0.001
RKFD, n (%)	206 (5.1)	73 (7.5)	79 (3.8)	54 (5.7)	< 0.001

SBP, Systolic Blood Pressure; DBP, Diastolic Blood Pressure; WC, Waist Circumference; BMI, Body Mass Index; WBC, White Blood Cell count; Hb, Hemoglobin; PLT, Platelet count; HbA1c, Glycated Hemoglobin A1c; FPG, Fasting Plasma Glucose; BUN, Blood Urea Nitrogen; eGFR, Estimated Glomerular Filtration Rate (creatinine-cystatin C); Cys C, Cystatin C; UA, Uric Acid; TC, Total Cholesterol; TG, Triglycerides; HDL-C, High-Density Lipoprotein Cholesterol; LDL-C, Low-Density Lipoprotein Cholesterol; hs-CRP, High-Sensitivity C-Reactive Protein; eGDR, Estimated Glucose Disposal Rate; CKO, Composite kidney outcome; CKD, Chronic Kidney Disease; RKFD, Rapid Kidney Function Decline

Regression modeling and predictive performance analysis

Over the four-year follow-up period, 255 CKO were recorded, accounting for 6.4% of the study population, with 90 incident CKD cases and 206 RKFD cases. Logistic regression analyses were conducted to examine the associations between changes in eGDR and outcomes (Table 2 and S4). First, we looked at the trajectory of eGDR changes using cumulative eGDR. With Quartile 1 as the reference group, the risks of CKO, CKD, and RKFD showed a stepwise decrease with increasing cumulative eGDR (P for trend < 0.05). Specifically, the greatest cumulative eGDR group (Quartile 4) had significantly lower CKO risk than the reference group (OR = 0.34 (0.23 ~ 0.50), P < 0.001), representing an approximately 60% risk reduction. When eGDR was treated as a continuous variable, a similar risk reduction trend emerged, with each 1-unit increase in eGDR associated with a 5% decrease in CKO risk (OR = 0.95 (0.94 ~ 0.96), P < 0.001). These findings persisted in Model 2 and Model 3 following additional adjustments for confounders.

Table 2.

Logistic regression analysis between eGDR and the primary outcome

Model	Model 1		Model 2		Model 3
	OR (95%CI)	P value	OR (95%CI)	P value	OR (95%CI)	P value
CKO
Quartile
1	1.00 (Reference)		1.00 (Reference)		1.00 (Reference)
2	0.69 (0.50 ~ 0.95)	0.022	0.73 (0.53 ~ 1.00)	0.051	0.75 (0.50 ~ 1.13)	0.171
3	0.42 (0.29 ~ 0.60)	< 0.001	0.46 (0.32 ~ 0.67)	< 0.001	0.53 (0.33 ~ 0.86)	0.010
4	0.34 (0.23 ~ 0.50)	< 0.001	0.37 (0.25 ~ 0.55)	< 0.001	0.47 (0.28 ~ 0.79)	0.004
Continuous	0.95 (0.94 ~ 0.96)	< 0.001	0.95 (0.94 ~ 0.97)	< 0.001	0.96 (0.94 ~ 0.99)	0.002
P for trend		< 0.001		< 0.001		0.002
Cluster
1	1.00 (Reference)		1.00 (Reference)		1.00 (Reference)
2	0.41 (0.31 ~ 0.56)	< 0.001	0.45 (0.33 ~ 0.61)	< 0.001	0.58 (0.37 ~ 0.92)	0.019
3	0.72 (0.52 ~ 0.99)	0.043	0.76 (0.55 ~ 1.05)	0.094	0.82 (0.54 ~ 1.25)	0.352

OR, Odds Ratio; CI: Confidence Interval; CKO, Composite kidney outcome

Model1: Crude

Model2: Adjust: Age, Gender

Model3: Adjust: Age, Gender, Marital status, Smoking, Drinking, SBP, DBP, BMI, WBC, Hb, PLT, FPG, BUN, UA, TC, TG, HDL-C, hs-CRP, Diabetes, Heart disease, Stroke, Hyperlipidemia, Antihypertensive treatment, Diabetes treatment, Lipid lowering treatment

To uncover different patterns of dynamic changes in insulin resistance, we employed the K-means clustering algorithm to analyze eGDR data obtained from two surveys. The optimal number of clusters was identified as three using the elbow rule and silhouette coefficient. As shown in Fig. 1A-C, participants were divided into three clusters:

Fig. 1.

Cluster analysis of eGDR change patterns from 2011 to 2015. (A) The elbow method for determining the optimal number of clusters. (B) Scatter plot of eGDR₂₀₁₁ and eGDR₂₀₁₅ by cluster. (C) Dumbbell plot of mean eGDR from 2011 to 2015 by different classes. (D-E) Ridgeline plots of eGDR distributions in 2011 (D) and 2015 (E) by cluster, illustrating the distinct characteristics and temporal evolution of each cluster’s metabolic phenotype

Cluster 1 (Low-stable)

This group exhibited persistently low eGDR levels throughout the study period, indicating chronically impaired glucose disposal capacity. The stable trajectory suggests these individuals may have established insulin resistance with limited potential for spontaneous improvement.

Cluster 2 (High-stable)

Characterized by consistently high eGDR values, this group maintained optimal insulin sensitivity over time. The stability of their metabolic profile suggests inherent protective factors against insulin resistance development.

Cluster 3 (Moderate-rapid decline)

This group showed intermediate baseline eGDR levels but experienced a pronounced decline during follow-up. The rapid deterioration in glucose disposal capacity identifies individuals at critical transition points who may benefit most from early intervention.

Compared to Cluster 1 (Low-stable), Cluster 2 (High-stable) exhibited a significantly lower risk of CKO (OR = 0.58 (0.37 ~ 0.92), P = 0.019). In Cluster 3 (Moderate-rapid decline), a trend toward reduced risk was observed, although this trend did not reach statistical significance (OR = 0.82 (0.54 ~ 1.25), P = 0.352). Secondary outcomes, such as the incidence of CKD and RKFD, followed a similar pattern: the group with persistently high eGDR had the lowest risk, the group with moderate to rapid eGDR decline had an intermediate risk, and the group with consistently low eGDR levels had the highest.

As shown in Fig. 2A-C, after adjusting for potential confounders, eGDR was independently associated with all study outcomes (P for overall < 0.05). Notably, this association exhibited a clear linear trend (P for nonlinear > 0.05), suggesting an inverse dose-response association between eGDR and the risk of adverse outcomes.

Fig. 2.

Dose-response relationships of cumulative eGDR with clinical outcomes and its predictive performance. (A) The linear relationship between cumulative eGDR and CKO. (B) The linear relationship between cumulative eGDR and CKD. (C) The linear relationship between cumulative eGDR and RKFD. (D) AUC values of variables for CKO. (E) AUC values of variables for CKD. (F) AUC values of variables for RKFD

Then, we assessed the predictive performance of cumulative eGDR for each outcome using ROC curves. As shown in Fig. 2D-F, cumulative eGDR demonstrated superior predictive performance for CKO, CKD, and RKFD compared to baseline eGDR. To further compare the predictive performance of eGDR with traditional insulin resistance indices, we introduced METS-IR as a reference. The results similarly showed that eGDR outperformed METS-IR in predicting all outcomes in this study. Furthermore, NRI and IDI analyses revealed that the improvement in predictive performance with cumulative eGDR was statistically significant (Table S5).

Subgroup analysis

Figure 3 shows the associations between cumulative eGDR, eGDR classes, and CKO, stratified by age, gender, and BMI. No significant interactions were found between subgroups and eGDR (P for interaction > 0.05), indicating that the association between eGDR and outcomes remained consistent across these subgroups.

Fig. 3.

Subgroup analysis of composite kidney outcomes by eGDR grouping methods​. (A) Forest plot of risk estimates by eGDR quartiles. (B) Forest plot of risk estimates by eGDR classification

WQS analyses

We also employed the WQS regression model to comprehensively analyze eGDR. After adjusting for all confounders in Model 3, Fig. 4 presents the weight distribution of the components of eGDR. At the baseline survey in 2011, hypertension (37.96%) and HbA1c (37.3%) had similar weights, while WC had the lowest contribution weight at 24.74%. After four years of follow-up, the weight of HbA1c grew to 42.73%, overtaking hypertension (38.86%) as the leading contributor to CKO, while the weight of WC reduced to 18.41%, remaining the third-ranked.

Fig. 4.

Estimated weights of eGDR components assigned by the WQS Model in 2011 and 2015. (A) Weight distribution of various components in 2011. (B) Weight distribution of various components in 2015

Discussion

Our study demonstrates that longitudinal eGDR trajectories outperform both traditional insulin resistance indices and single baseline measurements in predicting CKD risk, underscoring the value of dynamic metabolic monitoring. These findings support a paradigm shift from static to dynamic risk stratification in clinical practice. The feasibility of integrating cumulative eGDR into electronic health records is high, as it relies on routinely measured parameters—HbA1c, blood pressure, and waist circumference. For at-risk middle-aged and elderly individuals, annual eGDR calculation could become a standard preventive measure, where a declining or persistently low trajectory would serve as an early warning, triggering intensified lifestyle and renal function monitoring before significant eGFR decline occurs. The predominant contributions of blood pressure and glycemia to eGDR’s predictive weight reinforce the vascular-metabolic axis in CKD pathogenesis, supporting integrated management targeting all three components. Moreover, the linear eGDR–CKD relationship suggests that even modest improvements within normal metabolic ranges—such as slight reductions in waist circumference or blood pressure—can cumulatively lower CKD risk, offering tangible goals for preventive care.

Our results align with and extend previous research on insulin resistance and kidney disease. While Liu et al. reported associations between multiple insulin resistance indices and RKFD [34], and Luo et al. linked lower eGDR to renal function decline only in those with dysglycemia [35], our study demonstrates that eGDR trajectories predict CKD risk consistently across a general population, overcoming prior limitations of disease-specific or regional samples. Further supporting the value of longitudinal assessment, Chen et al. showed that incorporating variability and cumulative exposure of the TyG index improved CKD prediction [36], consistent with our findings. Moreover, eGDR outperformed traditional indices like METS-IR, likely due to its integration of multidimensional metabolic information, enhancing its suitability for dynamic risk assessment in clinical practice. Notably, the weighting of eGDR components may vary by outcome, as evidenced by Yao et al.‘s finding that waist circumference and hypertension predominated in stroke prediction, with HbA1c playing a lesser role [29], highlighting eGDR’s adaptability across conditions.

Several pathophysiological mechanisms may explain the link between insulin resistance and CKD. Current research indicates that insulin resistance may have a role in the initiation and advancement of CKD via many pathways. Initially, insulin resistance may provoke an intensification of systemic inflammatory response and oxidative stress, which can directly injure the kidneys [37]. Previous studies have shown that hyperinsulinemia can directly stimulate the renin-angiotensin system (RAS), causing intraglomerular hypertension and proteinuria, which subsequently exacerbates renal impairment [38]. Secondly, insulin resistance frequently coexists with several metabolic diseases, such as hyperglycemia, hypertension, and dyslipidemia [39], which collectively exacerbate the renal burden. Hyperglycemia can directly harm renal endothelial cells and cause fibrosis by facilitating the formation of advanced glycosylation end products (AGEs) [40], a process further corroborated in our WQS model research. Moreover, a bidirectional relationship exists between insulin resistance and CKD: insulin resistance facilitates the initiation and advancement of CKD through the previously mentioned mechanisms, while CKD patients frequently exhibit diminished insulin sensitivity [41], thereby aggravating metabolic disorders and resulting in a detrimental feedback loop. As a surrogate marker of insulin resistance, the decline in eGDR not only reflects a reduction in insulin sensitivity but may also indicate the progression of underlying pathophysiological processes, thereby providing an important basis for the early warning of CKD.

The strengths of this study lie in its utilization of national cohort data with a substantial and representative sample size, allowing for a fuller depiction of the relationship between eGDR and CKD. The longitudinal design also allowed us to evaluate the predictive value of dynamic changes in eGDR for various kidney outcomes, so overcoming the constraints of single-time point assessments. We further validated the robustness of our findings through multivariable adjustments and subgroup analyses. Nonetheless, it is essential to acknowledge the existing limitations of this study. First, while we utilized CHARLS data, certain important variables like medication dosages and detailed treatment regimens were unavailable, potentially affecting our assessment of renal function. Second, our exclusion criteria, while necessary for study design, may have introduced selection bias; specifically, the exclusion of a substantial number of participants due to missing follow-up data or key variables at baseline could limit the generalizability of our findings if these individuals systematically differed from the analyzed cohort. Third, the reliance on eGFR alone, without urine biomarkers (such as albumin-to-creatinine ratio) or renal imaging data, represents a significant limitation. This may have led to the misclassification of some participants, particularly those with early-stage chronic kidney disease or non-glomerular pathologies, and consequently to an underestimation of the true prevalence of declining renal function. Although we used both SCr and Cys C to estimate the glomerular filtration rate, and Cys C has been shown to be highly sensitive to early renal impairment [42], achieving a comprehensive assessment of renal function remains difficult. Finally, as this study focused on a middle-aged and elderly Chinese population, further validation in multi-ethnic cohorts—particularly those with differing metabolic risk profiles (e.g., populations with varying obesity prevalence or younger cohorts with early metabolic dysfunction)—is needed. Future research should develop automated eGDR tracking tools and test whether eGDR-guided management improves hard renal outcomes.

Conclusions

This study, based on nationwide cohort data, reveals the independent association between dynamic changes in the surrogate insulin resistance index eGDR and the risk of CKD in a middle-aged and elderly population. To our knowledge, this is among the most comprehensive longitudinal studies to characterize eGDR trajectories and their association with CKD development. The findings demonstrate that a sustained decline in eGDR is significantly associated with an increased risk of CKD, and its cumulative assessment value outperforms baseline levels and traditional insulin resistance indices in disease prediction. By integrating multidimensional metabolic parameters, eGDR provides a novel and practical biomarker for the early identification of individuals at high risk for CKD. Future research should validate the generalizability of eGDR in broader populations and develop personalized intervention models based on dynamic changes, aiming to optimize insulin resistance management strategies, effectively delay CKD progression, and improve patient outcomes.

Abbreviations

AGEs

Advanced glycosylation end products

ANOVA

Analysis of variance

AUC

Area under the curve

BMI

Body mass index

BUN

Blood urea nitrogen

CHARLS

China health and retirement longitudinal study

CKD

Chronic kidney disease

CKO

Composite kidney outcome

CKD-EPI

Chronic kidney disease epidemiology collaboration

Cys C

Cystatin C

DBP

Diastolic blood pressure

eGDR

Estimated glucose disposal rate

eGFR

Estimated glomerular filtration rate

ESRD

End-stage renal disease

FPG

Fasting plasma glucose

GVIF

Generalized variance inflation factor

Hb

Hemoglobin

HbA1c

Glycated hemoglobin

HEC

Hyperinsulinemic-euglycemic clamp

HDL-C

High-density lipoprotein cholesterol

hs-CRP

High-sensitivity C-reactive protein

IDI

Integrated discrimination improvement

IQR

Interquartile range

KDIGO

Kidney disease: improving global outcomes

LAP

Lipid accumulation product

LDL-C

Low-density lipoprotein cholesterol

METS-IR

Metabolic score for insulin resistance

NRI

Net reclassification improvement

OR

Odds ratio

PLT

Platelet count

RAS

Renin-angiotensin system

RCS

Restricted cubic spline

RCT

Randomized controlled trial

RKFD

Rapid kidney function decline

ROC

Receiver operating characteristic

SBP

Systolic blood pressure

SCr

Serum creatinine

TC

Total cholesterol

TG

Triglycerides

T1DM

Type 1 diabetes mellitus

T2DM

Type 2 diabetes mellitus

TyG

Triglyceride-glucose index

UA

Uric acid

WC

Waist circumference

WBC

White blood cell count

WQS

Weighted quantile sum

Author contributions

Weiyi Zhou : Conceptualization, Formal analysis, Investigation, Methodology, Writing – original draft. Guangdong Wang : Data curation, Software, Validation, Visualization, Writing – review & editing.Yaxin Zhang : Project administration, Resources, Supervision, Writing – review & editing.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

The data supporting the findings of this study are available in the CHARLS website (http://charls.pku.edu.cn/en).

Declarations

Ethics approval and consent to participate

Ethical approval for this program was granted by the Biomedical Ethics Review Board of Peking University, with approval numbers IRB00001052-11015 and IRB00001052-11014. All participants provided informed consent before inclusion in this program.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.