Association Between Estimated Glucose Disposal Rate and Chronic Kidney Disease in Elderly Patients with Type 2 Diabetes Mellitus: A Cross-Sectional Study in Zhejiang, China

Abstract

Purpose

To examine the association between estimated glucose disposal rate (eGDR) and study-defined chronic kidney disease (CKD) among elderly patients with type 2 diabetes mellitus (T2DM) in Zhejiang Province, China.

Methods

This cross-sectional study included 713 elderly T2DM patients (≥60 years) from Zhejiang Province, conducted between March and November 2018 as part of the China National Diabetic Chronic Complications Study. Study-defined CKD was operationally defined as an estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m² and/or a urinary albumin-to-creatinine ratio (UACR) ≥30 mg/g. Multivariable logistic regression was used to assess the association of eGDR with study-defined CKD. Restricted cubic spline (RCS) analysis examined the shape of the association. Generalized additive models evaluated associations with UACR and eGFR, and receiver operating characteristic (ROC) analysis assessed discrimination.

Results

Among 713 participants, 245 (34.36%) met the study-defined CKD criteria. Each 1-unit increase in eGDR was associated with lower odds of study-defined CKD after multivariable adjustment (OR = 0.78, 95% CI: 0.70–0.86; p < 0.001). Compared with the lowest eGDR tertile, the adjusted ORs (95% CIs) were 0.51 (0.35–0.76) for the middle tertile and 0.31 (0.20–0.48) for the highest tertile (both p < 0.001). RCS analysis demonstrated an approximately linear inverse association between eGDR and study-defined CKD (p for nonlinearity > 0.05). eGDR was inversely associated with UACR (ρ = −0.277, p < 0.001) whereas the association between eGDR and eGFR was not statistically significant. The discriminative ability of eGDR for study-defined CKD was modest (AUC = 0.65, 95% CI: 0.61–0.70). The direction of the association was generally consistent across subgroups, with no significant interactions.

Conclusions

Lower eGDR was associated with a higher prevalence of study-defined CKD in elderly patients with T2DM, with an approximately linear cross-sectional association. Prospective studies are needed to establish temporality and assess whether eGDR adds information beyond established CKD assessment.

Introduction

Type 2 diabetes mellitus (T2DM) is a major global public health epidemic, with an estimated 537 million adults affected worldwide in 2021, a number projected to rise to 783 million by 2045.¹ Among the most devastating complications of T2DM is chronic kidney disease (CKD), which not only increases the risk of end-stage renal disease (ESRD) but also markedly elevates cardiovascular morbidity and all-cause mortality.^2,3 In China, diabetic kidney disease (DKD) is an important renal complication of diabetes and contributes substantially to the burden of CKD.^4,5 The aging population further compounds this challenge, as elderly patients with T2DM exhibit a higher susceptibility to renal function decline due to longer disease duration, polypharmacy, and age-related physiological changes.^6,7

Insulin resistance (IR) is a fundamental pathophysiological hallmark of T2DM and has been independently linked to the development and progression of CKD.⁸ However, the conventional gold standard for assessing IR, the hyperinsulinemic-euglycemic clamp, is invasive, labor-intensive, and impractical for large-scale epidemiological studies.⁹ Surrogate markers such as the homeostasis model assessment of IR (HOMA-IR) are more feasible but require fasting insulin measurements, which are not routinely available in many clinical settings, particularly in primary care or large cohort studies.

The estimated glucose disposal rate (eGDR) has emerged as a validated, cost-effective, and non-invasive surrogate for IR, originally developed in type 1 diabetes and subsequently extended to T2DM.^10,11 eGDR is calculated using readily available clinical parameters: waist circumference (WC), hypertension status, and hemoglobin A1c (HbA1c) level.¹² A growing body of evidence has demonstrated that lower eGDR levels are associated with increased risks of cardiovascular disease, diabetic retinopathy, and peripheral neuropathy.^13–15 Nevertheless, the relationship between eGDR and CKD in elderly patients with T2DM remains insufficiently defined. Most existing studies have primarily focused on middle-aged or general populations, with limited evidence specifically addressing elderly individuals.

Inflammation may provide an additional biological link among IR, T2DM, and CKD. Recent work has evaluated eGDR together with systemic inflammatory indices, indicating close connections between metabolic dysfunction and inflammatory burden.¹⁶ Chronic low-grade inflammation is a recognized feature of diabetes,¹⁷ and innate immune activation and inflammatory signaling contribute to CKD and its cardiovascular complications.¹⁸ These observations provide further rationale for examining eGDR in relation to CKD.

T2DM represents an important public health concern in Zhejiang Province, eastern China.¹⁹ However, to our knowledge, no previous study has specifically examined the association between eGDR and CKD among elderly patients with T2DM in this population. In older adults, age-related changes in body composition and kidney function can affect both metabolic and renal measures.^20,21 These characteristics support the need for focused investigation of the association between eGDR and CKD in this age group.

Therefore, we conducted this cross-sectional study to evaluate the association between eGDR and study-defined CKD in elderly patients with T2DM in Zhejiang, China. We also examined the shape of the association, assessed the relationships of eGDR with UACR and eGFR separately, and described the discriminative ability of eGDR for study-defined CKD.

Materials and Methods

Study Design and Population

This cross-sectional study was conducted from March to November 2018 in Zhejiang Province, eastern China, as part of the China National Diabetic Chronic Complications Study (China DiaChronic Study). The parent study used a multistage stratified random-sampling design based on disease surveillance points and local diabetes management registration systems. Two urban districts and two rural counties were selected as study sites according to the national sampling framework. Four neighborhoods or villages were then selected within each study site, and participants were sampled according to prespecified age and sex strata.²²

Under the parent protocol, adults aged ≥18 years with diagnosed diabetes who had lived in the study area for at least 6 months during the preceding year were eligible. Pregnant women and individuals with severe health conditions that prevented participation, including those who were bedridden or had intellectual disability, were excluded. The Zhejiang component had a planned sample of 1920 participants with T2DM. Of these, 134 did not complete the required physical examinations or laboratory testing, and 30 did not complete the questionnaire survey, leaving 1756 participants (91.5% of the planned sample) with complete questionnaire, physical examination, and laboratory data. For the present analysis, all participants aged ≥60 years were selected, while 1043 participants aged <60 years were excluded. The final analytic sample therefore consisted of 713 elderly participants with T2DM. No additional sampling was performed for the present study. The participant selection process is shown in Figure 1.

Figure 1.

The detailed selection process of participants.

Data Collection

Data collection was conducted at local hospitals or community health service centers by trained staff from local Centers for Disease Control and Prevention (CDCs) and participating medical institutions. All investigators completed standardized training before fieldwork. Face-to-face interviews were conducted using Android tablets to collect demographic characteristics, lifestyle factors, and medical history. Questionnaire responses and physical examination results were entered directly into the tablet-based data collection system, which incorporated mandatory entry, predefined range checks, and logical consistency checks. Physical examinations followed a standard protocol. Height and weight were measured to the nearest 0.1 cm and 0.1 kg, respectively, and WC was measured to the nearest 0.1 cm at the midpoint between the lower rib margin and iliac crest. After a 5-minute seated rest, blood pressure was measured three times at 1-minute intervals using an automated OMRON device, and the average was used for analysis. As part of parent-study quality control, provincial supervisors re-examined at least 5% of participants to assess adherence to standardized field procedures.

After an overnight fast, venous blood and random urine samples were collected. Fasting plasma glucose (FPG) was measured in qualified local laboratories using the hexokinase method under the unified quality-control program of the parent study. Before FPG testing, participating laboratories verified repeatability and intermediate precision using uniformly issued quality-control materials. During the testing period, real-time quality control using low-, medium-, and high-concentration control materials was performed on each testing day. Other biochemical measurements were performed centrally at Guangzhou KingMed Diagnostics Group Co., Ltd. HbA1c was measured by high-performance liquid chromatography using a Bio-Rad D10 analyzer; total cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C) were measured enzymatically using a Roche cobas c701/702 analyzer. Serum uric acid (SUA), serum creatinine, and urinary creatinine were measured by enzymatic methods, and urinary albumin was measured by immunoturbidimetry. The parent study used standardized equipment and laboratory quality-control procedures; HbA1c, lipid, creatinine, urinary albumin, and urinary creatinine measurements were centralized.²² The parent study protocol did not specify whether laboratory technicians were blinded to participant characteristics. However, all biochemical measurements were performed using standardized automated procedures, and study-defined CKD status was subsequently derived from prespecified eGFR and UACR criteria.

Definition of Variables

For the purposes of this analysis, study-defined CKD was operationally defined as an estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73 m² calculated using the CKD-EPI equation and/or a urinary albumin-to-creatinine ratio (UACR) ≥ 30 mg/g.²³ Because these measurements were obtained at a single study visit, chronicity for at least 3 months could not be confirmed. Hypertension was defined as systolic blood pressure (SBP) ≥ 140 mmHg and/or diastolic blood pressure (DBP) ≥ 90 mmHg, or a self-reported history of hypertension;²⁴ and the eGDR was calculated as .²⁵ In this formula, waist circumference was expressed in centimeters, hypertension was coded as a binary variable (1 = present, 0 = absent), and HbA1c was expressed as a percentage (%). Sex was categorized as male or female; educational level was classified as primary school or below, secondary school or above; residence was dichotomized as urban or rural; current smoking was defined as daily or occasional cigarette use; alcohol consumption was defined as any intake within the past 30 days; dyslipidemia was defined as TC ≥ 6.22 mmol/L, TG ≥ 2.26 mmol/L, LDL-C ≥ 4.14 mmol/L, or HDL-C < 1.04 mmol/L, or a self-reported hospital diagnosis;²⁶ and body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared (kg/m²).

Statistical Methods

Continuous variables were expressed as mean ± standard deviation or median (interquartile range), depending on distribution, and categorical variables as number (percentage). Between-group comparisons used the independent-samples t-test for normally distributed continuous variables, the Mann–Whitney U-test or Kruskal–Wallis test for non-normally distributed continuous variables, and the chi-square test for categorical variables, as appropriate. Multivariable logistic regression models were fitted to evaluate the association of eGDR, modeled continuously and in tertiles, with study-defined CKD. Model 1 was unadjusted; Model 2 was adjusted for age and sex; and Model 3 was further adjusted for education, residence, dyslipidemia, serum uric acid, smoking, alcohol consumption, and diabetes duration. WC, hypertension, and HbA1c were not additionally entered into Model 3 because they are components of eGDR, which could introduce over-adjustment and structural collinearity. Odds ratios (ORs) and 95% confidence intervals (CIs) were reported. Multicollinearity among model predictors was assessed using variance inflation factors (VIFs), and model goodness of fit was evaluated using the Hosmer–Lemeshow test.

Restricted cubic spline (RCS) regression based on Model 3 was used to examine the shape of the association between eGDR and study-defined CKD, with four knots placed at the 5th, 35th, 65th, and 95^th percentiles of the eGDR distribution. Spearman rank correlation coefficients were calculated to assess the correlations of eGDR with UACR and eGFR. Generalized additive models (GAMs) with restricted maximum likelihood smoothing were used to examine the shapes of these relationships. Receiver operating characteristic (ROC) analysis was used descriptively to quantify the discrimination of eGDR for study-defined CKD; the area under the curve (AUC) and the Youden-index cut-off were calculated. The ROC analysis was not treated as a clinical prediction model, and no claim of incremental prediction beyond established CKD factors was made.

For exploratory subgroup analyses, eGDR was dichotomized at the Youden-index cut-off of 5.65, and analyses were stratified by age, sex, BMI, residence, diabetes duration, dyslipidemia, smoking, and alcohol consumption. The stratification variable was omitted from the corresponding adjusted model when applicable. Multiplicative interaction terms were evaluated with likelihood ratio tests. Among the 713 participants included in the final analysis, no missing data were present for any variable used in the present study. Questionnaire and physical examination data were entered directly into the tablet-based system with mandatory entry and logical consistency checks. Participants without the required biological samples for laboratory testing were excluded before selection of the final analytic sample, as shown in Figure 1. Therefore, the analysis was restricted to complete cases, and no imputation was performed. A two-sided p < 0.05 was considered statistically significant. All analyses were conducted using R software (version 4.2.1; R Foundation for Statistical Computing, Vienna, Austria).

Results

Participant Characteristics

As presented in Table 1, a total of 713 participants with T2DM were included, of whom 245 (34.36%) met the study-defined CKD criteria. Participants meeting the study-defined CKD criteria were older than those without study-defined CKD (p < 0.001). Compared with the group without study-defined CKD, those with study-defined CKD had higher BMI, WC, SBP, and DBP (all p < 0.01), along with a higher prevalence of hypertension (p < 0.001) and longer diabetes duration (p < 0.01). Glycemic indicators were less favorable in the study-defined CKD group, with higher FPG and HbA1c (both p < 0.001). Lipid profiles were also less favorable, characterized by higher TG and lower HDL-C (both p < 0.001), whereas TC and LDL-C did not differ significantly; accordingly, dyslipidemia was more prevalent in the study-defined CKD group (p < 0.001). SUA was higher in the study-defined CKD group (p < 0.001). As expected from the study-defined CKD criteria, participants meeting these criteria had lower eGFR and higher UACR (both p < 0.001). In contrast, sex, educational level, residence, smoking, and alcohol consumption were comparable between groups (all p > 0.05).

Table 1.

Participant Characteristics Classified by Study-Defined CKD (n=713)

Variables	Overall (n=713)	Participants Without Study-Defined CKD (n=468)	Participants with Study-Defined CKD (n=245)	Statistic	P value
Age, years, Mean ± SD	66.49 ± 3.90	66.13 ± 3.76	67.17 ± 4.08	−3.40a	<0.001
Sex, n(%)				0.23b	0.631
Female	352 (49.37)	228 (48.72)	124 (50.61)
Male	361 (50.63)	240 (51.28)	121 (49.39)
Educational level, n(%)				2.14b	0.144
Primary school or below	588 (82.47)	393 (83.97)	195 (79.59)
Secondary school or above	125 (17.53)	75 (16.03)	50 (20.41)
Residence, n(%)				0.88b	0.348
Rural	352 (49.37)	237 (50.64)	115 (46.94)
Urban	361 (50.63)	231 (49.36)	130 (53.06)
Duration of diabetes(years), Median(IQR)	7.00 (4.00, 13.00)	7.00 (4.00, 11.00)	8.00 (4.00, 15.00)	−2.60c	0.009
WC, cm, Mean ± SD	87.49 ± 9.17	86.42 ± 8.68	89.54 ± 9.73	−4.36a	<0.001
BMI, kg/m2, Mean ± SD	24.48 ± 3.13	24.14 ± 2.96	25.14 ± 3.33	−4.13a	<0.001
BMI category, n(%)				5.13b	0.024
< 24 kg/m2	318 (44.60)	223 (47.65)	95 (38.78)
≥ 24 kg/m2	395 (55.40)	245 (52.35)	150 (61.22)
SBP, mmHg, Mean ± SD	138.27 ± 17.02	135.80 ± 16.09	143.01 ± 17.78	−5.48a	<0.001
DBP, mmHg, Mean ± SD	75.39 ± 9.17	74.62 ± 8.67	76.86 ± 9.90	−3.12a	0.002
Hypertension, n(%)	554 (77.70)	341 (72.86)	213 (86.94)	18.39b	<0.001
TG, mmol/L, Median(IQR)	1.50 (1.11, 2.25)	1.38 (1.02, 2.04)	1.81 (1.30, 2.70)	−5.86c	<0.001
TC, mmol/L, Mean ± SD	4.57 ± 0.97	4.54 ± 0.90	4.63 ± 1.08	−1.10a	0.273
HDL-C, mmol/L, Mean ± SD	1.27 ± 0.37	1.31 ± 0.36	1.20 ± 0.36	4.06a	<0.001
LDL-C, mmol/L, Mean ± SD	2.72 ± 0.87	2.73 ± 0.84	2.71 ± 0.93	0.25a	0.806
Dyslipidemia, n(%)	374 (52.45)	217 (46.37)	157 (64.08)	20.23b	<0.001
FPG, mmol/L, Mean ± SD	7.72 ± 2.36	7.46 ± 2.00	8.21 ± 2.88	−3.64 a	<0.001
HbA1c, %, Mean ± SD	7.18 ± 1.37	7.04 ± 1.25	7.44 ± 1.54	−3.52a	<0.001
SUA, μmol/L, Mean ± SD	343.02 ± 101.37	329.60 ± 86.60	368.64 ± 120.94	−4.49a	<0.001
eGFR, mL/min/1.73 m2, Mean ± SD	79.04 ± 16.60	84.25 ± 10.99	69.09 ± 20.53	10.78a	<0.001
UACR, mg/g, Median(IQR)	11.97 (5.46, 32.35)	7.93 (4.32, 14.46)	52.69 (31.73, 119.38)	−17.49c	<0.001
eGDR, Mean ± SD	6.68 ± 1.92	7.02 ± 1.91	6.04 ± 1.77	6.84a	<0.001
Cigarette smoking, n(%)	162 (22.72)	110 (23.50)	52 (21.22)	0.48b	0.490
Alcohol consumption, n(%)	267 (37.45)	181 (38.68)	86 (35.10)	0.88b	0.349

Notes: ^aStudent’s t-test, ^bChi-square test, ^cMann–Whitney test.

Abbreviations: CKD, chronic kidney disease; SD, standard deviation, WC, waist circumference; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; TG, triglycerides; TC, total cholesterol; HDL-C, high density lipoprotein-cholesterol; LDL-C, low density lipoprotein-cholesterol; FPG, fasting plasma glucose; SUA, serum uric acid; eGFR, estimated glomerular filtration rate; UACR, urinary albumin-to-creatinine ratio; eGDR, estimated glucose disposal rate.

As shown in Table 2, across eGDR tertiles (T1: lowest eGDR; T3: highest eGDR), participants in T1 exhibited the most adverse cardiometabolic profile. Age, BMI, WC, SBP, and DBP were highest in T1 (all p < 0.05). Glycemic control was poorest in T1, with higher FPG and HbA1c (both p < 0.001). Similarly, TG and SUA were highest, whereas HDL-C and eGFR were lowest in T1 (all p < 0.05), while TC and LDL-C showed no significant differences. The prevalence of hypertension, dyslipidemia, and study-defined CKD was also highest in T1 (all p < 0.01), and smoking differed across tertiles (p < 0.01). In contrast, sex, educational level, residence, diabetes duration, and alcohol consumption were similarly distributed across tertiles (all p > 0.05).

Table 2.

Participant Characteristics According to eGDR Tertile (n=713)

Variables	Overall (n=713)	T1 (n=235)	T2 (n=243)	T3 (n=235)	Statistic	P value
Age, years, Mean ± SD	66.49 ± 3.90	66.80 ± 3.84	66.76 ± 3.87	65.90 ± 3.95	3.99a	0.019
Sex, n(%)					3.67b	0.160
Female	352 (49.37)	104 (44.26)	126 (51.85)	122 (51.91)
Male	361 (50.63)	131 (55.74)	117 (48.15)	113 (48.09)
Educational level, n(%)					2.60b	0.272
Primary school or below	588 (82.47)	194 (82.55)	207 (85.19)	187 (79.57)
Secondary school or above	125 (17.53)	41 (17.45)	36 (14.81)	48 (20.43)
Residence, n(%)					2.22b	0.329
Rural	352 (49.37)	123 (52.34)	111 (45.68)	118 (50.21)
Urban	361 (50.63)	112 (47.66)	132 (54.32)	117 (49.79)
Duration of diabetes(years), Median(IQR)	7.00 (4.00, 13.00)	7.00 (4.00,13.00)	8.00 (4.00,12.00)	7.00 (3.00,13.00)	0.05c	0.978
WC, cm, Mean ± SD	87.49 ± 9.17	94.98 ± 7.91	86.18 ± 5.69	81.37 ± 8.01	212.03a	<0.001
BMI, kg/m2, Mean ± SD	24.48 ± 3.13	26.61 ± 3.10	24.19 ± 2.13	22.66 ± 2.73	130.31a	<0.001
BMI category, n(%)					130.45b	<0.001
< 24 kg/m2	318 (44.60)	42 (17.87)	111 (45.68)	165 (70.21)
≥ 24 kg/m2	395 (55.40)	193 (82.13)	132 (54.32)	70 (29.79)
SBP, mmHg, Mean ± SD	138.27 ± 17.02	144.11 ± 17.00	140.48 ± 16.53	130.16 ± 14.32	48.16a	<0.001
DBP, mmHg, Mean ± SD	75.39 ± 9.17	77.90 ± 9.15	75.80 ± 9.24	72.46 ± 8.27	22.35a	<0.001
Hypertension, n(%)	554 (77.70)	235 (100.00)	239 (98.35)	80 (34.04)	385.77b	<0.001
TG, mmol/L, Median(IQR)	1.50 (1.11, 2.25)	1.87 (1.33.20.70)	1.50 (1.11.20.22)	1.26 (0.91.10.81)	59.83c	<0.001
TC, mmol/L, Mean ± SD	4.57 ± 0.97	4.67 ± 1.05	4.56 ± 0.95	4.48 ± 0.90	2.49a	0.083
HDL-C, mmol/L, Mean ± SD	1.27 ± 0.37	1.17 ± 0.32	1.27 ± 0.36	1.38 ± 0.38	20.32a	<0.001
LDL-C, mmol/L, Mean ± SD	2.72 ± 0.87	2.79 ± 0.93	2.72 ± 0.85	2.66 ± 0.83	1.23a	0.294
Dyslipidemia, n(%)	374 (52.45)	158 (67.23)	130 (53.50)	86 (36.60)	44.39b	<0.001
FPG, mmol/L, Mean ± SD	7.72 ± 2.36	8.81 ± 2.64	7.10 ± 1.90	7.28 ± 2.13	41.66a	<0.001
HbA1c, %, Mean ± SD	7.18 ± 1.37	8.02 ± 1.38	6.72 ± 1.08	6.81 ± 1.25	81.27a	<0.001
SUA, μmol/L, Mean ± SD	343.02 ± 101.37	358.31 ± 111.18	348.12 ± 98.61	322.46 ± 90.34	7.97a	<0.001
eGFR, mL/min/1.73 m2, Mean ± SD	79.04 ± 16.60	77.45 ± 17.42	78.20 ± 16.93	81.50 ± 15.12	4.00a	0.019
UACR, mg/g, Median(IQR)	11.97 (5.46, 32.35)	19.84 (7.81,54.83)	12.43 (5.79,33.86)	8.11 (4.01,18.93)	48.57c	<0.001
Study-defined CKD, n(%)	245 (34.36)	117 (49.79)	80 (32.92)	48 (20.43)	45.25b	<0.001
Cigarette smoking, n(%)	162 (22.72)	65 (27.66)	39 (16.05)	58 (24.68)	9.94b	0.007
Alcohol consumption, n(%)	267 (37.45)	90 (38.30)	97 (39.92)	80 (34.04)	1.87b	0.393

Notes: ^aANOVA, ^bChi-square test, ^cKruskal–Wallis test.

Abbreviations: eGDR, estimated glucose disposal rate; SD, standard deviation, WC, waist circumference; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; TG, triglycerides; TC, total cholesterol; HDL-C, high density lipoprotein-cholesterol; LDL-C, low density lipoprotein-cholesterol; FPG, fasting plasma glucose; SUA, serum uric acid; eGFR, estimated glomerular filtration rate; UACR, urinary albumin-to-creatinine ratio; CKD, chronic kidney disease.

Multivariable Regression Analysis of eGDR with Study-Defined CKD

Multivariable logistic regression models were used to examine the association between eGDR and study-defined CKD. ORs and 95% CIs were estimated for eGDR modeled both continuously and in tertiles, with T1, representing the lowest eGDR level, as the reference group (Table 3).

Table 3.

Association Between eGDR and Study-Defined CKD

eGDR	Model 1		Model 2		Model 3
	OR (95%CI)	P value	OR (95%CI)	P value	OR (95%CI)	P value
Continuous
Per 1 unit increase	0.74 (0.68–0.81)	<0.001	0.75 (0.68–0.82)	<0.001	0.78 (0.70–0.86)	<0.001
Tertiles
T1	1.00 (ref)		1.00 (ref)		1.00 (ref)
T2	0.49 (0.34–0.72)	<0.001	0.49 (0.33–0.71)	<0.001	0.51 (0.35–0.76)	<0.001
T3	0.26 (0.17–0.39)	<0.001	0.27 (0.18–0.40)	<0.001	0.31 (0.20–0.48)	<0.001
P for trend		<0.001		<0.001		<0.001

Notes: Model 1: unadjusted analysis (no covariates included); Model 2: adjusted for age, sex; Model 3: fully adjusted model, incorporating age, sex, education, residence, dyslipidemia, serum uric acid, smoking, alcohol consumption, and diabetes duration.

Abbreviations: eGDR, estimated glucose disposal rate; CKD, chronic kidney disease; OR, odds ratio; CI, confidence interval.

When eGDR was analyzed as a continuous variable, each 1-unit increase in eGDR was associated with lower odds of study-defined CKD in the unadjusted model (p < 0.001). This association remained statistically significant after adjustment for age and sex in Model 2 (p < 0.001) and after further adjustment for education, residence, dyslipidemia, serum uric acid, smoking, alcohol consumption, and diabetes duration in Model 3 (OR = 0.78, 95% CI: 0.70–0.86; p < 0.001).

When eGDR was analyzed in tertiles, a graded inverse association with study-defined CKD was observed. In the unadjusted model, participants in T2 and T3 had lower odds of study-defined CKD than those in T1 (both p < 0.001). The associations remained statistically significant after adjustment for age and sex. In the fully adjusted Model 3, the ORs for study-defined CKD were 0.51 (95% CI: 0.35–0.76) for T2 and 0.31 (95% CI: 0.20–0.48) for T3 compared with T1 (both p < 0.001). Significant trends across eGDR tertiles were observed in all three models (all p for trend < 0.001). No substantial multicollinearity was observed among the covariates, with VIFs ranging from 1.02 to 1.64. The Hosmer–Lemeshow test showed no evidence of poor model fit (χ² = 11.90, df = 8, p = 0.156).

Shape of the Association Between eGDR and Study-Defined CKD: RCS Analysis

To further explore the shape of the association between eGDR and study-defined CKD, RCS analysis was performed based on the fully adjusted model (Model 3), incorporating age, sex, education, residence, dyslipidemia, serum uric acid, smoking, alcohol consumption, and diabetes duration (Figure 2). The RCS curve demonstrated a progressive decline in the odds of study-defined CKD with increasing eGDR values, indicating an approximately linear inverse association between eGDR and study-defined CKD prevalence (p for nonlinearity > 0.05).

Figure 2.

Association between estimated glucose disposal rate (eGDR) and study-defined chronic kidney disease (CKD). This figure illustrates the association between eGDR and the odds of study-defined CKD, with 95% confidence intervals (CIs), based on restricted cubic spline models. The x-axis represents eGDR value, and the y-axis represents the odds ratios (ORs) for study-defined CKD. The solid line indicates the estimated association, and the shaded area represents the corresponding 95% CIs. The horizontal dashed line at or = 1.0 indicates no association. The model was adjusted for age, sex, education, residence, dyslipidemia, serum uric acid, smoking, alcohol consumption, and diabetes duration.

GAM Analysis of the Associations of eGDR with UACR and eGFR

GAMs were used to examine the relationships of eGDR with UACR and eGFR. As shown in Figure 3A, eGDR was inversely correlated with UACR (ρ = −0.277, p < 0.001), and the smooth-fitted curve showed a generally decreasing trend in UACR with increasing eGDR. In contrast, the correlation between eGDR and eGFR was small and not statistically significant (ρ = 0.064, p >0.05; Figure 3B). The smooth-fitted relationship between eGDR and eGFR likewise did not show a clear association in this study population.

Figure 3.

Generalized additive model (gam) analysis of the associations between estimated glucose disposal rate (eGDR) and renal function parameters. GAMs were employed to evaluate the associations of eGDR with urinary albumin-to-creatinine ratio (UACR) and estimated glomerular filtration rate (eGFR). (A) Association between eGDR and UACR (mg/g). (B) Association between eGDR and eGFR (mL/min/1.73 m²). The solid lines represent the fitted smooth curves, and the shaded areas denote the 95% confidence intervals.

ROC Curve Assessment and Subgroup Analysis

ROC curve analysis was performed to descriptively assess the discrimination of eGDR for study-defined CKD. The AUC was 0.65 (95% CI: 0.61–0.70, p <0.001), indicating modest discrimination (Figure 4). The optimal cut-off value based on the Youden index was 5.65. Participants were therefore categorized into low eGDR (<5.65) and high eGDR (≥5.65) groups for exploratory subgroup analyses.

Figure 4.

Receiver operating characteristic (ROC) curve of estimated glucose disposal rate (eGDR) for discriminating study-defined chronic kidney disease (CKD). The ROC curve illustrates the discriminatory capacity of the eGDR for discriminating study-defined CKD in elderly patients with type 2 diabetes mellitus. The diagonal reference line represents an AUC of 0.50, indicating no discriminatory ability.

Subgroup analyses were conducted according to age (<70 vs ≥70 years), sex (female vs male), BMI (<24 vs ≥24 kg/m²), residence (rural vs urban), diabetes duration (<10 vs ≥10 years), dyslipidemia (no vs yes), cigarette smoking (no vs yes), and alcohol consumption (no vs yes). Multivariable logistic regression models based on Model 3 were used to estimate adjusted ORs and 95% CIs within each subgroup.

As shown in Figure 5, participants with high eGDR had lower odds of study-defined CKD than those with low eGDR in the overall population (adjusted OR = 0.38, 95% CI: 0.27–0.54; p <0.001). The direction of the association was generally consistent across the examined subgroups. The association appeared stronger among participants aged ≥70 years (OR = 0.18, 95% CI: 0.08–0.40) than among those aged <70 years (OR = 0.45, 95% CI: 0.30–0.67); however, the interaction by age was not statistically significant. No statistically significant interactions were observed for age, sex, BMI, residence, diabetes duration, dyslipidemia, smoking, or alcohol consumption (all p for interaction >0.05). Given the exploratory nature of these subgroup analyses, subgroup-specific estimates should be interpreted cautiously.

Figure 5.

Subgroup analysis of the association between estimated glucose disposal rate (eGDR) and study-defined chronic kidney disease (CKD). Forest plot displaying the adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for study-defined CKD comparing high eGDR (≥ 5.65) vs low eGDR (< 5.65) across examined subgroups. Models were adjusted for age, sex, education, residence, dyslipidemia, serum uric acid, smoking, alcohol consumption, and diabetes duration, excluding the stratification variable when applicable. The left column displays the number of participants meeting the study-defined CKD criteria and total participants in each subgroup for the low eGDR and high eGDR groups. P values for the association within each subgroup and P values for interaction are presented in the right columns. The vertical dashed line at or = 1.0 represents no association.

Discussion

The prevalence of study-defined CKD among elderly patients with T2DM in this study was 34.36%, higher than the 11.41% reported in a community-based study of the general elderly population in China.²⁷ Given the substantial burden of diabetes in Zhejiang Province,²⁸ this finding supports the importance of characterizing metabolic factors associated with study-defined CKD in older adults with T2DM.

In the present study, lower eGDR remained associated with a higher prevalence of study-defined CKD after adjustment for measured demographic, lifestyle, and clinical covariates. The inverse association was observed when eGDR was modeled continuously and in tertiles, and RCS analysis suggested an approximately linear pattern within the observed range. Higher eGDR was also associated with lower UACR, whereas no statistically significant association with eGFR was detected. These findings describe cross-sectional relationships and do not establish whether lower eGDR preceded study-defined CKD or resulted partly from CKD-related metabolic changes.

The observed association is consistent with previous epidemiological and mechanistic evidence linking IR with kidney dysfunction in diabetes.^29–31 HOMA-IR is widely used as a surrogate measure of IR but requires fasting insulin measurement and is sensitive to preanalytical and analytical variation.³² By contrast, eGDR is derived from routinely collected clinical variables and has been used as a surrogate measure of insulin sensitivity in diabetic populations.³³

Previous studies have linked lower eGDR with macrovascular and microvascular complications, particularly in type 1 diabetes,^11,13 and more recent work has evaluated eGDR in T2DM and other cardiometabolic populations.³⁴ The present study does not identify a new biological relationship; rather, it provides population-specific evidence among elderly Chinese patients with T2DM and examines UACR and eGFR separately. A previous analysis from the same Zhejiang source population focused on central obesity, family history, and hypertension rather than IR or renal outcomes.³⁵ Moreover, the previous study examined hypertension as the primary outcome and did not assess IR using eGDR.³⁵ The current findings are therefore best viewed as confirmatory evidence in a more narrowly defined elderly population.

The approximately linear RCS pattern indicates that, within the observed eGDR range, the prevalence odds of study-defined CKD were progressively lower at higher eGDR values. Because eGDR and study-defined CKD were measured at the same time, this pattern cannot be interpreted as evidence that increasing insulin sensitivity would reduce future CKD occurrence. Prospective studies are needed to establish temporality and to determine whether changes in eGDR are associated with subsequent changes in kidney outcomes.

A notable finding was the different pattern observed for UACR and eGFR. eGDR was inversely associated with UACR, whereas no statistically significant association with eGFR was detected. UACR reflects glomerular albumin leakage and endothelial injury, whereas eGFR is influenced by age, hemodynamic factors, and compensatory changes.^36–38 In older adults, creatinine-based eGFR also requires careful interpretation because serum creatinine is affected by muscle mass, sarcopenia, and frailty. Therefore, the small nonsignificant correlation with eGFR in this sample should not be taken as evidence of absence of a relationship. The stronger cross-sectional association with UACR may be consistent with earlier microvascular injury, but this explanation remains a hypothesis that requires longitudinal confirmation.

In the exploratory subgroup analyses, the direction of the association between eGDR and study-defined CKD was similar across the examined strata, and no interaction test was statistically significant. Although the point estimate was lower among participants aged ≥70 years, the interaction by age was not significant. Because the study was not specifically powered for interaction testing and some subgroup estimates were based on relatively few participants meeting the study-defined CKD criteria, these results should not be interpreted as evidence of true differences between subgroups.

The ROC analysis yielded an AUC of 0.65, indicating only modest discrimination of study-defined CKD by eGDR alone. This performance is insufficient to support eGDR as a standalone screening or prediction tool. In addition, the present study did not compare eGDR with established CKD risk models or test whether adding eGDR improves discrimination, calibration, or risk classification beyond conventional clinical factors. eGDR may complement existing clinical assessment because it is based on routinely measured variables, but its incremental clinical value requires evaluation in independent prospective studies.

Several biological pathways could explain an association between IR and kidney abnormalities, although these mechanisms were not directly measured in the present study. IR is linked to chronic low-grade inflammation and endothelial dysfunction.^39,40 It is also associated with oxidative stress, advanced glycation end-products, and activation of the renin–angiotensin–aldosterone system.^41,42 Hemodynamic pathways, including glomerular hyperfiltration and increased intraglomerular pressure, may also contribute to glomerular injury.^43–45 These mechanisms are consistent with the observed association with albuminuria but cannot establish the direction of association in this cross-sectional study.

From a clinical perspective, the findings do not establish eGDR as a diagnostic or prognostic tool. Its main practical feature is that it can be calculated from routinely collected variables without fasting insulin measurement. Given the AUC of 0.65, eGDR should at most be considered a possible complementary marker alongside standard renal assessment, including UACR and eGFR. Prospective studies should test whether eGDR adds information beyond conventional CKD factors before its use for screening or risk classification is considered.

This study has several strengths. It focuses on an elderly population with T2DM, a group with a high burden of CKD. Data collection followed the standardized China DiaChronic Study protocol, and major biochemical measurements were performed centrally with prespecified quality-control procedures. The analysis also evaluated eGDR as both a continuous and categorical variable and separately examined UACR and eGFR, which allowed assessment of several related cross-sectional patterns.

Limitations of the Study

Several limitations should be considered. First, the cross-sectional design precludes causal or temporal inference. IR and CKD may have a bidirectional relationship, and CKD or its treatment may influence hypertension and glycemic measures that are components of eGDR. Reverse causation therefore cannot be excluded. Second, the study was conducted in one province in China, which may limit generalizability to populations with different demographic, clinical, or healthcare characteristics. Third, eGDR is an indirect measure of insulin sensitivity and does not replace the hyperinsulinemic–euglycemic clamp. Fourth, residual confounding may remain, including confounding by diet, physical activity, frailty, and use of kidney- or glucose-lowering medications such as SGLT2 inhibitors or GLP-1 receptor agonists. Fifth, study-defined CKD was classified from single eGFR and UACR measurements rather than persistence for at least 3 months, which may cause misclassification. In addition, creatinine-based eGFR can be affected by reduced muscle mass and frailty in older adults. Finally, the subgroup analyses were exploratory and may have had limited statistical power for interaction testing.

Conclusion

Lower eGDR was associated with a higher prevalence of study-defined CKD and higher UACR, whereas no statistically significant association with eGFR was detected. The association was approximately linear across the observed eGDR range. Because the study was cross-sectional, the temporal direction and causality of this relationship cannot be determined. Prospective studies are needed to determine whether eGDR provides additional value beyond established CKD assessment.

Ethics Statement

This study adhered to the principles of the Declaration of Helsinki. The parent national study received ethical approval from the Ethics Committee of Shanghai Jiao Tong University Affiliated Sixth People’s Hospital (Approval No: 2018–010), and all participants provided written informed consent. Zhejiang Province was included as one of the study sites. The present analysis used anonymized secondary data; therefore, the Institutional Review Board of the Zhejiang Provincial Center for Disease Control and Prevention waived the requirement for additional ethical approval in line with institutional regulations.

Funding Statement

This study was supported by Zhejiang Science and Technology Plan for Disease Prevention and Control (2026JKY116), Noncommunicable Chronic Diseases-National Science and Technology Major Project (2023ZD0509800, 2023ZD0509806) and the Healthy Zhejiang One Million People Cohort (K-20230085).

Disclosure

The authors report no conflicts of interest in this work.