Influence of metabolism-related comorbidities and insulin resistance on new onset of chronic kidney disease in a health check-up population: a two-stage retrospective cohort study

Abstract

Introduction

Limited research has focused on the prospective influence of insulin resistance (IR) on new-onset chronic kidney disease (CKD) in healthy screening populations. Therefore, we aimed to investigate how IR, assessed via the estimated glucose disposal rate (eGDR), and metabolism-related comorbidities influence new-onset CKD.

Research design and methods

This two-stage retrospective cohort study (cross-sectional and longitudinal analyses) used data from health check-up participants at the Chinese People’s Liberation Army General Hospital (2009–2021). The cross-sectional analysis included 83 346 participants with or without CKD; the longitudinal analyses included 13 738 participants without prior CKD who visited the hospital at least two times. The cross-sectional phase of this study analyzed the relationship between IR and CKD; the longitudinal phase analyzed the relationship between IR and new-onset CKD. The mediating role of metabolism-related comorbidities was also explored.

Results

In the cross-sectional analysis, 6.77% (n=5643) of patients had prior CKD. The eGDR was significantly higher in the non-CKD group than in the CKD group (9.16±2.11 vs 7.19±2.32, p<0.001). Higher eGDR was associated with lower CKD prevalence (OR: 0.91, 95% CI: 0.89 to 0.93, P for trend<0.001). In the cohort analysis, the average time to trigger endpoint events was 2.95±2.02 years, with 403 (2.93%) new-onset CKD cases reported. A linear correlation was observed between eGDR and new-onset CKD (p<0.001), with higher eGDR linked to reduced CKD risk (HR: 0.88, 95% CI: 0.82 to 0.96, P for trend=0.002). Mediation analysis revealed significant indirect effects of diabetes mellitus (17.1%), systolic blood pressure (22.0%), glycated hemoglobin (11.1%), and brachial–ankle pulse wave velocity (9.7%) (all p<0.05).

Conclusions

IR is independently linked to new-onset CKD, with blood glucose, blood pressure, and arterial stiffness mediating this relationship. These findings underscore the importance of managing IR and metabolic comorbidities to prevent CKD onset in at-risk populations.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Insulin resistance (IR) is associated with chronic kidney disease (CKD) in individuals with metabolic disorders, but its role in new-onset CKD among generally healthy populations remains unclear. Prior studies have primarily focused on patients with pre-existing CKD, highlighting the contribution of IR to disease progression rather than its predictive value for CKD onset.

WHAT THIS STUDY ADDS

• This study demonstrates that lower estimated glucose disposal rate (eGDR), an IR-related index, is independently associated with a higher risk of new-onset CKD in a health check-up population. Furthermore, metabolism-related comorbidities such as diabetes mellitus, hypertension, glycated hemoglobin, and arterial stiffness significantly mediate this association, thus highlighting the multifaceted role of IR in the development of CKD.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These findings suggest that monitoring and managing IR in apparently healthy individuals could aid in CKD prevention. Clinicians should consider eGDR as a potential risk marker for early CKD detection. Future research and public health strategies should focus on targeted interventions addressing IR and metabolic comorbidities to mitigate CKD risk.

Introduction

Chronic kidney disease (CKD) has emerged as a significant global health issue with a growing prevalence that poses substantial challenges to healthcare systems worldwide.¹ CKD is defined by markers of kidney damage, such as proteinuria or a glomerular filtration rate (GFR) of <60 mL/min/1.73 m² for 3 months or longer.² In clinical practice, once patients develop proteinuria and experience a decline in the GFR, many pathophysiological changes become irreversible, leaving limited treatment options. Therefore, early identification of high-risk individuals and delay in the onset of CKD are vital for the prevention and management of CKD.

The rising incidence of CKD may be linked to risk factors such as an aging population, lifestyle changes, and a high prevalence of metabolic comorbidities.^{3 4} Insulin resistance (IR) plays a crucial role in this process. IR is a common mechanism underlying various metabolism-related comorbidities and is prevalent among patients with CKD.⁵ IR is present in patients with both early-stage CKD^{6 7} and end-stage renal disease.^{8 9} Furthermore, IR and related metabolic factors contribute to the development of microalbuminuria¹⁰ and CKD,¹¹ accelerate the progression of renal function decline in patients with CKD,⁶ and increase the risk of cardiovascular events and all-cause mortality.^{7 12 13} Most large-scale clinical studies have primarily used various IR indices to assess IR in populations.¹¹ Notably, the estimated glucose disposal rate (eGDR) is regarded as a reliable IR index owing to its strong sensitivity and specificity, displaying a significant correlation with the GDR (typically presented by the steady-state glucose infusion rate) obtained using the IR gold standard glucose clamp technique.^{7 13}

Limited research has focused on IR levels in healthy screened populations and their role in new-onset CKD. Therefore, in this study, we aimed to examine the relationship between IR, assessed via the eGDR, and CKD in a health check-up population, emphasizing the predictive capacity of IR for CKD onset and its mediating effects in both cross-sectional and longitudinal cohort studies. The findings of this comprehensive analysis highlight the significant role of IR in new-onset CKD risk within the health check-up population and offer crucial clinical insights for the early prevention and management of CKD.

Materials and methods

Study participants

This was a two-stage retrospective cohort study. The first stage was a cross-sectional study involving individuals who visited the Chinese People’s Liberation Army (PLA) General Hospital for health check-ups between December 2009 and May 2021. In total, 83 346 participants were included in the primary analysis, consisting of 77 703 individuals without a history of CKD and 5643 with previous CKD. The differences in demographic characteristics, metabolism-related comorbidities, and IR levels were assessed between the CKD and non-CKD groups. The second stage was a longitudinal analysis involving 14 468 first-stage participants who visited the hospital at least twice. Data were collected and analyzed to evaluate the associations between new-onset CKD, eGDR, and metabolism-related comorbidities. The interval between the two visits was no less than 6 months. The confirmation time of new-onset CKD was based on the first diagnosis of CKD during follow-up. Medical information was collected from the participants during their last visit in the second stage. The enrollment details are shown in figure 1.

Figure 1. Flow chart of the study population. CKD, chronic kidney disease; eGDR, estimated glucose disposal rate; HbA1c, glycated hemoglobin; UACR, urine albumin-to-creatinine ratio.

Data collection and covariates

The following data were collected: (1) demographic and lifestyle information, including age, sex and smoking and drinking status; (2) disease history data, including coronary heart disease (CHD), hypertension (HTN), diabetes mellitus (DM), hyperuricemia, and hyperlipidemia; (3) body measurements, including body mass index (BMI), waist circumference (WC), heart rate, systolic blood pressure (SBP), and diastolic blood pressure (DBP); (4) laboratory findings, including serum creatinine (sCr), blood urea nitrogen (BUN), urinary albumin-to-creatinine ratio (UACR), serum uric acid (Ua), total cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDL-c), high-density lipoprotein cholesterol, fasting blood glucose (FBG), glycated hemoglobin (HbA1c), and C reactive protein (CRP); and (5) arterial stiffness (AS), measured by brachial–ankle pulse wave velocity (baPWV). We classified baPWV into three categories: normal AS (baPWV<1400 cm/s), moderately elevated AS (1400≤baPWV< 1800 cm/s), and severely elevated AS (baPWV≥1800 cm/s), based on established standards.¹⁴

Detailed medical history collection, data collection, and laboratory testing methods are outlined in online supplemental data.

Exposure variable

The eGDR (mg/kg/min), a validated surrogate measure of IR, was employed as the primary exposure variable in this study and was calculated using the following formula:

$${\displaystyle \mathrm{e}\mathrm{G}\mathrm{D}\mathrm{R}=21.158-(0.09\times \mathrm{W}\mathrm{C})-(3.407\times \mathrm{H}\mathrm{T}\mathrm{N})-(0.551\times \mathrm{H}\mathrm{b}\mathrm{A}1\mathrm{c}),}$$

where WC is waist circumference (cm), HTN is hypertension (yes=1/no=0), and HbA1c is glycated hemoglobin (Diabetes Control and Complications Trial, DCCT %).¹⁵ In the cross-sectional analysis, the population was divided into quartiles based on the eGDR values (Q1: 1.51–8.28; Q2: 8.28–10.00; Q3: 10.00–10.90; Q4: 10.90–13.40), with the lowest quartile (Q1, indicating the highest IR) serving as the reference group.

Outcomes of interest

In the retrospective cohort analysis, the primary endpoint was new-onset CKD, defined as the first confirmed diagnosis of CKD during the follow-up period, which also corresponds to the time of revisit. CKD was defined as albuminuria or estimated GFR (eGFR) of <60 mL/min/1.73 m² according to the Kidney Disease Improving Global Outcomes guideline.¹⁶ Albuminuria corresponded to a UACR of ≥30 mg/g.¹⁶ The eGFR was calculated using the 2021 Chronic Kidney Disease Epidemiology Collaboration formula based on sCr levels.¹⁷

All individuals were monitored from their initial observation date (between 2009 and 2021) until either the onset of the primary endpoint (CKD) or their last follow-up, whichever occurred first.

Statistical analysis

Data are presented as mean±SD for continuous variables and as frequencies and percentages for categorical variables. We used Welch’s t-test or analysis of variance for continuous variable group comparisons and the χ² test for categorical data analysis.

Non-linear relationships between changes in eGDR and CKD, eGFR, UACR (cross-sectional study), and new-onset CKD (longitudinal analyses) were analyzed using linear and Cox regression models with restricted cubic splines (RCS). RCS analysis was performed with four knots at the 5th, 35th, 65th, and 95th percentiles to flexibly model the association.

The cumulative incidence curve for new-onset CKD was created using the Kaplan-Meier method and compared using the log-rank test.

Cox proportional hazards regression models were used to evaluate the association between eGDR and CKD events. Multiple models, each adjusted for various covariates, were constructed to determine the effect of these factors on the observed association. The model results were compared to evaluate the consistency of effect estimates with different levels of covariate adjustment, where covariates included those with a variance inflation factor<10, offering insights into the robustness of the impact of eGDR on cumulative CKD events. Subgroup analyses using Cox regression models were performed based on sex, age, HTN, CHD, type 2 DM, BMI, HbA1c, and baPWV to identify potential effect modifiers and evaluate the robustness of the main findings across various patient populations.

We used a quasi-Bayesian approach with 100 simulations to perform a causal mediation analysis. The survival outcome, defined as the time to CKD diagnosis, was modeled using the Weibull distribution in the “survreg” function of R V.4.2.2. Mediation effects were estimated with the “mediation” package, which decomposes total effects into direct and indirect components. Cox proportional hazard models accounted for follow-up time and event status (CKD diagnosis), and CIs and p values were obtained using simulation-based inference.

Statistical analyses were performed using Statistical Package for the Social Sciences V.29.0 (IBM), R software (V.4.2.2; R Foundation for Statistical Computing, Vienna, Austria), and MSTATA (www.mstata.com). A two-sided p value of <<0.05 was established as the threshold for statistical significance.

Results

Demographic characteristics of the participants

In the first cross-sectional stage of the study, 83 346 participants were included, of whom 6.77% (n=5643) had a history of CKD. Table 1 presents a detailed comparison of the baseline characteristics between the two groups based on CKD status, showing the significant differences across various variables. The CKD group had a mean age of 50.24±9.55 years, which was higher than that of the non-CKD group (47.62±9.06), and showed higher BMI, blood pressure (SBP and DBP), and heart rate (all p<0.001). Additionally, the CKD group displayed elevated levels of metabolism-related and AS-related indicators, such as FBG, HbA1c, Ua, TC, TG, LDL-c, and eGDR, with significant between-group differences (all p<0.001). Renal-related indicators, including sCr level, BUN level, and UACR, were increased in the CKD group, whereas eGFR was decreased (all p<0.001).

Table 1. Patient demographics and baseline characteristics.

Characteristic	Chronic kidney disease			P value
	Overall n=83 346	No n=77 703	Yes n=5643
Age, years	47.80±9.12	47.62±9.06	50.24±9.55	<0.001
Sex				<0.001
Female	28 227 (33.87%)	26 559 (34.18%)	1668 (29.56%)
Male	55 119 (66.13%)	51 144 (65.82%)	3975 (70.44%)
Current smoking	26 390 (31.66%)	24 489 (31.52%)	1901 (33.69%)	<0.001
Current drinking	49 308 (59.16%)	45 880 (59.05%)	3428 (60.75%)	0.012
CHD				<0.001
No	81 681 (98.00%)	76 252 (98.13%)	5429 (96.21%)
Yes	1665 (2.00%)	1451 (1.87%)	214 (3.79%)
HTN				<0.001
No	58 848 (70.61%)	56 624 (72.87%)	2224 (39.41%)
Yes	24 498 (29.39%)	21 079 (27.13%)	3419 (60.59%)
Hyperuricemia				<0.001
No	68 134 (81.75%)	63 911 (82.25%)	4223 (74.84%)
Yes	15 212 (18.25%)	13 792 (17.75%)	1420 (25.16%)
Hyperlipidemia				<0.001
No	38 624 (46.34%)	36 877 (47.46%)	1747 (30.96%)
Yes	44 722 (53.66%)	40 826 (52.54%)	3896 (69.04%)
DM				<0.001
No	72 399 (86.87%)	68 832 (88.58%)	3567 (63.21%)
Yes	10 947 (13.13%)	8871 (11.42%)	2076 (36.79%)
BMI, kg/m2	25.17±3.44	25.05±3.38	26.83±3.83	<0.001
WC, cm	87.52±10.12	87.18±10.01	92.24±10.52	<0.001
HR, bpm	68.02±9.08	67.79±8.94	71.15±10.29	<0.001
SBP, mm Hg	122.09±15.81	121.16±15.24	134.95±17.80	<0.001
DBP, mm Hg	81.63±11.52	81.03±11.19	89.87±12.79	<0.001
sCr, μmol/L	70.47±14.56	70.29±13.58	72.96±24.20	<0.001
eGFR, ml/min/1.73 m2	104.95±11.33	105.19±10.83	101.73±16.42	<0.001
eGFR<60 mL/min/1.73 m2, n (%)	201 (0.24%)	0 (0%)	201 (3.56%)	<0.001
BUN, mg/dL	5.14±1.26	5.12±1.23	5.47±1.63	<0.001
UACR, mg/g	12.47±27.09	7.37±4.98	82.75±72.10	<0.001
UACR≥30 mg/g, n(%)	5572 (6.69%)	0 (0%)	5572 (98.74%)	<0.001
Ua, μmol/L	338.73±89.75	337.23±88.95	359.31±97.84	<0.001
TC, mmol/L	4.82±0.92	4.80±0.91	5.00±1.04	<0.001
TG, mmol/L	1.71±1.20	1.67±1.14	2.30±1.73	<0.001
LDL-c, mmol/L	3.03±0.80	3.03±0.79	3.10±0.89	<0.001
HDL-c, mmol/L	1.28±0.34	1.29±0.34	1.20±0.32	<0.001
FPG, mmol/L	5.85±1.40	5.76±1.23	7.01±2.60	<0.001
HbA1c, %	5.86±0.82	5.81±0.74	6.49±1.42	<0.001
CRP, mg/L	0.19±0.52	0.18±0.45	0.32±1.07	<0.001
baPWV, cm/s	1449±220	1439±215	1576±255	<0.001
eGDR, mg/kg/min	9.05±2.17	9.19±2.09	7.22±2.32	<0.001

baPWV, brachial–ankle pulse wave velocity; BMI, body mass index; bpm, beats per minute; BUN, blood urea nitrogen; CHD, coronary heart disease; CRP, C reactive protein; DBP, diastolic blood pressure; DM, diabetes mellitus; eGDR, estimated glucose disposal rate; eGFR, estimated glomerular filtration rate; FPG, fasting plasma glucose; HbA1c, glycated hemoglobin; HDL-c, high-density lipoprotein cholesterol; HR, heart rate; HTN, hypertension; LDL-c, low-density lipoprotein cholesterol; SBP, systolic blood pressure; sCr, serum creatinine; TC, total cholesterol; TG, triglycerides; Ua, uric acid; UACR, urinary albumin-to-creatinine ratio; WC, waist circumference.

The baseline characteristics of all participants categorized by eGDR quartiles in the second stage of the study (longitudinal cohort study) are presented in online supplemental table S1. Compared with Q4, lower eGDR (Q1) was associated with older age (49.72 vs 43.49 years), higher proportion of male sex (85.47% vs 34.33%), and poorer cardiovascular and metabolic status, as shown by elevated BMI (26.81 vs 21.46 kg/m²), SBP (134.46 vs 109.81 mm Hg), and CRP level (0.20 vs 0.11 mg/L). The eGFR was lower in Q1 compared with Q4 (103.19 vs 108.69 mL/min/1.73 m²). Additionally, Q1 had a higher prevalence of DM, HTN, and hyperlipidemia.

eGDR distribution in the non-CKD and CKD groups

The trend analysis of eGDR between the CKD and non-CKD groups showed that the average eGDR was significantly higher in the non-CKD group than in the CKD group (9.16±2.11 vs 7.19±2.32 mg/kg/min, p<0.001, table 1 and online supplemental table S2). The plot indicates that the eGDR was generally lower in the CKD group than in the non-CKD group (online supplemental figure S1).

Relationships between eGDR, CKD, eGFR, and UACR

In the cross-sectional study, online supplemental figure S2 depicts the associations between eGDR (mg/kg/min) and CKD risk, eGFR, and UACR, stratified by unadjusted (online supplemental figure S2A–C) and adjusted (online supplemental figure S2D–F) RCS analyses. In the unadjusted models, a significant non-linear relationship between eGDR and CKD risk was observed (online supplemental figure S2A; p<0.001, p non-linear=0.006). The OR for CKD decreased markedly, with an increasing eGDR, suggesting that higher eGDR is associated with a lower risk of CKD. With increasing eGDR, the level of eGFR increased sharply (p<0.001, p non-linear<0.001, online supplemental figure S2B), the level of UACR decreased significantly (p<0.001, p non-linear<0.001, online supplemental figure S2C), and the non-linear associations remained significant after adjusting for covariates (p<0.001, p non-linear<0.001). The eGDR was U-shaped with CKD risk, with the lowest risk of CKD seen at moderate eGDR (online supplemental figure S2D). The eGDR was consistently positively correlated with eGFR, but there was evidence that the relationship flattened as eGDR increased (online supplemental figure S2E). Likewise, the inverse relationship between eGDR and UACR was maintained, with UACR declining sharply at lower eGDR and flattening in the upper eGDR range (online supplemental figure S2F).

Higher eGDR was significantly associated with lower odds of CKD, in all models (online supplemental table S3). In model 1, each unit increase in eGDR lowered CKD odds (OR: 0.68, 95% CI: 0.67 to 0.69, p<0.001). This inverse relationship was also observed in model 2 (OR: 0.72, 95% CI: 0.71 to 0.73, p<0.001) and model 3 (OR: 0.91, 95% CI: 0.89 to 0.93, p<0.001). Furthermore, higher eGDR quartiles correlated with reduced CKD risk (p for trend<0.001).

Subgroup analysis showed significant associations between eGDR and CKD across various groups. In the overall population, eGDR was inversely linked to CKD (OR: 0.80, 95% CI: 0.78 to 0.81, p<0.001). A notable interaction was observed for baPWV (p for interaction<0.001), with the strongest inverse association in individuals with baPWV≥1800 (OR: 0.75, 95% CI: 0.72 to 0.79, p<0.001) (figure 2 and online supplemental table S4).

Figure 2. Subgroup analysis. HR (95% CI) for the association between estimated glucose disposal rate and chronic kidney disease using a logistic regression model. *Covariates included smoking, drinking, hyperuricemia, hyperlipidemia, heart rate, systolic blood pressure, diastolic blood pressure, blood urea nitrogen, uric acid, triglycerides, high-density lipoprotein cholesterol, and C reactive protein. baPWV, brachial–ankle pulse wave velocity; BMI, body mass index; CHD, coronary heart disease; DM, diabetes mellitus; HbA1c, glycated hemoglobin; HTN, hypertension.

Relationships between eGDR and new-onset CKD

In the cohort study, 13 738 participants without a history of CKD, who visited the hospital at least two times, were included. The interval between the first and last visits ranged from 1 to 11 years. During this follow-up period, 403 patients (2.93%) developed new-onset CKD, with an average time to endpoint events of 2.95±2.02 years. A lower eGDR (Q1) was associated with a higher new-onset CKD incidence (5.24% vs 1.39%) (online supplemental table S1).

Figure 3 depicts the associations between eGDR (mg/kg/min) and new-onset CKD using the multivariate Cox regression model; in figure 3A–C, after adjusting for different sets of covariates, a higher eGDR remained consistently associated with a lower HR for new-onset CKD (p<0.001 in figure 3A,B; p=0.003 in figure 3C). The p non-linear values suggested no significant evidence of non-linearity (figure 3A–C).

Figure 3. Association of eGDR with new onset of CKD using a multivariate Cox regression model of restricted cubic splines. (A–C) Model with four knots located at 5th, 35th, 65th, and 95th percentiles. The Y-axis displays the HR and 95% CI for new onset of CKD, comparing any eGDR value to the reference value at the 50th percentile. In (A) no covariates were adjusted. The Cox regression in (B) was adjusted for sex, age, smoking, drinking, coronary heart disease, diabetes mellitus, hyperuricemia, and hyperlipidemia. The Cox regression in (C) was adjusted for model B plus body mass index, brachial–ankle pulse wave velocity, heart rate, systolic blood pressure, diastolic blood pressure, blood urea nitrogen, uric acid, triglycerides, high-density lipoprotein cholesterol, fasting plasma glucose, and C reactive protein. Kaplan-Meier curve of the cumulative incidence of CKD based on quartiles of eGDR (D). The P value was generated based on the log-rank test. CKD, chronic kidney disease; eGDR, estimated glucose disposal rate.

Figure 3D presents the Kaplan-Meier analysis stratified by eGDR quartiles, which showed that participants in the lowest quartile (Q1), experienced the highest cumulative incidence of CKD over time, whereas those with a higher eGDR had lower event rates (log-rank p<0.0001).

A higher eGDR was consistently associated with a lower HR for new-onset CKD across models 1–3 (online supplemental table S5). Notably, in model 3 (fully adjusted), the continuous eGDR showed an HR of 0.88 (95% CI: 0.82 to 0.96, p=0.002), indicating a significant protective effect. The standardized form paralleled this finding (HR: 0.78, 95% CI: 0.67 to 0.91, p=0.002). Quartile analyses further highlighted the inverse relationship, with Q4 displaying the lowest HR (0.54, 95% CI: 0.34 to 0.87, p=0.011). The p value for the trend remained significant (p=0.006), suggesting a dose–response relationship. The results remained consistent after adjusting demographic factors, comorbidities, and metabolic parameters, highlighting the potential of eGDR as a reliable predictor of CKD onset (online supplemental table S5).

Subgroup analysis of the association between eGDR and new-onset CKD

In the subgroup analysis, a higher eGDR consistently correlated with a lower HR for new-onset CKD across various subgroups. In the overall population (n=13 738), a lower risk of new-onset CKD was observed (HR: 0.83, 95% CI: 0.77 to 0.88, p<0.001). Stratified by sex, both males (HR: 0.80, 95% CI: 0.74 to 0.87, p<0.001) and females (HR: 0.86, 95% CI: 0.75 to 0.99, p=0.035) exhibited reduced CKD risks, with no significant interaction between the sexes (p for interaction=0.835). Age subgroups showed consistent protective effects, particularly in individuals aged ≥40 to < 50 years (HR: 0.84, 95% CI: 0.76 to 0.94, p=0.001) and ≥60 years (HR: 0.77, 95% CI: 0.62 to 0.95, p=0.016), with no significant interaction across age (p for interaction=0.534).

HTN status did not significantly modify the association (p for interaction=0.455), with both non-hypertensive (HR: 0.82, 95% CI: 0.70 to 0.95, p=0.007) and hypertensive (HR: 0.77, 95% CI: 0.65 to 0.90, p=0.001) subgroups showing reduced CKD risks. Similarly, no significant interaction was observed for CHD or BMI (p for interaction=0.91 and 0.638, respectively). However, a significant interaction was observed for diabetes (p for interaction=0.008), with individuals without diabetes (HR: 0.84, 95% CI: 0.77 to 0.91, p<0.001) showing a stronger protective effect than those with diabetes (HR=0.95, 95% CI: 0.83 to 1.08, p=0.449). Additionally, a borderline significant interaction was also observed for baPWV (p for interaction=0.051), with the strongest protective effect in individuals with baPWV≥1800 (HR: 0.72, 95% CI: 0.59 to 0.87, p=0.001) (online supplemental figure S3 and table S6).

Mediation analysis

The total effect of eGDR on new-onset CKD risk was consistently significant across all mediators, with coefficients ranging from 1.1075 to 1.1481 (p<0.001). Significant indirect effects were observed for mediators, such as DM (coefficient: 0.2036, proportion: 17.1%, p<0.001), SBP (coefficient: 0.2467, proportion: 22.0%, p<0.05), HbA1c (coefficient: 0.1289, proportion: 11.1%, p<0.001), and baPWV (coefficient: 0.1137, proportion: 9.7%, p<0.001), indicating that these factors partially mediated the relationship between eGDR and CKD event (figure 4). Mediators such as sex, smoking, and BMI showed minimal or non-significant indirect effects, whereas direct effects remained significant for all mediators, suggesting a large independent impact of eGDR on CKD risk (online supplemental table S7).

Figure 4. Mediation analysis on associations between eGDR and new onset of CKD. (A) Indirect associations of eGDR and CKD events via DM. (B) Indirect associations of eGDR and CKD events via HbA1c. (C) Indirect associations of eGDR and CKD events via FBG. (D) Indirect associations of eGDR and CKD events via SBP. (E) Indirect associations of eGDR and CKD events via DBP. (F) Indirect associations of eGDR and CKD events via baPWV. baPWV, brachial–ankle pulse wave velocity; β, coefficient; CKD, chronic kidney disease; DBP, diastolic blood pressure; DE, direct effect; DM, diabetes mellitus; eGDR, estimated glucose disposal rate; FBG, fasting blood glucose; HbA1c, glycated hemoglobin; ie, indirect effects; Prop: proportion mediated; SBP, systolic blood pressure; TE, total effect.

Discussion

Our study showed that the average eGDR was significantly higher in the non-CKD group than in the CKD group, suggesting reduced insulin sensitivity in individuals with CKD. A follow-up study further identified a significant association between decreased insulin sensitivity and new-onset CKD, which was particularly pronounced in individuals without diabetes and those with higher baPWV. Mediation analysis indicated that diabetes, elevated blood pressure and AS mediated the relationship between eGDR and new-onset CKD.

IR can occur in both the early⁷ and uremic⁹ stages of CKD, and our data indicate that patients with CKD exhibit higher IR levels than those without CKD. Since this study was conducted in people who underwent health screening, most of the patients diagnosed with CKD were in the microalbuminuria stage, and relatively few individuals were in CKD stages 3–5. Therefore, no correlation analysis based on CKD stages was performed in this study. Common risk factors for IR in patients with CKD include obesity, reduced physical activity, sarcopenia, HTN, and dysregulated glucose and lipid metabolism.^{18 19} Skeletal muscles are the main sites of IR in patients with CKD.^{20 21} The muscles in patients with CKD become increasingly weak as the disease progresses, which may lead to IR by reducing the ability of the muscles to metabolize glucose. IR may disrupt muscle protein homeostasis, decreasing muscle mass and function.^{22 23} Moreover, patients with CKD exhibit several unique metabolic abnormalities such as accumulation of uremic toxins, metabolic acidosis, and vitamin D deficiency.¹⁹ From a molecular perspective, endothelial dysfunction, oxidative stress, inflammatory responses, activation of the renin–angiotensin–aldosterone system, muscle mitochondrial dysfunction, and metabolomic changes are all closely associated with IR.^{8 24}

The hyperinsulinemic–euglycemic clamp technique (HECT) is the gold standard for evaluating IR. However, its complex operation and time-consuming nature make it difficult to implement on a large scale. Simplified insulin sensitivity indices developed based on FBG, insulin levels, and intravenous glucose tolerance tests tend to underestimate the degree of IR compared with HECT.^{25 26} Additionally, fasting insulin measurement is not a routine test for non-diabetic patients, and some antidiabetic medications may interfere with insulin measurements.²⁷ The eGDR, calculated using WC, HbA1c, and HTN, makes it more suitable for clinical applications. Importantly, the eGDR demonstrates similar accuracy to the HECT clamp in assessing IR status. Studies indicate that IR exists prior to the onset of HTN and glucose/lipid metabolism disorders.^{28 29} IR shares common risk factors with CKD, including obesity, physical inactivity, sarcopenia, HTN, glucose/lipid metabolism disorders, and poor dietary habits. Accurate and convenient assessment of IR levels facilitates early identification of risk factors for CKD.

The relationship between IR and new-onset CKD has been previously demonstrated. The Uppsala Longitudinal Study of Adult Men Study suggested a negative correlation between IR and the occurrence of renal dysfunction, with a stronger correlation in individuals with normal FBG levels.⁶ A Swedish cohort study enrolling 22 146 patients with type 1 DM, with an average follow-up period of 4.8 years, showed that the incidence of CKD was 6.5% in those with an eGDR≥8 mg/(kg/min) and 46.6% in those with an eGDR<4 mg/(kg/min), indicating a significant increase in CKD incidence with increasing IR.¹¹ In this study, diabetes-stratified subgroup analyses indicated a stronger link between IR and new-onset CKD in patients without DM compared with those with DM. These results suggest that IR contributes to CKD development independent of its impact on glucose metabolism. As confirmed in our previous research using the HECT, IR is caused by a variety of metabolic diseases and is not solely due to disturbances in glucose metabolism.³⁰

In our previous study, we identified a non-linear relationship between homeostatic model assessment for IR and UACR with a saturation point.³¹ Similar findings were observed in studies exploring the relationship between IR and pre-diabetes and HTN in patients with abnormal glucose metabolism and non-alcoholic fatty liver disease.32,34 In this study, we assessed the linear and non-linear relationships between eGDR and CKD in both cross-sectional and cohort studies using the RCS. The results indicated that the non-linear relationship in the cross-sectional study was statistically significant, whereas the p value for non-linearity in the cohort study was not statistically significant. However, regardless of the study design, the overall trend showed that the CKD risk decreased with increasing eGDR. This was confirmed by logistic and Cox regression models and the subgroup analysis in both studies. However, the cause of this phenomenon remains unclear. Some researchers have attributed it to differences in participant selection and covariates. However, this explanation is not convincing. Additionally, the formulas used to assess IR in these studies were different. Another study applied multiple IR assessment formulas to explore the relationship between IR and CKD, indicating that the RCS curves of different formulas exhibit certain differences; however, this study was cross-sectional and did not use the eGDR formula.³⁵ Overall, based on a literature review and our findings, we speculate that when IR reaches or exceeds a certain threshold, its detrimental effects do not disappear but rather saturate. This suggests that early intervention for IR may be beneficial for patients. Therefore, clinical practice requires a comprehensive analysis of the effect of IR levels on patients.

We revealed that the relationship between eGDR and new-onset CKD is influenced by glucose metabolism and blood pressure and partially mediated by baPWV. The American Heart Association defines cardiovascular–kidney–metabolic (CKM) syndrome as adverse interactions between cardiovascular disease, kidney disease, and metabolic disorders.³⁶ Thus, IR may serve as a bridge to CKM syndrome. With an increase in IR levels, the proportion of patients with CKM syndrome stage 3 (including very high-risk CKD or subclinical cardiovascular disease) also increases.³⁷ Atherosclerosis is associated with various metabolic diseases, including HTN, diabetes, hyperlipidemia, and obesity, with IR being a central aspect of these metabolic disorders.^{38 39} A decline in eGFR is not significantly correlated with new-onset atherosclerosis; instead, renal function decline is a consequence of atherosclerosis.⁴⁰ Elevated IR levels correlate with a higher risk of AS (baPWV>1800 cm/s), microalbuminuria, and CKD.⁴¹ Hyperglycemia activates the advanced glycation end-products pathway, leading to endothelial injury, oxidative stress, and inflammatory responses, thereby accelerating atherosclerosis and glomerulosclerosis.^{42 43} IR stimulates the renin–angiotensin–aldosterone system, leading to water and sodium retention, elevated blood pressure, and intraglomerular HTN.^{44 45} Prolonged HTN exacerbates cardiac afterload and glomerular hyperfiltration, forming a vicious cycle of cardiac-kidney interactive damage. Renal blood flow, characterized by high perfusion and low resistance, makes the kidneys susceptible to damage from pulsatile pressure and flow.⁴⁶ Atherosclerosis reduces the ability of vascular walls to buffer against pressure, leading to high-velocity arterial blood flow that affects the fragile microvascular system of kidneys, resulting in endothelial dysfunction and oxidative stress, ultimately causing microvascular ischemia and renal impairment.^{47 48} Treatment with angiotensin-converting enzyme inhibitors and angiotensin receptor blockers has been shown to improve atherosclerosis and slow the progression of CKD.^{46 49}

Our results confirmed the association between IR and incident CKD and introduced a novel finding that baPWV mediates this relationship. Nonetheless, this study has some limitations. The HECT is recognized as the definitive method for assessing IR. Previous studies have established a strong correlation between eGDR and GDR measured using the HECT¹³; however, this claim has not been validated in the current population. Additionally, in large-sample clinical studies, the feasibility of the HECT is relatively low. However, because each simplified IR index has limitations, a combined assessment of multiple IR indices may improve the scientific rigor and accuracy of the research. At last, the study cohort was sourced from the PLA General Hospital Health Check Center, which may have led to a selection bias. Collecting nationwide data from various regions and conducting multicenter studies may yield more representative results.

In conclusion, this retrospective cohort study demonstrated an association between the IR index eGDR and the incidence of CKD, with a higher eGDR linked to a reduced risk of new-onset CKD. Furthermore, mediation analysis revealed that blood glucose, blood pressure, and AS mediated this relationship. These findings emphasize the need to manage IR and metabolic comorbidities to prevent CKD onset.

Data availability statement

Data are available upon reasonable request.