Abstract

Background and aims:

Insulin resistance, assessed by estimated glucose disposal rate (eGDR), is linked to atherosclerosis, yet evidence primarily comes from diabetic populations. The aim of this study was to investigate the association between eGDR and coronary artery calcification (CAC) in non-diabetic adults.

Methods:

In this cross-sectional study, 4750 participants aged 40–65 years without diabetes were enrolled from Jinling Hospital between 2022 and 2024. CAC was assessed via CT scans. Multivariable Logistic regression and restricted cubic splines were used to analyze the relationship of eGDR with CAC prevalence and severity.

Results:

The mean age was 47.98 ± 6.61 years; 79.31% were male, and 665 participants (14.0%) had CAC. After multivariable adjustment, lower eGDR was significantly associated with higher CAC risk. Each 1-unit increase in eGDR was associated with a 9% risk reduction (OR = 0.91, 95% CI: 0.84–0.98, P = 0.02). Compared to the lowest eGDR quartile (Q1), the odds ratios for Q2, Q3, and Q4 were 0.84 (0.66–1.07), 0.69 (0.52–0.92), and 0.63 (0.44–0.90), respectively (P for trend = 0.004). Tobit regression confirmed an inverse association between eGDR and CACS (β = -4.07, P = 0.0002). Multivariate ordered Logistic regression analysis revealed that after adjusting for multiple factors, eGDR was significantly associated with the severity of CAC, both as a continuous variable (mild CAC: OR = 0.90, 0.83–0.97; moderate-to-severe CAC: OR = 0.86, 0.78–0.95) and across quartiles (P for trend = 0.03). A nonlinear dose-response relationship was observed (P for nonlinearity = 0.01), with CAC risk increasing as eGDR fell below 10.8.

Conclusion:

Lower eGDR is independently associated with increased prevalence and severity of CAC in non-diabetic middle-aged Chinese adults.

Introduction

Atherosclerosis represents a major contributor to global morbidity and mortality, responsible for millions of cardiovascular disease (CVD)-related deaths each year (1). In China, an estimated 330 million people are affected by CVD, and its prevalence continues to rise, imposing a substantial public health burden (2). As the primary pathological basis of ischemic heart disease, atherosclerosis requires early detection of subclinical stages for effective prevention. Coronary artery calcification (CAC), a hallmark of advanced atherosclerosis, serves as a crucial biomarker for evaluating both the presence and severity of coronary artery disease (3). Multislice computed tomography (CT) provides a rapid and noninvasive imaging modality, recognized for its reliability, high sensitivity, and specificity in diagnosing CAC (4). This technique enables the quantification of CAC scores, which reflect the overall burden of coronary plaque. Importantly, elevated CAC scores have been shown to independently and incrementally predict future coronary events and clinical outcomes (5).

Insulin resistance (IR), a pathophysiological condition characterized by diminished sensitivity of target tissues to insulin, impairs glucose utilization and is a well-established risk factor for atherosclerosis (6). Given the adverse implications of IR, several methods have been developed to assess it. Although the hyperinsulinemic-euglycemic clamp is considered the gold standard for identifying IR (7), its clinical utility and feasibility in large-scale epidemiological studies are limited due to the time-consuming and labor-intensive nature of the procedure. Similarly, the Homeostasis Model Assessment of Insulin Resistance (HOMA-IR) is less suitable for large population-based cohorts because of its cost and operational complexity (8). As a result, the estimated glucose disposal rate (eGDR)—a measure derived from waist circumference (WC), hypertension status, and glycated hemoglobin (HbA1c)—has emerged as a reliable surrogate marker of IR (9). This method has demonstrated high accuracy when compared with the hyperinsulinemic-euglycemic clamp technique, making it a valuable tool for assessing insulin resistance in large patient populations (10).

Recent studies have utilized eGDR as a surrogate for insulin resistance in predicting stroke, coronary artery disease, and all-cause mortality (11–13). However, most of these studies have focused primarily on individuals with diabetes, which may overestimate or confound the role of IR. Previous research consistently indicates significant heterogeneity between diabetic and non-diabetic populations (14). Individuals with diabetes are more likely to develop additional health complications, face higher cardiovascular risk, and experience increased mortality. As also highlighted by Ren et al., the non-diabetic population exhibits greater sensitivity to eGDR (15). Therefore, the objective of this study is to investigate the association between eGDR and the prevalence and severity of CAC in a large sample of non-diabetic adults undergoing routine health examinations.

MethodsStudy design and population

This cross-sectional study included individuals aged 40 to 65 years who underwent health check-ups at the Department of Health Medicine, Jinling Hospital, Affiliated with Nanjing University, between January 2022 and December 2024, and who completed chest computed tomography (CT) examinations. For participants with multiple visits during this period, only the most recent record was included, resulting in an initial cohort of 6,788 subjects. All participants were recruited from various organizations across Jiangsu Province, representing diverse socioeconomic backgrounds.

As part of the check-up process, each subject was interviewed by an internist regarding lifestyle, medical history, and medication use. Anthropometric measurements and blood samples were collected by trained nurses. Laboratory analyses were performed uniformly by specialized technicians in the Laboratory Department under standardized quality control protocols. The following exclusion criteria were applied: (1) Missing blood pressure data (n = 228); (2) Missing waist circumference data (n = 689); (3) Missing glycated hemoglobin (HbA1c), fasting blood glucose (FBG), or postprandial blood glucose (PBG) data (n = 750); (4) Diagnosis of diabetes mellitus (n = 344); (5) History of coronary stenting or coronary artery bypass grafting (n = 21); (6) History of malignant neoplasms (n = 6). After applying these criteria, a total of 4,750 participants were included in the final analysis, as illustrated in Figure 1. The study protocol was approved by the Institutional Review Board of Jinling Hospital, Affiliated Hospital of Medical School, Nanjing University (No. 2024DZKY-015-01), and all participants provided written informed consent.

Flow chart of study participants.

Data collection and definition

The physician collected detailed information such as smoking and drinking habits, disease history, and medication use from the subjects during the internal medicine physical examination. Smoking status was defined as current smoking (yes/no), and drinking status was defined as current drinking (yes/no), both based on self-reported verbal inquiry during the clinical interview. Height and weight were measured using an SH-200G meter. Body mass index (BMI) was calculated by dividing weight in kilograms by height in metres squared. WC was measured by professional nurses. After a 10-minute rest period, blood pressure was measured in the non-dominant arm using an Omron sphygmomanometer. Venous blood samples were collected from subjects who had fasted for at least eight hours. FBG, triglyceride (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-c), low-density lipoprotein cholesterol (LDL-c), serum creatinine (SCR) and uric acid (UA) levels were measured using a Hitachi 7600 fully automated biochemistry analyzer. Venous blood was drawn two hours after breakfast to measure PBG. HbA1c levels were measured using ion exchange high-pressure liquid chromatography. Quality control of laboratory testing was performed in accordance with the Indicators of Medical Quality Control for Clinical Laboratory Specialties (2015 edition) of the National Health Commission of the People’s Republic of China.

Hypertension was defined as follows: self-reported hypertension based on physician diagnosis, and/or any use of antihypertensive medications, and/or SBP higher than 140 mmHg and/or DBP higher than 90 mmHg (16). Diabetes was defined based on a self-reported physician diagnosis, use of hypoglycemic drugs, or FBG higher than 7.0 mmol/L, or PBG higher than 11.0 mmol/L (17). Estimated glomerular filtration rate (eGFR) was calculated using the modified MDRD equation: eGFR(ml/min/1.73m2) =186 × (SCR) ^-1.154 × (age)^-0.203 × (0.742 female) × (1.233 Chinese) (18). The formula for calculating eGDR was as follows: eGDR (mg/kg/min) = 21.158 − (0.09 × WC) − (3.407 × hypertension) − (0.551 × HbA1c) [WC (cm), hypertension (yes = 1/no = 0), and HbA1c (%)] (19).

Assessment of CAC

Each participant underwent a chest CT scan using a 64-slice multi-slice CT machine (SIEMENS SOMATOM Definition Flash). CAC scores (CACS) were calculated using an automated software program and the Agatston scoring method (20). Participants were categorized as follows based on the CAC scores: no CAC, CACS = 0; mild CAC, 0<CACS ≤ 100;moderate CAC, 100<CACS ≤ 300; severe CAC, CACS>300 (21).

Statistical analysis

We used SAS software, version 9.4 (SAS Institute, Cary, NC) for statistical analysis. A two-sided P value of less than 0.05 was considered statistically significant. Study participants were categorized into four groups (Q1–Q4) according to quartiles of the eGDR. Continuous variables with normal distribution are presented as mean ± standard deviation (SD), while those with non-normal distribution are summarized as median (interquartile range) [M (Q1, Q3)]. Categorical variables are expressed as numbers and percentages. Trends across eGDR quartiles were assessed using linear regression for continuous variables and the Cochran–Armitage trend χ² test for categorical variables. Logistic regression was used to analyze the odds ratios (ORs) of the relationship between eGDR (continuous as well as categorical variables) and CAC. Multivariable Tobit regression models with left-censoring at zero were employed to examine factors associated with CACS, given the highly left-censored distribution of CACS. Multivariate ordered Logistic regression analysis was used to analyze the correlation between eGDR (continuous variable and categorical variable) and the severity of CAC. To allow for more flexibility in the model and visualize dose-response relationships, restricted cubic spline models with four knots at the 5th, 35th, 65th, and 95th percentiles of eGDR were constructed.To address potential confounding by smoking and drinking status, sensitivity analyses were performed with additional adjustment for these variables in the subset with available data (n = 3,360). Given the male predominance in our study (79.31%), sex-stratified analyses were conducted, and interaction terms (eGDR × sex) were included in the models to assess effect modification by sex. Multicollinearity among independent variables was assessed using variance inflation factors (VIF), with VIF < 5 considered indicative of no serious collinearity. Post-hoc power analyses were performed based on the observed sample sizes, event rates, and effect sizes to evaluate the statistical adequacy of our study, with power ≥80% considered adequate. Additionally, to validate the use of eGDR in non-diabetic populations, we conducted a sensitivity analysis using the triglyceride-glucose (TyG) index, calculated as ln[fasting triglycerides (mg/dL) × fasting glucose (mg/dL)/2], as an alternative insulin resistance surrogate.

ResultsParticipants characteristics

Table 1 presents the baseline characteristics of the study participants stratified by eGDR quartiles (Q1: 8.26 ± 0.40; Q2: 10.06 ± 0.18; Q3: 10.61 ± 0.17; Q4: 11.50 ± 0.46). A total of 4750 subjects (mean age: 47.98 ± 6.61 years) with 79.31% male were included in this study. The mean age, proportion of male, SBP, DBP, BMI, WC, levels of FBG, PBG, HbA1c, TC, TG, LDL-c and UA all decreased with increasing eGDR (all P for trends < 0.05). However, individuals with higher levels of eGDR tended to have higher HDL-c and eGFR (P for trend < 0.0001). A significant inverse relationship was observed between eGDR and the prevalence of CAC, with CAC prevalence decreasing across ascending eGDR quartiles. Furthermore, higher eGDR levels were associated with an increased proportion of mild CAC and a decreased proportion of severe CAC. The trend in CAC severity across eGDR quartiles was statistically significant (P for trend < 0.0001).

CharacteristicsOverallQuartiles of eGDRQuartile 1Quartile 2Quartile 3Quartile 4P for trendn47501183120211741191 eGDR10.11± 0.408.26 ± 0.4010.06 ± 0.1810.61 ± 0.1711.50 ± 0.46 < 0.0001 Age, years47.98 ± 6.6150.16  ± 7.3347.87 ± 6.2146.90 ± 6.0346.98 ± 6.27 < 0.0001 Male, n (%)3767 (79.31)1043 (88.17)1138 (94.68)1016 (86.54)570 (47.86)< 0.0001 SBP, mmHg120.56 ± 13.43130.31 ± 14.99119.69 ± 10.48117.51 ± 10.81114.74 ± 11.49 < 0.0001 DBP, mmHg74.26 ± 9.6780.39 ± 10.4274.15 ± 8.0972.33 ± 8.1870.18 ± 8.70 < 0.0001 BMI, kg/m224.45 ± 2.6526.58 ± 2.6225.29 ± 1.7623.93 ± 1.7322.01 ± 1.93 < 0.0001 WC, cm84.43  ± 8.8192.43 ± 8.4488.55 ± 2.5683.01 ± 2.6773.73 ± 5.32 < 0.0001FBG, mmol/L5.25 ± 0.495.46 ± 0.535.29 ± 0.485.18 ± 0.425.06 ± 0.42< 0.0001PBG, mmol/L6.21  ± 1.336.76 ± 1.466.34 ± 1.276.03 ± 1.205.71 ± 1.14< 0.0001HbA1c, %5.63 ± 0.365.78 ± 0.425.68 ± 0.315.58 ± 0.335.49 ± 0.30 < 0.0001TC, mmol/L5.10 ± 0.915.15 ± 1.005.10 ± 0.905.07 ± 0.895.07 ± 0.85 0.02TG, mmol/L1.24 (0.87 − 1.77)1.57 (1.09 − 2.27)1.38 (1.00 − 1.94)1.18 (0.85− 1.64)0.94 (0.71 − 1.32) < 0.0001HDL-c, mmol/L1.28 ± 0.301.19 ± 0.271.19 ± 0.241.29 ± 0.281.44 ± 0.32 < 0.0001LDL-c, mmol/L2.91 ± 0.712.98 ± 0.782.98 ± 0.702.90 ± 0.692.77 ± 0.66 < 0.0001UA, umol/L367.59 ± 84.36393.64 ± 84.37391.18 ± 76.37368.84 ± 77.79316.68 ± 75.14 < 0.0001eGFR, mL/min/1.73m2124.02 ± 20.19122.19 ± 21.40122.94 ± 19.09124.01 ± 19.12126.94 ± 20.74< 0.0001Smoking status0.27 Yes, n (%)1252 (26.36%)259 (21.89%)363 (30.20%)345 (29.39%)285 (23.93%) No, n (%)2108 (44.38%)475 (40.16%)543 (45.17%)550 (46.85%)540 (45.34%) Missing, n1390 (29.26%)449 (37.95%)296 (24.63%)279 (23.76%)366 (30.73%)Drink status< 0.0001 Yes, n (%)1207 (25.41%)320 (27.05%)372 (30.95%)333 (28.36%)175 (14.69%) No, n (%)2153 (45.32%)414 (35.00%)534 (44.42%)562 (47.87%)650 (54.58%) Missing, n1390 (29.26%)449 (37.95%)296 (24.63%)279 (23.76%)366 (30.73%)CAC, n (%)665 (14.00%)277 (23.42%)186 (15.47%)123 (10.48%)79 (6.63%)< 0.0001 Mild, n (%)412 (61.96%)159 (57.40%)110 (59.14%)91 (73.98%)52 (65.82%)< 0.0001 Moderate, n (%)167 (25.11%)78 (28.16%)52 (27.96%)17 (13.82%)20 (25.32%) Severe, n (%)86 (12.93%)40 (14.44%)24 (12.90%)15 (12.20%)7 (8.86%)

Baseline characteristics of participants stratified by quartiles of estimated glucose disposal rate.

BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; eGDR, estimated glucose disposal rate; FBG, fasting blood glucose; PBG, 2-hour postprandial blood glucose; HbA1c, glycosylated haemoglobin; HDL-c, high density lipoprotein cholesterol; LDL-c, low density lipoprotein cholesterol; TC, total cholesterol; TG, triglycerides; UA, uric acid; WC, waist circumference; eGFR, estimated glomerular filtration rate; CAC, coronary artery calcification.

Association of eGDR and risk of CAC

As shown in Table 2, each 1-unit increase in eGDR was associated with a 17% reduction in the risk of CAC after adjustment for age and sex (OR = 0.83, 95% CI: 0.78–0.88, P < 0.0001). This association, though attenuated, remained statistically significant after further adjustment for body mass index, blood pressure, blood lipid profiles, uric acid, blood glucose, and eGFR (OR = 0.91, 95% CI: 0.84–0.98, P = 0.02).

eGDRNo. of CAC/total NModel 1Model 2OR (95%CI)P valueOR (95%CI)P valueContinuouseGDR, per 1unit665/47500.83 (0.78-0.88)< 0.00010.91 (0.84-0.98)0.02CategoricalQ1277/1183ref< 0.0001ref0.004Q2187/12020.71 (0.57-0.88)0.84 (0.66-1.07)Q3123/11740.53 (0.42-0.67)0.69 (0.52-0.92)Q478/11910.44 (0.33-0.58)0.63 (0.44-0.90)

Association of eGDR with risk of CAC.

Model 1, adjusted for age and sex.

Model 2, additionally adjusted for body mass index, systolic blood pressure, diastolic blood pressure, total cholesterol, triglyceride, LDL-c, HDL-c, uric acid, eGFR, fasting blood glucose and 2h plasma glucose based on model 1.

Similar results were observed in the quartile-based analysis. Compared with the lowest eGDR quartile (Q1), the adjusted risks of CAC in Q2, Q3, and Q4 were significantly lower after controlling for age and sex, with reductions of 29% (OR = 0.71, 95% CI: 0.57–0.88), 47% (OR = 0.53, 95% CI: 0.42–0.67), and 56% (OR = 0.44, 95% CI: 0.33–0.58), respectively (P for trend < 0.0001). After additional adjustment for the full set of covariates, the association was attenuated but remained significant across eGDR quartiles (P for trend = 0.004).

Association of eGDR with severity of CAC

As presented in Table 3, Tobit regression (accounting for 86% zero values) was used to evaluate associations with CACS. In a model adjusted for all available covariates (Model 1), eGDR was significantly and inversely associated with CACS (β = -3.70, P = 0.004). After further adjustment for age, sex, BMI, and fasting blood glucose (Model 2), eGDR remained a significant negative predictor of CACS (β = -4.07, P = 0.0002), indicating that lower eGDR is associated with higher CACS. The significant scale parameter (Sigma) in both models (P < 0.0001) confirmed the appropriateness of the Tobit model.

CharacteristicModel 1Model 2βSEP valueβSEP valueeGDR-3.701.270.004-4.071.100.0002age1.940.21< 0.00012.010.20< 0.0001sex-14.743.65< 0.0001-13.183.25< 0.0001BMI1.500.600.011.180.570.04SBP0.170.140.21DBP-0.170.180.35TG0.601.790.74TC-2.865.020.57LDL-2.985.800.60HDL8.417.420.26FBG5.652.880.04996.812.630.01PBG0.991.040.34HbA1c3.633.860.35UA-0.0040.020.82eGFR0.050.060.41Model statisticsSigma (Scale Parameter)83.370.86< 0.000183.540.86< 0.0001

Multivariable Tobit regression analysis for factors associated with CACS. .

BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; eGDR, estimated glucose disposal rate; FBG, fasting blood glucose; PBG, 2-hour postprandial blood glucose; HbA1c, glycosylated haemoglobin; HDL-c, high density lipoprotein cholesterol; LDL-c, low density lipoprotein cholesterol; TC, total cholesterol; TG, triglycerides; UA, uric acid; WC, waist circumference; eGFR, estimated glomerular filtration rate. Data were presented as coefficient (β) and standard error (SE) from Tobit regression models with left-censoring at zero to account for the 86% zero values. Model 1 includes all available covariates as listed. Model 2 retains variables that were significantly associated with CACS (P < 0.05) in Model 1, except where theoretical considerations apply.

The association between eGDR as a continuous variable and CAC severity was presented in Table 4. After adjustment for age and sex, each 1-unit increase in eGDR was associated with 18% (OR = 0.82, 95% CI: 0.78–0.87, P < 0.0001) and 21% (OR = 0.79, 95% CI: 0.73–0.86, P < 0.0001) reduced risks of mild CAC and moderate-to-severe CAC, respectively. Following further multivariable adjustment for BMI, blood pressure, blood lipids, eGFR, uric acid, FBG, and PBG, the associations, though slightly attenuated, remained statistically significant, with corresponding ORs of 0.90 (95% CI: 0.83–0.97, P = 0.007) for mild CAC and 0.86 (95% CI: 0.78–0.95, P = 0.003) for moderate-to-severe CAC. The point estimates suggest a slightly stronger inverse association for moderate-to-severe CAC than for mild CAC, though the confidence intervals overlapped considerably.

OutcomeModel 1Model 2OR (95%CI)P valueOR (95%CI)P valueNo CACrefrefMild CAC0.82 (0.78-0.87)< 0.00010.90 (0.83-0.97)0.007Moderate-to-severe CAC0.79 (0.73-0.86)< 0.00010.86 (0.78-0.95)0.003

Ordered Logistic regression of eGDR with severity of CAC.

Model 1, adjusted for age and sex.

Model 2, additionally adjusted for body mass index, systolic blood pressure, diastolic blood pressure, total cholesterol, triglyceride, LDL-c, HDL-c, uric acid, eGFR, fasting blood glucose and 2h plasma glucose based on model 1.

Table 5 showed the association between categorical eGDR and CAC severity. After adjusting for age and sex (Model 1), a significant inverse dose–response relationship was observed across eGDR quartiles (P for trend = 0.0002). Compared with the lowest quartile (Q1), the risk of more severe CAC was reduced by 30% in Q2 (OR = 0.70, 95% CI: 0.56–0.86; P = 0.0009), by 49% in Q3 (OR = 0.51, 95% CI: 0.40–0.65; P < 0.0001), and by 56% in Q4 (OR = 0.44, 95% CI: 0.33–0.58; P < 0.0001). After additional adjustment for metabolic indicators (Model 2), the inverse trend remained significant (P for trend = 0.03). Risk reductions persisted in Q3 (OR = 0.68, 95% CI: 0.51–0.90; P = 0.007) and Q4 (OR = 0.64, 95% CI: 0.45–0.92; P = 0.01), corresponding to reductions of 32% and 36%, respectively, while the association in Q2 was attenuated and no longer significant (OR = 0.84, 95% CI: 0.66–1.06; P = 0.13).

eGDRModel 1Model 2OR (95%CI)P valueP for trendOR (95%CI)P valueP for trendQ1ref0.0002ref