This retrospective, population-based longitudinal cohort study is based on a general health check-up population from the Department of Health Medicine at the Chinese PLA General Hospital. From October 2009 to December 2023, a total of 21,642 participants aged 18 and above underwent general health check-ups. The exclusion criteria were as follows: (1) no carotid ultrasound examination (n = 2,012); (2) only one general health check-up (n = 4,393); (3) patients lacking necessary blood sample tests, WC, and baseline history of HTN and diabetes (n = 3,029); (4) patients with malignancy (n = 1,204); and (5) patient lacking follow-up carotid color ultrasonography (n = 3,644). In the end, a total of 7,360 participants were included in the final analysis and further divided into four subgroups on the basis of the quartiles (Qs) of the eGDR. The detailed inclusion and exclusion process is shown in Figure 1. This study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the Chinese PLA General Hospital (S2019-190-02). All patients provided written informed consent.

Clinical data were collected from the enrolled patients, and their general information and medical history were recorded. Researchers also measured participants' weight, height, and WC while they were wearing light clothing and not wearing shoes. The body mass index (BMI) was calculated as weight (kg)/height (m)², and obesity was defined as a BMI ≥ 28 kg/m². The definition of HTN is a systolic blood pressure ≥140 mmHg or a diastolic blood pressure ≥90 mmHg, current use of antihypertensive medication, or self-reported history of HTN. The definition of diabetes was FBG ≥7.0 mmol/L in the cohort examination or a self-reported history of diabetes diagnosed by a doctor (19). In addition, the biochemical parameters tested included HbA1c, total cholesterol (TC), TGs, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), FBG, high-sensitivity C-reactive protein (hsCRP), and blood creatinine. The eGDR (mg/kg/min) was calculated with the following formula: 21.158–(0.09^*WC)–(3.407^*HTN)–(0.551^*HbA1c) (20). The NHHR data is obtained using the formula for TC minus HDL-C, then divided by HDL-C. Fasting blood samples were collected in the morning via a Roche C8000 fully automated biochemical analyzer (Roche, Mannheim, Germany) equipped with corresponding reagents, calibrators, and quality control materials to analyze a series of biochemical parameters.

The ultrasound evaluation of the bilateral carotid arteries was manually performed by certified physicians from the Department of Health Medicine at the Second Medical Center of the PLA General Hospital, who were unaware of this study. The participants were examined via a high-frequency ultrasound probe (7.5–10.0 MHz). Abnormal carotid intima–media thickness (cIMT) is defined as a maximum cIMT value ≥0.9 mm, which is the maximum distance between the intima-media and the outer membrane of the lumen. Carotid plaques were defined as cIMT ≥ 1.5 mm, focal structures protruding into the arterial lumen ≥0.5 mm, or ≥50% of the surrounding cIMT value. Furthermore, the progression of CAS is defined as the emergence of new carotid artery stenosis, carotid artery plaques, or cIMT during the follow-up period compared with baseline. For individuals with combined carotid plaques and cIMT, baseline data and follow-up results are defined on the basis of the dominant manifestation (i.e., carotid plaques) (21).

Continuous variables were analyzed using ANOVA for normally distributed data and the Kruskal-Wallis H test for skewed distributions, with results expressed as mean ± standard deviation (SD) or median (minimum-maximum range) respectively. Categorical data are reported as frequency counts (percentages), and between-group comparisons were analyzed by chi-square (χ²) tests. Proportional hazards assumptions were confirmed using Schoenfeld residuals prior to Kaplan-Meier analysis. Survival probabilities across eGDR quartile groups were estimated using Kaplan-Meier analysis, with between-group statistical differences evaluated by log-rank tests. The adjusted covariates comprised demographic factors (age, sex), anthropometric measures (body mass index), lifestyle variables (smoking, alcohol consumption), clinical comorbidities (hypertension, diabetes), and biochemical parameters (triglycerides, hemoglobin, uric acid, high-sensitivity C-reactive protein). Covariate selection was based on clinical relevance and prior evidence from observational cohort studies.

The dose-response relationships of both eGDR and NHHR with CAS progression were modeled using restricted cubic splines (RCS) with five knots. The optimal knot positions in the restricted cubic spline (RCS) models were determined through Akaike Information Criterion (AIC) minimization. Kaplan-Meier and Cox regression models were used to analyze the associations of eGDR and NHHR with the progression of CAS.

Mediation analysis was used to investigate the potential mediating role of NHHR in the association between eGDR and CAS progression. Stratified analyses assessed effect heterogeneity across clinically relevant subgroups: age (< 60 vs. ≥60 years), sex, smoking status, and alcohol use, with multiplicative interaction terms evaluated through likelihood ratio tests. We also performed interaction analyses to assess the potential interactions between each subgroup and CAS progression.

We evaluated the robustness of our conclusions by conducting several sensitivity analyses. First, the analysis was repeated after excluding subjects with diabetes. Second, hypertension was redefined using a threshold of 130/80 mmHg, and eGDR was recalculated to reanalyze the data for the remaining subjects. The incremental predictive capacity of eGDR and NHHR for CAS progression was assessed through Harrell's C-statistic, with model discrimination improvements quantified using continuous net reclassification improvement (NRI) and integrated discrimination improvement (IDI) indices. All statistical analyses were performed using R software version 4.4.2. All inferential analyses employed two-tailed hypothesis testing with α = 0.05 as the significance threshold.

Among the 7,360 eligible participants, the average age was 49.3 ± 8.72 years, with 31.0% being female. The median eGDR was 8.71, and the median follow-up time was 30 months. A total of 3,751 patients (51.0%) experienced outcome events. The baseline characteristics of the individuals included were based on CAS progression (see Supplementary Table S1). A comparison of the baseline characteristics stratified by eGDR quartiles (Q1: 0.330–6.139; Q2: 6.141–8.711; Q3: 8.712–10.498; and Q4: 10.507–13.403) is shown in Table 1. The average age, male sex ratio, systolic blood pressure, diastolic blood pressure, BMI, WC, HGB, HbA1c, TG, LDL, UA, and hsCRP levels, current smoking status, and current alcohol consumption status all decreased with an increasing eGDR (all P < 0.001).

According to restricted cubic spline analysis, the eGDR in the population showed a non-linear relationship with CAS progression, whereas the NHHR showed a linear positive correlation with CAS progression. After fully adjusting for covariates, the eGDR still exhibited a non-linear relationship with CAS progression (overall P < 0.001, non-linear P < 0.001). After redefining HTN as ≥130/80 mmHg, the eGDR showed a linear negative correlation with CAS progression (Supplementary Figure S1). The restricted cubic spline plot shows that the cutoff value of NHHR is 2.89. When the NHHR were >2.89, the risk of CAS progression significantly increased (Figure 2). Some previous studies have also shown that the high-risk threshold of NHHR is around 2.8. Therefore, we selected 2.89 as the high-risk threshold for NHHR for subsequent analysis (22, 23). The restricted cubic spline plot shows that the cutoff value of eGDR is 8.73. When the eGDR were < 8.73, the risk of CAS progression significantly increased (Figure 2). However, since eGDR was nonlinearly correlated with the progression of carotid atherosclerosis, we finally selected a median of 8.71 as the high-risk threshold for subsequent analysis. As shown in Table 2 and Figure 3A, the Kaplan–Meier survival curve indicated that individuals with a lower eGDR had a greater risk of CAS progression. Compared with those with an eGDR in Q1, the HRs (95% CI) for CAS progression with an eGDR in Q2-4 were 0.82 (0.75–0.89), 0.78 (0.71–0.85), and 0.44 (0.40–0.49), respectively. After adjusting for factors such as age, sex, smoking status, alcohol consumption status, and UA levels, the results were similar. When jointly assessing the progression of CAS at the baseline eGDR index and NHHR values, the lower the population eGDR index and the higher the NHHR value were, the greater the risk of CAS progression was (Figure 3B, Table 3).

We used mediation analysis to further clarify the reciprocal mediating effects of the eGDR and NHHR on the progression of CAS (Figure 4, Supplementary Figure S2). According to the unadjusted model, the NHHR significantly mediated the effect of the eGDR on the risk of CAS progression (proportion = 19.54%). After fully adjusting for covariates, the NHHR still considerably mediated the effect of the eGDR on the risk of CAS progression (proportion = 7.50%). Similarly, the eGDR also plays an important mediating role in the impact of the NHHR on the progression of CAS (Supplementary Figure S2).

The results of the subgroup analysis revealed that in most of the prespecified subgroups, the relationship between eGDR combined with the NHHR and the risk of CAS progression was consistent with the main results (Figure 5). A significant impact on predictive performance was observed in the age subgroup (interaction P = 0.038).

In the sensitivity analysis, when we excluded diabetes patients defined by FBG and HbA1c measurements, no significant changes were observed in the results (Table 4). When we recalculated the eGDR using the redefined HTN criterion (≥130/80 mmHg), the results remained largely unchanged (Supplementary Table S2). To further reduce the impact of drug treatment on the outcome, we excluded the population using lipid-lowering drugs and hypoglycemic drugs. The result we obtained is consistent with the main result (Supplementary Table S3). Furthermore, to minimize potential selection bias as much as possible, we excluded participants with overly long follow-up periods and those with overly short follow-up periods. The results we obtained are similar to the main results, and these analyses prove the robustness of the main results (Supplementary Table S4).

Model 3 was used to construct the basic model (including age, sex, BMI, smoking status, alcohol consumption status, and TG). The inclusion of both the eGDR and the NHHR optimized the predictive ability of the basic model for CAS progression (Table 5). Moreover, adding the eGDR and NHHR improved the predictive ability for CAS progression (C statistics: 0.600 vs. 0.597, P < 0.001), and all the NRIs in the three models were significant (all P < 0.05; Table 5). In addition, we conducted the receiver operating characteristic curves analysis (ROC) to build the basic model (including age, sex, current smoking, current drinking, TC, HDL, TG, LDL, UA, hsCRP). The results show that the area under the curve (AUC) of the basic model is 0.648, the AUC including NHHR is 0.656, the AUC including eGDR is 0.663, and the AUC after including both eGDR and NHHR in the model is 0.670. eGDR+NHHR showed the highest predictive value among the three models and had statistically better discriminative performance (Supplementary Figure S3).

This study has several advantages. This was a large-scale longitudinal cohort study in a real-world setting. Repeated carotid ultrasound and biochemical index measurements allowed us to explore the impact of the eGDR and NHHR on the progression of CAS in adults. Our research has several limitations that need to be noted. First, this study was conducted at the Chinese PLA General Hospital, and the included research population was mainly the physical examination population, most of whom were of Han ethnicity. Therefore, the external validity of the results of this study may be limited, and caution is needed when applying them to other geographical regions or populations with different demographic characteristics. To confirm the general applicability of the findings of this study and enhance their external validity, future research should focus on validating these results in prospective cohorts of multi-center and more diverse populations. Second, as an observational study, we cannot determine the causal relationships between variables and outcomes. Although, we conducted a mediation analysis to approximate the causal relationship. This study can only confirm the association between eGDR and CAS. In future studies, whether improving eGDR can delay carotid atherosclerosis still needs to be verified through lifestyle intervention trials. Third, although the sample size of this study was relatively large, the number of elderly individuals was relatively small, which may affect the generalizability of our conclusions. Therefore, it is necessary to conduct statistical analysis on a broader population sample to further confirm our results. In addition, our eGDR and NHHR metrics are calculated from baseline levels.