Of the 17,708 people surveyed at baseline between June 2011 and March 2012, 6752 were recruited for analysis. Of these, 48% were male and 52% female. The median (IQR) for age, BMI, and WC for all participants were 58 (51,62), 23.0 (20.8,25.5), and 84 (78,91) respectively. Participants were categorized according to eGDR quartiles (Q) as shown in Table 1. Participants with higher eGDR quartiles were likely to be younger, female, married, agriculture hukou, less educated, have lower BMI, WC, SBP, DBP, WBC, HGB, CRP, FBG, TC, TG, LDL-c, UA, Scr, fewer comorbidities (such as hypertension, diabetes, and dyslipidemia), and lower new-onset CVD (including new-onset HD and new-onset stroke). Conversely, HDL-c and eGFR were significantly higher. History of smoking and alcohol consumption, along with depression symptoms were statistically significant among the groups; however, no significant differences were found in the history of cancer, sleep problems, platelet (PLT), and BUN.

Table S2 provides a comprehensive overview of baseline characteristics according to the occurrence of CVD, revealing no significant differences in hukou, marital status, education levels, smoking status, history of cancer, PLT, BUN, Scr, UA, and HGB (P > 0.05). The overall prevalence rates of hypertension, hyperglycemia, and dyslipidemia were 39%, 60%, and 69% respectively. Compared with participants without CVD, participants with new-onset CVD were older, female, had higher BMI, WC, SBP, and DBP, tended to have a history of alcohol consumption and comorbidities (hypertension, hyperglycemia, dyslipidemia, sleep problems, and depression), and had a higher prevalence of new-onset HD, new-onset stroke, and CKM stage 3, and higher WBC, CRP, TG, TC, LDL-c, FBG, and HbA1c, but there was lower HDL-c, eGFR, and eGDR.

During follow-up between 2013 and 2020, a total of 1495 (22%) participants developed CVD, including 1178 from new-onset HD, and 823 from new-onset stroke. Multivariable logistic regression models were utilized to estimate the relationship between the eGDR and the risk of primary and secondary outcomes, as presented in Table 2. After full multivariable adjustment in Model 3, eGDR was inversely linked to the risk of new-onset CVD, new-onset HD, and new-onset stroke. The eGDR was transformed from a continuous variable to a categorical variable based on quartiles. Compared with Q1, Q4 had a decreased risk of new-onset CVD (OR = 0.43, 95% CI: 0.36–0.50), new-onset HD (OR = 0.64, 95% CI: 0.50–0.81), and new-onset stroke (OR = 0.64, 95% CI: 0.50–0.81) in the unadjusted model. The negative association of eGDR with the incidence of CVD, HD, and stroke remained steady after adjusting for age, sex, BMI, hukou, marital status, and education level in Model 1. In Model 2, eGDR was inversely linked to the risk of new-onset CVD and new-onset HD compared with Q1; for stroke, although no significant difference was found in Q2, a negative relationship was still observed in Q3 and Q4. OR and 95% CI from Q1 to Q4 were 1.00 (reference), 0.81 (0.68, 0.96), 0.72 (0.58, 0.88), and 0.74 (0.58, 0.94) respectively, for the occurrence of CVD; 1.00 (reference), 0.81 (0.67,0.97), 0.72 (0.57,0.90), and 0.75 (0.58,0.97) respectively, for the occurrence of HD; 1.00 (reference), 0.91 (0.74,1.12), 0.80 (0.62,1.04), and 0.71 (0.52,0.97) respectively, for the occurrence of stroke after adjusting for all potential covariates. Moreover, the findings revealed that the risk of new-onset CVD, HD, and stroke decreased with increasing eGDR (from Q1 to Q4) (P for trend < 0.05). After adjusting for all covariates in all individuals with CKM syndrome stages 0–3, the RCS curves showed an approximately inverse “L” correlation between eGDR and occurrence of CVD and HD (All P for overall < 0.001, All P for nonlinear = 0.005) (Fig. 2A and B), while there was a negative linear correlation between eGDR and the risk of new-onset stroke (P for overall = 0.026, P for nonlinear = 0.098) (Fig. 2 C). In participants with CKM syndrome stage 2, eGDR exhibited a negative linear relationship with the risk of new-onset CVD (P for overall = 0.055, P for nonlinear = 0.093) (Fig. 2D), new-onset stroke (P for overall = 0.181, P for nonlinear = 0.651) (Fig. 2F), and a U-shaped relationship with the risk of new-onset HD (P for overall = 0.057, P for nonlinear = 0.032) (Fig. 2E). Among participants with CKM stage 3, eGDR displayed a decreasing linear relation with the risk of new-onset CVD (P for overall = 0.055, P for nonlinear = 0.093) (Fig. 2G) and HD (P for overall = 0.003, P for nonlinear = 0.102) (Fig. 2H), while exhibiting a U-shaped correlation with the risk of new-onset stroke (P for overall = 0.102, P for nonlinear = 0.037) (Fig. 2I).

Mediation analyses revealed that BMI partially mediated the association between eGDR and outcome (Fig. 3). Specifically, in the unadjusted model, the indirect effects of eGDR mediated by BMI were found to be associated with CVD, HD, and stroke, accounting for 19.22%, 28.01%, and 16.58%, respectively. In the fully adjusted model, the proportions mediated through BMI were 41.98%, 43.05%, and 43.23% for CVD, HD, and stroke, respectively.

BMI mediated the effect of eGDR and new-onset CVD, HD, and stroke in different adjusted models

BMI body mass index; eGDR estimated glucose disposal rate; CVD Cardiovascular diseases; HD heart diseases; βIndir Beta coefficients of Indirect effect; βdir Beta coefficients of direct effect; βtotal Beta coefficients of Total effect; PM proportion mediated

To further assess the association between eGDR and the risk of CVD, HD, and stroke, several subgroup analyses and interactions were performed. The analyses were stratified by age, sex, BMI, smoking status, alcohol consumption, dyslipidemia, diabetes mellitus, and CKM stage. The results of the stratified analysis are shown in Table 3 and Table S3-S4. After adjusting for multiple factors, there was no significant interaction between the aforementioned subgroup characteristics and eGDR quantile for CVD, HD, and stroke (P for interaction > 0.05).