C-reactive protein-triglyceride glucose index in evaluating cardiovascular disease and all-cause mortality incidence among individuals across stages 0–3 of cardiovascular–kidney–metabolic syndrome: a nationwide prospective cohort study | Cardiovascular Diabetology

Baseline characteristics

Table 1 displays baseline CVD risk factors stratified by CTI index tertiles. Finally, the final analysis included 5723 participants for CVD and 5847 participants for all-cause mortality. Individuals in the highest CTI tertile exhibit marked differences versus the lowest on various indicators. People in the top CTIs were the elderly and had high levels of BMI, SBP, DBP, fast blood glucose, HbA1c, LDL-C, TG, TC, serum creatinine, TyG, CRP, CTI. In addition, the prevalence of comorbidities such as hypertension, dyslipidemia, and diabetes was notably higher in people with elevated CTI levels. Baseline characteristics were compared between participants with and without CVD and all-cause death in Tables S4 and S5, while the detailed baseline characteristics comparison.

Association of CTI index with CVD and total mortality in CKM syndrome patients

As shown in Table 2, multivariable-adjusted analysis revealed a graded cardiovascular risk profile associated with CTI. Among CKM stage 0–3 individuals, the highest CTIQ quartile (Q3) showed 52% increased CVD risk versus Q1 in the crude model (HR 1.52, 95% CI (1.44, 1.59), P < 0.0001), which remained significant after adjusting for demographic (age, gender, smoking, drinking status, marital status) and clinical confounders (BMI, eGFR, hypertension, diabetes, dyslipidemia, Hypertension treatment, Diabetes treatment, dyslipidemia treatment,), with 18% excess risk in Model 2 (HR 1.18, 95% CI 1.12, 1.24, P < 0.0001; P-trend < 0.001). Meanwhile, the highest quartile (Q3) showed adjusted hazard ratios of 1.35 for TyG (95% CI 1.26–1.44; P < 0.001) and 1.43 for standardized CRP values (CRP-SD) (95% CI 1.34–1.52; P < 0.001). Both were lower than the CTI’s highest quartile HR of 1.52 (95% CI (1.44, 1.59), P < 0.0001) (Table S7). For all-cause mortality, each unit increase in continuous CTI corresponded to 60% higher risk (Model 2: HR 1.60, 1.29–1.99, P < 0.0001). Stratified analysis demonstrated a 111% mortality increase in Q3 versus Q1 (Model 2: HR 2.11, 1.35–3.30, P = 0.001), whereas Q2 showed non-significant association (P = 0.52), indicating a threshold effect of CTI.

RCS and threshold effect analysis

We performed RCS analysis and threshold analysis to verify the association of CTI and CVD prevalence and mortality rates. For CVD incidence, standard Cox regression revealed non-linear association between CTI and CVD incidence (HR 1.075–1.224, P < 0.0001) (Fig. 2). Two-piecewise linear regression revealed a critical inflection point at CTI = 8.602: Below this threshold, each unit increase in CTI significantly elevated CVD risk (Model 2: HR 1.153, 95% CI 1.060–1.254, P < 0.001). Conversely, above the IP, CTI showed no significant association with CVD incidence in the fully adjusted Model 2 (HR 0.997, 95% CI 0.954–1.042, P = 0.89). The log-likelihood ratio test (P < 0.0001) confirmed the superiority in segmented model, highlighting a nonlinear dose–response pattern between CTI and CVD risk (Table S8). For all-cause mortality, standard Cox regression showed a notable linear trend between CTI and overall mortality (HR 1.486–1.674, P < 0.0001) (Fig. 2). While two-piecewise linear regression suggested a potential risk transition at CTI = 8.606 (HR 1.427–1.779, P ≤ 0.021 above the threshold; no significant association below the threshold, HR 1.245–1.443, P ≥ 0.345), the log-likelihood ratio tests (P ≥ 0.829 for all models) indicated no statistically significant improvement in model fit with segmentation. Thus, despite localized risk differences near the inflection point, the overall relationship predominantly aligns with a linear pattern (Table S9).

Kaplan–Meier (K–M) survival curves

The Kaplan–Meier (K–M) survival curves showed an elevated CVD incidence or overall mortality in the high CTI group. The log-rank test’s P values for the Q2 and Q3 groups, all below 0.05, confirm a higher risk compared to the Q1 group (Figure S1).

Subgroup analyses

To delve deeper into the connection between CTI and the likelihood of CVD or all-cause mortality, researchers conducted subgroup and interaction analyses on various variables, which include age, gender, tobacco use, alcohol consumption, marital status, education attainment, diabetes statuse, hypertension, dyslipidemia, glucose levels (including NGR, Pre-DM, and DM), and CKM syndrome (0–3 stages). For CVD incidence, sex stratification revealed markedly higher CVD risk in males (HR 1.627, 95% CI 1.451–1.824; P < 0.0001) compared to females (HR 1.359, 95% CI 1.223–1.510; P < 0.0001), with a pronounced interaction effect (P = 0.02). A dose–response relationship was evident, as higher CTIQ quartiles (Q3 vs. Q1) consistently correlated with elevated CVD risk across all strata (P for trend < 0.0001). Notably, smoking status (P = 0.029) and marital status (P < 0.0001) exhibited significant interactions, where current smokers (HR 1.722, Q2 vs. Q1) and married individuals (HR 1.809, Q3 vs. Q1) showed heightened vulnerability. Conversely, no interaction was observed for age, hypertension, or dyslipidemia (P > 0.05). Education level further modulated risk, with individuals above junior high school education displaying the steepest CTIQ-associated risk gradient (HR 2.935, Q3 vs. Q1; P < 0.0001) (Figure S4, Table S12). For overall mortality, lower education (e.g., “Junior high school and below”: HR 1.656, P < 0.001) and non-drinkers (HR 1.774, P < 0.0001) showed elevated mortality. CTI consistently predicted mortality across most subgroups (e.g., males: HR 1.538; females: HR 1.628, both P < 0.01), despite nonsignificant interactions for sex (P = 0.783) and age (P = 0.658). Notably, pre-diabetic individuals exhibited a non-significant trend toward heightened risk (HR 1.497, P = 0.254), warranting further investigation. The most pronounced interaction emerged in CKM strata (P < 0.001), which revealed CKM = 0 individuals have an exceptionally high risk (HR 8.225, 95% CI 2.558–27.964, P < 0.001). Therefore, the future study assessed the connection between CTIQ and overall death in the 0–3 stage of CKM group: CKM Stage 0 exhibited an extreme hazard ratio (Q3 vs. Q1: HR 19.611, 95% CI 2.251–171.300, P = 0.004), but with wide confidence intervals (Figure S5, Table S13).

AUC and ROC

For CVD risk, CRP and CTI retained comparable performance (AUC = 0.66 and 0.61, respectively), whereas TyG again performed at chance level (AUC = 0.5). Similarly, for all-cause mortality, CRP demonstrated moderate predictive utility (AUC = 0.66, 95% CI 0.61–0.66), just ahead of CTI (AUC = 0.61, 95% CI 0.56–0.61), while TyG showed no discriminative capacity (AUC = 0.5, 95% CI 0.45–0.5). The findings indicate that CTI could outperformed CRP and the TyG index in assessing overall mortality risk stratification (Figure S6).

Sensitivity analyses

To check the stability of our findings, we conducted multiple sensitivity analyses. Firstly, in order to tackle the issue of missing data and to limit the possibility of bias, we employed multiple imputations. Subsequent analysis revealed that the correlation among CTI and CVD incidence and overall mortality was in accordance with the basic results (Table S10). Secondly, the application of logistic regression models to investigate the connection between CTI and CVD incidence, as well as all-cause mortality, yields consistent results. (Table S11). Thirdly, the analysis of the piecewise Cox regression model confirmed result stability (Tables S12–S13). Fourthly, we also assessed the correlation among CTI and CVD incidence and overall mortality stratified by sex, age, and glucose level (grouped into NGR, Pre-DM, and DM). (Table S14). Furthermore, we explored additional analyses to analyze the associations of TyG and CRP standardized values (CRP-SD) with both CVD incidence and all-cause mortality, thereby validating the robustness of our primary findings (Table S7). Moreover, analyses stratified by sex and CKM stage (0–3) revealed consistent associations between CTI and adverse outcomes, as evidenced by Kaplan–Meier curves (all log-rank P < 0.05) (Figures S2 and S3). We excluded those individuals who died within 2 years of baseline to reduce potential bias from early terminal events (Table S6). Finally, RCS analyses (4 knots located at Harrell’s recommended percentiles) were also adopted to further assess the nonlinear associations of CTI and the risk of CVD and death (Table Figure S7).