A cohort of 1400 patients from the Third People’s Hospital of Chengdu, Sichuan, China, who underwent PCI between July 2018 and December 2020, were included in the study. Plasma creatinine and cystatin C measurements were recorded for each participant. The study excluded patients with: (1) prior coronary artery bypass grafting (CABG) surgery; (2) structural heart disease necessitating surgical or percutaneous treatment; (3) severe hepatic/respiratory dysfunction; (4) advanced malignancies with limited life expectancy; (5) in-hospital mortality; and (6) more than 10% of critical medical data missing.

Demographic characteristics, medical background, and key clinical parameters were extracted from electronic health records. Fasting venous blood samples were analyzed for lipid profiles, fasting blood glucose, homocysteine, renal biomarkers (serum creatinine, cystatin C), brain natriuretic peptide (BNP), and cardiac troponin T (cTnT) using standardized assays. The bSS was computed using an online calculator available at https://syntaxscore.org/. Preprocedural angiograms were independently assessed by two blinded evaluators, with discrepancies resolved by a third arbitrator. All data were meticulously entered into a specialized computer database, which was then rigorously checked for quality to ensure precision and reliability. eGFR values were calculated using the 2021 CKD-EPI equation, which integrates both creatinine and cystatin C measurements. This method yields a more accurate estimation of the glomerular filtration rate (GFR) than equations relying solely on creatinine [25].

Follow-up assessments were conducted at 1, 3, 6, and 12 months post-discharge, followed by annual evaluations via telephone interviews or clinic visits. Trained personnel prospectively documented clinical events throughout the follow-up period. The primary endpoint was major adverse cardiovascular events (MACEs), defined as a composite of all-cause mortality, nonfatal myocardial infarction (MI), or ischemia-driven unplanned revascularization. Secondary endpoints included all-cause mortality, cardiac death, MI, unplanned revascularization and nonfatal stroke. All endpoints underwent rigorous adjudication by an independent committee based on standardized criteria, including review of medical records and imaging data.

The sample size was calculated based on the expected incidence of MACEs over two years (16%), with a target power of 0.8, a two-sided significance level ($\alpha$ = 0.05), and an assumed hazard ratio (HR) of 1.2 per 10-unit decrease in eGFR. The standard deviation (SD) of eGFR was assumed to be 15 units, and a 5% loss to follow-up was accounted for. Using a Cox proportional hazards regression model for continuous variables, the required sample size was calculated to be approximately 700 participants.

Continuous variables were presented as mean $\pm$SD or median (interquartile range [IQR]) based on distributional normality. Group comparisons were performed using appropriate statistical methods based on the data distribution. Categorical variables were reported as counts (percentages) and analyzed using $\chi$² tests or Fisher’s exact tests as appropriate. Missing data were addressed using multiple imputation with fully conditional specification to preserve statistical validity and mitigate potential estimation bias. This calculation ensures adequate statistical power to detect the specified effect size of eGFR on MACE risk. To better illustrate the association between kidney function decline and adverse cardiovascular outcomes, we transformed eGFR values by taking the negative and dividing by 10 for inclusion as continuous variables in analyses.

The correlation between the bSS and eGFR was explored using Spearman’s correlation analysis. Logistic regression analysis was adopted to analyze the association between risk factors including eGFR and the angiographic severity of coronary artery disease (CAD) (bSS $\ge$22). The dose-response association between eGFR, bSS, and adverse outcomes in ACS patients were analyzed using restricted cubic spline (RCS). Patients were divided into two groups based on predefined thresholds for eGFR (60 vs. $\ge$60 mL/min/1.73 m²) and bSS (22 vs. $\ge$22), and subsequently into four groups combining eGFR and bSS to observe differences in outcome events. Cumulative event rates were estimated using Kaplan-Meier curves, with between-group differences assessed via log-rank tests. Time-to-event Cox regression models were used to evaluate associations between eGFR, bSS, and cardiovascular outcomes. In order to assess the consistency of the predictive utility of eGFR and bSS across diverse patient subgroups, we conducted predefined analyses stratified by age ($\le$65 vs. $>$65 years), sex (male vs. female), body mass index (BMI) (24 vs. $\ge$24 kg/m²), hypertension, diabetes mellitus, acute myocardial infarction (AMI) status, and smoking history (all categorized as yes/no). Interaction terms were incorporated into multivariable Cox regression to evaluate potential effect modifications between these variables and primary outcomes, thereby strengthening the interpretability of the findings.

Mediation analysis was performed using the ‘mediation’ package in the R Programming Language 4.4.1 (R Foundation for Statistical Computing, Vienna, Austria). The analysis evaluated the mediating role of coronary artery lesion complexity (bSS as a continuous variable) in the association between renal dysfunction (eGFR as a continuous variable) and MACEs (time-to-event endpoint), adjusted for multiple covariates in separate multivariable Cox regression models. Bayesian methods were applied for the estimation of mediation effects. A significant mediating effect was established when meeting all three criteria: (a) statistically significant indirect effect (95% credible interval excluding null); (b) significant total effect (p 0.05 in Cox regression); (c) positive proportion-mediated estimate ($>$0% of total effect). We systematically evaluated the robustness of the mediation effects through adjustments for multiple covariates in separate multivariable Cox regression models. Specifically, we replaced eGFR values with those derived solely from creatinine to assess robustness, as well as analyzed the mediation effect with all-cause mortality as the outcome event.

Multiple confounding factors were adjusted in separate multivariable Cox regression models, with variable selection guided by either statistical significance (p 0.05) in preliminary univariate analyses or established clinical relevance to either prognosis or lesion complexity. Model I was adjusted for age, sex, BMI, hypertension (HTN), diabetes mellitus (DM), smoking, previous PCI, total cholesterol (TC), triglycerides (TG), low density lipoprotein cholesterol (LDL-C), high density lipoprotein cholesterol (HDL-C); Model II was adjusted for age, BMI, DM, heart rate, diuretics, Isu, aspirin, AMI, left ventricular ejection fraction (LVEF); Model III was adjusted for Model I plus Model II. All statistical analyses in the present study were performed using the R Programming Language 4.4.1, and MedCalc 19.1 (MedCalc Software Ltd, Ostend, Belgium). Statistical significance was defined as two-tailed p 0.05.

The final analytic cohort meeting the eligibility criteria included 1400 patients (Supplementary Fig. 1), with a mean age of 67.1 $\pm$ 11.3 years. Baseline characteristics stratified by the incidence of MACEs are presented in Table 1. Participants experiencing MACEs were older and had higher levels of heart rate, cTnT, BNP, serum creatinine, cystatin C, Hcy, as well as higher bSS. They also exhibited lower LVEF, BMI, and eGFR calculated using a combined creatinine-cystatin C method. Compared to patients without MACEs, the MACEs group showed higher rates of AMI, chronic obstructive pulmonary disease (COPD) and DM, increased use of insulin and diuretics, but lower aspirin utilization at discharge. A greater proportion of these patients had eGFR values below 60 mL/min/1.73 m² [85 (37.1%) vs. 282 (24.1%)] or a bSS of 22 or greater [78 (34.1%) vs. 201 (17.2%)], compared to those who did not experience MACEs (all p-values were 0.05). Baseline characteristics of the four groups classified by eGFR (60 vs. $\ge$60 mL/min/1.73 m²) and bSS (22 vs. $\ge$22) are detailed in Supplementary Table 1.

Spearman’s correlation analysis identified a statistically significant negative correlation between eGFR and bSS (r = –0.20, p 0.001, Supplementary Fig. 2). Participants were categorized into four groups based on different eGFR levels: T1: eGFR 30 mL/min/1.73 m²; T2: 30 $\le$ eGFR 60 mL/min/1.73 m²; T3: 60 $\le$ eGFR 90 mL/min/1.73 m²; T4: eGFR $\ge$90 mL/min/1.73 m². Participants with lower eGFR levels (38.2% vs. 26.8%, 18.2% vs. 14.2%, p 0.010, Fig. 1) had a higher proportion of patients with a bSS $\ge$22.

Fig. 1.

The proportion of complex lesions (bSS $\mathrm{\ge }$22) in ACS patients based on eGFR groups. Note: T1: eGFR 30 mL/min/1.73 m²; T2: 30 $\le$ eGFR 60 mL/min/1.73 m²; T3: 60 $\le$ eGFR 90 mL/min/1.73 m²; T4: eGFR $\ge$90 mL/min/1.73 m². bSS, baseline SYNTAX score; eGFR, estimated glomerular filtration rate; ACS, acute coronary syndrome.

Multivariate logistic regression analysis revealed that each 10-unit decrease in eGFR was independently associated with an increased likelihood of complex lesions (bSS $\ge$22), with an odds ratio of 1.121 (95% CI: 1.050–1.197, p = 0.001, Supplementary Table 2). This association was adjusted for potential confounders, including age, sex, hypertension, BMI, DM, smoking, TC, TG, LDL-C, HDL-C, prior PCI, and homocysteine levels. RCS curve analysis demonstrated a dose-response relationship between eGFR and the risk of complex lesions in ACS patients, after adjusting for the aforementioned variables (p for nonlinearity = 0.925, Supplementary Fig. 3).

Over a median follow-up period of 31.03 (27.34, 35.06) months, 229 cases (16.4%) of MACEs were documented, including 99 cases (7.0%) of all-cause death, 41 cases (2.9%) of MI, and 123 cases (8.9%) of unplanned revascularization. The incidence of MACEs, all-cause death, and cardiac death increased with decreased eGFR and elevated bSS (Fig. 2 and Supplementary Table 3, Supplementary Figs. 4,5). Patients were further stratified into four groups based on eGFR and bSS. Kaplan-Meier curves indicated that individuals with both lower eGFR (60) and higher bSS ($\ge$22) exhibited the highest incidences of MACEs, all-cause death, and cardiac death (Supplementary Fig. 6).

Fig. 2.

Cumulative incidence of MACEs stratified by eGFR (A) and bSS (B). bSS, baseline SYNTAX score; eGFR, estimated glomerular filtration rate; MACEs, major adverse cardiovascular events.

Univariate Cox regression showed that eGFR, age, BMI, AMI, DM, heart rate, bSS, LVEF, diuretics, and insulin use were risk factors for MACEs, whereas aspirin use was associated with a reduced risk (Supplementary Table 4). After adjusting for multiple confounding factors, each 10-unit decrease in eGFR (Model I: HR 1.125, 95% CI: 1.056–1.197, p 0.001; Model II: HR 1.095, 95% CI: 1.029–1.166, p = 0.004; Model III: HR 1.093, 95% CI: 1.024–1.166, p = 0.007) and 1-unite increase in bSS (Model I: HR 1.037, 95% CI: 1.024–1.050, p 0.001; Model II: HR 1.035, 95% CI: 1.021–1.049, p 0.001; Model III: HR 1.034, 95% CI: 1.020–1.048, p 0.001) were associated with an increased risk of MACEs in ACS patients (Table 2). RCS curve analysis showed eGFR exhibited a negative dose-response relationship with MACEs (non-linear p = 0.087), while bSS demonstrated a positive relationship (non-linear p = 0.006) (Fig. 3). Similar trends were observed when all-cause death was considered as the outcome (Supplementary Table 5, Supplementary Fig. 7).

Fig. 3.

Dose-responsive relationship of the eGFR (A) and bSS (B) with MACEs. The RCS curves are derived from a Cox regression adjustment model, which includes factors such as age, sex, BMI, HTN, DM, smoking, previous PCI, TG, TC, LDL-C, HDL-C, HR, diuretics, Isu, aspirin, AMI, LVEF. RCS, restricted cubic spline; bSS, baseline SYNTAX score; eGFR, estimated glomerular filtration rate; MACEs, major adverse cardiovascular events; BMI, body mass index; HTN, hypertension; DM, diabetes mellitus; PCI, percutaneous coronary intervention; TG, triglyceride; TC, total cholesterol; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; HR, hazard ratio; AMI, acute myocardial infarction; LVEF, left ventricular ejection fraction.

Subgroup analyses were performed to determine if the predictive power of eGFR and bSS was consistent across different demographic groups and comorbidities (Fig. 4). After adjusting for various confounding variables, we found that both eGFR and bSS were significant indicators of MACEs across various subgroups. In the selected subgroup, no significant association was found with the risk of MACE (all interaction p-values were $>$0.05). Comparable trends were observed when all-cause death was considered as outcome (Supplementary Fig. 8).

Fig. 4.

Subgroup Analyses for MACEs. The hazard ratio (HR) indicates that each 10-unit decrease in eGFR and each 1-unit increase in bSS correspond to an increased risk. All models were adjusted for age, sex, BMI, HTN, DM, smoking, previous PCI, TG, TC, LDL-C, HDL-C, Heart rate, diuretics, Isu, aspirin, AMI, LVEF. bSS, baseline SYNTAX score; eGFR, estimated glomerular filtration rate; MACEs, major adverse cardiovascular events; BMI, body mass index; HTN, hypertension; DM, diabetes mellitus; PCI, percutaneous coronary intervention; TG, triglyceride; TC, total cholesterol; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; AMI, acute myocardial infarction; LVEF, left ventricular ejection fraction.

As illustrated in Fig. 5, the mediation analysis demonstrated that the complexity of coronary artery lesions (bSS) mediated the relationship between renal function (represented by eGFR) and the occurrence of MACEs across various adjusted models. Specifically, bSS accounted for mediation proportions of 16.49%, 14.76%, 12.87%, and 11.95% in unadjusted Model, adjusted Model I, adjusted Model II, and adjusted Model III analyses, respectively. Furthermore, we tested the robustness of our findings by substituting eGFR values derived solely from creatinine. The mediation proportions were 20.71%, 20.30%, 17.20%, and 15.66% across the respective analyses (Supplementary Fig. 9). Similar mediating effects were noted when endpoint was defined as all-cause death (Supplementary Fig. 10).

Fig. 5.

Causal mediation analysis quantified bSS-mediated pathways in eGFR-MACE associations across covariate-adjusted models. (A) represents the unadjusted Model; (B) represents the adjusted Model I; (C) represents the adjusted Model II; (D) represents the adjusted Model III. Model I was adjusted for age, sex, BMI, HTN, DM, smoking, previous PCI, TG, TC, LDL-C, HDL-C; Model II was adjusted for age, BMI, DM, Heart rate, diuretics, Isu, aspirin, AMI, LVEF; Model III was adjusted for Model I plus Model II. bSS, baseline SYNTAX score; eGFR, estimated glomerular filtration rate; MACE, major adverse cardiovascular event; BMI, body mass index; HTN, hypertension; DM, diabetes mellitus; PCI, percutaneous coronary intervention; TG, triglyceride; TC, total cholesterol; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; AMI, acute myocardial infarction; LVEF, left ventricular ejection fraction.

While this study provides valuable insights, it also presents certain limitations that warrant consideration. First, the single-center observational design limits causal inference between eGFR, bSS, and cardiovascular outcomes, and residual confounding may persist despite adjustment for known risk factors. Nevertheless, this limitation has been partially mitigated through subgroup analyses and sensitivity analyses. Secondly, although the 2021 CKD-EPI equation (combining creatinine and cystatin C) improved the accuracy for estimating GFR, reliance on a single in-hospital eGFR measurement may introduce bias due to fluctuations in short-term renal function. Future studies should incorporate repeated eGFR measurements during follow-up or use urinary biomarkers (e.g., albumin-to-creatinine ratio) to better capture trends in renal function [42]. Finally, since we exclusively used a Chinese cohort, validation across diverse ethnic populations is needed to confirm generalizability.