Introduction

Contrast-induced acute kidney injury (CI-AKI) refers to acute renal impairment resulting from the administration of contrast agents during angiography or other medical procedures1. It is now recognized as the third most common cause of hospital-acquired acute kidney injury, significantly increasing the risks of short- and long-term morbidity and mortality2,3. Since no definitive therapeutic strategies are currently available for CI-AKI, its pathogenesis is potentially modifiable, early identification of high-risk individuals and implementation of preventive measures are critically important4. The exact mechanisms underlying CI-AKI remain unclear, however, existing research highlights a strong association with inflammatory activation5,6,7.

Remnant cholesterol (RC) reflects the remnants of lipoproteins at various stages of triglyceride breakdown and conversion, representing the total cholesterol content within all triglyceride-enriched lipoproteins (TRLs)8. As an indicator of lipid metabolism abnormality, RC has been shown to induce the release of inflammatory cytokines and activate inflammatory signaling pathways, which are associated with cardiovascular diseases and renal dysfunction9,10,11.

Previous research has examined the relationship between traditional lipid markers, both total cholesterol (TC) and triglycerides (TG), and CI-AKI12,13,14. The association between RC and CI-AKI following percutaneous coronary intervention (PCI) remains unclear. This study therefore aims to investigate this association in patients with ST-segment elevation myocardial infarction (STEMI) undergoing PCI.

Methods

Subjects and research design

This retrospective analysis focuses on patients diagnosed with STEMI at admission who had complete medical records and underwent PCI treatment between January 2020 and December 2023. Their baseline and clinical characteristics were obtained from the electronic medical record system of the Affiliated Hospital of Xuzhou Medical University. The study cohort was subsequently divided into CI-AKI and non-CI-AKI groups based on postoperative outcomes.

Exclusion Criteria: (1) Patients with incomplete clinical information; (2) Heart failure with New York Heart Association (NYHA) functional status of Class III or higher; (3) Severe hepatic dysfunction; (4) Acute infections; (5) History of malignancies; (6) Use of other radiographic contrast agents or nephrotoxic substances—including but not limited to aminoglycoside antibiotics (e.g., gentamicin, amikacin, tobramycin), immunosuppressants with elevated or newly initiated trough concentrations (e.g., cyclosporine, tacrolimus), and nephrotoxic traditional/herbal medicines such as preparations containing aristolochic acid (Fig. 1).

Fig. 1
figure 1

Study flowchart. Note STEMI, ST-segment elevation myocardial infarction; CI-AKI, Contrast-induced acute kidney injury.

Full size image

Data acquisition and PCI protocol

Sample Testing: All blood specimens and laboratory indicators were tested and assessed in the central and clinical laboratories of Xuzhou Medical University Affiliated Hospital. Left ventricular ejection fraction (LVEF) was assessed via echocardiography performed within 2 days of admission.

PCI Procedure: PCI procedures were carried out by skilled interventional cardiologists. Standard catheters, guidewires, balloons, and stents were used, with a radial artery approach in adherence to clinical guidelines. A non-ionic, low-osmolar contrast agent (iodixanol) was selected, and intravenous infusion of 0.9% saline at 1 mL/kg/h was administered for 12 h post-contrast use.

Definition of the indicators

(1) The 2019 European Society of Cardiology and European Atherosclerosis Society (ESC/EAS) guidelines recommend RC was calculated using the formula RC = TC minus LDL-C and HDL-C15. (2) CI-AKI was defined as a ≥ 26.5 μmol/L increase in serum creatinine within 48 h after PCI, or a serum creatinine level ≥ 1.5 times baseline, or a urine output < 0.5 ml/kg/h for 6 consecutive hours16.

Statistical analysis

Continuous data are presented as mean ± standard deviation or median with interquartile range, based on their normal or non-normal distribution, and were compared using the independent samples *t*-test or Mann–Whitney U test, respectively. Categorical variables are expressed as counts and percentages and were compared using the chi-square test. Collinearity was first screened with Pearson correlation, retaining the clinically more relevant variable among correlated pairs, and multicollinearity was subsequently assessed by calculating the variance inflation factor (VIF) after model construction. Independent risk factors for CI-AKI were identified using Boruta analysis followed by logistic regression. The dose–response relationship between RC and the probability of CI-AKI was analyzed using restricted cubic splines (RCS).The predictive performance of RC was evaluated using the receiver operating characteristic (ROC) curve. Furthermore, the incremental value of adding RC to an established CI-AKI risk model was quantified by calculating the integrated discrimination improvement (IDI) and net reclassification improvement (NRI).All analyses were performed using R software (version 4.3.1). Specific packages included: Boruta (v8.0.0) for feature selection, glm for logistic regression, rms (v6.7-0) for RCS modeling, pROC (v1.18.4) for ROC analysis, and PredictABEL (v1.2-5) for IDI/NRI calculations. A two-sided P-value < 0.05 was considered statistically significant.

Results

Baseline characteristics

The incidence of CI-AKI after PCI was 15.1% (195/1288). A comparison of baseline characteristics between the CI-AKI and non-CI-AKI groups is presented in Table 1. Patients who developed CI-AKI were significantly older, more likely to be female, smokers, and to have type 2 diabetes mellitus (T2DM). They also presented with a significantly higher resting heart rate, fasting plasma glucose, neutrophil count, and a greater prevalence of chronic kidney disease and higher Killip classification. Additionally, levels of D-dimer, red blood cell distribution width, glycosylated hemoglobin, total cholesterol, triglycerides, hs-CRP, hs-TnT, NT-proBNP, lactate dehydrogenase, and RC were higher in the CI-AKI group. However, the CI-AKI group had a lower estimated glomerular filtration rate (eGFR) and left ventricular ejection fraction (LVEF). These differences were statistically significant (P < 0.05).

Table 1 Patient characteristics.
Full size table

Risk factors for CI-AKI

The dependent variable was defined as whether CI-AKI occurred, with the 19 indicators serving as the independent variables. Feature selection was performed using the Boruta algorithm, a wrapper method built on a random forest ensemble.

It first augments the original feature set with “shadow features” created by permuting each real variable, trains a random forest on the combined set, and then compares the importance score of every real feature with the highest importance achieved by any shadow feature. Features whose importance consistently exceeds this “shadow maximum” are tagged “important”, those that repeatedly fall below are tagged “irrelevant”, and the remainder are marked “tentative”. The procedure iterates until all features are classified or a stopping limit is reached, producing an objective, threshold-free subset of variables that carry unique predictive information for the target. In the Boruta algorithm report, variables such as RC in the green zone are highlighted as significant contributors to the model. Those in the yellow zone are tentatively linked to potential adverse outcomes, whereas variables in the red zone are considered unimportant (Fig. 2). The Boruta algorithm identified total cholesterol (TC), blood urea nitrogen (BUN), lymphocytes (LYM), NT-proBNP, type 2 diabetes mellitus (T2DM), glycated hemoglobin (HbA1c), fasting plasma glucose (FPG), eGFR, LVEF, and RC as important features for CI-AKI in STEMI patients undergoing PCI.

Fig. 2
figure 2

Importance of potential risk factors of contrast-induced acute kidney injury ranked by Boruta algorithm. The horizontal axis is the name of each variable, and the vertical axis is the Z value of each variable. The box plot shows the Z value of each variable during model calculation. The green boxes represent important variables, the red boxes represent unimportant variables, and the yellow boxes represent potentially important variables. Note T2DM, Type 2 diabetes mellitus; NTproBNP, N-terminal pro-B-type natriuretic peptide; HR, Heart rate; RDW, Red cell distribution width; CKD, Chronic kidney disease; LDH, Lactate dehydrogenase; NEUT, Neutrophil; BUN, Blood urea nitrogen; LYM, Lymphocyte; FPG, Fasting plasma glucose; eGFR, Estimated glomerular filtration rate; TC, Total cholesterol; LVEF, Left ventricular ejection fraction; RC, Remnant cholesterol; HbA1c, Glycated hemoglobin A1c.

Full size image

Logistic regression analysis using the above factors revealed that BUN, FPG, LVEF, and RC are significant independent predictors of CI-AKI (Table 2). The variance inflation factor for all covariates in the final model ranged from 1.03 to 1.93, indicating no substantial multicollinearity.

Table 2 Association of patient characteristics with CI-AKI: univariate and multivariate regression analysis.
Full size table

Relationship between RC levels and CI-AKI

RCS analysis demonstrated a significant nonlinear association between RC and the risk of CI-AKI (P for overall < 0.001; P for nonlinear < 0.001) (Fig. 3). Using RC = 0.637 mmol/L as the reference point, the risk of CI-AKI was lower when RC levels were below the reference, while a gradual increase in CI-AKI risk was observed with higher RC levels. Notably, the odds ratio (OR) rose markedly when RC exceeded approximately 1.0 mmol/L, confirming that elevated RC is an independent risk factor for CI-AKI..

Fig. 3
figure 3

Dose–response relationship between remnant cholesterol and contrast-induced acute kidney injury from RCS analysis; and the shaded area around the solid line represents the 95% confidence interval of the curve.

Full size image

The receiver operating characteristic analysis

ROC analysis demonstrated that the AUC of BUN was 0.656, LVEF was 0.689, FPG was 0.646, and RC was 0.717 (P < 0.05). RC had a cutoff value of 0.47 ng/mL, which yielded a sensitivity of 77.4% and a specificity of 60.7%. (Table 3, Fig. 4).

Table 3 ROC curve for CI-AKI.
Full size table
Fig. 4
figure 4

Receiver operating characteristic analysis (ROC) for contrast induced acute kidney injury (CI-AKI). (A) ROC of remnant cholesterol for CI-AKI; (B) ROC of left ventricular ejection fraction for CI-AKI; (C) ROC of fasting plasma glucose for CI-AKI; (D) ROC of blood urea nitrogen for CI-AKI.

Full size image

Construct models

Informed by the outcomes of multivariable analysis, a traditional model including BUN, LVEF, and FPG was constructed. ROC analysis indicated that this traditional model had significant value for predicting CI-AKI, with an AUC of 0.739 (95% CI 0.703–0.775, P < 0.001).The incorporation of RC into this model significantly increased the predictive performance, yielding an AUC of 0.803, (95% CI 0.772–0.834, P < 0.001) (Table 4, Fig. 5).

Table 4 ROC curve of RC for CI-AKI.
Full size table
Fig. 5
figure 5

Receiver operating characteristic analysis of combined parameters for contrast-induced acute kidney injury. Note LVEF, Left ventricular ejection fraction; BUN, Blood urea nitrogen; FPG, Fasting plasma glucose; RC, Remnant cholesterol.

Full size image

Net reclassification index and integrated discrimination index

The incremental predictive value of adding RC to the baseline model was further quantified using the net reclassification index (NRI) and integrated discrimination index (IDI). These findings showed NRI was 0.618 (95% CI 0.470–0.766) and IDI was 0.053 (95% CI 0.036–0.076), both achieving statistical significance (P < 0.001). These findings confirm that the inclusion of RC significantly enhances the model’s ability to stratify risk for CI-AKI (Table 5).

Table 5 Discrimination accuracy and reclassification of RC for CI-AKI.
Full size table

Discussion

This study demonstrates that RC is an independent risk factor for CI-AKI in STEMI patients undergoing PCI, exhibiting a significant nonlinear dose–response relationship. Furthermore, incorporating RC into a clinical prediction model significantly enhanced its discriminatory power for CI-AKI. These findings align with contemporary research establishing elevated RC as a significant residual risk factor for adverse outcomes in STEMI patients17,18,19. For instance, the ACCORD trial prospectively demonstrated that each standard deviation increase in RC was associated with an approximately 7% higher risk of cardiovascular events20. The pathophysiological link between lipids and renal injury has long been hypothesized, notably by Moorhead et al.’s “lipid nephrotoxicity hypothesis”, which implicated aberrant lipid metabolism in glomerular and tubulointerstitial damage21. Our results provide compelling clinical evidence supporting this mechanistic link in the context of CI-AKI.

Inflammation is a well-established critical driver in the pathogenesis of CI-AKI, and specific inflammatory markers can predict its occurrence in ACS patients following PCI10,11,22. A Mendelian randomization study confirmed the causal association between RC and inflammation, suggesting that RC may contribute to CI-AKI through inflammation-mediated pathways9. The non-linear relationship observed in our RCS analysis also indicates that RC is a modifiable target. Recent studies suggest that RC can infiltrate arterial walls, causing localized microvascular lesions: (1) Inflammatory Activation and Atherogenesis: RC lipoproteins infiltrate arterial walls and are taken up by macrophages, forming foam cells. This process activates PI3K/Akt and MAPK signaling, inducing pro-inflammatory cytokines (e.g., IL-6, TNF-α) and exacerbating vascular inflammation and oxidative stress. RC further amplifies this cascade by modulating oxidized LDL and NF-κB-dependent receptors, promoting atherosclerosis and renal ischemia23,24,25.(2) Pro-thrombotic Effects: In the hypercoagulable state of STEMI, RC upregulates PAI-1 to enhance platelet aggregation and amplifies the coagulation cascade, thereby reducing renal perfusion and increasing hypoxia26,27,28.(3) Impaired Renal Repair: The pre-existing renal vulnerability from elevated RC is compounded by PCI, as RC impairs the tubular stress response and intrinsic self-repair mechanisms, exacerbating contrast-induced damage29.

Numerous studies have shown that FPG, LVEF, and BUN are independent factors for CI-AKI30,31,32. This research, using a combination of the Boruta algorithm and logistic regression analysis, identified these same factors as independent predictors of CI-AKI in STEMI patients following PCI. This study developed a traditional model that included FPG, LVEF, and BUN. The ROC analysis indicated that this traditional model achieved an AUC of 0.739 for CI-AKI. After integrating RC, the AUC significantly increased to 0.803, enhancing the model’s ability to identify CI-AKI. Further analysis using IDI and NRI confirmed that incorporating RC significantly improved the CI-AKI risk prediction model. As an easily accessible and calculable clinical biomarker, RC facilitates improved risk stratification for CI-AKI, enabling the timely identification of high-risk STEMI patients for targeted preventive measures.

Study strengths and limitations

This retrospective study possesses several methodological strengths. First, it explores the nonlinear associations between RC and CI-AKI in the STEMI-PCI population, utilizing RCS analyses with multiple knot configurations. Second, the integration of machine learning–driven feature selection, specifically using the Boruta algorithm, with conventional logistic regression helps to enhance the objectivity of variable selection. This approach may offer a modest methodological improvement compared to previous predictive models. Third, incorporating RC significantly improves risk stratification, as evidenced by IDI = 0.053 and NRI = 0.618, thereby highlighting its immediate clinical applicability. This is particularly noteworthy, in light of RC’s cost-effectiveness and its derivability from routine lipid panels.

Several limitations warrant consideration. First, the single-center, single-ethnicity cohort may limit the generalizability of the findings to broader populations. Second, despite adjusting for established confounders, residual confounding from unmeasured variables, such as dietary patterns, may still persist. Third, mechanistic insights into RC’s potential nephrotoxic effects remain speculative until experimental validation is obtained. Finally, external validation across diverse healthcare settings is necessary before clinical implementation.

Conclusion

RC is an independent predictor of CI-AKI in patients with STEMI undergoing PCI. Incorporating RC significantly enhances the predictive model for CI-AKI.