The association of albumin-corrected anion gap and acute kidney injury in heart failure patients: a competing risk model analysis

Abstract

Background: The combination of heart failure (HF) and acute kidney injury (AKI) increases the mortality of patients. It is critical to identify HF patients who may have a high risk for AKI. Albumin-corrected anion gap (ACAG) is a new indicator, but there are no studies on ACAG and the risk of AKI in HF patients.

Methods: Data for HF patients was obtained from the MIMIC-IV database. Receiver operating characteristic (ROC) analysis and decision curve analysis (DCA) were employed to evaluate the clinical value of ACAG in predicting AKI risk. Logistic regression analysis and restricted cubic spline (RCS) curve were conducted to explore the relationship between ACAG and AKI. A competing risk model was developed to further investigate the relationship between ACAG on AKI.

Results: The study analyzed 5,972 HF patients, with 49.82% (2886/5972) suffering from AKI. The prediction performance of ACAG on AKI was good (AUC:0.656). Continuous ACAG was associated with AKI after adjusting for various significant variables (Model 1: OR = 1.094, 95%CI: 1.078-1.110; Model 2: OR = 1.150, 95%CI: 1.133-1.166; Model 3: OR = 1.035, 95%CI. 1.017-1.054). All High ACAG groups showed a higher risk of AKI (all P < 0.001). ACAG was also linked to in-hospital mortality (P < 0.001). The competing risks model revealed that high ACAG was still a risk factor for AKI when in-hospital mortality served as a competing risk event (P < 0.001).

Conclusion: High ACAG was associated with the risk of AKI in HF patients. Clinicians can risk-stratify HF patients by combining ACAG levels.

Keywords: Acute kidney injury; Albumin-corrected anion gap; Competing risk; Heart failure; Prognosis.

Fig. 1

Flow chart of patient selection

Fig. 2

LASSO regression analysis results for 34 variables. (A) LASSO coefficient of variables. (B) Cross-validation curve. The 1-SE Criteria (indicated by the right dotted line) was employed to select the optimum λ value in the Lasso regression. The standard error (SE) in this study was 0.01

Fig. 3

AKI prediction performance of single ACAG and ACAG integrated SOFA or APSIII. (A) ROC comparison among single ACAG and ACAG integrated SOFA or APSIII. (B) The DCA of single ACAG and ACAG integrated SOFA or APSIII. Abbreviation: Acute kidney injury, AKI; Albumin-corrected anion gap, ACAG; Sequential Organ Failure Assessment, SOFA; Acute Physiology Score III, APSIII; Receiver operating characteristic, ROC; Decision curve analysis, DCA

Fig. 4

The RCS results between ACAG and AKI in different adjusted models. (A) In model 1: adjusted for the variables of severity of disease: SOFA, APSIII, and AKI stage. (B) In model 2: adjusted for the variables of demographic, comorbidities, and medication use: gender, alcohol use, MI, malignant cancer, diabetes, hypertension, CCBs, and diuretics. (C) In Model 3: adjusted for laboratory indicators: BUN, creatinine, PCO₂, pH, PO₂, TBil, WBC, and lactate. Abbreviation: Restricted cubic spline, RCS; Odds ratio, OR; Albumin-corrected anion gap, ACAG; Acute kidney injury, AKI; Sequential Organ Failure Assessment, SOFA; Acute Physiology Score III, APSIII; Myocardial infarct, MI; Calcium channel blockers, CCBs; Blood urea nitrogen, BUN; Partial pressure of carbon dioxide, PCO₂; Pondus Hydrogenii, pH; Partial pressure of oxygen, PO₂; White blood cell, WBC; Total bilirubin, TBil

Fig. 5

The relationship between ACAG and in-hospital mortality. (A) In the whole HF population. (B) In AKI patients. (C) In non-AKI patients. Abbreviation: Albumin-corrected anion gap, ACAG; Heart failure, HF; Acute kidney injury, AKI

Fig. 6

The competing risk model by Nelson-Aalen cumulative risk curve. (A) The competing risk model about AKI and in-hospital mortality without considering ACAG. (B) The competing risk model about AKI and in-hospital mortality in low and high ACAG. Abbreviation: Albumin-corrected anion gap, ACAG; Acute kidney injury, AKI