This observational, single-center respective study, was approved by the Fuwai Hospital Ethics Committee (2021-1496). Due to the retrospective design of this study, signed informed consent was waived. All patient data were anonymized.

Patients 18 years and older undergoing a single VA ECMO run for PCS between January 2009 and December 2020 were included. The exclusion criteria were: (1) heart transplantation patients; (2) patients with preoperative ECMO or undergoing more than one ECMO run; and (3) patients with chronic hemodialysis.

The decision for implantation or weaning of VA ECMO was decided by the ECMO management team, which consisted of a cardiac surgeon, intensive care specialist, and a perfusionist. Nearly 90% of peripheral cannulations were performed percutaneously (through the femoral vein and artery), which allows for chest closure and can reduce mediastinal bleeding. Central aortic cannulation was performed through the sternum with the cannulas already in place for cardiopulmonary bypass (CPB) for the patients who developed an aortic dissection or pulmonary dysfunction. The anticoagulation management of VA ECMO is summarized in the Supplementary Files. The goal after VA ECMO support was to reduce intravenous inotropes to allow optimal myocardial recovery. Cardiopulmonary recovery was assessed daily by clinical, echocardiographic, and hemodynamic measurements to define the optimal weaning time.

The primary endpoint was the incidence of AKI $\ge$3 in patients with PCS supported by VA ECMO and its impact on the postoperative outcomes. Patients with and without AKI $\ge$3 were divided into two groups. The outcomes evaluated included in-hospital death, septicemia, liver dysfunction, pneumonia, cerebrovascular accidents, atrial fibrillation, and malignant arrhythmias. In addition, we sought to identify the risk factors for developing AKI $\ge$3 using patient demographic data, the Charlson Comorbidity index, laboratory tests prior to ECMO, the type of surgery, intraoperative factors, and the time of ECMO implantation.

AKI was defined according to the KDIGO guideline. The definition of AKI $\ge$3 was the stage 3 AKI evaluated by the KDIGO guideline within 7 days of VA-ECMO implantation, in which the increase in serum creatinine was more than 3.0 times baseline or increased to more than 4.0 mg/dL, or the patient required continuous renal replacement therapy (CRRT). The indications for initiation of CRRT included volume overload, progressive metabolic acidosis (pH 7.1~7.2 or serum bicarbonate level 12~15 mmol/L), and severe electrolyte abnormalities (potassium level $>$6.5 mmol/L, hyponatremia, hypernatremia, and hyperphosphatemia). Liver dysfunction was defined as the increase of total bilirubin in serum greater than 171 $\mu$mol/L, or an international normalized ratio greater than 1.5, or prothrombin coagulative activity lower than 40%. A cerebrovascular accident was diagnosed as cerebral hemorrhage or cerebral embolism based on cranial CT scans. Malignant arrhythmias were defined as ventricular arrhythmias that caused hemodynamic disturbances in a short period of time and required defibrillation or medication to restore sinus rhythm.

The data are expressed as mean $\pm$ standard deviation (SD) in continuous variables with Gaussian distributions and n (%) in categorical variables. For the skewed variables, the median (interquartile range, IQR) was used. The student’s t test was used for investigating the impact of AKI on ICU stay, hospital stay, and mechanical ventilation time. Potential risk factors for developing AKI $\ge$3 were evaluated by the univariate logistic regression analysis. Candidate risk factors (p 0.1) in the univariate analysis were entered into the multivariable logistic regression model. The tolerance and variance inflation factors (VIF) were calculated to evaluate the collinearity of the potential variables. When the tolerance 0.1 or the VIF $>$10, there is collinearity among the variables. A p-value of 0.05 was set for statistical significance. All of the analyses were performed using the SPSS statistics software (version 26.0, IBM, Corp., Armonk, NY, USA).

A total of 136 patients with a mean age of 53.6 $\pm$ 13.9 years (66.9% male) were identified for the study after excluding 62 patients undergoing a heart transplantation. Demographic characteristics of the study population is summarized in Table 1. The median time of ECMO support was 114.3 (44.1, 150.5) hours. VA ECMO was initiated in the operation room in 95 patients and in the intensive care unit (ICU) in 41 patients. Blood products were transfused in 81 patients intraoperatively and 125 patients postoperatively. 16 patients received central cannulation; the other 120 patients received peripheral cannulation. 80 patients (58.8%) developed AKI $\ge$3. CRRT was used in 93 patients, including all the AKI $\ge$3 patients and 13 patients of AKI 3. 36 patients recovered after the treatment of CRRT, 34 in the AKI $\ge$3 group.

Complications that VA ECMO patients experienced included infections, and involved the circulatory, digestive and other systems. 18.4% (25/136) of patients suffered septicemia. 25.7% (35/136) patients developed pneumonia. 52.2% (71/136) patients experienced malignant arrhythmias and 57 patients had atrial fibrillation. Laboratory tests showed that 30 patients had liver dysfunction, and 8.1% (11/136) of patients suffered a cerebrovascular accident. The in-hospital mortality was 59.6% (81/136) (Table 1). Patients were grouped according to the occurrence of AKI $\ge$3. 80 (58.8%) patients developed AKI $\ge$3 according to the KDIGO stage 3. Binary logistic regression results showed that there was no significant difference in the in-hospital mortality between the two groups (62.5% vs. 57.5%, p = 0.559). Patients with AKI $\ge$3 required significantly longer mechanical ventilation (200.9 [128.0, 534.8] hours vs. 78.9 [13.0, 233.0] hours, p 0.001). The ICU stay and hospital stay of patients with AKI $\ge$3 were significantly longer than that of patients with AKI 3 (384 [182, 648] hours vs. 216 [48, 456] hours, p = 0.001; 25.0 [15.3, 46.6] days vs. 13.4 [7.4, 38.4] days, p = 0.022, respectively).

To investigate the potential risk factors for AKI $\ge$3, demographic characteristics, the Charlson Comorbidity index, preoperative laboratory tests, intraoperative factors and variables at the time of VA ECMO implantation were entered into the univariate analysis. There was no difference among all the demographic and preoperative factors between the two groups (Table 2). Before ECMO implantation, the mean arterial blood pressure of the AKI $\ge$3 group was lower than that of the AKI 3 group (58 $\pm$ 13 vs. 65 $\pm$ 14 mmHg, p = 0.020) (Table 3). The time of ECMO implantation (intraoperatively or postoperatively) was significantly different between the two groups (82.1% vs. 61.3%, p = 0.011). Previous studies have reported that age, lactate and ejection fraction (EF) may be risk factors for the poor outcome of patients receiving VA ECMO after PCS [6, 8]. Age, cross-clamp time, CPB time, the timing of ECMO implantation, mean artery pressure (MAP), lactate concentration before ECMO, and preoperative EF were entered into multivariable logistic regression analyses. The results showed that the timing of ECMO implantation was an independent risk factor for the development of AKI $\ge$3 (p = 0.036) (Table 4). Intraoperative implantation of ECMO may decrease the risk of AKI $\ge$3 (OR = 0.298, 95% CI = 0.096–0.925). The tolerance and variance inflation factors showed that there was no collinearity among these variables.