This study was performed in the ICU of a single center. Data from patients who received ECMO therapy were retrieved from the electronic medical records system and reviewed by an attending physician. The study was carried out in accordance with the guidelines of the Declaration of Helsinki and approved by the Ethics Committee of Bengbu Medical University (Protocol No. 2025 [039]).

The inclusion criteria were as follows: (1) received VA-ECMO support between July 2020 and January 1, 2025, and (2) received VA-ECMO support for $\ge$48 hours. Patients were excluded if one of the following conditions was met: (1) age 18 years old, (2) incomplete data, (3) had any infection before receiving ECMO support, (4) patients who died within the first 48 hours, or (5) received venovenous (V-V) ECMO or combined VA and V-V ECMO.

Critical care specialists and cardiologists were responsible for the timing of VA-ECMO initiation and weaning. These specialists performed the procedures for implantation, daily management, and weaning according to the recommendations of the Extracorporeal Life Support Organization (ELSO). The femoral vein and femoral artery are the sites of cannulation in our center. For patients undergoing VA-ECMO cannulation, prophylactic antibiotic therapy was administered as a single intravenous injection of a second-generation cephalosporin at the time of ECMO implantation. Subsequently, depending on serum lactate levels and organ function, the ECMO flow rate was adjusted to ensure adequate tissue perfusion. To maintain the activated clotting time within 180–220 seconds, a standardized heparin anticoagulation protocol was used.

The definition of NIs was derived from the criteria established by the U.S. Centers for Disease Control and Prevention in 1988 [13]. The occurrence of NIs was evaluated in instances where the initiation of VA-ECMO occurred more than 24 hours prior and within 48 hours subsequent to the discontinuation of VA-ECMO. Specifically, the following types of infections were considered: respiratory tract infection (RTI), blood stream infection (BSI), urinary tract infection (UTI), and surgical site infection (SSI). RTI was suspected if at least one of the following clinical manifestations was detected: a temperature of $\ge$38.0 °C or 36.0 °C, a leukocyte count $>$12 $\times$ 10⁹/L or 4 $\times$ 10⁹/L, purulent excretion from the lower respiratory tract, positive sputum culture, and new or persistent pulmonary infiltrates observed on chest imaging. BSI is characterized as the detection of bacterial or fungal pathogens in one or more blood cultures. UTI is characterized by the manifestation of clinical symptoms or indicators, alongside positive outcomes from routine urinalysis, coupled with a confirmed positive urine culture. SSI is identified as an infection affecting the dermal layer, subcutaneous tissue, or muscle that arises within 30 days post-surgery. This condition is marked by one or more of the following: purulent discharge from the surgical site drainage, affirmative culture results from the drainage fluid, or a diagnosis of infection made by the surgeon through clinical assessment.

During VA-ECMO application, adverse events included neurological complications such as cerebral haemorrhage, ischaemic stroke, and hypoxic ischaemic encephalopathy; cardiovascular events like malignant arrhythmia, cardiogenic shock, and septic shock; as well as gastrointestinal bleeding, limb ischaemic necrosis, multiple organ dysfunction syndrome, and disseminated intravascular coagulation.

The following information was recorded retrospectively for each patient: demographic data, including age, sex, height, weight, diagnosis, previous medical history, and personal history. We also collected scores such as the Acute Physiology and Chronic Health Evaluation II (APACHE II) and the Sequential Organ Failure Assessment (SOFA). Additionally, data on the duration of ECMO support, IABP support, CRRT support, mechanical ventilation, and the use of tracheotomy and bronchoscopy were documented. The maximal vasoactive-inotropic score (VIS) and laboratory test results within 48 hours of ECMO initiation—such as white blood cell and platelet counts, procalcitonin levels, liver and renal function, N-terminal pro-brain natriuretic peptide (NTproBNP), and troponin (TnI)—were also recorded. Microbiological data, including the type and site of isolated bacteria, presence of adverse events, CVC duration, ICU stay, and in-hospital mortality were noted.

The analyses were performed using SPSS version 26.0 (IBM, Armonk, NY, USA), and R software, version 4.4.2 (The R Foundation for Statistical Computing, Vienna, Austria). Continuous variables are presented as the median (interquartile range, IQR) or the mean (standard deviation, SD), whereas categorical variables are presented as numbers (percentages). For continuous variables, t-tests and Mann–Whitney U tests were used for normally distributed and skewed variables, respectively. The $\chi$² test or Fisher’s exact test was performed to compare binary outcomes. Odds ratios (OR) and 95% confidence intervals (CI) were calculated. A univariate logistic regression analysis was conducted to uncover the determinants linked to NIs and in-hospital mortality rates. Nonlinear associations were explored using restricted cubic splines (RCS). A two-piecewise linear regression model was then constructed to determine the inflection point. A two-sided p-value 0.05 was considered to indicate statistical significance.

In our cohort, gram-negative bacteria were the predominant microorganisms causing NIs in patients supported by VA-ECMO. This finding is consistent with previous studies [14, 15, 16]. Xia et al. [17] have identified Acinetobacter baumannii, Klebsiella, and Pseudomonas aeruginosa as prevalent pathogens responsible for NIs. Patients often require mechanical ventilators, CRRT, and intravenous catheters at the same time during ECMO support. Sites of endotracheal tubes and cannulas are often colonized by microorganisms. Moreover, ECMO patients with compromised immune systems and mucosal barrier damage are prone to infections [18]. However, the effectiveness of prophylactic antibiotic therapy in the prevention of NIs remains inconclusive [19, 20]. Thus, identifying associated factors for NIs and adopting appropriate treatment strategies are critical. In our study, associated factors for NIs included blood transfusion, long ECMO duration, CVC duration, and ICU stay. The need for blood transfusion was an indication that the patient was in a more severe condition. However, a blood transfusion is essential for patients experiencing haemorrhage and disseminated intravascular coagulation. Moreover, the association between blood transfusion and adverse outcomes has been attributed to immunosuppression in the recipient, which can lead to NIs. This finding is consistent with previous studies [21]. This finding has been confirmed by similar studies [22, 23]. A longer duration of ECMO not only increases the risk of prolonged exposure to invasive devices, but also leads to extended ICU and hospital stays. These factors are associated with an increased risk of NIs [24]. However, a prolonged ECMO duration is likely a strong marker of underlying critical illness severity and complexity rather than an independent causal factor for acquiring NIs. In contrast, the persistent association with ICU length of stay, even after the landmark adjustment, strengthens its role as a robust, independent risk factor, highlighting the importance of cumulative exposure to the hospital environment and invasive procedures. These identified associated factors for NIs are likely strong markers of underlying critical illness severity and prolonged exposure to invasive care, rather than independent causal agents of infection. For example, longer ECMO support may reflect more complex primary conditions (e.g., cardiogenic shock or respiratory failure) that inherently increase vulnerability to infections due to immune suppression or compromised barriers. Blood transfusion often signifies complications like hemorrhage or anemia, which are common in critically ill patients and may serve as indicators of disease burden rather than direct causes of infection. Similarly, extended CVC use and ICU stay highlight the intensity of medical interventions and overall patient acuity, which collectively elevate the risk of exposure to pathogens. Therefore, these factors should be interpreted as proxies for the cumulative impact of critical illness and invasive care, underscoring the need for integrated management strategies to mitigate infection risks in this vulnerable population. In contrast to previous studies, we estimated the inflection points of ECMO duration and ICU stay to predict the occurrence of NIs using RCS analysis. The relationship between ECMO duration and NIs incidence was a positive linear correlation. This indicates that prolonged ECMO increases the risk of NIs. The occurrence of NIs increased significantly when ICU stay exceeded approximately 24 days. These findings provide valuable insights for the prevention and management of VA-ECMO-associated NIs.

Previous studies have demonstrated that NIs patients have a lower ECMO weaning survival rate and a higher risk of in-hospital mortality compared with non-NIs patients [25, 26]. Wang J and colleagues reported that the incidence of NIs was associated with a 3.726-fold higher risk of in-hospital death in ECMO-supported patients [27]. In our study, we did not find a statistically significant association between NIs and in-hospital mortality in univariate analysis (p = 0.052). However, this lack of significance may be attributed to limited statistical power due to our small sample size, as well as potential survivor bias, wherein only patients with longer survival times could develop NIs. Additionally, survivor bias could have influenced our results, as patients who died early in the ICU stay (median 13.88 days) may not have been exposed to NIs risks, thereby diluting the observed effect. Instead, we identified that older age, a history of diabetes, and hepatic or renal dysfunction were significantly associated factors for in-hospital mortality. This finding underscores that a patient’s baseline pathophysiological status prior to ECMO initiation is critical for determining outcomes. Furthermore, a shorter ICU stay was also associated with mortality. This is not contradictory, as most deaths in our cohort occurred early during the ICU stay (within a median of 13.88 days), likely due to severe initial clinical deterioration. Therefore, we speculate that patients with a worse clinical status at baseline are at a higher risk of early death despite VA-ECMO support. Conversely, for those who survive this critical early phase, a prolonged ICU stay may subsequently increase the risk of developing NIs.

Our study has several limitations. First, its retrospective, single-center design and small sample size may limit the generalizability of our findings to the broader population of VA-ECMO patients who develop NIs in the ICU. The small sample size may have increased the risk of Type II errors, meaning we might have missed a true association between NIs and mortality. Furthermore, the higher mortality in the non-NIs group could be partly explained by severe initial clinical deterioration, which led to early death before NIs could develop. Second, due to the limited sample size and number of outcome events in our cohort, we were unable to construct stable multivariate models to adjust for potential confounders. Therefore, the identified associated factors are based on univariate analyses and should be interpreted with caution. Future studies with larger sample sizes are warranted to independently validate these associations. Third, the diagnosis of NIs was based on the Centers for Disease Control (CDC) 1988 criteria. While these criteria were rigorously applied, they are less specific than contemporary definitions. However, the consistent application of the same criteria across all patient groups ensures the internal validity of the comparative analyses presented herein. Future studies utilizing current standardized definitions are warranted to confirm our findings. Finally, this study did not analyze the duration of mechanical ventilation, an important potential risk factor for NIs, due to the unavailability of these data in our records. This may limit the comprehensiveness of our risk factor assessment for NIs. Despite these limitations, our findings offer valuable insights, particularly in demonstrating the use of RCS to explore nonlinear correlations between variables and outcomes. We believe that future large-scale, multicenter studies are warranted to validate our findings and provide more robust evidence on the characteristics and impact of NIs in VA-ECMO patients.