Extracorporeal Membrane Oxygenation-Related Nosocomial Infection after Cardiac Surgery in Adult Patients

Introduction The evaluation of extracorporeal membrane oxygenation-related nosocomial infection (ECMO-related NI) in a homogeneous cohort remains scarce. This study analyzed ECMO-related NI in adult patients who have undergone cardiac surgery. Methods From January 2012 to December 2017, 322 adult patients who have received ECMO support after cardiac surgery were divided into the infection group (n=131) and the non-infection group (n=191). ECMO-related NI was evaluated according to demographic data, surgical procedures, and ECMO parameters. Results The incidence of ECMO-related NI was 85.4 cases per 1000 ECMO days. Acinetobacter baumannii was the most common pathogen causing blood stream infection and respiratory tract infection. Prolonged duration of surgery (P=0.042) and cardiopulmonary bypass assist (P=0.044) increased the risk of ECMO-related NI. Body mass index (odds ratio [OR]: 1.077; 95% confidence interval [CI]: 1.004-1.156; P=0.039) and duration of ECMO support (OR: 1.006; 95% CI: 1.003-1.009; P=0.0001) were the independent risk factors for ECMO-related NI. Duration of ECMO support > 144 hours (OR: 2.460; 95% CI: 1.155-7.238; P<0.0001) and ECMO-related NI (OR: 3.726; 95% CI: 1.274-10.895; P=0.016) increased significantly the risk of in-hospital death. Conclusion Prolonged duration of ECMO support was an independent risk factor for NI. Surgical correcting latent causes of cardiopulmonary failure and shortening duration of ECMO whenever possible would reduce susceptibility to NI.

in a range of 180-200 s. ECMO flow was initially set with a rate of 100 ml/kg/min and the flow level was tailored to achieve adequate tissue perfusion evaluated by lactate level and endorgan function. The management protocols of ECMO followed the Extracorporeal Life Support Organization (ELSO) guidelines [7] .

Definition of ECMO-Related NI
NI occurring 24h after initiation and 48h after discontinuation of ECMO was defined as ECMO-related NI [8] . The type of ECMOrelated NI included blood stream infection (BSI), respiratory tract infection (RTI), urinary tract infection (UTI), and sternal wound infection (SWI). The diagnosis and definition of ECMO-related NI were based on the Centers for Disease Control and Prevention definitions for NI [9] . Isolation of pathogenic microorganisms was related to clinical symptoms, typical inflammation characteristics in blood samples, and radiographic findings.

Anti-Infective Therapy
Patients received antibiotic prophylaxis with secondgeneration cephalosporins at initiation of ECMO support. If ECMO-related NI occurred, antibiotic was adjusted according to the culture results. The conventional dose of antibiotic was started and regulated according to renal function and drug concentration monitoring.

Statistical Analysis
Statistical analysis was performed using an extensively admissive software program, the SAS software (SAS Institute Inc 2013. SAS/ACCESS® 9.4 Interface to ADABAS: Reference. Cary, NC: SAS Institute Inc.). Continuous data and categorical data were expressed as mean ± standard deviation and percentage, respectively. Variance test was used to address nonpaired samples for comparing normally distributed parameters, and Wilcoxon rank sum test was used for comparing nonparametric variables. The qualitative variables were compared using chisquared test or Fisher's exact probability test. The multiple logistic regression analysis was conducted to identify the risk factors of ECMO-related NI, BSI, and in-hospital death. P<0.05 was considered to be statistically significant.

INTRODUCTION
Extracorporeal membrane oxygenation (ECMO) is one of the most important life support techniques for patients with severe but reversible cardiac and/or pulmonary failure. Despite advance in technique and management of ECMO, nosocomial infection (NI) is still one of severe complications of ECMO that remarkably increases morbidity and mortality [1] .
Comparing with other patients in intensive care unit (ICU), the patients undergoing ECMO are with increasing possibility of NI due to cannulation disrupting the skin barrier and allowing pathogens invasion and the hollow fiber structure of the oxygenator contributing to pathogens establishment [2,3] . Several studies have analyzed the risk factors and pathogens of ECMOrelated NI [4][5][6] , but different studies have obvious heterogeneity from baseline characteristics including populations (pediatric vs. adult patients), number of patients, concomitant diseases, causes of cardiac or pulmonary failure, ECMO cannulation access (percutaneous vs. intrathoracic), and ECMO support mode (venoarterial [VA] vs. venovenous). Such heterogeneity may affect the incidence, risk factors, and pathogens of ECMO-related NI.
In addition, the data on ECMO-related NI in adult patients after cardiac surgery remain scarce.
Therefore, we focused on ECMO and cardiac surgery parameters and investigated a single-center large sample homogenous group to analyze the incidence, risk factors, and pathogens of ECMO-related NI in adult patients who have undergone cardiac surgery and VA-ECMO support.

Patient Population
From January 2012 to December 2017, a total of 35,327 adult patients underwent cardiac surgery at the Beijing Anzhen Hospital, Capital Medical University, People's Republic of China. Among those patients, 463 suffering from severe but reversible cardiac and/or pulmonary failure received ECMO support. This study's inclusion criteria were: (1) age ≥ 18 years; (2) intraoperative or postoperative ECMO support; (3) no infection identified by specimen culture (blood, sputum, urine, or wound secretion, etc.) before ECMO support; (4) VA-ECMO support ≥ 48 hours.
According to the inclusion criteria, 322 patients were recruited to this study. Baseline characteristics, surgical parameters, ECMO parameters, and in-hospital outcomes were reviewed from medical records. This is a retrospective cohort study in accordance with the Ethical Guideline of the Committee on Human Experimentation of our institution, and the informed consent was obtained from the patients.

ECMO Procedures
The indications for ECMO were identified by extracorporeal life support physicians and cardiac surgeons. VA-ECMO mode with cannulation via femoral artery and vein was a predominant procedure duo in all patients who underwent cardiac surgery. Systemic anticoagulation was achieved by continuous intravenous heparin infusion to maintain activated clotting time

Baseline Characteristics and In-Hospital Outcomes of ECMO Support
The total ECMO support time in 322 patients was 1675.34 days. The average age of this cohort was 57.19±11.85 years, and male patients accounted for 74.22% (n=239). Patients in the infection group underwent longer duration surgeries (447.92±206.07 min vs. 402.35±191.35 min, P=0.042) and received higher percentage of cardiopulmonary bypass (CPB) assist (58.02% vs. 46.60%, P=0.044) than those in the non-infection group. Detailed baseline characteristics were showed in Table 1.
Comparing with the non-infection group, the patients in the infection group underwent longer ECMO support (153.53±105.33 hours vs. 105.21±71.79 hours, P<0.001) with lower ECMO weaning rate (56.49% vs. 70.16%, P=0.012); details were presented in Table  2. Regarding major complications and outcomes, acute renal failure requiring continuous renal replacement therapy (42.75% vs. 31.41%, P=0.037 ) was more common and survival rate to discharge was lower (41.98% vs. 53.92%, P<0.001) in the infection group than in the non-infection group (Table 3).

Incidence and Pathogenic Microorganisms of NI
Among 322 patients, 131 suffered from a total of 143 infectious events. These included 40 cases of BSI, 76 cases of RTI, 10 cases of UTI, and 17 cases of SWI. The incidence of ECMO-related NI was 85.4 cases per 1000 ECMO days. The incidences of ECMO-related NI and BSI were gradually increased with prolonging duration of ECMO support (Figure 1). The receiver operating characteristic curve showed that a cutoff value was 92 hours (sensitivity 0.794, specificity 0.497), which indicated a significantly increased risk of NI with ECMO support > 92 hours ( Figure 2).
Gram-negative bacteria were predominant pathogens and accounted for 64.01% of the total culture-positive pathogens, Gram-positive bacteria for 27.75%, and fungi for 8.24%. Acinetobacter baumannii was the most common pathogen of    Table 4.

Multivariate Analysis of Risk Factors for NI and In-Hospital Death
Multivariate logistic regression analysis ( Figure 3) Figure 5 revealed that age > 60 years, ECMO support > 144 hours, ECMO-related NI, and BSI increased significantly the risk of in-hospital death.

DISCUSSION
To the best of our knowledge, this investigation was the largest single-center observational study analyzing characteristics of ECMO-related NI in adult patients who underwent cardiac surgery. In this study, 40.68% (131/322) of the patients suffered from NI during ECMO support with a NI incidence of 85.4 cases/1000 ECMO days. However, the latest literature reported that the rate of NI was 30.4% during ECMO support [7] , and the incidence of ECMO-related NI ranged from 11.92 to 43.3 cases/1000 ECMO days [8,10] . The ECMO population was often companied by severe comorbidities, patients in our study were exposed to invasive procedures and suffered from severe surgical complications before ECMO support, which may increase the risk of ECMO-related NI. Moreover, the patients often required prolonging time of mechanical ventilation, central venous catheters, and urinary catheter, or there was an additional needing for intra-aortic balloon pump during ECMO support or continuous renal replacement therapy due to deleterious hemodynamic conditions after cardiac surgery, which also contributed to high incidence of ECMO-related NI in our study.
We evaluated causes of NI and found that prolonged duration of ECMO support was an independent risk factor for ECMO-related NI. Our result was in line with several previous studies [11,12] . And our results also showed that BMI was an

RTI
Regarding the type of ECMO-related NI, BSI was the most common in most studies, followed by RTI, UTI, and SWI [19,20] . But our study showed that RTI was predominant, followed by BSI, SWI, and UTI. Most patients were assisted by invasive mechanical ventilation before ECMO support and required prolonging mechanical ventilation due to severe heart failure, which may explain the difference of infection types between our ECMO cohort and the general ECMO cohort. Moreover, antibiotic prophylaxis may affect infection types. Despite of lacking studydemonstrated benefits, antibiotic prophylaxis is commonly used at ECMO centers, according to surveys of ELSO member sites [21,22] .
The pathogens species causing ECMO-related NI change with therapy chronology and antibiotics renewing. ELSO reported that the majority of culture-positive pathogens in patients who have undergone ECMO were coagulase-negative Staphylococci, Candida albicans, and Pseudomonas aeruginosa [23] . In our study, the Gram-negative bacilli were the most common pathogens, which were often isolated species in the ICU settings. And the most common pathogen causing BSI and RTI was Acinetobacter baumannii, which was conditioned pathogen. Severe morbidities after cardiac surgery might disturb the immune system and break ecological balance of the conditioned pathogen and the host, resulting in infections related to conditioned pathogen. In independent risk factor for ECMO-related NI. Some studies [13,14] suggested that obesity was an independent risk factor for NI. The precise mechanism of increasing risk of NI with increasing BMI remains unknown, but obesity is closely associated with impaired immune responses [15] , which may disturb the body's defense against infections.
Moreover, prolonged surgical time and CPB assist during surgery increased the risk of ECMO-related NI. Prolonged surgical time increased open chest duration, which increased the possibility of pathogens invasion. And CPB caused ischemiareperfusion injury of lung and induced systemic inflammatory response, which may disturb immune response, and then increase susceptibility to infections [16,17] . In addition, prolonged surgical time and CPB assist remarkably increased risk of multiple organ failure. And multiple organ failure was an independent risk factor for infections [18] . However, the multivariate regression analysis did show that prolonged surgical time and CPB assist were not independent risk factors for ECMO-related NI. Patients who have undergone prolonged surgical time and CPB assist were more susceptible to suffer from severe multiple organ failure, and then lost chance of ECMO support or death occurred within 48 hours of ECMO support, which might explain the abovementioned difference in our results.  addition, Candida was the main fungal pathogen in both BSI and RTI, which was associated with prolonged use of wide-spectrum antibiotics during ECMO support.

Limitations
This study had limitations regarding susceptibility to inherent bias from retrospective data from a single center. But a homogeneous cohort and relatively large number of patients allowed us to draw strong conclusions.

CONCLUSION
The prolonged duration of ECMO support was an independent risk factor for NI after cardiac surgery. Therefore, the best strategies to prevent ECMO-related NI in these patients were as follows: timely surgery correcting causes of cardiopulmonary failure -such as myocardial ischemia due to susceptible acute occlusion of grafts, unsatisfactory repair of heart valves, and delayed cardiac tamponade to shortening duration of ECMO support -and considering to wean mechanical ventilation whenever possible, when patients meet protocols of extubation.