https://orcid.org/0000-0002-1789-603X
Abstract
Over the past 10 years, there has been a rapid expansion in the use of extracorporeal membrane oxygenation (ECMO) in the care of patients with refractory cardiac or respiratory failure. Infectious diseases clinicians must reconcile conflicting evidence from limited studies as they develop practices at their own institutions, which has resulted in considerably different practices globally. This review describes infection control and prevention as well as antimicrobial prophylaxis strategies in this population. Data on diagnostics and treatment for patients receiving ECMO with a focus on diagnostic and antimicrobial stewardship is then examined. This review summarizes gaps in the current ECMO literature and proposes future needs, including developing clear definitions for infections and encouraging transparent reporting of practices at individual facilities in future clinical trials.
Extracorporeal membrane oxygenation (ECMO) usage in critically ill adults has expanded significantly over the past decade following publication of the CESAR and EOLIA studies and extensive ECMO use during the H1N1 and coronavirus disease 2019 (COVID-19) pandemics [1, 2]. The ECMO circuit is arranged in 2 main configurations, venoarterial and venovenous, to provide either cardiac or respiratory support, respectively. ECMO circuits include ≥1 large bore cannula (eg, 15–31 French) in a major vessel, an artificial membrane lung, a blood pump, and tubing to connect the components (Figure 1). Depending on the configuration, cannulas can be placed in the neck, groin, or through an open chest. While ECMO circuit components may be rapidly exchanged in the event of mechanical failure, the cannulas are generally left in place for the entirety of a patient's ECMO course because few blood vessels are large enough to accommodate them and they are technically difficult to safely replace, especially when the patient is fully dependent on the ECMO circuit.
A, B, Schematic of an extracorporeal membrane oxygenation (ECMO) circuit in venoarterial (A) and venovenous (B) configurations. Abbreviations: AO, aorta; CO2, carbon dioxide; LA, left atrium; LV, left ventricle; O2, oxygen; PA, pulmonary artery; PV, pulmonary vein; RA, right atrium; RV, right ventricle. (Source for A and B: R. H. Bartlett, MD, with permission.) C, Chest radiograph of a patient receiving venovenous ECMO, illustrating the inherent difficulties in radiologic diagnosis of new nosocomial pneumonia.
While the rates of hospital-acquired infections have been shown to be high in this population, practice patterns for ECMO vary widely depending on the center [3, 4]. In response to the need for additional modalities of respiratory support during the COVID-19 pandemic, many centers started ECMO programs worldwide [5]. With the rapid expansion of ECMO centers, both academic and community infectious diseases clinicians have had to confront the limited literature on ECMO infections to develop protocols at their centers to minimize the impact of hospital-acquired infections on patients receiving ECMO, as well as appropriately work up and treat infections once they are suspected. This article evaluates and summarizes the most recent evidence regarding nosocomial hospital-acquired infections during ECMO to describe current understanding, provide guidance for current practices, and explore research priorities for future studies.
EPIDEMIOLOGY
There is extensive epidemiologic literature concerning ECMO infections in adults. Patients receiving ECMO are at high risk for infection due to a number of factors (Figure 2). The largest studies evaluating the rates and risk factors for infection have used the Extracorporeal Life Support Organization (ELSO) registry, which includes data from more than 200 000 patients from around the world [6]. However, centers report all positive cultures with an organism initially isolated after cannulation, a definition with obvious limitations [7]. Using this definition from worldwide ECMO centers, the incidence of nosocomial infection in adult patients has been reported at 30.6 per 1000 ECMO days [7]. While these database studies involve large numbers of patients, there are concerns of their external validity because the most common organisms isolated included coagulase-negative Staphylococcus species and Candida spp., which both often represent colonization or contamination when isolated from the blood and lungs or urine, respectively. Of note, organizations that track nosocomial infections, such as the National Health Safety Network, specifically exclude patients receiving ECMO from calculations for ventilator-associated events (VAEs) as well as central line–associated bloodstream infections (CLABSIs), which makes it difficult to understand the true incidence of these infections across centers.
Factors associated with increased infection rate in patients receiving extracorporeal membrane oxygenation (ECMO).
An alternative approach for understanding epidemiology has been single-center studies, which have mostly focused on incidence, risk factors and outcomes. However, without standardized definitions for infections, the definitions used and therefore incidence rates vary widely [4]. With differing protocols for antimicrobial prophylaxis, infection control, and use of surveillance cultures, comparing the rates between individual centers is difficult. Furthermore, time to infection is rarely reported, which leads to time-dependent bias [4]. In this literature, the greatest risk factor for infection in patients receiving ECMO is consistently the duration of ECMO [8]. However, patient factors, such as having COVID-19, a traumatic injury, or burn, have been previously described as additional risk factors for infection, with reported rates as high as 107 infections per 1000 patient-days [9, 10]. The association of nosocomial infection with mortality rates in patients receiving ECMO has been mixed, with most large multicenter ELSO registry studies showing an association between infection and mortality rates, which is generally not seen in smaller, single-center studies [9, 11].
INFECTION PREVENTION AND CONTROL
Prophylaxis
Current ELSO guidance is for centers not to give patients receiving ECMO antimicrobial prophylaxis to prevent infections, other than the standard periprocedural antimicrobial prophylaxis. Despite this guidance, many centers currently routinely use antimicrobial prophylaxis, with a 2022 Japanese study showing almost 40% of ECMO centers administering ongoing antimicrobial prophylaxis to patients after cannulation [12]. Pericannulation prophylaxis may have benefits in certain populations. For example, in a propensity-matched study of 7300 Japanese patients receiving ECMO, prophylaxis with a cephalosporin or glycopeptide within 48 hours of cannulation was associated with lower mortality rates (risk difference, 3.7%; number needed to treat, 27) and decreased nosocomial pneumonia rate (risk difference, 2.4%; , 41) [13]. However, this study did not address the utility of prophylaxis beyond 48 hours after cannulation and had several methodologic limitations, including the definitions of infection used and unavailability of important details regarding cannulation strategy and duration of ECMO. Single-center studies have also shown that ongoing protocolized antimicrobial prophylaxis after cannulation reduced broad-spectrum antimicrobial use in this population but had no effect on patient-centered outcomes [14]. Single-center studies of predominantly venoarterial ECMO showed no benefit to ongoing prophylactic antimicrobial regimens in either reducing mortality or infection rates [15, 16].
With the widely divergent results of retrospective studies of prophylactic antibiotics in the literature to date, there is currently no patient outcome benefit demonstrated with antimicrobial prophylaxis after cannulation in patients receiving ECMO. It is important that studies describing ECMO infections detail the antimicrobial prophylaxis strategy used when reporting infection rates. Our own recommendation is to administer 1 dose of prophylactic pericannulation antibiotics to prevent skin commensals infecting the incision site before open cannulation but not to administer antimicrobials before percutaneous cannulation.
Infection Prevention
A variety of methods to reduce the incidence of nosocomial infections for patients receiving ECMO have been attempted. While there are best practices for CLABSIs, catheter-associated urinary tract infections (CAUTIs), VAEs, Clostridioides difficile, and skin and soft-tissue infections, it remains unclear whether these infectious bundles are sufficient in the ECMO population. Future efforts to determine additional strategies to prevent nosocomial infections are needed. Surveys of ECMO centers show widely different practices in infection control, with most centers using chlorhexidine gluconate washes at the time of cannulation but less than half using chlorhexidine-impregnated dressings or isopropyl alcohol on circuit access ports [17]. Single-center studies have shown that daily 2% chlorhexidine gluconate and 70% isopropyl alcohol washes of exposed circuit components was associated with a reduction in bacteremia [18]. Some manufacturers, however, recommend avoiding using alcohol to clean polycarbonate components due to a risk of cracking. One recent ECMO-specific study compared 2 centers that performed a combination of a 4-times-daily course of oral amphotericin, colistin, and an aminoglycoside, body daily washes with 4% chlorhexidine, and a 5-day course of nasal mupirocin on admission to the intensive care unit to a center that performed standard care [19]. This multifaceted decontamination strategy was associated with a decreased risk of nosocomial infections (risk ratio, 0.42 [95% confidence interval, .23–.60), as well as a reduction in the acquisition of multidrug-resistant organisms (0.13 [.03–.56]), but no difference in mortality rates [19]. The long-term effects of this strategy on the ecology of intensive care units and patient's microbiota are unknown.
To prevent CLABSI, CAUTI, VAEs, and skin and soft-tissue infection, local infection prevention and controlled protocols should be formulated. At a minimum, the following infection control practices for patients receiving ECMO are recommended. CAUTI prevention strategies should include site/catheter care, securement devices, discontinuation of catheterization when no longer needed, appropriate urine culture diagnostic stewardship, and nursing protocols for catheter access and care [20]. CLABSI prevention strategies should include catheter insertion checklist, standardized insertion practices, discontinuation when no longer needed, site and catheter care, appropriate blood culture diagnostic stewardship, and nursing protocols for line access and care [21]. Strategies to prevent VAEs should include diagnostic stewardship of respiratory cultures and appropriate aspiration prevention strategies, including head of bed elevation, incentive spirometry, and oral care [22]. Finally, Clostridioides difficile prevention strategies should include diagnostic stewardship for testing, correct and early isolation of patients who have compatible syndromes, and avoiding unnecessary antimicrobial exposures in these patients [23].
In conjunction with all these efforts, it is essential that hospitals taking care of patients receiving ECMO have a robust infection prevention and control program. Most centers have near-real-time active electronic data managing systems for institutional management of nosocomial infections. As the numbers of patients needing ECMO support increase, programs ought to consider similar electronic data managing systems designed specifically for the ECMO patient population.
DIAGNOSIS
There are established immunologic changes that occur when a patient is on ECMO (Figure 3). In addition to the inflammation associated with critical illness, interactions between the patient's blood and the ECMO circuit include activation of the coagulation cascade, complement activation, as well as activation of neutrophils, and release of interleukin 8 and tumor necrosis factor α [24]. While there is cooling of the blood when outside the body and release of interleukin 6, which often acts as an anti-inflammatory mediator, patients receiving ECMO often have significant ongoing inflammation leading to elevated sepsis scores, regardless of whether or not they have an underlying infection [25]. Clinicians should be aware of their local ECMO policy regarding temperature management. Most ECMO programs use a thermoregulatory protocol where the ECMO circuit heater is adjusted to maintain normothermia and patients are unable to develop pyrexia. A less commonly used approach is where the circuit has a set temperature, which is not adjusted, and in which a patient may develop pyrexia.
Immunologic changes associated with extracorporeal membrane oxygenation favor inflammation. Abbreviations: IL-6, interleukin 6; IL-8, interleukin 8; TNF, tumor necrosis factor.
Practice patterns vary significantly across ECMO centers in diagnostics, with approximately one-third of centers performing routine surveillance cultures, as it has been suggested that clinical criteria alone can miss a significant number of true infections [3, 26]. These strategies for routine surveillance, however, are not recommended by ELSO because their utility and cost effectiveness have not been established. Several studies have evaluated the use of biomarkers for infection in patients with ECMO, such as white blood cell count, C-reactive protein, and procalcitonin. However, these tests have performed inconsistently when retrospectively applied to patients receiving ECMO and are limited by low specificity [4, 27].
A proposed algorithm for diagnosing ECMO infections is shown (Figure 4). With the inflammation associated with ECMO and the low specificity of individual biomarkers, there is no current specific combination of biomarkers with a high degree of certainty in determining whether a patient receiving ECMO has a secondary infection. Therefore, inflammatory markers such as C-reactive protein or procalcitonin should not be used to guide infectious workups. The most common infections in this cohort are bacterial and yeast bloodstream infections and bacterial respiratory infections [4]. Therefore, these should be the major disease processes investigated when a patient is believed to have an infection without a clear source [7].
Proposed algorithm for workup of infections in adult patients receiving extracorporeal membrane oxygenation (ECMO). Abbreviations: HAP, hospital-acquired pneumonia; VAP, ventilator-associated pneumonia.
In contrast, the rates of urinary tract infections are low in this population and urine cultures frequently isolate contaminates or colonizing organisms. Therefore, clinicians should generally defer obtaining urine microscopy and culture as part of an infectious workup unless the patient has urinary symptoms or risk factors for infection [27]. Similarly, additional testing for mold should not be routinely performed unless a patient has risk factors for a fungal infection because rates of fungal infection in adult populations are significantly lower than rates of bacterial respiratory infections [28]. There have been several recent outbreaks of nontuberculous mycobacteria, as well as Ralastonia, Pseudomonas, and Burkholderia species in water-based thermoregulation devices used during ECMO, and it is reasonable to evaluate for mycobacteria and atypical organisms when there is clinical suspicion and alternative causes have been ruled out [29–31]. As patients will commonly have elevated inflammatory markers with no clear infectious syndrome and negative cultures, it is reasonable to stop empiric antimicrobials at 48–72 hours and monitor patients for development of more focal signs or symptoms to guide further evaluation.
The most common infection in patients receiving venovenous ECMO in several reported studies is ventilator-associated pneumonia, with rates as high as 88% of patients with COVID-19 at one center [32]. There are no best practices for the type of culture to obtain, and sputum, tracheal aspirate, and bronchoscopy are all reasonable to use in this setting, although quantitative bronchoalveolar lavage has been used most frequently in studies of ECMO-associated pneumonia [33]. Of note, chest radiographs may not be a reliable marker of infection in this population with severe acute respiratory distress syndrome receiving venovenous ECMO (Figure 1), and thus it is reasonable to monitor for alterations in both extracorporeal gas exchange and mechanical ventilation requirements as an indication for respiratory cultures. Studies of procalcitonin, C-reactive protein, and leukocytosis are not reliably associated with positive respiratory cultures [27, 34].
Furthermore, the diagnostic utility of serum fungal antigen tests has been reported to be limited by low sensitivity compared with culture for invasive aspergillosis [35]. The lack of sensitivity and specificity for various markers compounded with the lack of clear definitions leads to a wide practice in the diagnosis and interrater agreement on classifying infections [36]. There are currently no criteria to differentiate infection from colonization, and thus the general approach should be only to culture patients with a high pretest probability of respiratory infection, with consideration of using quantitative cultures as recommended in some international guidelines [37].
Bloodstream infections are common in patients receiving ECMO. With concern for bacteremia or candidemia, a set of blood cultures should be drawn from at least 2 lines in the patient, if possible. We do not recommend accessing the ECMO circuit to obtain additional blood cultures due to the risk of contaminating the circuit [38]. Positivity rates vary significantly by study with varying culturing techniques [26]. There is also wide geographic variation in the most common organisms, with various American and European studies showing gram-positive bacteria as the most common cause of bacteremia, while gram-negative bacteria are more common in other countries [9, 39].
TREATMENT
There is limited evidence to guide therapy in patients receiving ECMO because few published studies provide granular data on both the therapies used as well as patient outcomes. This is compounded by the fact that patients receiving ECMO have dynamic renal function and hepatic function. These patients are also often on renal replacement therapy and have suboptimal cardiac function. In addition, there is evidence that many antibiotics have altered pharmacokinetics-pharmacodynamics during ECMO, with reduced drug availability, especially for highly protein bound or lipophilic antimicrobials [40]. Antimicrobials that are not protein bound or lipophilic are dosed similarly to dosing in other critically ill patients [41]. Therapeutic drug monitoring should be used if available, particularly in patients with complex conditions, such as those with renal or liver dysfunction, at extremes of weight, or receiving drugs with very narrow therapeutic indices. Regardless of infection type, a large area of clinical need is to correlate pathogen characteristics, treatment course with dosing information, and patient outcomes. While awaiting these data, treatments should be individualized for each patient, balancing the most likely site of infection and the adequacy of source control, as well as host factors.
Despite the high rate of VAEs in patients receiving ECMO, there are no trials to guide the optimal duration of treatment in this population or guide appropriate antibiotics. This gap is likely strongly influenced by the lack of clear definitions of pneumonia in this population. Without any trials examining outcomes, strategies to treat nosocomial infections should be 7 days of culture-guided antimicrobials for most patients, with consideration for longer courses based on patient and pathogen factors as well as response to therapy. In patients whose conditions rapidly improve, shorter courses may be sufficient. With the high antimicrobial use in this patient population and the risk of future infections, understanding the best balance of clinical improvement with the prevention of antimicrobial-resistant pathogens is a critical future study need.
In bloodstream infections, there is particular concern for biofilms forming on the ECMO cannulas, which are technically difficult or impossible to replace. As cannulas are generally retained when a patient has a bloodstream infection, there is concern for deep infection and a lack of source control. ECMO catheters were shown to be colonized with bacteria at a 5-fold higher rate in patients with bacteremia than in those without infection in one study [42]. Studies in patients with Enterococcus spp., Staphylococcus aureus, and Candida bloodstream infections showed that 30%–40% of patients who remained cannulated after completing antimicrobials for their infection had a subsequent episode of bacteremia or fungemia [43, 44]. Furthermore, exchange of circuits is not often performed due to infection because it has been shown to increase risk of future infections [4]. With concern for cannulas as a source of future episodes of bacteremia, an open area of clinical investigation is the use of biofilm-active agents such as rifampin in this population or, alternatively, suppressive antibiotics are used after initial treatment until after decannulation. While gram-negative bacteria tend not to form biofilms on cannulas, persistent culture positivity is associated with high mortality rate and is not correlated with clinical markers [45]. Therefore, all bloodstream infections should be monitored with repeated cultures in adult patients receiving ECMO.
RESEARCH GAPS
While there have been many advances in ECMO use in adults over the past decade, many of the recent studies evaluating infection have been small retrospective cohort studies, which vary significantly in terms of ECMO practices, patient populations, and the definitions of infections used. With these studies, several gaps need to be systematically answered in order to improve care for patients receiving ECMO. The largest gap in ECMO research and the first question that must be answered is the creation of standardized definitions for infections reported in different studies, as well as in the ELSO Registry. Once these definitions are established, the creation of prospective, multicenter observational studies with diverse patient populations is possible to guide best practices, which can lead to randomized controlled trials. Furthermore, reporting of ECMO infections should be standardized, with information on prophylaxis used, as well as culturing practices at each center, and outcomes should be reported with information on antimicrobial selection and duration. With the growth of ECMO globally, there is a need for infectious diseases clinicians to help develop key metrics to guide care in this population.
Notes
Disclaimer. The views expressed herein are those of the authors and do not necessarily reflect the official policy or position of the Defense Health Agency, Brooke Army Medical Center, the Department of Defense, or any agencies under the US government.
References
Author notes
J. E. M. and A. S. contributed equally to this work.
Potential conflicts of interest. The authors: No reported conflicts of interest.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
