The development of extracorporeal support devices has been fueled by the need to treat critically ill patients failing all conventional therapeutic options. In the last years, evidence regarding their safety and feasibility has grown, and their impact on the improvement of survival has also been demonstrated.1,2

In the setting of the intensive care unit, the most common types of extracorporeal support include extracorporeal membrane oxygenation (ECMO), and ventricular assist devices (VAD).

ECMO support is frequently percutaneous and not requiring surgery, but is associated with the need of bed riding, a larger exposure of blood to foreign surfaces, and, as it is used as a rescue therapy, with multiple organ damage. On the opposite, left VAD (LVAD) requires extensive surgery, and the tunneling of cannulas and drivelines through the chest of the patient. Furthermore, VAD is indicated in patients with preserved lung function, as the circuit does not include the oxygenator. Intracorporeal VAD is indicated in patients who are candidates to receive heart transplant or as destination therapy. On the contrary, paracorporeal VAD is employed for patients who could experience myocardial recovery.

Since the first experiences with prolonged extracorporeal support, infection has been unanimously pointed out as one of the most frequent and adverse complications.3–5

Indeed, infections occurring on extracorporeal support are often difficult to treat, especially if they involve the foreign device. Moreover, few data are available on the distribution and the penetration of antibiotics on their surfaces.

As both ECMO and VAD are increasingly used as a bridge to further therapeutic options, including heart and lung transplants, the presence of infection, however, jeopardizes patient's eligibility to receive these therapies, with consequent poor prognostic implications.

Despite the heavy burden of infection, paradoxically, few definite criteria and guidelines have been developed for the prevention and management of this complication during extracorporeal support. A marked variability in antimicrobial prophylaxis and infection surveillance practices among Extracorporeal Life Support Organization (ELSO) centers has also been reported.6

The different epidemiology of pathogens in the different intensive care units also contributes to make the task more difficult.

A comprehensive and rigorous picture of infection epidemiology, however, is the only basis on which protocols for infection prevention and management can be developed.

Aim of our study was to analyze rates and causal organisms of infections occurring in patients receiving ECMO or VAD, either paracorporeal or implantable, in a national tertiary referral center over a two-years period, and to assess the differences between these types of extracorporeal support setting.

Seventy-five patients received extracorporeal support in the study period: 60 patients were treated with ECMO (37 VA ECMO and 23 VV ECMO), and 15 patients were implanted a VAD. Fourteen patients (13 on VA ECMO and 1 on VV ECMO) received extracorporeal support for less than 48h and were excluded from the analysis. Therefore 61 patients were included in our analysis: 46 patients treated with ECMO (24 with VA ECMO and 22 with VV ECMO) and 15 patients with VAD.

Patients’ characteristics, demographic and clinical data are shown in Table 1. Twenty-eight of the 61 patients (46%) developed extracorporeal related infection. Patients with infection had longer ECMO support duration (p=0.03), VAD support duration (p=0.01), intensive care unit (ICU) stay (p=0.002) and hospital stay (p=0.03) than patients without infection (p=0.03). Notably, patients with infection had a higher SOFA score at the time of implant than patients without infection (10.6±3.2 vs 8.3±3.6, p=0.03)

Overall hospital mortality was 55.7%. Although patients with infection died more often than patients without infection (67.9% vs 45.5%), the difference in mortality did not reach a statistical significance (p=0.1).

The rate of infection was 57 per 1000 days of ECMO, 41 per 1000 days of VAD, and 50 per 1000 days mechanical circulatory support (i.e. ECMO and VAD considered all together).

The overview of infections occurring in patients treated with ECMO is presented in Table 2.

Eighteen out of the 46 patients supported with ECMO (39%) developed nosocomial ECMO-related infection: 7 patients with VV ECMO, and 11 patients with VA ECMO.

Four of the 18 patients with infection (22%) developed a polymicrobial infection. In particular, 28 micro-organism were isolated: 13 Gram-negative bacteria (46%), 10 fungi (36%), and 5 Gram-positive bacteria (18%).

The most frequent pathogens responsible for infection were Candida spp. (29%), followed by Klebsiella spp. (21%).

Pneumonia was the most frequent type of infection (11 episodes), followed by bloodstream infection (8 episodes), surgical site infection (2 episodes), and urinary tract infection (1 episode).

Micro-organisms responsible for infection in patients with VAD are listed in Table 3. Ten out of 15 patients (67%) developed nosocomial infection following VAD implantation, and 28 isolates were recorded: 14 Gram-negative bacteria (50%), 9 Gram-positive bacteria (32%), and 5 fungi (18%).

Seven out of the 10 patients with infection (70%) had a polymicrobial infection. Twelve micro-organisms were isolated in the context of a VAD-related infection, while the remaining 16 were responsible for non-VAD infections. No VAD-specific infection was observed.

The most frequently involved micro-organisms were Staphylococcus spp. (18%), and Pseudomonas aeruginosa (14%). In contrast to the epidemiology of ECMO-related infections, Candida spp. were found only in the 11% of the infective episodes.

Pneumonia was the most common type of infection in patients supported with VAD (7 episodes), followed by bloodstream infection (5 episodes), urinary tract infection (5 episodes), and mediastinitis (2 episodes).

In the total of patients undergoing mechanical support, Candida spp. were the predominant pathogens, as they were isolated in 11 episodes (20%).

SOFA score (OR=12.33, 95% CI 1.15–132.36, p=0.04) and VAD support (OR=1.27, 95% CI 1.04–1.56, p=0.02) were shown to be significantly associated with infection at logistic stepwise regression analysis. Gram-negative bacteria infections were associated with VAD support analysis (OR=4.11, 95% CI 1.20–14.12, p=0.02).

Moreover, independent significant predictors of the duration of ECMO support were the following: infection (regression coefficient=3.99, 95% CI 0.93–7.05, p=0.02), BMI (regression coefficient=0.46, 95% CI 0.09–0.83, p=0.02), fungal infection (regression coefficient=4.96, 95% CI 1.42–8.44, p=0.009), and obesity (regression coefficient=10.47, 95% CI 1.77–19.17, p=0.02).

The Kaplan–Meyer curves of time free from infection were not significantly different between the 3 groups (VV ECMO, VA ECMO, and VAD (p=0.2), between ECMO, either VV or VA, and VAD (p=0.2), and between VA ECMO and VAD (p=0.08)).

Our study shows that infection is a very common complication in patients undergoing extracorporeal support. This is not surprising, as patients requiring ECMO or VAD support are a sick population of immunocompromised hosts with underlying comorbidities, often undergoing prolonged hospitalization and multiple drug regimens. Nevertheless, the inter-hospital comparison between different types and the incidence of infections in patients undergoing extracorporeal support have been limited by the fact that publications available on this topic were based on heterogeneous definitions of ECMO related and VAD related infections. Furthermore, micro-organisms’ epidemiology greatly varies among different intensive care units.

Patients or candidates to receive such therapies should not be regarded as a homogeneous category. Indeed, we observed a remarkable difference in infection rates and causal organisms between ECMO and VAD patients.

Incidence of infections occurring during ECMO treatment in literature ranges from 3.5% to 45.1%.11–15

In a query recently conducted on the 20,741 ECMO patients of the Extracorporeal Life Support Organization (ELSO), adult patients were found to be more vulnerable to infection compared to neonatal and pediatric population (infection rate of 30.6 per 1000 days of ECMO) with Candida spp being the predominant pathogen in the adult population, being responsible for 15% of infections.3

Furthermore, Sun et al. found that bloodstream infection is the most common infection, with Candida spp, Stenotrophomonas maltophilia, and Pseudomonas aeruginosa being the predominant blood isolates.16

Similarly, we confirmed that fungal infection is a frequent complication of ECMO, and we also found that it is associated with a longer duration of the ECMO support. Indeed, we showed that a high incidence of infections by Candida species was found in various studies, and that antimicrobial prophylaxis differs between centers as no guidelines exist.6

The incidence of Candida spp. infections that we reported in our population of patients supported with ECMO (29%) is higher compared to the 15% of the patients included in the ELSO registry.3 This observation further strengthens the consideration that the same pathogen may have a different impact in different clinical settings. The mortality attributable to Candida infection is known to be high in literature, ranging from 5% up to 49% according to the different types of studies (retrospective vs prospective), patients (ICU vs. non-ICU, comorbidity), and healthcare settings.17–19 Whether Candida spp. infection may play an impact on mortality also in ECMO patients is an intriguing issue, but we cannot answer this question based on our data, as we observed no statistically significant difference in mortality between patients with and without infection and due to the small sample size of our population.

VAD support too has been demonstrated to be associated with a high rate of both device-correlated and not device-correlated infection, the most feared being endovascular infection which is associated with increased mortality.20–22 In particular, Gordon et al. reported a bloodstream infection rate of 7.9 per 1000 VAD days.21

Recently, a multidisciplinary working group of the International Society for Heart and Lung Transplantation (ISHLT) provided a panel of standard international definitions including not only major infections but also comprehensive details of all aspects of VAD-specific and VAD-related infection in these patients.8

Therefore, we classified infection occurring in the patients of our study according to their definitions, which distinguish between VAD-specific, VAD-related, and non-VAD infections.

Interestingly, we did not observe VAD-specific infections during hospital stay, or at the 3-months follow-up. VAD-specific infections are a severe complication with a significant impact on the prognosis, but do not occur early after VAD implantation in our experience. On the contrary, we found that more than a half of patients receiving VAD support developed nosocomial infection before hospital discharge. The consequences of such infections on the prognosis of the patients and of the VAD, together with the costs of prolonged ICU and hospital stay, have probably been underestimated in the past. Notably, patients with paracorporeal VAD resulted to be less prone to develop infection than those with implantable devices in our experience, as no VAD-related infection was registered.

Gram-positive bacteria have been found to be the predominant pathogens responsible for infection in patients with VAD,23 and Staphylococcus species were the predominant pathogens also in our study population. However, VAD support was found to be associated with an increased occurrence of infections Gram-negative pathogens in our analysis, thus confirming that the burden of Gram-negative bacteria infection remains heavy also in these patients.

It should be also noticed that we reported a higher infection rate as compared to previous studies concerning mechanical circulatory support. The difference observed may have different explanations, including a slight heterogeneity of infection definition among different centers, bias in data collection, and a different degree of experience of different centers in dealing with patients treated with mechanical circulatory support which leads to a different incidence of complications.

Despite the few data available on this topic, however, all these studies show that infection is to be considered a very common complication during mechanical circulatory support.

Furthermore, patients with extracorporeal related infection require longer ICU stay and hospital stay than patients without infection, with subsequent consequent increasing costs. As a consequence, the need for prolonged antimicrobial therapies may lead to organ toxicity and the development of resistant microbiological strains. This phenomenon is also exacerbated by the fact that in the real world of ECMO and VAD it is difficult to discriminate as which patients are free from infection without any doubt, and, as a consequence, antibiotics are extensively used.

In the light of the data shown, we changed antimicrobial prophylaxis in patients undergoing mechanical circulatory support and we now routinely administer broad spectrum antibiotics and fluconazole.

However, the identification of the optimal prophylaxis for ECMO treatment is not trivial, as no randomized trials have ever addressed this issue.

A recent survey among ECMO centers6 reported that forty-two percent of centers routinely administer antibacterial prophylaxis to all ECMO patients. Among centers with standardized protocol or antibacterial prophylaxis based on physician preference, less than half use a single agent. On the contrary, only a minority of centers reported administering routine (2%) or selective antifungal agents (13%) to ECMO patients.

This study presents some limitations. First of all, we report data from a relatively small population of heterogeneous patients. Second, all data came from a single center. Third, patients usually received antibiotic prophylaxis and this might have influenced our results.

In conclusion, mechanical circulatory support related infections are frequent, difficult to treat, and lead to prolonged hospitalization. The epidemiology of infection is different among ECMO and VAD, suggesting the need for an appropriate antibiotic prophylaxis according to the type of extracorporeal support.

The work was supported by departmental funds.

The authors have no conflicts of interest to declare.