Bloodstream infection (BSI) is one of the most common complications during extracorporeal membrane oxygenation (ECMO) in patients treated with venoarterial ECMO for cardiac support.Reference Schmidt, Brechot and Hariri 1 Compared with patients receiving venoarterial ECMO, patients receiving venovenous ECMO for respiratory support are more commonly exposed to antibiotics and corticosteroids.Reference Brodie and Bacchetta 2 In addition, the duration of venovenous ECMO support is longer than for venoarterial ECMO for cardiac support,Reference Thiagarajan, Barbaro and Rycus 3 which affects the rate of BSI.Reference Schmidt, Brechot and Hariri 1 , Reference Burket, Bartlett, Vander Hyde and Chenoweth 4 Therefore, the features of BSI in patients receiving venovenous ECMO might be different from those in patients receiving venoarterial ECMO. The number of patients receiving venovenous ECMO support in clinical practice is growing.Reference Thiagarajan, Barbaro and Rycus 3 However, data on the epidemiology and clinical relevance of BSI during venovenous ECMO remain limited.Reference Kutlesa, Santini and Krajinovic 5 , Reference Grasselli, Scaravilli and Di Bella 6 Therefore, we investigated the incidence, etiologies, risk factors, and clinical outcomes of BSI in patients receiving venovenous ECMO for severe acute respiratory failure in a medical ICU.
METHODS
This retrospective cohort study included 121 adult patients who received venovenous ECMO for >48 hours at Samsung Medical Center in South Korea between January 2012 and December 2016. We used the ECMO database of our hospital, in which the clinical and laboratory data of patients who received ECMO were prospectively registered, and we conducted a retrospective review of patient hospital charts to supplement these data.
Routine surveillance blood culture during ECMO is not mandatory in our hospital. However, cultures of blood as well as other appropriate specimens are performed when a new episode of infection is clinically suspected. Bloodstream infection was defined as satisfying 1 of 2 criteriaReference Garner, Jarvis, Emori, Horan and Hughes 7 , Reference Horan, Andrus and Dudeck 8 : (1) the isolation of a recognized pathogen from blood culture or (2) the presence of clinical manifestation and common skin contaminant isolated from at least 2 blood cultures. Infections were classified as primary BSI when the organism isolated from blood culture was not related to infection at another site or when the BSI was related to the catheter.Reference Garner, Jarvis, Emori, Horan and Hughes 7 If the isolated organism from blood culture matched an organism identified from another site, it was classified as a secondary BSI. Only infections occurring >48 hours after initiation of ECMO until decannulation were considered as ECMO-related infections in this study.
RESULTS
Overall, the median age of patients receiving venovenous ECMO for respiratory support was 60 years (range, 51–67 years), and 33 patients (27%) were male. Moreover, 36 patients (30%) were immunocompromised: 20 (17%) had hematological malignancies or solid tumor treated with chemotherapy; 10 (8%) had undergone solid organ transplantation; and 6 (5%) had undergone high-dose or long-term corticosteroid treatment. Bacterial pneumonia was the most common pulmonary condition causing respiratory failure (41%), followed by acute exacerbation of interstitial lung disease (17%) and viral pneumonia (16%).
The median duration of mechanical ventilation (MV) prior to the initiation of ECMO was 2 days (range, 0–6 days), and corticosteroid treatment was administered to 59 patients (58%). Femoro-jugular and femoro-femoral cannulations were used in 91 patients (75%) and 27 patients (22%), respectively. Every patient had an arterial catheter, and 109 (90%) patients also had a central venous catheter. Broad-spectrum antibiotics and antifungal agents were used in 111 patients (92%) and 33 (27%) patients during ECMO support, respectively. The median duration of ECMO was 14 days (range, 7–26 days).
Bloodstream infections occurred during ECMO support in 21 patients treated with venovenous ECMO (17%; 8.5 episodes per 1,000 ECMO days), and the median time from ECMO initiation to occurrence of BSI was 6 days (range, 4–19 days). However, the frequency of BSI over the duration of venovenous ECMO did not change (P=.337; for trends, see Supplemental Figure 1). Primary BSI developed in 4 patients (19%), and half of these were associated with central venous catheter infections. The most common source of secondary BSI was respiratory infections (62%), followed by intra-abdominal infections (10%) and urinary tract infections (10%). Staphylococcus aureus and Acinetobacter baumannii were the most commonly identified bacteria. Of the 5 cases of fungemia, Candida tropicalis was identified in 3 patients and Candida albicans was identified in 2 patients.
Comparisons of clinical characteristics between patients with or without BSI are shown in Table 1. Duration of arterial catheterization was longer in patients with BSI, but the duration of central venous catheterization was not different from patients without BSI. In addition, there was no significant difference in the proportion of patients using steroids or broad-spectrum antibiotics during ECMO support. However, cumulative blood transfusions during ECMO were significantly higher in patients with BSI than in those without BSI. After adjusting for potential confounding factors, duration of arterial catheterization (adjusted odds ratio [aOR], 1.25; 95% confidence interval [CI], 1.03–1.52) and total units of transfused bloods (aOR, 1.01; 95% CI, 1.00–1.02) were independently associated with BSI (Table 2).
TABLE 1 Comparison of Clinical Characteristics Between Patients With or Without Bloodstream Infection During Venovenous Extracorporeal Membrane Oxygenation
NOTE. BSI, bloodstream infection; CVC, central venous catheter; ECMO, extracorporeal membrane oxygenation; ILD, interstitial lung disease; MV, mechanical ventilation; SOFA, sequential organ failure assessment; VV, venovenous.
a Unless otherwise specified.
b Radiation therapy-induced pneumonitis, pulmonary tuberculosis, diffuse alveolar hemorrhage, and airway occlusion by tumor mass or blood clot.
TABLE 2 Predictors of Bloodstream Infection During Extracorporeal Membrane Oxygenation
NOTE. CVC, central venous cathether; ECMO, extracorporeal membrane oxygenation; VV, venovenous.
In patients with BSI, the successful weaning rate of ECMO was significantly lower (25% vs 60%; P=.003) than in patients without BSI and the median duration of ECMO was longer (23 days [range, 10–32 days] vs 13 days [range, 7–23 days]; P=.010) . In addition, we detected trends toward increased mortality in the ICU (76% vs 55%; P=.073) and during hospitalization (76% vs 58%; P=.120). However, there was no difference in length of stay in the ICU or at the hospital in general (Supplemental Table 1).
DISCUSSION
The reported incidence rates of BSI occurring in patients requiring ECMO support for circulatory failure or mixed population range widely, from 3.4% to 21.1%,Reference Schmidt, Brechot and Hariri 1 , Reference Burket, Bartlett, Vander Hyde and Chenoweth 4 , Reference Aubron, Cheng and Pilcher 9 – Reference Thomas, Hraiech and Cassir 16 Although the incidence of BSI in our study was similar to that reported previously, a recent study that included 100 adult patients treated with venovenous ECMO and who underwent routine blood cultures showed an incidence rate of 35%,Reference Kutlesa, Santini and Krajinovic 5 which is relatively high compared to our study but might be associated with routine surveillance blood culture.
Inconsistent with previous reports,Reference Schmidt, Brechot and Hariri 1 , Reference Kutlesa, Santini and Krajinovic 5 , Reference Grasselli, Scaravilli and Di Bella 6 , Reference Aubron, Cheng and Pilcher 9 – Reference Hsu, Chiu, Huang, Kao, Chu and Liao 11 a prolonged duration of ECMO support was not associated with BSI in our study. However, the longer duration of ECMO might be the result of an adverse effect by BSI rather than a risk factor for BSI.Reference Grasselli, Scaravilli and Di Bella 6 In this study, patients with BSI had almost double duration of venovenous ECMO. Using the trends test, we found no significant change in BSI rate across the duration of venovenous ECMO. ECMO patients generally have multiple indwelling catheters including central vein catheters and radial arterial catheters, in addition to the cannulas used for the ECMO circuit. We found that the risk of BSI increased independently with the duration of arterial catheterization. Therefore, arterial catheters should be examined as a potential source of BSI in patients receiving venovenous ECMO support.Reference O’Horo, Maki, Krupp and Safdar 17 In addition, a higher number of transfusions was independently associated with BSI. Although the association of blood transfusion with nosocomial infection in critically ill patients has been demonstrated previously,Reference Taylor, O’Brien and Trottier 18 no study has specifically addressed the potential association of blood transfusion and BSI in patients receiving ECMO. Patients with prolonged ECMO support have a higher probability of being transfused.Reference Butch, Knafl, Oberman and Bartlett 19 However, multicollinearity was not found in the regression model (variance inflation factor, 1.080).
Gram-negative organisms such as A. baumannii, Stenotrophomonas maltophilia, and Pseudomonas aeruginosa are common BSI pathogens.Reference Kutlesa, Santini and Krajinovic 5 , Reference Sun, Ko and Tsai 10 , Reference Kim, Yeo and Yoon 15 Consistent with data obtained in patients received venovenous ECMO,Reference Kutlesa, Santini and Krajinovic 5 , Reference Sun, Ko and Tsai 10 , Reference Kim, Yeo and Yoon 15 gram-negative bacteria accounted for 38% of BSI in this study. These results were likely related to the facts that cases suspected of contamination by common skin contaminants were not diagnosed as BSIs and that 81% of BSIs were secondary BSIs associated with pulmonary, intra-abdominal, or urinary tract infection. Pulmonary infection was most commonly associated with BSI in our study. Schmidt et alReference Schmidt, Brechot and Hariri 1 also demonstrated that ventilator-associated pneumonia was the most common source of BSI and that P. aeruginosa was the most commonly isolated organism from blood culture in adult cardiogenic shock patients with venoarterial ECMO.
The impact of BSI on clinical outcomes is controversial.Reference Schmidt, Brechot and Hariri 1 Some studies have demonstrated significantly increased mortality rates in pediatric patients with BSI compared with patients without BSI,Reference Steiner, Stewart, Bond, Hornung and McKay 13 but other studies involving patients receiving venovenous ECMO showed that BSI had no effect on mortality.Reference Kutlesa, Santini and Krajinovic 5 , Reference Aubron, Cheng and Pilcher 9 In this study, the ECMO-related outcomes, weaning rates, and duration of ECMO support were significantly worse in patients with BSI than in those without BSI. In addition, a trend toward higher mortality among patients with BSI was detected. Therefore, more effective infection control measures are warranted to reduce infections and to improve outcomes in patients receiving ECMO.
In summary, the incidence of BSI during venovenous ECMO support was as high as 17% in adult patients with severe acute respiratory failure. Longer duration of arterial catheterization and more blood transfusions were independently associated with BSI, which was associated with poor clinical outcomes.
ACKNOWLEDGMENTS
Financial support: This work was supported by Samsung Medical Center (grant no. OTX0002901).
Potential conflicts of interest: All authors report no conflicts of interest relevant to this article.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2018.90
