Extracorporeal Membrane Oxygenation–Associated Infections: Carefully Consider Cannula Infections!
Infectious complications, which include ECMO cannula infections, are among the most common complications (2). In the series by Grasselli et al (1), ECMO cannula infection occurred in only one patient. This seems a relatively small number that deserves scrutiny.
First, in their definition, authors restricted infections to those that occurred within 48 hours after decanulation, as reported previously (3). However, this delay of 48 hours is debatable. Indeed, by analogy, the delay for postoperative complications is 30 days (4) and our own experience with these patients and this technique is that infections—clearly related to the technique—may occur much later. This clinical impression was confirmed in our recent series of 33 ECMO cannula Escherichia coli infections (5). We observed that a third of these episodes occurred in a median of 8 days (5–10 d) after ECMO discontinuation. Noticeably, only one occurred within this 48-hour window. All but one had local signs (such as scar dehiscence, pus effusion, cellulitis), which clearly connects the technique to the complications. It could be therefore interesting to provide the prevalence of ECMO cannula infection in the study by Grasselli et al (1) considering both definitions.
Second, the vessel and cannulation technique may explain the discrepancies between ours and the results by Grasselli et al (1). Indeed, most of the vessel and cannulation techniques in the study by Grasselli et al (1) (87%) were percutaneous veno-venous cannulation, whereas most of our patients had a surgical veno-arterial cannulation.
Third, Grasselli et al (1) report a high rate of primary bloodstream infections (33/152), which may legitimately question about their origin. One may argue that without evidence of same pathogen found in another organ, these bloodstream infections should be related to ECMO cannula, especially in view of the prolonged use of ECMO (25.5 d [10.75–54 d]). For instance, in their previous work, Schmidt et al (3) reported 47 episodes of bloodstream infection of which half (n = 23) originated from the lungs, five from the ECMO cannula, and only five were isolated. In our study (5), four infections were bacteremic. Hence, taken together these results suggest that approximately 12% of cannula-related infections are bacteremic. Although the definition of ECMO cannula infection is still unclear, recent data suggest that these should be considered up to 30 days, especially in case of a surgical cannulation. Waiting for a consensual definition, it may be worth reporting the prevalence of ECMO cannula infection using both definitions.