Candida bloodstream infection under veno-arterial ECMO therapy

Cavayas et al. recently described invasive fungal infections in patients under extra-corporeal membrane oxygenation (ECMO) of the Extracorporeal Life Support Organization registry [1]. They found a 1.2% prevalence of Candida bloodstream infections (BSI). However, we highlighted the heterogeneity of this mixed population treated with veno-venous (VV) and veno-arterial (VA) ECMO and the scarce available data, in particular the delay from ECMO to infection [2].

As no specific report is available on Candida BSI under VA-ECMO, we investigated the incidence and timing of this complication in our large database of 150 VA-ECMO (January 2013 to January 2017) who survived more than 24 h (Table 1). Our surveillance protocol includes systematic blood culture (BC) once daily, since ECMO implantation up to 5 days after support withdrawal. Of the 2163 BC samples collected, either as routine or “on-demand” by the attending physician, 192 were positive; after exclusion of contaminants, 117 BC (61%) were related to bacterial infection in 46 patients. Only 7 (0.04%) were positive for yeasts, for a total of 5 BSI episodes in 4 patients. BSI rate was 43 cases/1.000 days of ECMO support. All yeast BSI were positive for Candida spp. In all cases, candidemia occurred in the third week after VA-ECMO implantation: delay between ECMO implantation and first positive BC was 17 [16–19] days. In one patient, candidemia occurred 4 days after ECMO withdrawal. Weaning from ECMO occurred in all 4 patients after 20 [17–21] days; moreover, only the patient with candidemia after ECMO withdrawal survived. Among the 46 patients with bacterial BSI, 27 died.

Table 1 Patients characteristics

Prevalence of invasive Candida disease ranges from 0 to a third of patients under ECMO [3, 4]. However, studies included both VV and VA-ECMO, without clear distinction between Candida BSI and other forms of Candida infections and without data on the timing of BSI occurrence. Moreover, whether BC were performed systematically or on-demand remains unknown. While it was difficult to draw conclusions with these heterogeneous populations, our results highlight that Candida BSI is rare and occurs late during the ECMO course. On the opposite, early septic shock under VA-ECMO is frequent and largely due to bacteria [5]. Consequently, antifungal therapy should not be part of the first-line empiric antimicrobial therapy in case of septic shock occurring within the first 2 weeks of ECMO support, unless indicated for other conditions (i.e., prolonged febrile neutropenia or tertiary peritonitis). Sepsis under prolonged mechanical support should raise the possibility of invasive Candida infection.