We are grateful to Messika et al (1) for their interest in our work (2), recently published in Critical Care Medicine. Infections are among the most frequent complications in extracorporeal membrane oxygenation (ECMO) patients and have a significant impact on clinical outcome (3). Since a number of confounding factors make the recognition of an infection in ECMO patients particularly challenging, the application of precise diagnostic criteria is paramount. We analyzed only microbiologically confirmed infections, after multidisciplinary evaluation based on rigorous criteria. Messika et al (1) highlights the low frequency of cannula-related infections in our patients: indeed, we reported only one case, confirmed by the presence of local erythema and purulent drainage associated with positive culture of the drainage. At variance, in a series of 220 adult patients supported with veno-arterial ECMO for cardiogenic shock published by Schmidt et al (4), cannula infections accounted for 10% of all infectious episodes. Several factors may explain our lower rate of cannula infections. First, the vast majority of our patient received veno-venous ECMO for respiratory failure, and in “all” cases (and not in 87% of them as reported by Messika et al ), cannulation was performed with a fully percutaneous approach (5). We strongly believe that percutaneous cannulation performed with a rigorous aseptic technique is of crucial importance in preventing cannula infections. Nevertheless, we acknowledge that surgical exposure of the vessel may be necessary in emergencies (such as veno-arterial ECMO support for circulatory failure), thereby increasing the risk of infectious complications. Notably, in the study of Schmidt et al (4), 55% of the patients received central intrathoracic cannulation, which was associated with a higher infectious risk. Second, we included in the risk analysis only the infectious episodes occurring from the moment of ECMO start to 48 hours after decannulation. Based on the findings of a retrospective analysis of 33 cannula infections (6), of whom a third occurred after a median interval of 8 days after ECMO discontinuation, Messika et al (1) suggests to consider as ECMO-related all cannulation site infections observed up to 30 days after decannulation. After reviewing all infectious episodes occurred during the entire ICU stay, we identified only one “late” infection of the cannulation site, diagnosed 3 days after ECMO discontinuation. In our experience, virtually all infections of the cannulation site occur within the first 2–3 days from decannulation: hence, at least for percutaneously inserted cannulas, this observation period seems adequate. Third, we included in the inferential analysis only the first infection episode, but for descriptive purpose, we analyzed also the subsequent infections: as reported in the Online Supplement of our article (2), we identified two further (i.e., occurring not as first infectious episode) infections of the cannulation sites which were diagnosed after a median interval of 40 days from ECMO start. Finally, Messika et al (1) underline the high rate of primary bloodstream infections in our population, suggesting that a significant proportion of them may be cannula related. However, in all patients with positive blood cultures, the presence of signs of catheter-related infection was excluded by means of a thorough evaluation of all catheter insertion sites.