Predictors of Invasive Fungal Infection in Patients Receiving Extracorporeal Membrane Oxygenation: A Single‑Center Retrospective Cohort Study

Abstract Background. Invasive fungal infections (IFI) are considered as one of the significant complications of prolonged extracorporeal membrane oxygenation (ECMO). Published estimates of the incidence of IFI in ECMO patients vary widely, ranging from 1–2% to 30–40%. A key clinical problem is the distinction between cases of true mycotic infection, confirmed microbiologically or histologically, and cases of clinical suspicion, in which antifungal therapy begins empirically and is not confirmed. The consensus definitions of EORTC/MSGERC 2019/2020 distinguish the categories of "proven", "probable" and "possible" IFI, FUNDICU 2024 does not recommend using the definition of "possible" IFI for research purposes. Only the "proven" and "probable" IFI categories are considered a valid basis for clinical and epidemiological studies. To determine the frequency, etiological spectrum, and independent predictors of confirmed IFI — defined as "proven" or "probable" — in adults on ECMO for > 48 hours, and to describe patients who received empirical antifungal therapy without microbiological confirmation. Methods. This retrospective single‑centre cohort study included 235 consecutive adult patients who received ECMO for > 48 hours between 2019 and 2026. Confirmed IFI was defined as a diagnosis from the "proven" or "probable" categories according to the EORTC/MSGERC 2019/2020 criteria (for patients with classical risk factors for IFI in the host) or FUNDICU 2024 (for the rest). Cases of the "possible" category were not included in the main analysis and were analyzed separately. Univariate and multivariable logistic regression analyses were applied. Results. Confirmed IFI was diagnosed in 29 of 235 patients (12.3%): invasive candidiasis in 21 (8.9%), invasive aspergillosis in 7 (3.0%), and cryptococcosis in 1 (0.4%). The diagnosis was proven in 23 patients and probable in 6. A further 58 patients (24.7%) were diagnosed with possible IFI, 55 of whom received empirical antifungal therapy without microbiological confirmation. In multivariable logistic regression, the independent predictors of confirmed IFI were lymphocytopenia duration (adjusted odds ratio [aOR] 1.02 per day; 95% CI 1.00–1.05; p = 0.037) and the need for renal replacement therapy (RRT; aOR 12.23; 95% CI 1.55–96.40; p = 0.017); veno-venous ECMO showed a borderline association (aOR 3.80; 95% CI 0.98–14.75; p = 0.053). In the multivariable model, viral pneumonia was paradoxically associated with reduced odds of a confirmed diagnosis (aOR 0.14; 95% CI 0.04–0.50; p = 0.002), reflecting the predominance of the "possible" IFI category with empirical treatment without verification in this particular group of patients. Conclusions. In patients on prolonged ECMO, confirmed IFI develops in approximately 12.3%, with invasive candidiasis predominating. Prolonged lymphocytopenia and the need for RRT are the independent risk factors. The ratio between the number of confirmed cases (29) and the number of patients with clinical suspicion and/or empirical antifungal therapy (87) is 1:3 and highlights the significant proportion of unnecessary prescribing of antifungal drugs. Managing ECMO patients requires both clinical vigilance for fungal infection and rigorous microbiological confirmation to avoid unnecessary antifungal use and its associated toxicity.