Abstract
Background/Objectives: Reported infection rates during extracorporeal membrane oxygenation (ECMO) vary widely, partly because case definitions are applied inconsistently. We synthesized the prevalence and prognostic association of nosocomial and bloodstream infection (BSI) acquired during adult ECMO. Methods: PubMed/MEDLINE, Embase, Scopus and CENTRAL were searched through December 2025 for observational studies of at least 10 adults on ECMO with an explicit infection definition. We classified case definitions by the criteria actually applied. We pooled crude mortality odds ratios (ORs) and patient-level BSI prevalence in studies applying laboratory-confirmed criteria, assessed risk of bias with ROBINS-I and the JBI checklist, and rated certainty with GRADE. Results: Forty-seven reports (46 cohorts; 11,573 patients) were included. Pooled BSI prevalence was 25.8% (95% CI 17.3–36.5%; 7 studies, 1193 patients), 18.2% in pre-pandemic and 39.2% in COVID-19 cohorts (post hoc). Infection was associated with higher crude odds of death (OR 1.65, 1.33–2.04; prediction interval 0.95–2.85; 24 studies), but analyses modeling infection as time-varying were inconsistent, including no association in a multistate model (adjusted hazard ratio 1.00, 0.79–1.26), and small-study effects were suggested. Certainty was very low for both outcomes. Conclusions: Roughly one in four adults on ECMO developed laboratory-confirmed BSI, more often in COVID-19 cohorts. Whether infection acquired on ECMO increases mortality remains uncertain.