ECMO infections are a frequent complication of prolonged critical illness, but their reported frequency and relationship with survival depend heavily on how infection is defined and analyzed. A systematic review by Alzibdeh and colleagues brings together the adult literature on nosocomial and bloodstream infection. It identifies a substantial infection burden while showing why pooled estimates should be interpreted cautiously: the authors rated the certainty of both principal outcomes as very low.
Study Design and Population
The review searched PubMed/MEDLINE, Embase, Scopus, and CENTRAL through December 2025. It included observational studies with at least 10 adults receiving extracorporeal membrane oxygenation and an explicit infection definition. The final evidence base comprised 47 reports from 46 cohorts, covering 11,573 patients.
Two separate analyses addressed different questions. The mortality analysis pooled unadjusted comparisons between infected and uninfected groups. The bloodstream-infection analysis was narrower: it required laboratory-confirmed case definitions, a patient-level denominator, and the number of patients experiencing at least one bloodstream infection. Only seven studies, involving 1,193 patients, qualified for that prevalence estimate.
Why Infection Definitions Matter
The investigators classified studies according to the diagnostic criteria actually applied, rather than accepting a reference to a surveillance standard as proof of consistent use. Their highest-definition tier required laboratory-confirmed bloodstream-infection logic, including rules for interpreting common skin commensals. Other studies used less complete culture-based criteria or a molecular-testing composite.
This distinction matters for comparison between ECMO programs. Patients with an infection, infection episodes, positive cultures, and ECMO runs are different counting units. A report based on recurrent episodes cannot simply be combined with a report counting each infected patient once. Likewise, an infection window beginning after ICU admission does not necessarily identify only infections acquired after cannulation.
Bloodstream Infection: A Substantial but Variable Burden
Across the seven qualifying studies, pooled patient-level bloodstream-infection prevalence was 25.8% (95% confidence interval 17.3–36.5%). This estimate applies to the selected laboratory-defined subset, rather than to every patient in the full review.
Between-study variation was considerable: the 95% prediction interval was 5.7–66.5%, showing how widely prevalence could differ in a comparable new cohort. Approximately 76% of patients in this analysis received venovenous ECMO, and 60% came from cohorts in which at least half the patients had COVID-19.
A post hoc comparison estimated prevalence at 18.2% in pre-pandemic cohorts and 39.2% in COVID-19 cohorts. These groups also differed in indication, treatment, surveillance, and duration of support. The contrast therefore generates a research question; it does not isolate a causal effect of COVID-19 or establish an expected infection rate for an individual center.
ECMO Infections and Mortality: Association Requires Context
The mortality analysis included 24 studies and 4,250 patients, runs, or procedures. It combined bloodstream infection, broader nosocomial infection, and one invasive-fungal-disease exposure. Infection was associated with higher crude odds of death: odds ratio 1.65 (95% confidence interval 1.33–2.04). The prediction interval, 0.95–2.85, included no association.
An odds ratio of 1.65 does not mean that infection caused 65% more deaths, and it is not a risk ratio. Patients who acquire an infection may differ in illness severity, organ support, immunosuppression, and time spent at risk. Infection can also occur late in an ECMO run, making a simple comparison of patients who were ever infected with those who were never infected difficult to interpret.
Analyses treating infection as an exposure that changes over time gave inconsistent results. One multistate analysis found no association between healthcare-associated infection and in-hospital death after adjustment (hazard ratio 1.00; 95% confidence interval 0.79–1.26). Other analyses evaluated different infection syndromes and mortality endpoints. These findings, together with signals of small-study effects, leave the mortality attributable specifically to infection unresolved.
Risk Factors, Microbiology, and Antimicrobial Prophylaxis
Longer ECMO duration was the most consistently reported correlate of infection. However, longer support creates more time for infection to develop, and infection may itself delay recovery. Duration should therefore be interpreted as part of a changing clinical course rather than as an isolated causal predictor. Other reported associations included illness severity, renal replacement therapy, transfusion, and immunosuppression, with inconsistent adjustment across studies.
The microbiological reports included Gram-negative organisms, Gram-positive organisms, and Candida species, with resistant organisms described in many cohorts. Different studies counted patients, isolates, cultures, or episodes, preventing reliable pooled estimates for individual pathogens or antimicrobial resistance.
Antibacterial prophylaxis practices also varied and were closely linked to ECMO indication and configuration. This review could not separate the effect of prophylaxis from differences in case mix, and it does not establish a preferred prophylactic antibiotic or antifungal regimen.
Relevance for Perfusion and ECMO Teams
A practical implication of this synthesis is the need for infection data that can be interpreted alongside the ECMO timeline. For local audit, teams can specify the diagnostic criteria, distinguish patient counts from recurrent episodes, and record when suspected infection occurred relative to ICU admission and cannulation. Documenting culture collection practices and antimicrobial exposure would make comparisons between periods and programs more informative.
For clinical interpretation, a positive culture should be considered together with the organism, sampling context, infection site, and the patient’s course. The review supports coordinated attention to infection prevention, diagnostic stewardship, and antimicrobial review, but it did not test a surveillance schedule, prevention bundle, or treatment protocol. Its pooled prevalence is a description of selected studies, not a performance threshold for ECMO services.
Limitations
All included evidence was observational. All comparisons contributing to the pooled crude mortality analysis were judged at serious risk of bias. In six of the seven prevalence studies, cultures were obtained on clinical suspicion rather than through systematic surveillance. Differences in infection definitions, onset windows, ECMO configuration, pandemic era, and outcome timing further limit comparability.
The review was registered after searching and study selection. Several methodological decisions and subgroup analyses were made later, although these changes were disclosed and sensitivity analyses were reported. Some adult-predominant cohorts included, or may have included, younger patients, and several source reports used runs or procedures instead of individual patients. The resulting very low certainty means that both the frequency estimate and the mortality association may differ substantially across clinical settings.
Original Article
Alzibdeh A, Alamami A, Zara S, et al. Infectious Complications in Adults Receiving Extracorporeal Membrane Oxygenation: A Systematic Review and Meta-Analysis of Nosocomial and Bloodstream Infection, Prevalence, Risk Factors, and Clinical Outcomes. Medical Sciences. 2026;14(6):653. doi:10.3390/medsci14060653.
The authors reported no funding for this study and no conflicts of interest. The original article is available under a Creative Commons Attribution license.
