The technical premise of the review is straightforward. ECMO circuits consist of cannulae, a centrifugal pump, and an oxygenator membrane that oxygenate blood outside the body. Venoarterial (VA) configurations support the heart, venovenous (VV) configurations support the lungs, and extracorporeal cardiopulmonary resuscitation (ECPR) deploys the technology during cardiac arrest. In all cases, ECMO is intended as a bridge—to recovery, to transplantation, or to durable mechanical circulatory support such as a left ventricular assist device. When those destinations prove unreachable, the device becomes what clinicians grimly call a bridge to nowhere, and the question shifts from how to sustain life to how to end support humanely.

The epidemiology underscores why this matters. According to the most recent ELSO Registry International Summary cited in the review, survival ranges from roughly 60 percent among patients on VV-ECMO to 49 percent on VA-ECMO, dropping to just 31 percent for patients cannulated during ECPR. A 2019 single-center study found that withdrawal was requested in just over 25 percent of supported adults, and in every one of those cases the patients lacked decision-making capacity, leaving surrogates to decide. Only 19 percent of those withdrawal cases had documentation explicitly noting futility. A 2022 retrospective cohort of 150 patients found that nearly half underwent withdrawal of life-sustaining therapy, at a median of day five of ECMO support, with 25 percent withdrawn within the first 72 hours.

Registry-scale data sharpen the picture further. A 2025 observational study of more than 23,000 ELSO registry patients found that older age, higher body mass index, cardiac arrest before ECMO initiation, and renal failure were all associated with increased odds of death after liberation from support—and that 43.7 percent of deaths in the cohort occurred in the context of withdrawal. Meanwhile, an analysis of 411 ECPR cases between 2007 and 2017 found that more than half involved early withdrawal, defined as within 72 hours of cannulation, a finding the authors flag as concerning because post-arrest prognostication may not yet be reliable that soon. Predictors included low pH at initiation and at 24 hours, higher ventilatory requirements, lower mean arterial pressure, and ECMO-related complications.

A central inflection point, the review argues, is the moment a patient is deemed ineligible for transplantation or durable mechanical support. Eligibility criteria vary across professional societies and centers. For lung transplantation, guidelines from the International Society for Heart and Lung Transplantation, the American Thoracic Society, and the European Respiratory Society generally weigh whether a patient has a greater than 50 percent risk of dying within two years without transplant against an greater than 80 percent likelihood of five-year post-transplant survival, with absolute contraindications including unresectable malignancy with high recurrence risk, uncontrolled infection, and active substance use. For heart transplantation, the American College of Cardiology lists systemic illness with life expectancy under two years and irreversible pulmonary hypertension as absolute contraindications. Crucially, eligibility is not static: many patients are cannulated precisely because a formal assessment was impossible beforehand, and eligibility can be lost mid-course through stroke, infection, multi-organ failure, or deconditioning.

The ethical core of the paper lies in its treatment of consent. The authors build on a well-known 2014 argument by Abrams and colleagues that the effort invested in informed consent before ECMO determines how easily consensus on withdrawal can be reached later. Yet a 2019 international survey found that 45 percent of physicians sometimes, rarely, or never discussed the possibility of ECMO withdrawal with patients or families. The problem is structural: ECMO is often initiated emergently—during active CPR, post-cardiotomy failure, or acute cardiogenic shock—when meaningful pre-procedural consent is impossible. In the DeMartino cohort, 82 percent of withdrawn patients had been cannulated as a bridge to decision and 26 percent during active CPR.

To address this, the University of Washington team proposes a two-phase consent model for emergent cannulations. Phase one, at the time of cannulation, frames ECMO as a time-limited intervention initiated to allow prognostic clarification. Phase two, within 24 to 48 hours, consists of a formal goals-of-care conversation once the patient has stabilized and the trajectory is clearer. For urgent but non-emergent cannulations, the full scope of informed consent should be addressed beforehand, including what constitutes an unacceptable outcome and under what circumstances support would be discontinued. The authors warn, citing a 2023 ethics analysis, that failing to establish exit criteria at the outset is what transforms ECMO from a bridge into a de facto destination therapy—creating ethically intractable situations when capacitated patients later refuse withdrawal.

When withdrawal becomes appropriate, the review endorses the Compassionate ECMO Discontinuation framework published by Machado and colleagues in 2022, which the authors identify as the only structured, interdisciplinary, published tool designed specifically for ECMO withdrawal. It organizes the process into four domains: end-of-life discussions, clinical preparation and pharmacological management, technical circuit discontinuation, and bereavement support. On the pharmacological side, comfort medications—typically opioids and benzodiazepines—should be administered at least ten minutes before circuit disruption so they can circulate before flow is reduced, with vasoactive drugs weaned concurrently or beforehand and nurses equipped with standing orders and clear titration parameters. The ethical principle of double effect, the authors note, provides support for sedation intended to relieve dyspnea and pain even when it may foreseeably hasten death.

The technical mechanics differ by configuration. VV-ECMO withdrawal can proceed by sequential reduction of sweep gas flow, reducing assisted oxygenation, or by simultaneous cessation of pump flow and sweep, with the clamped circuit left in place until death. For fully supported VA-ECMO patients, hemodynamic collapse is expected and may be rapid; the central question is whether to reduce flow gradually over 15 to 30 minutes—allowing time for comfort medication titration and family preparation—or more quickly, which accelerates death but compresses the window for symptom control. Family meetings beforehand should prioritize anticipatory guidance—what the process will look like, how long survival is expected, what the patient will appear to experience—rather than merely documenting consent, because research on family experience identifies shock at the speed of death and confusion about the withdrawal sequence as key sources of trauma.

The review closes with a call for institutional policies in the absence of national or international guidelines, and a research agenda spanning validated pre-cannulation prediction models—including exploration of artificial intelligence and machine learning—clearer definitions of futility, and structured staff debriefings within 24 to 72 hours of death to address the well-documented moral distress that accompanies ECMO withdrawal, cited in 82 percent of ECMO ethics consultations in one series. Until robust evidence arrives, the authors argue, the proposed consent template—covering the intended bridge strategy, prognostication, defined withdrawal conditions, expected complications, and surrogate identification—offers ECMO centers a practical tool to align expectations, reduce decisional regret, and ensure that when the bridge can no longer carry a patient, the crossing ends with compassion rather than improvisation.

Subject of Research: Withdrawal of extracorporeal membrane oxygenation and a proposed ECMO-specific consent framework for end-of-life care

Article Title: When ECMO can no longer bridge: a narrative review and proposed consent template

Article References: Felix, A., Zurbuchen, R., Ferreri, J., Huebner, E., Pauldine, R., & Bartz, R. (2026). When ECMO can no longer bridge: a narrative review and proposed consent template. Frontiers in Anesthesiology, 5, Article 1944591. https://doi.org/10.3389/fanes.2026.1944591

Image Credits: AI Generated

DOI: 10.3389/fanes.2026.1944591

Keywords: ECMO, withdrawal of life-sustaining treatment, informed consent, palliative care, bridge to transplant, ECPR, ELSO registry, moral distress, compassionate discontinuation, critical care, goals of care, transplant eligibility