Chapter question: The machine was started on a patient who could not be asked. It is still running. The heart has not recovered, the brain may or may not have, and nobody has scheduled a moment at which anyone decides anything. How does an ECPR run end well — and what does the circuit become when it fails?
Evidence search date: 15 September 2026.
Primary sources: ELSO Red Book 6th edition — Chapter 32 and the ethics and end-of-life material. ECPR and Resuscitative ECMO (Shinar and Badulak). The 2023 AAN/AAP/CNS/SCCM consensus guideline on death by neurologic criteria, which for the first time addresses determination during extracorporeal support.
External evidence: the mode-of-death literature (four cohorts, one registry analysis, one SAVE-J II secondary analysis totalling 1,660 patients), the apnoea-testing literature on ECMO (one systematic review, one multicentre series, one Harlequin-specific study, one 2026 national-directive analysis), the ECMO decision-making ethnography, the ECPR organ-donation literature (one 307-patient cohort, one systematic review of 254 donors, one integrated ECPR-and-donation programme), and the normothermic regional perfusion literature including its ethical scoping review.
Everything external in this chapter comes from a structured abstract. Nothing was retrieved in full text. Every database row seeded from this chapter is marked Verified = No.
What this chapter covers — and what it does not
This is the last chapter of Part IV, and it is the only one about ending. Chapter 23 produced a probability. This chapter is about what is then done with it — and about three things that happen in no other ECMO chapter, because they happen almost only after ECPR.
This chapter owns | Owned elsewhere — cite, do not re-argue |
How ECPR runs actually end: the four modes of death and their four different clocks | Futility and non-beneficial ECMO as a concept — Chapter 74 |
Why a withdrawal decision is structurally hard to make on a circuit — the escalation trap | The procedure of withdrawing ECMO, and dying well on a circuit — Chapter 75 |
The ECPR consent asymmetry: started without consent, stopped only with it | Symptom control, family support and bereavement — Chapter 76 |
Determination of death by neurologic criteria on ECMO — the apnoea test the circuit sabotages | Resource allocation, and whether a unit's criteria are a dial — Chapter 77; Chapter 20 §20.9 |
Organ donation after failed ECPR, and normothermic regional perfusion | ECMO as a bridge to transplantation for the patient on it — Chapter 72, Chapter 82 |
Recognising the ECPR run that cannot be maintained at all | Weaning and decannulation of a recovering patient — Chapter 18; the checklist — Chapter 91 |
Prognostic input to a decision | How that prognosis is generated — Chapter 23; cerebral monitoring — Chapter 28 |
The ECPR bridge to nowhere specifically | Bridge-to-decision as a strategy — Chapter 70; goals of ECMO — Chapter 69 |
Pitfall · Three subjects in this chapter have no other home in this book, and that is deliberate
A tracker query run while preparing this chapter established that no chapter in Understanding ECMO owns determination of death by neurologic criteria on a circuit, organ donation after ECMO, or normothermic regional perfusion. They are placed here rather than in Part XI for a physiological reason, not an editorial one:
- Brain death is an ECPR disease. Roughly one ECPR patient in six dies brain dead, against a small fraction of other ECMO patients — the difference being the arrest, not the circuit.
- The apnoea test is sabotaged by the sweep gas, which is a circuit problem and belongs beside the circuit.
- Normothermic regional perfusion is, mechanically, an ECPR circuit with the aortic arch vessels tied off. A book that explains ECPR and then sends the reader elsewhere to learn what the same machine does after death has split one subject in two.
Chapters 74 to 77 own the ethics. This chapter owns the physiology and the process.
24.1 How ECPR runs actually end
Almost every clinician's mental model of ECPR failure is neurological: the patient survives the arrest, the brain does not, and the family is asked to agree to stop. That model describes a minority of deaths, and it describes the slowest of them.
Evidence · Low certainty · What actually kills ECPR patients
Four cohorts, adjudicated independently, converge.
- 274 consecutive ECPR cases, single centre over nine years; hospital survival 25.9%. Of 203 deaths: therapy-resistant shock 105 (51.7%), anoxic brain injury 69 (34.0%). Patients dying of shock were older (63.2 against 54.3 years), more often in-hospital arrests (64.4% against 29.9%) and had shorter times to ECMO (52.3 against 69.3 minutes) — all P below 0.01.
- 210 ECPR patients from a prospective registry, causes adjudicated by two independent investigators against a predefined framework (interrater agreement 0.81). Of 152 deaths: refractory shock 75 (49.3%), neurological injury 69 (45.3%), rearrest 7 (4.6%), comorbidity 1 (0.6%).
- 147 ECPR runs for refractory in-hospital arrest: 89 (60.5%) died during mechanical circulatory support — that is, on the machine. Fifty-six (38.1%) were weaned; 24 (16.3%) of those died before discharge. Overall survival 22.4%, with 20.4% reaching a good neurological outcome.
- 307 refractory out-of-hospital arrests treated with ECPR in one metropolitan centre: 256 (83%) died in hospital, 33% of them from brain death.
Two conclusions hold across all four. Refractory shock kills at least as many ECPR patients as the brain does. And the majority of ECPR deaths occur with the circuit still running, which means somebody must decide to stop it.
The mode of death is not merely a label. Each mode has its own clock, and the single most useful table in this literature is the one that separates them.
Mode of death | Median survival after cannulation | What it looks like at the bedside |
Refractory shock | 0 days (IQR 0–2) | Flow cannot be maintained, lactate will not clear, vasopressor requirement climbs against full support. The decision is often made by the physiology rather than by anyone in the room |
Brain death | 2 days (IQR 1–4) | Brainstem reflexes disappear early and completely. The question becomes one of determination, not prediction (§24.5) |
Non-neurological withdrawal | 3 days (IQR 1–8) | Multiorgan failure, uncontrollable bleeding, an unreconstructable heart with no destination |
Neurological withdrawal without brain death | 9 days (IQR 2–14) | The commonest single modality. The patient is not dead, will not recover, and the circuit is working perfectly |
Clinical pearl · The four clocks tell you which conversation to prepare
In the 95-patient series that produced those figures, 51 (53.7%) had died by day 28, and neurological withdrawal without brain death was both the commonest modality and the latest. Refractory shock was the earliest, with a median survival of zero days.
The practical consequence is a triage of conversations rather than of patients:
- Day 0 to 1 — prepare the family for death from shock. This is the likeliest death in the first 24 hours and it will not wait for a family meeting. Say so before it happens.
- Day 1 to 4 — prepare for the possibility of brain death, which is a determination and not a negotiation, and which requires a process that takes days to arrange if it has not been started (§24.5).
- Day 2 onwards — the long conversation begins, because the modality that will account for most of the remaining deaths declares itself slowly and is the only one in which anybody has a real choice.
A unit that runs only one kind of end-of-life conversation is prepared for the least common and latest of the four.
Danger · Most withdrawal decisions are made before Chapter 23's assessment can begin
A secondary analysis of 1,660 out-of-hospital ECPR patients found that 510 (30.7%) had a decision to withhold or withdraw life-sustaining therapy. The number of such decisions was highest on the first day, and the median was two days after intensive care admission.
The reasons given were: perceived unfavourable neurological prognosis 300/510 (58.8%); perceived unfavourable cardiac or pulmonary prognosis 105/510 (20.5%); inability to maintain extracorporeal support 71/510 (13.9%); complications 10 (1.9%); exacerbation of pre-existing comorbidity 7 (1.3%). Thirty-day survival was 36/506 (7.1%) with a withdrawal decision against 386/1140 (33.8%) without.
Now set that beside Chapter 23 §23.5 and §23.9, whose central practical conclusion was that the prognostic clock starts when sedation clears, which on a circuit is later than 72 hours — and that at 72 hours in the largest prospective algorithm cohort, half of all patients were still indeterminate.
The commonest single reason for stopping ECPR is a neurological prognosis, and the commonest day on which it is invoked is the first. Nothing in Chapter 23 supports a neurological prognosis on day one in a sedated patient on a circuit. Either these decisions are being made on grounds other than the ones recorded, or they are being made on grounds the evidence does not support. Both possibilities are serious, and the registry cannot distinguish them.
Physiology · Why refractory shock, not the brain, is the early killer — and why the machine hides it
Venoarterial ECMO delivers flow. It does not deliver a working heart, and it does not reverse the metabolic debt already incurred. After a prolonged low-flow interval, the patient arrives on the circuit with a vasoplegic, acidotic, capillary-leaking circulation, a stunned myocardium and, frequently, an ischaemic gut. The pump conceals the first consequence of that state — a falling blood pressure — while doing nothing about the state itself.
This is why the early trajectory must be read from variables the pump does not set: lactate clearance, vasopressor requirement at a fixed flow, mixed venous saturation trend, and the ability to maintain flow without progressive volume loading. A mean arterial pressure of 70 mmHg on a centrifugal pump at maximal vasopressor support is not reassurance; it is the machine answering a question nobody asked.
Chapter 13 owns the haemodynamics and Chapter 22 owns the first 24 hours. The point here is narrower: the mode of death that arrives first is the one the circuit is best at masking.
24.2 The decision that is never made
This is the organising idea of the chapter, and like Chapter 23's it is structural rather than clinical.
Every account of how ECMO should be stopped assumes that there is a moment at which stopping is decided. Somebody weighs the prognosis, somebody speaks to the family, a decision is reached. The one study that went and watched found that this moment does not reliably exist.
Evidence · Moderate certainty · What actually happens is escalation, not decision
A focused ethnography in two academic cardiothoracic intensive care units: 380 hours of direct observation, 34 weekly interviews with the families of 20 ECMO patients, and 13 interviews with unit clinicians.
The finding, in the authors' own words: "Following ECMO initiation, treatment was escalated as complications mounted until the patient either could be decannulated or interventional options were exhausted. Families were well-informed about treatment and prognosis but played minimal roles in shaping the trajectory of care."
And the conclusion that matters most: "Discussion between clinicians and families about prognosis and goals was frequent but did not occasion decision-making moments."
The authors offer this to explain something otherwise puzzling — why communication interventions designed to strengthen surrogate participation have had so little measurable effect on ECMO trajectories. They have little effect because they are aimed at a decision point that the structure of care does not produce.
Physiology · Why the circuit generates escalation rather than decision
The escalation trap is not a failure of character. It is built into what the machine does.
- Every complication has a next intervention. The left ventricle distends, so it is vented (Chapter 15). The limb goes dark, so a distal perfusion cannula is placed (Chapter 17). The kidneys fail, so a haemofilter is spliced in (Chapter 55). The oxygenator clots, so the circuit is changed (Chapter 32). At each step there is something to do, the something is technically satisfying, and doing it is never the moment to ask whether any of it is going anywhere.
- The circuit removes the physiological signal that normally ends treatment. In conventional critical care, an irrecoverable patient declares themselves by dying despite maximal therapy. On venoarterial ECMO the patient cannot declare themselves at all, because the machine is doing the declaring. The natural end point has been engineered away, and nothing has been engineered to replace it.
- The trajectory is set upstream. The ethnography's authors conclude that "a more comprehensive understanding of upstream factors that predispose courses of critical care is needed." In ECPR the upstream factor is the cannulation decision itself, made in minutes under compressions — which is exactly Chapter 20's subject, and exactly why Chapter 20 §20.8 argued for a refusal log.
The bedside consequence: decision points on a circuit do not occur. They have to be manufactured, in advance, and written down. §24.11 sets out how.
Clinical pearl · A decision point is a date, a question and a named person — anything less is a conversation
The difference between a family meeting and a decision point is that a decision point specifies, in the notes, before it arrives:
- when it will happen (a date and time, not "in a few days");
- what question it will answer (not "how is he doing" but "is there a destination, and has the neurological assessment become possible");
- what information must be available by then (the sedation-cleared examination, the imaging, the serial biomarker, the transplant or device assessment);
- who will be in the room, including who from outside the ECMO team;
- and what happens if the answer is no — stated in advance, because a plan made after the answer is a justification.
A time-limited trial is simply a decision point with the clock started at initiation. In one international survey of ECMO clinicians, a timeline for withdrawal was not adhered to in 87% of cases. The instrument is known; it is not used.
24.3 The asymmetry: begun without consent, stopped only with it
ECPR inverts the consent structure of essentially every other treatment in medicine, and the inversion is the source of most of the distress this chapter describes.
Stage | Who decides | Time available |
Conventional resuscitation is begun | Nobody consents. Implied consent, universally accepted | Seconds |
Conventional resuscitation is stopped | The treating team. Termination of resuscitation is a medical decision with limited family input, and this is not controversial | Minutes |
ECPR is begun | Nobody consents. The patient is in arrest; there is no time and no capacity | Minutes |
ECPR is stopped | Unclear, and jurisdiction-dependent. In some legal systems, withdrawal of life-sustaining treatment requires the consent of a substitute decision-maker, irrespective of the treating physician's judgement of futility | Days to weeks |
Danger · The moment perfusion moves from hands to pump, the decision moves from the team to the family — and nobody announces it
The clearest published statement of this problem comes from a Canadian ECPR programme. Their eligibility was tight — age 65 or under, previously healthy, witnessed non-traumatic arrest, early compressions, presumed cardiac, toxic or hypothermic cause — and the authors note that ECPR duration is typically under two days, with survival around 15%.
They then describe treating a young patient who sustained a severe anoxic brain injury. The family was unwilling to agree to withdrawal of mechanical circulatory support despite the futility of ongoing treatment, causing considerable moral distress for the care team. The relevant law in that jurisdiction, following a Supreme Court decision, mandates the consent of a substitute decision-maker before withdrawal of life-sustaining treatment even where the treating physician judges further treatment futile.
The authors state the structural point precisely: while manual cardiopulmonary resuscitation is a time-limited intervention whose termination is the team's responsibility, once the role of systemic perfusion shifts from chest compressions to ECMO flows, the role of the family in decisions about continuation becomes less clear.
Nothing clinical changes at that moment. Everything about who is allowed to stop does. No consent conversation marks the transition, because there was no consent conversation at all.
Evidence · Low certainty · What consent for ECMO actually looks like even when there is time
A cross-sectional survey of 14 departments across seven United States hospitals found the following.
- Mean time spent consenting for ECMO: 7.5 minutes (95% CI 5–10) for unstable patients, against 20 minutes (95% CI 15–30) for stable patients (P = 0.0001).
- Only 29% of respondents reported that patients or surrogates always possess informed consent for ECMO.
- 92% of departments have absolute exclusion criteria — most commonly older age (43%, with cut-offs from 60 to 75 years), active malignancy (36%) and elevated body mass index (29%).
- 29% of departments do not always offer the option of withdrawing ECMO to patients or surrogates.
- And for the defining scenario — a patient who cannot be liberated from ECMO and is ineligible for heart or lung transplantation — 36% of departments would recommend removal from ECMO and 64% would continue.
Seven and a half minutes is the figure for a patient who is merely unstable. In ECPR it is zero. And the 36-against-64 split is the bridge-to-nowhere disagreement quantified: on the single question this chapter exists to address, ECMO departments are divided roughly one to two.
Clinical pearl · Consent for ECPR is not a signature; it is a sequence of conversations that begins after the machine is running
A 2026 framework for starting ECMO makes two moves this chapter adopts.
- It distinguishes offering ECMO from initiating it. Whether support is technically feasible and medically indicated is a clinical determination the team makes; whether it is initiated then incorporates patient and family values. In ECPR the first happens in minutes and the second cannot happen at all — so the second is owed retrospectively, and owed promptly.
- It insists that informed consent be extended beyond a single time point into frequent, iterative conversations across the whole course, emphasising ECMO as a temporary, goal-directed intervention.
The practical form for ECPR: the first family conversation is not a consent discussion but a notification, and it should say so. "We could not ask you, and we could not ask him. We started because it was the only thing that could work. Now we have to decide together whether to continue, and I will come back to that question with you on a fixed schedule." That sentence creates the decision point that §24.2 showed does not otherwise arise.
Controversy 1 — If nobody could consent to starting, who is entitled to insist on continuing?
The case that the family's authority is the same as in any other withdrawal. Withdrawing and withholding life-sustaining treatment are widely held to be ethically and legally equivalent. Once a treatment is running, stopping it requires the same justification as any other withdrawal, and in some jurisdictions the same consent. A patient on ECMO is a patient on life support; nothing about the pump changes the standing of the person who speaks for them. On this reading, the distress described above is the ordinary distress of disagreement, and the answer is better communication and, where needed, the courts.
The case that ECPR is different in kind. ECPR was begun under implied consent for a resuscitation, in a patient whose wishes were entirely unknown, by a team exercising the same authority under which they would have stopped compressions ten minutes later. If the team retained the authority to terminate resuscitation at minute 40, it is not obvious how they lost it at minute 41 because a cannula went in. The commentary quoted above makes exactly this point. And a widely cited bioethical exchange argues that the equivalence thesis itself is doing work it cannot bear: if withholding and withdrawing really are identical, then a unit that would withhold ECMO from a patient it knew could not reach a destination is obliged to withdraw immediately from every patient who unexpectedly lands on a bridge to nowhere — a conclusion almost nobody accepts, and which suggests the premise is too strong.
Where this book lands. The asymmetry is real, it is not resolvable at the bedside, and pretending otherwise is what produces the moral distress. Four positions follow.
- The authority question must be settled before the programme starts, not during a case. It is a question of local law and institutional policy, and the answer differs between jurisdictions. A unit that does not know its own answer will discover it during its worst week.
- Where the family's consent is legally required to stop, the time to shape that consent is day one, not day nine — which is precisely what §24.2's manufactured decision points are for.
- The disagreement is not evidence of a bad family. It is the predictable consequence of a treatment begun without them, and framing it as obstruction guarantees the confrontation.
- This chapter does not adjudicate futility. That is Chapter 74, and the procedural machinery — ethics consultation, dispute-resolution forums, institutional accountability — belongs to Chapter 74 and Chapter 77.
What would settle it. Nothing empirical; it is a question of law and value, and the law differs. What can be settled empirically is whether structured early decision points reduce late disagreement. No such trial exists in ECPR.
[VERIFICATION REQUIRED] — the legal position described is reported second-hand through a commentary abstract; readers must confirm the law in their own jurisdiction rather than rely on this paragraph.
24.4 The run that cannot be maintained
Before any question of prognosis arises, a proportion of ECPR runs simply fail as runs. This is the earliest failure mode, the least discussed, and the one in which the decision is least contested — because the physiology makes it.
Failure | What is seen | Where it is owned |
Cannulation never succeeds | Access cannot be obtained, or is obtained in the wrong vessel, or the vessel is destroyed. In one integrated programme, cannulation was successful in 49 of 58 attempts (84%) — meaning roughly one in six did not get on the machine at all | Chapter 21 owns the technique. This chapter owns what follows: this is a termination of resuscitation, and it should be conducted and documented as one |
Flow cannot be achieved or maintained | Chatter, repeated suction events, a circuit that will not deliver target flow despite volume and repositioning. Usually tamponade, massive haemorrhage, profound vasoplegia with an empty circulation, or a cannula in the wrong place | Chapter 34 owns cannula problems; Chapter 14 owns initial management. The decision that no further flow is obtainable is this chapter's |
Support cannot be sustained | The recorded reason for 71 of 510 withdrawal decisions (13.9%) in the largest series was explicitly "inability to maintain extracorporeal cardiopulmonary support" | This chapter. It is the only withdrawal reason in that series that is a statement of fact rather than of judgement |
Refractory shock on full support | Rising lactate and vasopressor requirement at adequate flow, with no reconstructable cause. Median survival zero days (IQR 0–2) | This chapter, with the haemodynamics in Chapter 13 and the unloading decisions in Chapter 15 |
Catastrophic structural pathology found after cannulation | Aortic dissection, unsurvivable haemorrhage, a myocardial rupture. The cause of arrest turns out to be one that ECPR cannot bridge | Chapter 22 §22.8 owns the diagnostic workup. This chapter owns the conclusion drawn from it |
Clinical pearl · Name the failure mode out loud, because the four are not the same conversation
"He is not doing well" collapses five different situations into one, and the family hears only the fifth. Compare:
- "We cannot get the machine to run. There is nothing more to try." This is a termination of resuscitation. It is the team's decision, it is defensible, and it should be stated as a fact.
- "The machine is running perfectly and his circulation is still failing." Refractory shock. The prognosis is measured in hours and the family should be brought in now, not tomorrow.
- "The machine is running, his heart is not recovering, and there is no operation or transplant that would help." A bridge to nowhere (§24.8). Days to weeks, and this is where §24.2's decision points earn their keep.
- "The machine is running, his heart is recovering, and we are worried about his brain." Chapter 23's question, and the clock has probably not started.
- "He has no brain function at all, and we now need to confirm that formally." A determination, not a decision (§24.5).
Only the third and fourth are negotiations. Treating all five as negotiations is how a team ends up asking permission for something that is not a choice — and treating the third as a fact is how a team ends up in court.
24.5 Determination of death by neurologic criteria on a circuit
Roughly one ECPR patient in six dies brain dead. In the largest ECPR synthesis, brain death accounted for 17% (12–23%) of patients; in the Milan cohort, 33% of in-hospital deaths were from brain death. This is therefore not a rare curiosity but one of the four commonest ways an ECPR run ends — and it is the one the circuit most directly obstructs.
Physiology · Why the sweep gas breaks the apnoea test
Determination of death by neurologic criteria requires three findings: coma, absent brainstem reflexes, and apnoea — demonstrated by disconnecting ventilation and showing that no respiratory effort occurs despite a carbon dioxide tension high enough to drive the medullary chemoreceptors maximally.
On ECMO the test's premise is removed. Carbon dioxide clearance during extracorporeal support is governed almost entirely by sweep gas flow across the membrane (Chapter 6, Chapter 27). Disconnecting the ventilator therefore does not cause carbon dioxide to rise: the membrane goes on clearing it as efficiently as before. The stimulus the test depends on is being continuously removed by the machine that is keeping the patient alive.
The corollary defines every workable protocol: to perform an apnoea test on a circuit, the sweep gas must be reduced — not the blood flow. Sweep governs carbon dioxide; blood flow governs oxygen delivery and haemodynamics. Cutting blood flow to raise carbon dioxide would trade a confirmed test for a hypoxic, hypotensive patient, which is how the test acquired its reputation for danger.
On venoarterial support there is a second problem, and it is Chapter 16's problem wearing a different hat: with competing native and circuit circulations, the carbon dioxide the brainstem sees is not necessarily the carbon dioxide the sampling site reports. The test is only as valid as the site the blood came from (§24.6).
Evidence · Low certainty · How often the test is attempted, and how often it works
The systematic review. Twenty-two studies, 177 patients on ECMO undergoing brain-death assessment. Only 88 (50%) underwent an apnoea test at all. Where the method was reported, it was most often decreasing the sweep flow (14 studies, n = 42, 48%), then providing carbon dioxide through the ventilator (2 studies, n = 6, 7%), then through the oxygenator (1 study, n = 1, 1%); the method was not reported at all for 39 patients (44%). The test was non-confirmatory in 19 of 88 (22%) because of haemodynamic instability, hypoxia, insufficient carbon dioxide rise, or unreliability. A total of 157 ancillary tests were performed — electroencephalography 62%, computed tomography angiography 22%, transcranial Doppler 6%, nuclear cerebral blood flow 5%, cerebral angiography 4%. Forty-seven patients (53% of those with a confirmatory apnoea test) underwent ancillary testing anyway, and only 21 patients — 12% of the whole cohort — were declared brain dead on a confirmatory apnoea test alone.
The counter-example. A single-centre series of 169 consecutive brain-dead patients, 25 of them on ECMO — the largest ECMO cohort reported — using a technique combining applied positive end-expiratory pressure with subsequent recruitment. No apnoea test was aborted and no severe complication occurred; the test was completed in every patient. Fluid boluses were needed in under 10% and vasoactive escalation in under 3%. Severe hypoxia (arterial oxygen tension under 40 mmHg) occurred in 2.4% of non-ECMO and 8% of ECMO tests (P = 0.063), and was commoner in patients hypoxic at baseline (11.1% against 4.8%, P = 0.002).
The multicentre protocol. Thirty-one patients determined brain dead on ECMO (28 venoarterial, 3 venovenous): median sweep gas 3.5 L/min (3.0–4.0) before the test and 0.5 L/min (0.5–1) during it (P below 0.01), with blood flow held constant. Carbon dioxide and pH met criteria in both the right radial artery and the post-oxygenator circuit. All patients also had at least two positive ancillary tests. Twenty-five (81%) went on to donate organs.
Read together: the test fails about one time in five in pooled practice and essentially never in the hands of a unit with a protocol. The variable is not the patient. It is whether a protocol exists.
Clinical pearl · The elements every published ECMO apnoea protocol shares
Individual protocols differ; the following is what they agree on, and it is enough to build a local one. This is a summary of published practice, not a guideline, and it must be reconciled with national determination criteria before use.
- Pre-oxygenate through both lungs and circuit. Fraction of inspired oxygen 1.0 on the ventilator and on the sweep gas, before anything is reduced.
- Reduce the sweep gas. Do not reduce the blood flow. Reported targets cluster at 0.5 to 1 L/min, with the multicentre series using a median of 0.5. One venovenous protocol went as low as 200 mL/min. Stepwise reduction is described in paediatric practice.
- Keep the lungs open. The technique with the best completion record applied positive end-expiratory pressure throughout and recruited afterwards. Apnoeic oxygenation through a collapsing lung is where the hypoxia comes from.
- Sample from the right radial artery on venoarterial support, and understand what the other sites mean (§24.6).
- If carbon dioxide will not rise enough, add carbon dioxide rather than abandoning the test. Exogenous carbon dioxide into the circuit, or a carbogen mixture, is described; in one five-patient series the carbogen method hit the target carbon dioxide in all five with no adverse events.
- If oxygenation fails, increase the sweep back up and the blood flow, then treat and retry — an aborted test is not a negative test.
- Have the ECMO team present. One 2026 national directive now requires that apnoea testing in potential donors on extracorporeal support be accompanied by physicians familiar with the circuit.
24.6 Where the blood came from, and what to do when the test cannot be done
Two technical questions decide whether a determination on a circuit stands up: which arterial sample counts, and what may substitute for the apnoea test when it fails.
Evidence · Low certainty · The sampling site matters, and it matters more as the test proceeds
A systematic study in 27 venoarterial patients undergoing brain-death determination compared two ways of accounting for competing circulations during apnoea testing.
- Variant 1 — simultaneous samples from the right and left radial arteries. Median carbon dioxide difference 0.90 mmHg (95% CI 0.7–1.3) at the start of the test, and essentially unchanged at 1.1 mmHg (0.9–1.8) at the end.
- Variant 2 — simultaneous samples from the right radial artery and the post-oxygenator circuit. Median difference 3.3 mmHg (1.5–6.0) at the start (P = 0.001 against variant 1), rising to 9.9 mmHg (3.5–19.2) at the end (P = 0.002).
The authors conclude that simultaneous analysis from right and left distal arterial lines is the method of choice, because it reduces the risk of adverse effects such as severe respiratory acidosis while the test is being performed.
The finding is Chapter 16's Harlequin physiology reappearing at the worst possible moment. The post-oxygenator sample is not a proxy for what the brainstem is exposed to, and the discrepancy is smallest exactly when it least matters and largest exactly when the test is being read.
Danger · A right radial sample alone can understate the true stimulus, and a post-oxygenator sample can overstate it
On peripheral venoarterial support with any native ejection, the right radial artery is perfused predominantly by blood from the native heart and lungs, and the lower body predominantly by the circuit. During an apnoea test the native lungs are not ventilated, so carbon dioxide rises in the native circulation — which is the circulation that reaches the brain. That is the point of the test.
Two errors follow.
- Reading only the post-oxygenator sample reports the carbon dioxide the membrane has just cleared, which by the end of the test may be nearly 10 mmHg lower than the value at the right radial artery. A test declared to have failed on that basis may in fact have succeeded.
- Assuming the right radial sample is the whole story ignores that the mixing point moves with native ejection and with flow. The published solution is not a calculation but a measurement: take both radial samples at once and know how far apart they are.
And if the mixing point is unknown, the test result is unknown. Chapter 16 owns the physiology; this section owns its one irreversible application.
Controversy 2 — When the apnoea test cannot be completed, what may stand in its place on a circuit?
The case for ancillary testing as a routine partner. The apnoea test is non-confirmatory in roughly one attempt in five in pooled practice, and half of ECMO patients assessed for brain death never underwent it at all. In real practice ancillary testing is already the norm: 157 ancillary tests were performed in a cohort of 177 patients, and only 12% were declared on a confirmatory apnoea test alone. In a two-strategy multicentre protocol, requiring at least two positive ancillary tests raised the completion rate of determination and, with it, the proportion of families able to proceed to donation.
The case against relying on ancillary tests, and it is newly stronger. Ancillary tests were validated in patients with a native circulation. A 2026 update to one national directive removed cerebral blood flow-based methods from the permitted ancillary tests in potential donors on venoarterial ECMO, on the explicit grounds that they are insufficiently validated in that setting. In the single-centre analysis published alongside that change, cerebral circulatory arrest had been demonstrated in three cases before the update; afterwards, a flat-line electroencephalogram was obtained in four, and in one case the electroencephalogram was inconclusive because of artefact. An earlier series makes the physiological objection directly: cerebral flow studies may fail to document absent arterial flow because of the ischaemic nature of the injury itself. And in a tertiary-centre review, the seven patients with absent brainstem reflexes who never completed determination were precisely those in whom ancillary tests were inconsistent with the examination, with imaging, or with each other.
Where this book lands. Attempt the apnoea test. It is more reliable on a circuit than its reputation, and ancillary testing on a circuit is less reliable than its reputation.
- Build the protocol before you need it. The difference between a 22% failure rate and a 0% failure rate in the published series is not patient selection; it is whether a written technique existed.
- Treat an aborted test as uninterpretable, not negative. Restore sweep and flow, correct the problem, repeat.
- Use ancillary testing to support a determination, not to rescue one. Where the examination and imaging point the same way, a concordant ancillary test adds confidence. Where they do not, an ancillary test is not the tie-breaker.
- Know which ancillary tests your own jurisdiction still permits on venoarterial support — this is an area where national directives are actively changing, and the direction of change is restrictive.
- Electroencephalography on a circuit carries the artefact problem of Chapter 23 §23.3 and Chapter 28. An inconclusive recording is an inconclusive recording.
What would settle it. Validation of ancillary modalities specifically in patients on venoarterial support, with the mixing point characterised. None exists, which is the stated reason the directive above was tightened.
[VERIFICATION REQUIRED] — every source in this controversy is known through a structured abstract, and the determination criteria that apply are national. Nothing in this section substitutes for local law.
24.7 The withdrawal that pre-empts the determination
There is a sequence error in this field that is rarely named, and it costs both families and recipients.
Danger · Withdrawing before completing the determination converts a death into a decision
Two single-centre series describe the same thing.
- In a tertiary centre applying standardised neuromonitoring, eight ECMO patients (2.7%) met criteria for death by neurologic criteria, six of them with an apnoea test. A further seven patients (2.3%) had absent brainstem reflexes but never completed determination, because life-sustaining treatment was withdrawn before the evaluation was finished. In those seven the apnoea test was never performed.
- In an earlier review of 87 adults on ECMO, three (3.4%) lost all brainstem reflexes. The apnoea test was not performed in any of them because it was deemed "difficult," leading to withdrawal of ECMO and intensive care instead.
In both series, roughly as many patients had treatment withdrawn for a brain that had probably already died as were declared dead. The consequences are not academic.
- The family is asked to authorise a death that had already occurred, and carries that decision afterwards.
- Organ donation becomes far less likely, and where it occurs it follows a harder pathway.
- The record shows a withdrawal, so the unit's own audit understates how many of its ECPR deaths were neurological — which is one reason registry mode-of-death data and single-centre mode-of-death data disagree (§24.1).
The rule that follows is simple and is the most actionable sentence in this chapter: in a patient with absent brainstem reflexes, complete the determination before discussing withdrawal. The determination is not a formality appended to a decision. It is the alternative to one.
Clinical pearl · The determination takes longer to arrange than to perform — so start arranging it at the first absent reflex
The apnoea test itself takes minutes. Everything around it does not: a second examiner, a neurophysiologist for ancillary testing, the perfusionist who will manage the sweep, a repeat examination at the interval local criteria require, and — if donation may follow — a conversation with the donation service that must be kept separate from the determination (§24.9).
Trigger the process on the finding, not on the conclusion. The moment a comatose ECPR patient is documented to have absent pupillary and corneal reflexes off sedation, the determination pathway should be activated in parallel with the prognostic pathway of Chapter 23 — even though most such patients will not turn out to be brain dead. The cost of activating it and stopping is an hour of somebody's time. The cost of not activating it is the sequence error above.
24.8 The ECPR bridge to nowhere
A bridge to nowhere is a patient supported by a therapy from which there is no exit — who would die immediately if it stopped, and who has no route to recovery, transplantation or a durable device. Chapter 70 owns bridge-to-decision as a strategy and Chapter 74 owns futility as a concept. What belongs here is why ECPR produces this situation more readily than any other indication, and what is specific about it.
Why ECPR is the highest-risk route to a bridge to nowhere | Consequence |
The destination was never assessed. In elective venoarterial ECMO the team asks "recovery, transplant or device?" before cannulating. In ECPR there is no time to ask anything | The exit strategy is determined after the fact, and sometimes there is none. Chapter 20 §20.5 selects for arrest characteristics, not for destination characteristics |
The patient's wishes are entirely unknown, and often remain unknown — the arrest may have been the first medical event of their life | There is no prior expression of values to appeal to, in either direction |
Two organs can fail the exit independently. The heart may not recover, or the brain may not, and a patient can be barred from a device or a transplant by the second while being technically supportable by the first | The commonest ECPR bridge to nowhere is a recovering circulation in a devastated brain — which is also the modality with the longest median survival (§24.1) |
Transplant and device candidacy are frequently decided by people who were not in the room at cannulation. In one survey, 55% of ECMO clinicians reported that transplant surgeons are not involved in initiation decisions | A destination assumed at 3 a.m. can be withdrawn at the first formal assessment, converting a bridge into a bridge to nowhere without any change in the patient |
Clinical pearl · Ask the destination question on day one, in writing, and put a name against each answer
The general literature on temporary mechanical circulatory support converges on one recommendation that transfers cleanly to ECPR: continuously evaluate every patient against all four possible exits — native cardiopulmonary recovery, transplantation, a durable device, and palliative discontinuation — and do it from the beginning rather than when weaning fails.
The ECPR-specific form is a four-line entry in the notes within the first 24 hours:
- Recovery: plausible or not, and on what evidence (echocardiographic trend, the cause of arrest, whether it was treated).
- Transplant: who will assess, and by when. If nobody has been asked, write that.
- Durable device: the same.
- If none of the above: what the plan is, and the date of the decision point at which it becomes the plan.
The fourth line is the one that is almost never written, and it is the only one that prevents a bridge to nowhere. Writing it on day one costs nothing; discovering on day nine that nobody ever wrote it costs a family a fortnight.
Pitfall · "Prolonged support means futility" is not true in general, and the general truth does not transfer to ECPR
Long venovenous ECMO runs were once regarded as self-evidently futile, and that belief has been substantially overturned: prolonged runs for respiratory failure can have survival not very different from short ones, and one commentary on the newer data concluded that a cut-off of 21 days may be too short to declare failure, with family expectations more honestly set at six to twelve weeks.
Do not import that reassurance into ECPR. The venovenous argument works because the lung can take months to heal and the patient can be bridged to transplantation. After ECPR the two clocks are different: the myocardium either recovers within days or declares a need for a device or a transplant, and the brain's window is measured in days to weeks, not months. A prolonged ECPR run is not automatically futile either, but it is not reassured by the venovenous data.
The honest statement is that neither duration proves anything on its own. What proves something is whether a destination exists.
24.9 Organ donation after failed ECPR
This is the part of ECPR that almost no consent conversation mentions and almost no programme counts, and it is numerically the largest single output of a mature ECPR service.
Evidence · Low certainty · How many people an ECPR programme actually helps
The cohort. 307 adults with refractory out-of-hospital arrest treated with ECPR at one metropolitan cardiac arrest centre over nine years. Ninety-five per cent witnessed, 66% shockable, median low-flow 70 minutes (IQR 58–81). 256 (83%) died in hospital, 33% from brain death.
The donation. Fifty-eight patients (19%) donated at least one solid organ — 53 (17%) after determination of brain death and 5 (1.6%) after determination of circulatory death — contributing 167 solid organs, a median of 3.0 (IQR 2.5–4.0) organs per donor.
The arithmetic. Twenty-nine patients survived with a favourable neurological outcome. One hundred and sixty-seven organs went to potential recipients. A total of 196 individuals possibly benefited from the programme, and a composite outcome of survival-with-good-neurology or donation of at least one organ was reached by 87 patients (28%).
The independent check. A systematic review of organ procurement after ECPR identified 17 studies covering 254 patients with unfavourable outcomes, from whom 689 grafts were reported and 469 organ recipients were transplanted. The ratio is the same shape from a completely different denominator.
A programme that counts only its survivors is reporting a minority of the people its work reached.
Danger · The selection dial has a second output, and it moves in the opposite direction
Chapter 20 §20.9 established that a unit's inclusion criteria are a dial: tighten them and the survival rate rises, because the denominator changes. The Milan cohort measured what else moves when that dial is turned.
Solid organ donation fell from 19% of the whole cohort to 16% when only patients with a low-flow interval under 60 minutes were considered, and to 11% when the criteria were tightened further to low-flow under 60 minutes with an initial shockable rhythm.
Read that carefully. The selection criteria that make an ECPR programme look better — and that genuinely do improve the survival of the patients it cannulates — simultaneously reduce the number of organs it yields, by roughly half. The patients excluded by tightening are disproportionately the ones who would have died brain dead and donated.
This is recorded here because it is true, and it is immediately followed by the warning in the controversy below, because the inference most readily drawn from it is one this book rejects.
Controversy 3 — Should the possibility of organ donation influence any part of the ECPR pathway?
The case that it should, at the programme level. Refusing to count donation does not make it stop happening; it merely makes it unaccounted for. A service that treats 307 patients, saves 29 and enables 167 grafts has done something that a survival rate of 9% does not describe. Health systems allocate ECPR resources on cost per outcome, and if the outcome ledger omits five-sixths of the beneficiaries the allocation is made on a false number — which is Chapter 77's argument arriving from an unexpected direction. Some systems go further and build an integrated pathway: in one programme, patients under 65 with witnessed non-asystolic arrests, a reversible cause and high-quality resuscitation lasting under 60 minutes were triaged to ECPR, and those who did not meet those criteria were considered instead for uncontrolled donation after circulatory determination of death following a ten-minute no-touch period — a design that yielded six of 18 ECPR patients (33%) surviving with full neurological recovery and 44 kidneys transplanted from the others.
The case that it should not, at the bedside. The moment a donation consideration can influence whether a particular patient is cannulated, or how long support continues, or how a prognostic conversation is framed, the programme has acquired a conflict of interest that no amount of good faith neutralises — and public trust in both ECPR and transplantation is the asset most easily destroyed and least easily rebuilt. A 2026 ECPR-specific ethical framework is unambiguous on the remedy, and its central element is strict separation between resuscitative and donation-related decision-making, supported by transparent prognostication, explicit safeguards around treatment intent, early ethics involvement and structured communication. A systematic review of the ethics of this exact scenario found that of 17 studies, only seven addressed ethical aspects at all, and emphasised that ECPR is initiated to save patients with a likely favourable neurological prognosis; donation is considered only secondarily, after an unfavourable outcome.
Where this book lands. Both, on different levels, with a wall between them.
- The donation yield belongs in the programme's annual accounting and nowhere near the cannulation decision. Chapter 20's criteria are set by who might survive. They are not adjusted upward or downward by who might donate.
- The separation must be structural, not attitudinal. Different people, different conversations, different times. The clinician who says the prognosis is hopeless must not be the clinician who raises donation, and must not be in the room when it is raised.
- The determination comes first and completely (§24.7). Donation follows a death; it never precedes or hastens one.
- Do not let the donation figure become a defence of a weak programme. A service with poor survival and good donation numbers has a survival problem, not a donation success.
- Tell families the whole arithmetic when they ask what ECPR achieves — including, honestly, that most of the people it helps are not the patient in front of them.
What would settle it. Whether structural separation actually preserves trust is an empirical question about public attitudes, and it has not been studied in ECPR specifically. The one adjacent data point is encouraging but narrow: in an ancillary study of the INCEPTION trial, 81.3% of 32 surviving patients and bereaved relatives supported waived and deferred consent procedures in emergency ECPR research, and valued being given information later, once the emotional burden had eased.
[VERIFICATION REQUIRED] — every source here is a structured abstract; the no-touch interval, donation criteria and permissible pathways are set nationally and differ substantially.
24.10 Normothermic regional perfusion — the same machine, the opposite purpose
An ECMO book has to explain this, because the device is the one the reader has spent twenty-three chapters learning.
Physiology · Normothermic regional perfusion is an ECPR circuit with the arch vessels tied off
In donation after circulatory death, organs suffer warm ischaemia between the cessation of circulation and cold preservation. Normothermic regional perfusion re-establishes warm, oxygenated, machine-driven flow to the donor organs in situ after death has been determined, so that they are reperfused, metabolically corrected and functionally assessable before recovery.
Mechanically it is familiar: a centrifugal pump, a membrane oxygenator, a heat exchanger at 37 °C, drainage and return cannulae. In thoracoabdominal perfusion the chest is opened, the aortic arch vessels are ligated, and the circuit is run centrally, with the heart weaned off the pump and assessed by echocardiography before procurement. In abdominal-only perfusion, a specific technique is used to avoid restoring circulation to the brain after the death determination.
The ligation is the entire ethical architecture expressed as a surgical step. It is what distinguishes a perfusion that preserves organs from a resuscitation that would undo the determination on which the whole procedure rests. The same hardware that Chapter 21 used to restore cerebral flow is here configured explicitly to prevent it.
What normothermic regional perfusion achieves | Evidence |
More organs per donor | In a national controlled-donation analysis, 3.3 organs transplanted per donor with regional perfusion against 2.6 without. Adjusted odds of the organ being transplanted rose roughly threefold for liver (P below 0.0001; 95% CI 2.20–4.29), 1.5-fold for kidney (P = 0.12) and 1.6-fold for pancreas (P = 0.0611) |
Better graft outcomes | Twelve-month liver graft survival superior, with a 51% lower risk-adjusted hazard of transplant failure (HR 0.494). Kidneys had a 35% lower chance of delayed graft function (OR 0.65, 0.465–0.901) and an expected 12-month estimated glomerular filtration rate 6.3 mL/min/1.73 m² better |
Thoracic organ recovery that was previously impossible | A single high-volume centre attempted heart recovery in 200 donors and accepted 176 (88%) using thoracoabdominal perfusion; recipient outcomes did not differ across four successive eras despite progressively more aggressive donor and recipient selection |
Correction of the donor's metabolic state before assessment | In one protocol series, lactate fell from 9.4 ± 1.5 to 5.3 ± 2.7 mmol/L and potassium from 6.5 ± 1.8 to 4.2 ± 0.4 mmol/L during perfusion (both P below 0.001), with all 18 hearts recovered and transplanted |
A workable route for the uncontrolled donor — the ECPR-relevant one | Reviews of uncontrolled donation after circulatory death report lower short-term graft survival than donation after brain death, but long-term outcomes comparable to controlled donation and to donation after brain death |
Evidence · Moderate certainty · Prolonged resuscitation does not, by itself, ruin the organs
This is the objection most often raised at the bedside after a failed ECPR run, and it has been tested directly.
A systematic review and meta-analysis searched to January 2025 and included 33 studies covering 72,994 donors, asking whether solid organs from donors who had undergone cardiopulmonary resuscitation did worse than organs from donors who had not.
- In 24 studies comparing brain-dead donors with and without preceding resuscitation, outcomes did not differ for any organ.
- Uncontrolled donation after circulatory death, compared with donation after brain death, showed lower long-term liver graft survival (OR 0.51, 0.32–0.83) and lower short-term but not long-term kidney survival (OR 0.64, 0.36–1.15).
- Two studies comparing controlled with uncontrolled circulatory-death kidneys found no difference in long-term survival (OR 0.73, 0.27–1.99).
The conclusion is the one that should be carried to the bedside: organs from donors who received cardiopulmonary resuscitation performed comparably at longest follow-up to organs from donors who did not. The pathway matters more than the resuscitation.
Pitfall · The ethics of regional perfusion are unsettled, and a chapter that presented them as settled would be misleading
A scoping review identified 71 documents engaging substantively with the ethics of normothermic regional perfusion, grouped into six overlapping debates: compatibility with the dead donor rule; risk of harm to the donor; consent requirements; risks to stakeholder trust; implications for justice; and the benefits to stakeholders. Its finding was explicit: "We found no agreement on the ethical permissibility of NRP."
A neurological analysis frames the three questions most sharply: whether the brain is sufficiently reperfused through collateral circulation to permit re-emergence of consciousness or pain perception; whether the resumption of cardiac activity nullifies the prior determination of death; and whether specific authorisation for this procedure is required rather than general donation consent.
A counter-argument in the medical-ethics literature rejects the framing itself, contending that the debate should not be a binary contest with the dead donor rule but should instead turn on whether the donor is harmed — and that, performed to protocol, regional perfusion does not violate the obligations owed to donors.
This book does not adjudicate. It records that the practice is expanding rapidly, that the outcome data are good, that the ethical question is genuinely open, and that a European consensus document exists setting out the minimum ethical, logistical and technical requirements for a programme. A unit contemplating this pathway should start from that document and from its own national framework, not from this table.
Clinical pearl · What the intensivist needs to know, as distinct from what the retrieval team needs to know
The intensivist's role in this pathway is narrow and almost entirely front-loaded:
- Complete the determination properly (§24.5, §24.7). Everything downstream depends on it and nothing downstream can repair it.
- Keep the separation (§24.9). The determination and the donation approach are different conversations by different people.
- Know your own system's pathway before a case arises — whether uncontrolled donation is permitted at all, what no-touch interval applies, whether regional perfusion is used, and who to telephone. In a pathway measured in minutes, this cannot be researched during the case.
- Understand that uncontrolled donors are metabolically different from controlled ones. In a multicentre comparison, perfusion duration was longer in uncontrolled donors, transaminases rose during perfusion, blood flow fell progressively despite increasing fluid, and pH stayed lower and glucose higher despite larger doses of bicarbonate and insulin. That is the ischaemic debt of the arrest, still being paid.
The retrieval team owns the rest. This section exists so that an intensivist is not learning the words for the first time during the conversation with a family.
24.11 Running the end of an ECPR run
§24.2 showed that decision points do not arise on a circuit. This section builds them.
Step | What to do | Why it is in this order |
1. Write the four exits on day one | Recovery, transplant, device, none — each with a named person and a date by which the answer is expected | The fourth line is the only thing that prevents a bridge to nowhere, and it is the line never written (§24.8) |
2. Name the failure mode before every conversation | Five different situations, five different conversations (§24.4) | Treating a termination of resuscitation as a negotiation asks permission for something that is not a choice |
3. Activate the determination pathway on the finding, not the conclusion | The first documented absence of brainstem reflexes off sedation triggers it, in parallel with prognostication | Withdrawal before completed determination converts a death into a decision, and removes the donation pathway (§24.7) |
4. Reduce the sweep, never the blood flow | Target 0.5 to 1 L/min, lungs kept open with positive end-expiratory pressure, both radial samples simultaneously | Sweep governs carbon dioxide; flow governs oxygen and haemodynamics. This is the whole protocol in one line (§24.5) |
5. Treat an aborted test as uninterpretable | Restore sweep and flow, correct hypoxia or hypotension, repeat. Add carbon dioxide if it will not rise | A 22% pooled failure rate falls to nearly zero where a written technique exists (§24.5) |
6. Keep determination and donation strictly separate | Different people, different conversations, different times. The clinician who gives the prognosis does not raise donation | The separation must be structural, not attitudinal — it is the only protection for a trust that cannot be rebuilt (§24.9) |
7. Manufacture the decision point | A date, a question, the information required, the people present, and what happens if the answer is no — all written before it arrives | Escalation is the default trajectory of a circuit. Nothing stops it except an appointment (§24.2) |
8. Hand over cleanly at the boundary | Prognosis is Chapter 23. Futility is Chapter 74. The withdrawal procedure is Chapter 75. Symptom control and the family are Chapter 76 | This chapter recognises and sequences. It does not adjudicate, and it does not perform |
Clinical pearl · What to say when a family asks "are you giving up on him?"
Three answers are true and are rarely offered.
- "We are not deciding whether to treat him. We are deciding whether this machine can still get him anywhere." It relocates the question from the patient's worth to the treatment's purpose, which is where it belongs and where the family can actually engage with it.
- "If his brain has already died, that is not something you or I will decide. We will confirm it, and I will show you how." For the one ECPR patient in six for whom this is the answer, the most merciful thing a team can do is remove the family from a decision they were never required to make.
- "The machine will go on working. That is what makes this hard, and it is why we set a date to talk rather than waiting for something to change." This is §24.2 said out loud, and it is the honest explanation for why the conversation is happening on a day when nothing visibly happened.
The worst sentence available is "there is nothing more we can do," because on a circuit it is visibly untrue. There is always something more to do. That is the problem.
24.12 The errors that recur
Error | Why it is made | What to do instead |
Waiting for a decision point to arrive | Every other part of critical care has them. The patient deteriorates despite therapy and the situation declares itself | On a circuit the situation cannot declare itself, because the machine is doing the declaring. Put a date in the notes on day one. The one study that watched found frequent discussion and no decision-making moments (§24.2) |
Withdrawing before completing the brain-death determination | The apnoea test is believed to be impossible or dangerous on a circuit, and withdrawal feels like the kinder shortcut | Complete the determination first. In two series, roughly as many patients had treatment withdrawn with absent brainstem reflexes as were declared dead — asking families to authorise a death that had already happened, and closing the donation pathway (§24.7) |
Reducing the ECMO blood flow to raise carbon dioxide | It is the intuitive lever, and it does raise carbon dioxide | Reduce the sweep gas, never the blood flow. Sweep governs carbon dioxide; flow governs oxygenation and haemodynamics. Cutting flow is how the test acquired its reputation for causing instability (§24.5) |
Reading the post-oxygenator sample as the apnoea test result | It is the sample nearest to hand and the circuit is where the gas exchange is happening | Take simultaneous right and left radial samples. The post-oxygenator value diverged from the right radial by a median of 9.9 mmHg by the end of the test, against 1.1 mmHg between the two radial sites (§24.6) |
Quoting a neurological prognosis on day one | Day one is when the family asks, and when the team's own impression is strongest | The commonest recorded reason for withdrawal is an unfavourable neurological prognosis and the commonest day is the first — but Chapter 23 showed the clock starts at sedation clearance and that half of patients remain indeterminate at 72 hours. Say what is not yet knowable (§24.1) |
Assuming there is a destination because nobody said there was not | Cannulation happened in minutes; the transplant and device teams were not there | Write the four exits with a named person against each. In one survey, 55% of clinicians reported transplant surgeons are not involved in initiation decisions (§24.8) |
Treating every ending as the same negotiation | "Withdrawal" is one word, so it feels like one act | Five failure modes, five conversations. Only two of them are negotiations; one is a determination and two are statements of fact (§24.4) |
Letting the donation question touch the cannulation or continuation decision | The donation arithmetic is genuinely impressive, and the resource argument is genuinely real | The yield belongs in the annual report and nowhere near the bedside. Strict structural separation — different people, different conversations, different times (§24.9) |
Counting only survivors when reporting what a programme achieves | Survival is the trial endpoint and the registry field | One 307-patient programme produced 29 neurologically favourable survivors and 167 transplanted organs. A survival rate describes a minority of the people the work reached — and the tightened criteria that improve it halve the rest (§24.9) |
Saying "there is nothing more we can do" | It is the standard phrase, and it is meant kindly | On a circuit it is visibly false; the machine is working and there is always another intervention. Say instead what the treatment can no longer reach (§24.11) |
24.13 Key points
- Refractory shock kills at least as many ECPR patients as the brain does, and it kills them first. Across four cohorts, shock accounted for around half of deaths and neurological injury for a third to a half. Median survival is zero days for refractory shock, two for brain death, three for non-neurological withdrawal, and nine for neurological withdrawal without brain death — four modes, four clocks, and four different conversations a unit must be ready to have (§24.1).
- Most ECPR deaths occur with the circuit still running. In one refractory in-hospital series, 60.5% died during mechanical support. Somebody therefore has to stop it, which makes the process by which that happens a clinical subject and not an ethical footnote (§24.1).
- Decision points do not arise on a circuit; they have to be manufactured. The one ethnography that observed 380 hours of ECMO care found that treatment was escalated as complications mounted until the patient could be decannulated or options were exhausted, and that discussion with families "did not occasion decision-making moments." Every complication has a next intervention, and the machine removes the physiological signal that normally ends treatment (§24.2).
- ECPR inverts the consent structure of medicine: it is begun without consent and, in some jurisdictions, may be stopped only with it. Nothing clinical changes at the moment perfusion passes from hands to pump, but the authority to stop may. Units that have not settled this question in advance will settle it during their worst week (§24.3).
- The commonest recorded reason for stopping is a neurological prognosis, and the commonest day for it is the first. In 1,660 patients, 30.7% had a withdrawal decision, at a median of two days, with 58.8% citing perceived unfavourable neurological prognosis — against Chapter 23's finding that the prognostic clock starts at sedation clearance and that half of patients are indeterminate at 72 hours (§24.1).
- The apnoea test is sabotaged by the sweep gas, and rescued by lowering it. Carbon dioxide clearance on a circuit is governed by sweep, so disconnecting the ventilator does not raise it. Reduce the sweep to 0.5 to 1 L/min and hold the blood flow constant, keep the lungs open, and add exogenous carbon dioxide if needed. The pooled non-confirmatory rate is 22%; in a unit with a written technique and 25 ECMO patients, no test was aborted at all (§24.5).
- The sample site decides whether the test is valid. In venoarterial patients, the difference between right radial and post-oxygenator carbon dioxide rose to a median of 9.9 mmHg by the end of the test, against 1.1 mmHg between the two radial arteries. Chapter 16's Harlequin physiology reappears at the one moment it cannot be tolerated: take both radial samples simultaneously (§24.6).
- Withdrawal before completed determination converts a death into a decision. In two series, as many patients had treatment withdrawn while showing absent brainstem reflexes as were formally declared dead — in one, because the apnoea test was deemed "difficult" and simply not done. The family is then asked to authorise a death that had already occurred, the donation pathway closes, and the unit's own audit misclassifies the death. Complete the determination before discussing withdrawal (§24.7).
- Organ donation is numerically the largest single output of a mature ECPR programme, and the criteria that improve survival halve it. In 307 patients: 29 neurologically favourable survivors, 58 donors, 167 solid organs, 196 individuals possibly helped. Tightening low-flow and rhythm criteria reduced donation from 19% to 11%. The yield belongs in the programme's accounting and must be kept structurally separate from every bedside decision (§24.9).
- Normothermic regional perfusion is an ECPR circuit with the aortic arch vessels tied off — the same pump and membrane, configured to prevent the cerebral reperfusion that Chapter 21 existed to restore. It increases organs per donor from 2.6 to 3.3 and improves graft outcomes, organs from resuscitated donors perform comparably at long-term follow-up, and its ethics remain formally unresolved, with a 71-document scoping review finding no agreement on its permissibility (§24.10).
Cross-references
Inherited from earlier chapters
- Chapter 19 — the ECPR trials, and the fact that all three randomised the decision to cannulate and none randomised anything afterwards. Every figure in §24.1 is downstream of that design.
- Chapter 20 — selection, and §20.9's selection loop: the criteria list is the dial between a unit's survival figure and its competence. §24.9 shows that the dial has a second output nobody counts. §20.8's refusal log and this chapter's withdrawal decisions are the two ends of one policy.
- Chapter 21 — cannulation. §24.4 begins where it ends: roughly one attempt in six in one programme never achieved access at all, and that is a termination of resuscitation.
- Chapter 22 — the first 24 hours. The patient arrives at this chapter with gases titrated, a temperature chosen and a cause found or excluded.
- Chapter 23 — prognostication. Chapter 23 produces a probability and stops. This chapter is what happens next — and §24.1 records that in practice the commonest withdrawal reason invokes a prognosis on a day when Chapter 23's assessment cannot yet have begun.
- Chapter 16 — differential hypoxaemia, which returns in §24.6 as the reason a post-oxygenator blood gas cannot validate an apnoea test.
- Chapter 6 and Chapter 27 — sweep gas and carbon dioxide clearance, the physiology on which the entire apnoea protocol of §24.5 rests.
Handed forward
- Chapter 28 — Cerebral Monitoring. Electroencephalography on a circuit, including the artefact problem that made one determination inconclusive in §24.6.
- Chapter 34 — Cannula Problems, and Chapter 13 — VA ECMO Haemodynamics, for the failures of §24.4 that are mechanical or circulatory rather than decisional.
- Chapter 69 — Goals of ECMO, and Chapter 70 — Bridge-to-Decision. The four-exit assessment of §24.8 is their instrument; this chapter only insists it be written on day one.
- Chapter 72 and Chapter 82 — bridge to transplantation, for the destination this chapter can only ask about.
- Chapter 74 — Futility and Non-beneficial ECMO. This chapter recognises and sequences. Chapter 74 adjudicates, and owns the dispute-resolution machinery that Controversy 1 deliberately declines to build.
- Chapter 75 — Withdrawal of ECMO. The procedure itself: how flow is reduced, what is given, what the room looks like, and how a person dies well on a circuit.
- Chapter 76 — Palliative Care, and Chapter 59 — Mobilization and Rehabilitation, for the two destinations that are not death.
- Chapter 77 — Resource Allocation and ECPR Ethics. The donation arithmetic of §24.9, the moral distress documented in §24.2 and §24.3, and the argument that a programme's true output is not its survival rate.
- Chapter 68 — Paediatric Considerations. The paediatric apnoea-testing series, the paediatric ECPR ethics literature, and the finding that 10% of children who ultimately survived to discharge were still comatose on day three are all held there.
References
How to read this list
Every external reference in this chapter was read as a structured abstract. None was retrieved in full text. Digital object identifiers are reproduced as they appeared in the search record and have not been independently resolved; treat them as leads rather than as verified citations. Numerical values are those printed in the abstract. Every database row seeded from this chapter carries Verified = No.
A standing caution specific to this chapter: determination of death, permissible ancillary tests, donation pathways and no-touch intervals are set nationally and are actively changing. Nothing here substitutes for the criteria in force where the reader practises.
How ECPR runs end
- Zotzmann V, et al. Mode of death after extracorporeal cardiopulmonary resuscitation. Membranes. 2021. DOI 10.3390/membranes11040270. 274 consecutive ECPR cases; therapy-resistant shock 51.7% and anoxic brain injury 34.0% of 203 deaths.
- Farkasovska K, et al. Reasons for death in patients receiving ECPR for refractory out-of-hospital cardiac arrest. Resuscitation. 2025. DOI 10.1016/j.resuscitation.2025.110615. 210 ECPR patients, 152 deaths adjudicated by two investigators (kappa 0.81); refractory shock 49.3%, neurological injury 45.3%.
- Walker H, et al. Timing and mode of death following ECPR: a single-centre retrospective analysis. Heart, Lung and Circulation. 2025. DOI 10.1016/j.hlc.2025.04.072. 95 patients; the four median survival times that drive §24.1; four patients (7.8%) donated organs.
- Sokolov T, et al. Outcomes of extracorporeal cardiopulmonary resuscitation for refractory in-hospital cardiac arrest. Artificial Organs. 2025. DOI 10.1111/aor.15028. 147 patients; 89 (60.5%) died during mechanical circulatory support.
- Naito H, et al. Prevalence, reasons, and timing of decisions to withhold or withdraw life-sustaining therapy for out-of-hospital cardiac arrest patients with ECPR. Critical Care. 2023. DOI 10.1186/s13054-023-04534-2. SAVE-J II secondary analysis, 1,660 patients; 30.7% with a withdrawal decision at a median of two days. See the contradiction flagged in the chapter status.
- Carlson JM, et al. Early withdrawal of life-sustaining therapy in ECPR: results from the ELSO Registry. Resuscitation. 2022. DOI 10.1016/j.resuscitation.2022.07.038. 411 patients undergoing withdrawal at family request; 55.5% within 72 hours of cannulation. Direction used; the reported adjusted estimates are not — see the chapter status.
- Le Guen M, et al. Extracorporeal life support following out-of-hospital refractory cardiac arrest. Critical Care. 2011. DOI 10.1186/cc9976. 51 patients, median delay to support 120 minutes, two survivors (4%). Mode-of-death counts are internally inconsistent — see the chapter status.
Decision-making, consent and the escalation trap
- Hadler RA, et al. Escalation and withdrawal of treatment for patients on extracorporeal membrane oxygenation. Annals of Surgery. 2021. DOI 10.1097/sla.0000000000004838. Focused ethnography, two academic cardiothoracic intensive care units, 380 hours of observation. The organising source for §24.2.
- Piscitello GM, et al. Clinician ethical perspectives on extracorporeal membrane oxygenation in practice. American Journal of Hospice and Palliative Care. 2021. DOI 10.1177/10499091211041079. 14 departments across seven hospitals; the 7.5-minute consent figure and the 36-against-64 split on the bridge to nowhere.
- Grunau B, MacRedmond R, Gill J. A promising therapy in jeopardy. Circulation. 2019. DOI 10.1161/circulationaha.118.035674. The clearest published statement of the ECPR consent asymmetry, and the jurisdiction in which substitute-decision-maker consent is required before withdrawal.
- Moynihan KM, et al. Starting ECMO: a review of the ethical issues and recommended approach. Chest. 2026. DOI 10.1016/j.chest.2026.07.5210. The offering-versus-initiating distinction and iterative consent, adopted in §24.3.
- Siddiqui S, et al. What are the real issues in providing extracorporeal membrane oxygenation support: a survey. Journal of Intensive Care Medicine. 2024. DOI 10.1177/08850666241245933. 61 of 305 responses; four recurring dilemmas including moral distress about when to stop. Reported twice with differing figures — see the chapter status.
- Olive JK, et al. Bridge to decision, bridge to nowhere: ethical and clinical challenges in temporary mechanical circulatory support. Journal of Cardiac Failure — Intersections. 2026. DOI 10.1016/j.yjcafi.2025.11.006. The four-exit framework and the case for time-limited trials.
- Rubin J, et al. The human and humanity that differentiate withholding from withdrawing life-sustaining therapy: an ECMO bridge to nowhere. The American Journal of Bioethics. 2023. DOI 10.1080/15265161.2023.2201206. The equivalence-thesis argument used in Controversy 1. Held largely for Chapter 74.
- Eussen SEDM, et al. Attitudes of patients and family members towards deferred and waived consent in ECPR research: an ancillary study of the INCEPTION trial. Resuscitation Plus. 2026. DOI 10.1016/j.resplu.2026.101239. 32 of 38 questionnaires returned; 81.3% supported alternative consent procedures.
- Rajšić S, et al. Ethical considerations for patients requiring extracorporeal cardiopulmonary resuscitation. Journal of Cardiothoracic and Vascular Anesthesia. 2025. DOI 10.1053/j.jvca.2025.07.032.
- Schou A, et al. Ethics in extracorporeal life support: a narrative review. Critical Care. 2021. DOI 10.1186/s13054-021-03689-0. Three ethical discourses, one of which is death on extracorporeal support and its role in donation.
Determination of death by neurologic criteria on a circuit
- Greer DM, et al. Pediatric and adult brain death / death by neurologic criteria consensus guideline. Neurology. 2023. DOI 10.1212/wnl.0000000000207740. 85 recommendations, and the first major guideline to include guidance for determination in the context of extracorporeal membrane oxygenation. [VERIFICATION REQUIRED] — the ECMO-specific recommendations themselves were NOT retrieved and are not quoted anywhere in this chapter.
- Migdady I, et al. The use of apnea test and brain death determination in patients on extracorporeal membrane oxygenation: a systematic review. The Journal of Thoracic and Cardiovascular Surgery. 2020. DOI 10.1016/j.jtcvs.2020.03.038. 22 studies, 177 patients; 50% underwent apnoea testing, 22% non-confirmatory, only 12% declared on apnoea testing alone.
- Su Y, et al. Practice of apnea testing and determination of brain death during extracorporeal membrane oxygenation: a multicenter retrospective analysis. ASAIO Journal. 2025. DOI 10.1097/mat.0000000000002561. 31 patients; sweep reduced from a median of 3.5 to 0.5 L/min with blood flow held constant; 81% donated organs.
- Giani M, et al. Apnea test during brain death assessment in mechanically ventilated and ECMO patients. Intensive Care Medicine. 2015. DOI 10.1007/s00134-015-4105-6. 169 consecutive patients including 25 on ECMO; no test aborted, no severe complication.
- Salih F, et al. Brain death determination in patients with venoarterial extracorporeal membrane oxygenation: a systematic study to address the Harlequin syndrome. Journal of Critical Care. 2024. DOI 10.1016/j.jcrc.2024.154545. 27 patients; the sampling-site divergence that governs §24.6.
- Englbrecht J, et al. Determination of brain death in potential donors on ECMO support. Die Anaesthesiologie. 2026. DOI 10.1007/s00101-026-01695-x. The 2026 national directive update removing cerebral blood flow-based ancillary tests in potential donors on venoarterial support.
- Zhao DX, et al. Challenges in determining death by neurologic criteria in extracorporeal membrane oxygenation: a single-centre experience. Perfusion. 2023. DOI 10.1177/02676591231187548. Eight patients determined; seven further patients with absent brainstem reflexes in whom withdrawal pre-empted the evaluation.
- Muralidharan R, et al. The challenges with brain death determination in adult patients on extracorporeal membrane oxygenation. Neurocritical Care. 2011. DOI 10.1007/s12028-011-9516-9. Three of 87 adults lost all brainstem reflexes; the apnoea test was not performed in any, and support was withdrawn instead.
- Busl KM, et al. Apnea testing for the determination of brain death: a systematic scoping review. Neurocritical Care. 2020. DOI 10.1007/s12028-020-01015-0. 87 articles; most modifications are not standardised or endorsed by guidelines.
- Madden M, et al. Carbogen for apnea testing during the brain death declaration process in subjects on extracorporeal membrane oxygenation. Respiratory Care. 2019. DOI 10.4187/respcare.06378. Five subjects; the carbogen target was met in all, with no adverse events.
- Smilevitch P, et al. Apnea test for brain death determination in a patient on extracorporeal membrane oxygenation. Neurocritical Care. 2013. DOI 10.1007/s12028-013-9845-y. Sweep reduced from 3 to 1 L/min with oxygen delivery increased; carbon dioxide rose over 20 mmHg without a fall in oxygen tension.
- Faria C, et al. Apnea testing for brain death confirmation in venovenous ECMO patients with very low sweep flow. Critical Care Science. 2025. DOI 10.62675/2965-2774.20250373. A 200 mL/min sweep protocol in three of 93 patients.
- Sady É, et al. Apnea test for brain death diagnosis in adults on extracorporeal membrane oxygenation: a review. Revista Brasileira de Terapia Intensiva. 2020. DOI 10.5935/0103-507x.20200048. 17 publications; the practical strategies summarised in §24.5.
- Harrar DB, et al. Clinical determination of brain death in children supported by extracorporeal membrane oxygenation. Neurocritical Care. 2019. DOI 10.1007/s12028-019-00700-z. Eight children, 14 examinations; the circuit was modified in 13 of 14. Held largely for Chapter 68.
- Jarrah RJ, et al. Developing a standard method for apnea testing in the determination of brain death for patients on venoarterial extracorporeal membrane oxygenation: a pediatric case series. Pediatric Critical Care Medicine. 2014. DOI 10.1097/pcc.0000000000000006. Three children; two completed, one invalidated by haemodynamic instability. Held for Chapter 68.
Organ donation after ECPR, and regional perfusion
- Bonizzoni MA, et al. Organ donation after extracorporeal cardiopulmonary resuscitation for refractory out-of-hospital cardiac arrest in a metropolitan cardiac arrest centre in Milan, Italy. Resuscitation. 2024. DOI 10.1016/j.resuscitation.2024.110214. 307 patients; 29 favourable survivors, 167 organs, 196 possible beneficiaries, and the fall in donation from 19% to 11% as criteria tighten. The most important single reference in §24.9.
- Rajsic S, et al. Ethical aspects of organ donation following extracorporeal cardiopulmonary resuscitation with unfavorable neurologic outcomes: a systematic review. Journal of Cardiothoracic and Vascular Anesthesia. 2025. DOI 10.1053/j.jvca.2025.10.003. 17 studies, 254 patients, 689 grafts, 469 recipients; only seven studies addressed ethics.
- Rajsic S, et al. Organ donation after ECPR: ethical challenges and clinical implications. Critical Care. 2026. DOI 10.1186/s13054-026-06072-z. The ECPR-specific ethical model whose central element — strict separation of resuscitative and donation-related decision-making — this chapter adopts. The chapter's principal currency find.
- Roncon-Albuquerque R, et al. An integrated program of ECMO-assisted cardiopulmonary resuscitation and uncontrolled donation after circulatory determination of death in refractory cardiac arrest. Resuscitation. 2018. DOI 10.1016/j.resuscitation.2018.10.016. 58 patients; the ten-minute no-touch period, 33% full neurological recovery among ECPR patients, and 44 kidneys transplanted. Comparative figures are reported in a direction inconsistent with their own framing — see the chapter status.
- Sandroni C, et al. Does cardiopulmonary resuscitation before donor death affect solid organ transplant function? A systematic review and meta-analysis. Resuscitation. 2025. DOI 10.1016/j.resuscitation.2025.110654. 33 studies, 72,994 donors; organs from resuscitated donors performed comparably at longest follow-up. The answer to the commonest bedside objection.
- Jochmans I, et al. Consensus statement on normothermic regional perfusion in donation after circulatory death. Transplant International. 2021. DOI 10.1111/tri.13951. The European consensus, including minimum ethical, logistical and technical requirements. [VERIFICATION REQUIRED] — the recommendations themselves were not retrieved and none is quoted.
- Oniscu GC, et al. Improved organ utilization and better transplant outcomes with in situ normothermic regional perfusion in controlled donation after circulatory death. Transplantation. 2022. DOI 10.1097/tp.0000000000004280. 3.3 against 2.6 organs per donor, and the graft-outcome figures in §24.10.
- Murphy NB, et al. Ethical issues in normothermic regional perfusion in controlled organ donation after determination of death by circulatory criteria: a scoping review. Transplantation. 2024. DOI 10.1097/tp.0000000000005161. 71 documents, six themes, and the finding of no agreement on ethical permissibility.
- Kirschen MP, et al. Beyond the final heartbeat: neurological perspectives on normothermic regional perfusion for organ donation after circulatory death. Annals of Neurology. 2024. DOI 10.1002/ana.26926. The arch-vessel ligation and the three neurological questions.
- Liebman J. Does normothermic regional perfusion harm donors after circulatory death? Journal of Medical Ethics. 2025. DOI 10.1136/jme-2025-110947. The harm-based reframing of the dead-donor-rule debate.
- Miñambres E, et al. Improving the outcomes of organs obtained from controlled donation after circulatory death donors using abdominal normothermic regional perfusion. American Journal of Transplantation. 2017. DOI 10.1111/ajt.14214. 27 donors; describes a specific methodology to avoid restoring circulation to the brain after death determination.
- Williams AM, et al. 200 cases of cardiac donation after circulatory death utilizing normothermic regional perfusion: the four-year Vanderbilt experience. The Journal of Heart and Lung Transplantation. 2025. DOI 10.1016/j.healun.2025.05.007. 176 of 200 hearts accepted (88%) with stable recipient outcomes across four eras.
- James L, et al. Donation after circulatory death heart transplantation using normothermic regional perfusion: the NYU protocol. JTCVS Techniques. 2022. DOI 10.1016/j.xjtc.2022.11.014. 18 donors; the lactate and potassium correction figures in §24.10.
- Lazzeri C, et al. Management of normothermic regional perfusion performance in uncontrolled versus controlled donation after circulatory death: a multi-center investigation. Journal of Clinical Medicine. 2025. DOI 10.3390/jcm14197053. Uncontrolled donors show a more severe and only partly reversible metabolic derangement. Denominators are internally inconsistent — see the chapter status.
- Pionnier Y, et al. Solid organ transplantation originating from uncontrolled donation after circulatory death in Europe: a narrative review. Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine. 2024. DOI 10.1186/s13049-024-01305-y. Long-term outcomes comparable to controlled donation and donation after brain death; and the warning that programmes tend to overestimate the number of potential donors.
- Supady A, et al. Extracorporeal cardiopulmonary resuscitation for refractory cardiac arrest. The Lancet Respiratory Medicine. 2025. DOI 10.1016/s2213-2600(25)00122-5. A current overview stating that organ donation is possible after ECPR in patients with brain death or, per local regulation, irreversible post-anoxic injury.
Textbooks
- ELSO Red Book, 6th edition — Chapter 32, and the ethics and end-of-life material. Consulted for the neurological complication and brain-death figures used in §24.5. [VERIFICATION REQUIRED] — page numbers not confirmed.
- Shinar Z, Badulak J. ECPR and Resuscitative ECMO — consulted for programme structure and the arrival examination; the ECPR-specific end-of-life material is thin. [VERIFICATION REQUIRED].
- ISCCM Manual of RRT and ECMO in ICU, and Taha AR, Caridi-Scheible M, Leiendecker E, et al., ECMO: A Practical Guide to Management — consulted; not separately quoted in this chapter. [VERIFICATION REQUIRED].
Chapter status
Drafted and audited 15 September 2026. Ten-pass quality control completed: clinical, physiology, evidence, citation, numerical, safety, contradiction, redundancy, bedside utility and literature-currency passes. This chapter completes Part IV.
A scope finding that changed the chapter. A tracker query established that no chapter in this book owned determination of death by neurologic criteria on a circuit, organ donation after ECMO, or normothermic regional perfusion. All three were placed here on physiological grounds — brain death is overwhelmingly an ECPR event, the apnoea test is defeated by the sweep gas, and regional perfusion is mechanically an ECPR circuit with the arch vessels ligated. Chapters 74 to 77 keep the ethics; this chapter takes the physiology and the process.
The organising insight is drawn from the one study that observed rather than surveyed: over 380 hours in two units, treatment was escalated as complications mounted until the patient could be decannulated or options were exhausted, and discussion with families "did not occasion decision-making moments." Every complication on a circuit has a next intervention, and the machine removes the physiological signal that normally ends treatment. Decision points therefore do not occur; they must be manufactured and written down in advance. A second structural claim follows: ECPR is begun without consent and, in some jurisdictions, may be stopped only with it — an inversion that occurs the moment perfusion passes from hands to pump, and which nothing in the record marks.
Three controversies were set out rather than smoothed. §24.3 asks who is entitled to insist on continuing a treatment nobody consented to starting, and declines to resolve a question that is one of law rather than evidence. §24.6 asks what may substitute for an apnoea test that cannot be completed, and finds the evidence moving against ancillary testing on venoarterial support just as practice relies on it most. §24.9 asks whether organ donation may influence any part of the ECPR pathway, and separates the programme level from the bedside with a wall.
The numerical and contradiction audits flagged six problems and corrected none. (1) The ELSO withdrawal analysis reports an adjusted odds ratio of −3.1 with a confidence interval of 2.18–2.8 — a negative odds ratio whose point estimate lies outside its own interval — and describes a higher peak inspiratory pressure as associated with early withdrawal at an odds ratio of 0.84 (0.71–1.00). Only the direction of that paper is used; none of its adjusted estimates is. (2) The SAVE-J II withdrawal analysis is reported twice by the same authors with different denominators (1,660 and 510 against 1,651 and 506; reason denominators of 510 against 477) — the fourth instance in Part IV of the same-authors-different-numbers pattern, after the INCEPTION figures in Chapter 19, the SAVE-J II temperature analysis in Chapter 22 and the CARES Delphi in Chapter 22. (3) A 51-patient cohort reports mode-of-death counts that sum to 60, and describes 43 deaths as 47%. (4) An integrated ECPR-and-donation programme reports its comparative figures in a direction inconsistent with its own framing, so only its design and its transplant yield are used. (5) A multicentre regional-perfusion series states 99 donors and then analyses 44 plus 45. (6) A mode-of-death cohort reports lactate in mg/dL at values interpretable only as mmol/L, and the same survey of ECMO clinicians is published twice with 60 and 61 respondents and 68% and 70% for the same item.
A safety finding recorded as the chapter's most actionable sentence: in two independent series, as many ECMO patients had life-sustaining treatment withdrawn while showing absent brainstem reflexes as were formally declared dead — in one series because the apnoea test was deemed "difficult" and simply not attempted. Complete the determination before discussing withdrawal.
The redundancy audit removed futility as a concept, which belongs to Chapter 74; the withdrawal procedure itself, which belongs to Chapter 75; symptom control and bereavement, which belong to Chapter 76; resource allocation and the moral-distress literature, which belong to Chapter 77; bridge-to-decision as a strategy, which belongs to Chapter 70; and all prognostication, which belongs to Chapter 23. Two long-standing Part IV gaps remain open at the close of the Part: the ELSO ECPR interim guidance was never retrieved, and the ECMO-specific recommendations of the 2023 consensus guideline on death by neurologic criteria were identified but not obtained. Neither is quoted anywhere in Part IV.