Chapter question: This heart looks better. How do I find out whether it can carry the circulation on its own — and why does answering that question correctly still leave so many of these patients dead?
Evidence search date: 11 September 2026.
Primary source: ELSO Red Book 6th edition Chapter 29 — Weaning and Decannulation in Adult Cardiac Failure — the dedicated chapter, including Tables 29-1, 29-2 and 29-3 and the three weaning techniques. Supported by ISCCM Manual Ch 38, Taha Ch 11, and the Red Book's paediatric weaning chapter for the trial-off mechanics.
Two reviews retrieved in full this session: the 2022 Critical Care systematic review of weaning parameters (DOI 10.1186/s13054-022-04249-w) and the 2023 Perfusion practice-oriented review of patient-centred weaning (DOI 10.1177/02676591221115938). Around them, a 2026 meta-analysis of 37 studies and 3,458 patients and twelve cohort studies read as structured abstracts.
This chapter completes Part III. It is also the chapter where the book has to say something uncomfortable about its own subject: the entire predictive literature optimises an endpoint that is not survival.
What this chapter covers — and what it does not
This chapter owns | Deferred to |
The distinction between weaning and surviving; preconditions before any trial; the three weaning techniques and how they differ physiologically; how long to sit at minimum flow; the haemodynamic and echocardiographic criteria and their hierarchy; whether ejection fraction predicts anything; the right ventricle, where the weaning study is weakest; non-echo signals; weaning a vented ventricle; failing a trial and repeating it; the decannulation procedure; post-decannulation care; and the point at which failure to wean becomes a destination decision | Echocardiographic technique and image acquisition → Chapter 26
The sweep-gas-off trial — a VV-only manoeuvre → Chapter 10
Haemodynamic monitoring in depth → Chapter 25
LV venting decisions → Chapter 15
Conversion to VV for differential hypoxaemia → Chapter 16
Vascular repair and late stenosis → Chapter 17
Choosing a destination — recovery, decision, transplant, durable device → Chapters 70–73
Futility, one-way weaning as an end-of-life decision, withdrawal → Chapters 74–75
ECPR-specific prognostication → Chapters 23–24
The bedside checklist → Chapter 91 |
18.1 Two questions that are routinely treated as one
Danger — "successfully weaned" does not mean "survived", and the gap is enormous
A meta-analysis of 11 studies and 653 patients found pooled successful weaning of 45% (95% CI 39–50%) and in-hospital mortality of 46.6% (95% CI 33–60%). Those two numbers describe overlapping populations: a large share of the patients whose cannulas came out did not leave hospital.
The 2023 Perfusion review states the problem directly, and it is the single most important sentence in this chapter's source material: "a strikingly high mortality up to 70% after initial successful weaning raises concerns about the validity of current weaning strategies."
This matters because of how the literature is built. Almost every study in §18.5 defines its outcome as removal of the circuit without further mechanical support or transplant over the following 30 days. That is an endpoint about the device. A patient can satisfy every echocardiographic criterion in this chapter, be decannulated, and die three weeks later of the multi-organ failure that put them on the circuit in the first place — and be counted as a success.
So read every threshold below for what it is: a predictor that the heart will tolerate having the machine removed. Not a predictor that the patient will live.
Clinical pearl — ask the two questions separately and in this order
1. Can this circulation function without the circuit? A physiological question, answered by the weaning trial. This chapter.
2. Should this patient's support end here — by recovery, by escalation to a durable device or transplant, or by stopping? A destination question, answered by the team and the family. Chapters 70–75.
The two are separable and they fail in different directions. A patient who cannot wean may still have an excellent outcome — via a durable device or a transplant. A patient who weans beautifully may still be dying, and decannulation simply removes the thing that was making the dying visible.
The Red Book's own framing makes the dependency explicit: weaning assessment may need to start earlier than physiology alone would suggest — in a patient with irreversible ventricular dysfunction who is ineligible for transplant or durable support, where the plan is a one-way wean, and in a patient accumulating ECMO-related complications, where the risk of staying on the circuit has begun to exceed the risk of coming off it.
The decision to test is therefore not purely physiological, and pretending otherwise is how teams drift into supporting a circulation with no destination.
18.2 Preconditions — what must be true before anyone touches the flow
The Red Book's Table 29-1 sets out the prerequisites. They are unglamorous and they are where most premature trials go wrong.
Precondition | Detail |
The problem that caused the shock has been treated or has resolved | Revascularisation done, myocarditis settling, tamponade drained, pulmonary embolism treated, arrhythmia controlled. The ISCCM manual and Taha both insist on this first: "the primary cause of cardiogenic shock must be resolved before one begins weaning." Weaning a heart whose insult is ongoing is a test of the insult, not of the heart |
Other organs have stabilised | Adequate time for organ function to recover. A patient in established multi-organ failure will fail the trial for reasons that have nothing to do with the myocardium |
Safe ventilation and gas exchange | For VA ECMO, FiO₂ at or below 0.6, with an arterial gas showing adequate systemic oxygenation and a normal PaCO₂ or adequate metabolic compensation. The lungs must be able to take the transpulmonary flow the trial is about to send through them — which is the same physiology as Chapter 16 |
Adequate anticoagulation | Non-negotiable, and the reason is mechanical: every weaning technique reduces flow, and reduced flow is how circuits clot (§18.4) |
Low vasoactive support | Defined by the Red Book as epinephrine under 0.05 µg/kg/min, dobutamine under 5 µg/kg/min, milrinone under 0.375 µg/kg/min, norepinephrine under 0.05, or vasopressin under 0.02 U/min |
Preload optimised | The Red Book advises a fluid bolus to optimise ventricular preload before weaning assessment. Reducing circuit flow returns volume to a ventricle that has been chronically underfilled by the drainage cannula; a hypovolaemic patient will fail a trial they would otherwise pass |
Pitfall — the same chapter gives two different vasoactive-inotropic score thresholds
Red Book Table 29-1, listing preconditions, states that a vasoactive-inotropic score under 10 may be used as the definition of low-dose support. Red Book Table 29-2, listing predictors of successful weaning, footnotes that a vasoactive-inotropic score under 30 was used in one study.
Both appear in the same chapter with the same formula. Certainty: very low for either number. They are not strictly contradictory — one is a precondition to attempt and the other a condition observed in a study population — but they are three-fold apart and are easily read as interchangeable.
There is also a unit inconsistency in the low-dose definition as printed: epinephrine, dobutamine and milrinone are given in µg/kg/min, while norepinephrine is printed as "under 0.05 µg/min" — which for an adult is a dose no infusion pump would deliver. It is almost certainly µg/kg/min. It is reproduced here as printed with the discrepancy flagged, not silently corrected.
What to take to the bedside: use the drug-by-drug list rather than a composite score, and treat "low" as meaning a patient you would be comfortable extubating the vasopressors from — not a number you can argue about.
18.3 Three ways to test, and they are not equivalent
All three techniques do the same physiological thing — reduce circuit flow, which simultaneously raises ventricular preload and lowers LV afterload — and then ask whether the heart copes. They differ in one respect that decides which one answers your question: how much load the right ventricle actually sees.
Technique | How it is done | RV loading | Main risk |
1. Stepwise flow reduction — the default | Flow reduced in 0.5–1 L/min steps, or as a percentage of baseline (for example 66% then 33%), over 10–15 minutes per step or gradually over several hours. Echocardiography needs 10–15 minutes at a step to show LV and RV change. Final step: minimum possible rate, typically 0.5–1 L/min | Incomplete. The Red Book names this as the key caveat: "full RV loading is not achieved for accurate RV assessment" | Circuit thrombosis at low flow (§18.4) |
2. Arteriovenous bridge | A length of tubing with three-way taps joins the arterial and venous limbs. Patient flow is reduced gradually with a releasable clamp; when haemodynamics hold at low flow, both cannulae are isolated by turning the taps off to them, and heparin flush lines are added or the cannulae flushed manually. Blood is diverted through the bridge and circuit flow is kept high to avoid clot | Full. The patient is genuinely off the circuit, so LV and RV can both be assessed independently of ECMO. It also reveals whether a hypoxic patient needs reconfiguring from VA to VV (Chapter 16) | Air entrainment during circuit manipulation; thrombus near the three-way taps; bridge flow must be monitored |
3. Pump-controlled retrograde trial off (PCRTO) | Pump speed is reduced until systemic pressure exceeds the pump's, and blood flows backwards through the arterial cannula, around the circuit, and back through the venous cannula. Typically under 1000 RPM to achieve about 0.5–1.0 L/min of retrograde flow, creating a left-to-right shunt. The pump now acts as a brake — raising RPM reduces the retrograde flow. The flow probe may need reversing. Circuit oxygen is removed, which allows native lung function to be assessed independently, and the distal perfusion cannula is disconnected and put on a heparin flush | Full, and with an added load: the circuit imposes extra resistance. "If hemodynamics do not change even with the additional resistance offered by the circuit, the patient is most likely to wean" | Systemic flow steal. The shunt must be regulated or the patient develops RV failure or high-output failure. Titrate RPM slowly to avoid an abrupt fall in LV afterload |
Physiology — why "no circuit manipulation" is not automatically the safer choice
The stepwise flow reduction is popular because, as the Red Book says, "it is easy to do at the bedside without any circuit manipulation." That is a real advantage and it is also the source of its central weakness.
Leaving the drainage cannula running means the right ventricle never sees its true preload. Some proportion of venous return is still being taken by the circuit, so the RV is being tested at a discount — and then the patient is decannulated, the discount is withdrawn, and the RV meets the full return for the first time. Every source in §18.5.4 agrees the right ventricle is where weaning assessment is least reliable, and this is the mechanical reason why.
The bridge and PCRTO solve exactly this: both deliver full RV loading. PCRTO goes further and adds load, which is why its proponents argue it provides "some reassurance that the patient has recovered sufficiently to decannulate."
Certainty: low. The Red Book notes that the AV bridge evidence is "based on small, single-center studies," and that for PCRTO "only two small, single-center studies have been conducted in adults so far... although the results are encouraging, further validation is required."
The practical reading: if the question is whether the left ventricle can eject, a flow-reduction trial answers it. If the question is whether the right ventricle can carry the venous return — which is the question in RV failure, in pulmonary hypertension, and before any LVAD decision — a flow-reduction trial is the wrong instrument.
18.4 Controversy 1 — how long can the patient sit at minimum flow?
Controversy 1 — How long should the final, minimum-flow stage of a weaning trial last?
The question. The trial's final stage puts the circuit at 0.5–1 L/min. Low flow in a large, thrombogenic circuit is exactly the condition that produces clot. But a short observation may not reveal a heart that fails after twenty minutes rather than after five. How long is right?
The published answers span an order of magnitude.
— Red Book Ch 29 (isolated RV failure protocol): "leave at minimum acceptable blood flow for under 15 minutes only to prevent circuit thrombosis."
— A 2025 single-centre protocol: daily "short (30 minutes)" trials, requiring two successful weanings separated by 24 hours before removal.
— Taha: "a low blood flow is maintained in the ECMO for approximately 40–60 minutes."
— A 57-patient study defined its weaning test as flow under 2 L/min for at least 60 minutes.
— The Red Book's paediatric chapter: "it is not uncommon for ECLS trial-off periods to last for 1–2 hours before decannulation."
What the evidence actually shows. There is none. No study compares trial durations, and none of these figures is derived from an outcome comparison. Certainty: very low for every number above. They are institutional practice reported as method.
But the disagreement is not arbitrary, and reading it carefully dissolves most of it. The sources differ because they are not all describing the same circuit state.
— The Red Book's under-15-minutes applies to a circuit still carrying the patient's blood at 0.5–1 L/min — genuinely stagnant, genuinely thrombogenic, and the clock is a clot clock.
— The 60-minute and 1–2-hour figures come from protocols using a bridge, where the patient is isolated and circuit flow is deliberately kept high through the bridge. There is no stagnation to time.
— The 30-minute daily trials and the 2 L/min tests sit in between, at flows that are low for the patient but not near-zero for the circuit.
The variable that actually matters is not elapsed time. It is flow through the oxygenator, and the Red Book's paediatric chapter states the governing rule that the adult chapter leaves implicit: "ECLS flow should not be reduced below the lowest flow rate established for the circuit's oxygenator." It adds that thrombus risk depends on anticoagulation strategy, circuit size, existing clot burden and circuit complexity — four patient- and circuit-specific variables that no fixed duration can accommodate.
Where practice actually sits. Most units run short trials at genuinely low flow, and longer ones only when the circuit is protected by a bridge or by PCRTO's continued circulation.
What would resolve it. A comparison of trial durations with circuit thrombosis and weaning-decision accuracy as co-primary outcomes. None exists.
What to take to the bedside. Do not read a duration off a protocol written for a different circuit state. Ask three questions instead: what is the minimum flow this oxygenator is rated for; is the patient fully anticoagulated right now; and does this circuit already have visible clot? If the answer to the last is yes, the trial is short and the decision is made on what you see in the first ten minutes. If you need a long assessment — and for the right ventricle you often do — use a technique that keeps the circuit moving (§18.3) rather than extending a low-flow period.
18.5 What to measure during the trial
18.5.1 The haemodynamic floor
Parameter | Threshold | Note |
Mean arterial pressure | At or above 60 mmHg on low vasoactive support | In a 2022 review, MAP at the time of weaning was independently associated with success (OR 1.05, p=0.009). In a meta-analysis, weaned patients had MAP higher by a mean difference of 20.15 mmHg (13.8–26.4) |
Pulsatility | Present and increasing as flow falls | Chapter 13's weaning signal — "pulse contour increases when flow is decreased" — becomes a criterion here. Pulse pressure below 30 mmHg was independently associated with weaning failure. In meta-analysis, weaned patients had pulse pressure higher by 12.7 mmHg (7.3–18) |
Central venous pressure | Under 15 mmHg, or a rise of less than 5 mmHg during the trial | A rising CVP as flow falls is the right ventricle telling you it cannot take the returning volume |
Lactate | Not rising | Lactate clearance in the first 12 hours predicted weaning outcome (AUROC 0.72), and lactate at 24 hours was independently associated with success (OR 0.52, p=0.018). Initial and pre-weaning values were inconsistent |
18.5.2 Echocardiography — and which parameter actually performs
Evidence — the 2026 meta-analysis, and the parameter that beats all the clever ones
A meta-analysis searching six databases to December 2025 pooled 37 studies and 3,458 patients, using bivariate random-effects modelling for diagnostic accuracy. The hierarchy it produced:
Parameter | AUC (95% CI) |
Aortic valve opening status | 0.88 (0.82–0.93) |
LVOT velocity–time integral | 0.85 (0.81–0.88) |
Tissue Doppler mitral annular systolic velocity | 0.81 (0.76–0.86) |
Left ventricular ejection fraction | 0.79 (0.75–0.83) |
Provisional thresholds, derived mainly from studies using a 48-hour decannulation definition: LVEF 20–25%, LVOT VTI at or above 10 cm, TAPSE at or above 17 mm — and the authors state these "should be interpreted as exploratory rather than confirmatory."
Certainty: low, and the authors say so. The pooled studies are the same heterogeneous retrospective literature described below; a meta-analysis of biased studies is a precise summary of biased studies.
But the ordering is the finding worth carrying. The parameter at the top is not a strain index or a coupling ratio. It is whether the aortic valve opens — the single observation this book has been pointing at since Chapter 13 §13.9, where the conclusion was "target the aortic valve opening, not the number," and again in Chapter 15 §15.3, where valve opening was the sensitive sign of distension while chamber size was not.
Three chapters arrived at the same answer from three different problems. On VA ECMO, the aortic valve is the highest-yield thing to look at.
Evidence — the conventional triad, and where it came from
The criteria in every textbook come from a 51-patient study published in 2011. Patients underwent flow reduction to under 1.5 L/min under clinical and Doppler monitoring; 38 tolerated at least one trial and 20 were ultimately weaned. Every successfully weaned patient had, at minimal flow:
Aortic VTI at or above 10 cm · LVEF above 20–25% · lateral mitral annulus peak systolic velocity (TDSa) at or above 6 cm/s.
These separated weaned from non-weaned patients better than anything else tested, and the authors closed by asking for validation.
Certainty: low. Fifty-one patients, one centre, twenty events. Fifteen years later the 2022 systematic review found LVOT VTI is still "the most widely used parameter to track LV recovery" — while noting that the common threshold it found was above 9.5 cm, and that "the threshold itself and the conditions under which it is measured vary significantly across studies, making comparisons difficult." Reported discrimination for VTI ranged from AUROC 0.85 to 0.74 in different studies.
The triad is a floor, not a target, and it was never validated as one. A patient who fails it will almost certainly fail decannulation. A patient who meets it has cleared the bar that 20 patients in 2011 cleared.
Clinical pearl — the newer parameters all share one idea: measure the change, not the value
The conventional triad is measured at minimal flow. The parameters that outperform it are measured as a response to changing the flow, and that is not a coincidence — a value at one loading condition describes a state, whereas a change across two describes reserve.
— Tissue Doppler response. In a 92-patient multicentre study, flow was reduced by 30–50% for 15 minutes; 64 of 92 were weaned. Improvement in lateral e′ velocity and tricuspid annular S′ velocity during the flow study predicted success, and the model using them was "much higher" in predictability than the conventional triad. The 2022 systematic review reports this combination at AUROC 0.93.
— Left atrial reservoir strain. In a 120-patient prospective registry, the relative increase in reservoir strain during the trial was 39.5% in those weaned versus −1.2% in those who failed (p under 0.001); adjusted OR 1.02 per 1% (1.01–1.03); adding it to conventional criteria moved discrimination from AUC 0.75 to 0.83 (p=0.010).
— Total isovolumic time, a load-independent performance index: its improvement in the first 48 hours was "the strongest predictor" in one analysis.
— Corrected LV ejection time indexed to wedge pressure. In a 50-patient study, LVETc/PAWP above 15.9 was an independent predictor (OR 0.82, 0.71–0.94, p=0.005; AUROC 0.82).
The bedside translation does not require any of these to be available in your unit. It requires the habit: scan at baseline flow, change the flow, scan again, and compare. The delta is the information.
18.5.3 Controversy 2 — does ejection fraction predict anything?
Controversy 2 — Is left ventricular ejection fraction a useful weaning criterion?
The question. LVEF above 20–25% is in every protocol, including the Red Book's. Two well-conducted studies reach opposite conclusions about whether it has any independent predictive value.
Position A — LVEF is the only thing that independently predicts. An 85-patient referral-hospital series found 61% successful weaning, and on multivariable analysis LVEF was the only independent predictor (hazard ratio 0.938, 0.888–0.991, p=0.02). LVEF above 33.4% was the optimal cutoff (AUC 0.808, sensitivity 93%, specificity 72%) and was associated with survival at discharge of 60% versus 20% (p under 0.001) — that is, it predicted the endpoint that actually matters.
Position B — LVEF predicts nothing. A 76-patient observational study testing longitudinal function and cardiac time intervals found that t-IVT under 14.4 s/min, LVOT VTI above 12.3 cm, MAPSE above 8.9 mm, TAPSE above 16 mm and E/e′ under 15.5 were all associated with weaning success and free survival afterwards — while "LVEF did not predict the weaning success and survival at any time-point (p=0.230)", despite changing dynamically during the trial.
What the evidence actually shows. The 2022 systematic review supplies the mechanism that reconciles them, and it is one word: load. It notes that LVEF, LVOT VTI, LV systolic velocity, strain and strain rate are all load-dependent — and a patient on VA ECMO is in the most artificial loading conditions in medicine, with preload removed by the drainage cannula and afterload imposed by the return cannula (Chapter 13 §13.3).
Ejection fraction is the most load-dependent of the four, because it is a ratio of volumes and both volumes are being manipulated. At high circuit flow an underfilled ventricle can show a flattering ejection fraction while ejecting almost nothing; the 2026 meta-analysis ranks LVEF last of four at AUC 0.79, and the systematic review reports that absolute LVEF and LVEF change were independently associated with success only at 48 hours, with a mean improvement of 9.0% (p=0.001) in weaned patients — that is, the trajectory carried the signal, not the value.
The resolution this book proposes: the two studies disagree because they measured LVEF under different loading conditions and asked it to do different jobs. As a single value at minimal flow, it is a weak and load-contaminated parameter — Position B is right. As a trajectory over the first 48 hours of support, and at a cutoff far above the conventional one, it carries real prognostic information — Position A's 33.4% is nowhere near the textbook 20–25%, and that gap is the clue.
Certainty: moderate that LVEF is load-dependent and a poor single-timepoint discriminator. Low that any specific cutoff is useful.
What to take to the bedside. Stop using "LVEF above 20–25%" as a gate to pass and start using it as a floor below which you should not be attempting decannulation at all. The useful questions are how much has it improved since cannulation, and does the valve open — and for the second, you do not need to measure anything.
18.5.4 The right ventricle — where the weaning study is weakest
Danger — you are assessing the right ventricle under conditions it will never face again
Three problems compound here.
1. The load is wrong. In a flow-reduction trial the drainage cannula is still taking venous return, so full RV loading is not achieved (§18.3). The RV passes a test it was never given.
2. No single parameter works. The Red Book: "No one parameter can predict this. Clinicians must depend on serial hemodynamic and echocardiography evaluation of the RV." Its Table 29-2 accordingly uses a count rather than a threshold: fewer than 2 of — RV end-diastolic dimension at or above 35 mm, TAPSE under 1.5 cm, tricuspid annular S′ under 10 cm/s, poor RV ejection fraction, severe tricuspid regurgitation.
3. TAPSE is unreliable in exactly the patients who need it most. The Red Book states that TAPSE is normally reduced after lung transplantation and cardiac surgery, and "is not reliable in the assessment of weaning of RV support in these patient groups." Post-cardiotomy patients are a large share of the VA ECMO population.
And the stakes are asymmetric. RV assessment during the weaning study is what determines, in a patient heading for a durable left ventricular assist device, whether they will also need right ventricular support. Getting that wrong does not produce a failed trial; it produces an LVAD implanted into a circulation that cannot fill it.
If the RV is the question, use a bridge or PCRTO, and assess it serially rather than once.
Evidence — the right-heart parameters that have actually been tested
Three-dimensional RV ejection fraction. In 46 patients studied before the first intent of decannulation, 3D RVEF had the highest area under the curve of any parameter (0.90, p under 0.001) with a cutoff of 24.6%. Worse function — RVEF at or below 24.6% — carried a hazard ratio for 30-day all-cause mortality of 15.86 (95% CI 3.56–70.73, p under 0.001). Certainty: low — 46 patients, retrospective, and 3D RV volumetry is not available in most units.
RV coupling to the pulmonary circulation. In 79 prospectively enrolled patients scanned at a median of 3 days after initiation and a median flow of 3.2 L/min, indices formed by dividing RV function by RV systolic pressure — tricuspid annular S′/RVSP, TAPSE/RVSP and RV free-wall strain/RVSP — all performed well, and S′/RVSP was significantly better than the conventional triad. The Red Book reports the derived threshold as S′/RVSP above 0.33, and notes it is measured at maintained flow, not at minimum flow.
Certainty: low — single centre, 79 patients, awaiting external validation.
Why coupling ratios are conceptually right even if the numbers are provisional. A contractility measure alone (TAPSE, S′) tells you what the ventricle is doing; dividing by the pressure it is doing it against tells you what it is doing relative to its load — which is the only question that survives the loading artefacts described in §18.5.3. This is the right-sided analogue of the argument that ejection fraction is load-contaminated, and it is the same reasoning Chapter 15 applied to the left ventricle through the pressure–volume area.
18.5.5 The signals that are not echocardiography
Signal | What it shows | Caveats |
End-tidal CO₂ | Chapter 13 established it as a surrogate for transpulmonary blood flow. Here it becomes a recovery signal: an increase of 5 mmHg or more above previous mean values across two consecutive 12-hour periods occurred in all successfully weaned patients, rising from 9 mmHg after cannulation to 21 mmHg at 24 hours (p=0.04). Taha notes the rise preceded changes in haemodynamic monitoring and cardiac index | Depends on pulmonary dead space, which is increased after cardiac arrest, and varies with ventilator settings. It is free and continuous, which is its whole value |
Lactate and its clearance | Clearance in the first 12 hours (AUROC 0.72); value at 24 hours independently associated (OR 0.52) | Initial and pre-weaning values were inconsistent across studies |
CK-MB and troponin | Lower peak CK-MB predicts success — peak CK-MB under 183 U/L predicted weaning with sensitivity 86% and specificity 71%. Peak troponin associated in infarct patients (p=0.003) | These are markers of how much myocardium was lost, not of recovery. They are prognostic from day one and do not change the daily decision |
Natriuretic peptides and other cardiac markers | Do not use them for this. The ISCCM manual reports that troponin I, NT-proBNP, proadrenomedullin and copeptin are all raised in refractory cardiogenic shock on VA ECMO, but "their kinetics do not predict any cardiac recovery during the first week of the ECMO support" | One of the few clearly negative findings in this literature, and worth knowing |
Microcirculation | Skin blood flow at or above 34 perfusion units by laser Doppler predicted weaning success with AUROC 0.93; perfused small-vessel density was higher in weaned patients (p=0.002) | Research tools. Recorded because the discrimination is striking and because they measure the thing the whole exercise is about — whether tissue is being perfused |
Right-heart pressures | Successfully weaned patients had lower right atrial to pulmonary capillary wedge pressure ratio and higher pulmonary artery pulsatility index at 48 and 72 hours | Requires a pulmonary artery catheter, whose cardiac output number remains uninterpretable on VA support (Chapter 13 §13.5). The pressures are usable; the output is not |
18.6 Four situations that change the trial
Pitfall — weaning a ventricle that is being vented
Chapter 15 may have placed a left atrial or left ventricular vent. The Red Book is explicit about what that does to a weaning trial: "Weaning is complicated in the presence of an LV vent because it further reduces LV preload... when the ECMO flow is reduced, the venting cannulas will still be accessing blood from the LA or LV, and the LV preload will not be optimized."
And the effect is not uniform: "the effect on LV preload of a transapical LV vent is greater because the cannula is shorter and broader. Therefore, even at lower ECMO flow, more blood may be drained from LV than the RA."
Read that last clause carefully. In a vented patient at low circuit flow, the vent can become the dominant drain — so you are not testing a ventricle with restored preload at all. Every LV parameter in §18.5 is measured under conditions the vent is still distorting.
The practical consequence is that the vent has to be part of the weaning protocol, not a fixed background condition. Chapter 15 §15.6.1 noted that a transseptal cannula is explanted with the circuit and gives no support during weaning, whereas a micro-axial pump can outlast it — that difference now determines whether your weaning trial is testing the heart or testing the vent.
Situation | What changes |
Isolated right ventricular failure | The Red Book gives a separate protocol (Table 29-3) and it is markedly more conservative: stepwise reduction to 2.0 L/min, then stable haemodynamics on low vasoactive support for at least 24 hours at that flow before going further, then echo-guided reduction of 0.5 L/min every 15 minutes to minimum. RV criteria: no RV dilatation with reduction in LV end-diastolic volume; recruitment of RV systolic function to normal or mild dysfunction; TAPSE above 10 mm; no worsening tricuspid regurgitation; increment in RVOT VTI — for which the Red Book states no cut-off value has been defined and that the criteria are "based on authors' personal experience." For an oxygenated RVAD, turn off fresh gas flow for more than 2 hours and check a gas first |
The patient who is hypoxic rather than hypotensive | A trial that reveals adequate cardiac function but inadequate gas exchange is not a failed wean — it is an indication to reconfigure from VA to VV (Chapter 16 §16.10). The Red Book notes the AV bridge is particularly useful here because it separates the two questions. And Chapter 16's paradox applies: differential hypoxaemia is itself evidence of cardiac recovery, so the patient in whom it appears is often the patient closest to weaning |
The post-cardiac-arrest patient | In a 100-patient series, cardiac arrest patients were far less likely to be weaned: only 20.3% of arrest patients were successfully removed, against a much higher rate in those cannulated for shock alone (p under 0.001). Overall weaning in that cohort was 31% at a median of 140 hours. The 2022 systematic review deliberately excluded ECPR-only studies, because neurological prognosis changes the decision entirely (Chapters 23–24) |
The patient with no destination | The Red Book describes the one-way wean: in irreversible dysfunction with ineligibility for transplant or durable support, clinicians may proceed "after optimizing as many potentially reversible conditions as possible." This is an end-of-life decision wearing the clothes of a physiological test, and it should be named as one — it belongs to Chapters 74 and 75, and it should never be undertaken without the conversation those chapters describe having already happened |
18.7 Failing the trial is not the same as failing to wean
Clinical pearl — a failed first trial is a common route to a successful second one
A 57-patient study using a weaning test of flow under 2 L/min for at least 60 minutes reported that 36 (63.2%) were eventually weaned, but only 31 (54.4%) after the first test. Of those who failed the first test, 3 of 7 were weaned after a second test, and 3 of 4 after a third.
The same study identified what makes failure likely: pre-existing ischaemic heart disease (OR 9.6, 1.1–83); pre-test LVEF at or below 25% and/or post-test LVEF at or below 40% (OR 11, 0.98–115); post-test systolic blood pressure at or below 120 mmHg (OR 33, 3–385); and duration of support beyond 7 days (OR 24, 2–269). Note how wide those intervals are — certainty: very low, 57 patients.
A separate series found failed first tests are frequent even in selected patients, describing the test as failing "in less than 40% of patients considered suitable for weaning."
Two things follow. First, a failed trial is information about today, not a verdict — unless the failure was catastrophic, the correct response is to identify what failed (preload, rhythm, the right ventricle, the lungs), fix it, and retest. Second, the duration signal cuts the other way: every day of support adds complication risk (Chapters 15–17) and predicts weaning failure. Repeat trials are worth doing and they are not worth doing indefinitely, and the point at which that balance tips is a destination conversation, not a physiological one.
18.8 Decannulation
Danger — decide whether you could put this patient back on before you take them off
The Red Book's instruction is one sentence and it is the most important operational line in the chapter: "patient suitability for re-instituting VA ECMO always needs to be made before decannulation."
Some patients fail decannulation despite everything in §18.5. If the answer to "would we re-cannulate this patient?" is no, then decannulation is a one-way decision and everyone in the room — including the family — should know that before the cannulas come out, not afterwards.
The 2023 Perfusion review makes the same argument as a system requirement: teams should have "a bailout strategy when weaning fails" and should implement predetermined contingency plans rather than attempting urgent re-cannulation without preparation.
Step | Detail |
1. Hold at partial support | After a successful weaning study the patient is left on partial VA ECMO support while awaiting decannulation — the study and the removal are separate events |
2. Isolate and observe | The cannula is disconnected from the circuit, flushed with heparinised saline, and kept clamped for a period — usually 15 minutes. Decannulation proceeds if the patient remains stable |
3. Remove the arterial cannula — usually surgically | Surgically placed cannulae must come out surgically. For percutaneously placed femoral arterial cannulae, the Red Book states open surgical removal is commonly performed — because the cannula is large and patients may need vascular patch repair. Some require embolectomy to prevent limb ischaemia from distal embolisation of residual thrombus (Chapter 17) |
4. Treat it as a vascular operation | Chapter 17 §17.10 records that technical problems during explantation are an independent predictor of late arterial stenosis at the cannulation site, and that dissection, pseudoaneurysm and retroperitoneal bleeding occur in 7–14% of patients across placement and removal combined. How the cannula comes out determines a complication that may not declare itself for months |
5. Continue limb surveillance afterwards | Bilateral regional oximetry and examination do not stop at decannulation (Chapter 17 §17.5) |
Pitfall — the sweep-gas-off trial has no place here, and this is the fifth chapter to say so
Chapters 10, 11, 12 and 13 all warn of it; this is where the warning must be repeated, because decannulation is the moment somebody reaches for the VV habit.
On VV ECMO, turning the sweep gas off makes the circuit inert and is the definitive weaning trial (Chapter 10). On VA ECMO it does the opposite: the pump keeps circulating blood, but that blood is no longer being oxygenated, so the circuit becomes a device that actively pumps deoxygenated blood into the aorta — Chapter 13 §13.4's reverse harlequin effect, described by its source as "something that should NEVER be done."
Note the one legitimate exception, and why it is not an exception at all: during PCRTO the circuit oxygen is deliberately removed (§18.3). That is safe precisely because the flow is retrograde — the circuit is no longer delivering into the aorta. The rule is about the direction of flow, not about the gas.
18.9 After the cannulas are out
Danger — the period after decannulation is when a large share of these patients die
Return to §18.1. Pooled successful weaning is 45%; pooled in-hospital mortality is 46.6%; and one review reports mortality of up to 70% after initially successful weaning. Reported weaning success across the 2022 systematic review's 47 studies ranged from 30 to 75% — a spread that reflects differing definitions as much as differing practice.
What kills them is mostly not the heart. By the time a patient is decannulated they have typically had days to weeks of shock, anticoagulation, transfusion, immobility and vascular instrumentation. The complications of Chapters 15, 16 and 17 do not resolve when the circuit is removed, and several — late arterial stenosis, neurological injury, deconditioning — declare themselves afterwards.
Three practical consequences.
1. Decannulation is not the end of ECMO care. Chapter 10 established what follows removal on the venous side — post-decannulation systemic inflammatory response, reported in around 60% of patients, and deep vein thrombosis. Both apply here.
2. The patient who weaned is not a step-down patient. Haemodynamic deterioration in the first 48 hours is the commonest definition of weaning failure in the literature, and it happens in a patient who has just met every criterion.
3. This is the argument for the destination conversation happening before weaning, not after it. The 2023 review's recommendation is that patients at high risk of weaning failure should be identified early, "enabling timely transportation to an advanced heart failure center" — that is, the decision about where this patient is going should be made while there is still time to move them.
18.10 When the heart does not recover
Evidence — failure to wean is a referral, not a verdict
The Red Book: "Despite treatment optimization, some patients cannot be weaned and require further evaluation for durable MCS or HTx."
The pivotal assessment is the one §18.5.4 says is hardest: "One of the primary considerations in patients eligible for LVAD is RV function, to determine whether the patient needs RVAD (temporary or durable) along with LVAD. Therefore, determining RV function on echocardiography during the weaning study is critical."
This is the practical reason the weaning study matters even when it fails. A trial that a patient fails still generates the data that determines what device they get. A left ventricle that will not eject and a right ventricle that copes points to an LVAD; both failing points to biventricular support or transplant.
And where neither is possible, the Red Book states the remaining path plainly: "In patients deemed ineligible for long-term MCS, clinicians must decide about a one-way wean from VA ECMO or palliative care measures, after discussions with patients and families."
One therapy has been formally tested for improving weaning: levosimendan, the subject of a systematic review and meta-analysis in adult cardiogenic shock patients on VA ECMO. That review was not retrieved in this session and its results are deliberately not stated here — see the open items. Taha separately records that "the value of intra-aortic balloon pumps during VA ECMO weaning has not been evaluated."
18.11 The errors that recur
Error | Correction |
Treating "successfully weaned" as "survived" | They are different endpoints and the gap is up to 70%. The literature predicts device removal, not survival |
Starting a trial before the cause is treated | You are testing the insult, not the heart |
Trialling a hypovolaemic patient | Reducing flow returns volume to a chronically underfilled ventricle. Give the preload bolus first |
Assessing the right ventricle during a flow-reduction trial | The drainage cannula is still unloading it. If the RV is the question, use a bridge or PCRTO |
Reading a minimum-flow duration off someone else's protocol | The durations differ because the circuit states differ. Ask what your oxygenator is rated for and whether there is clot in the circuit |
Using LVEF above 20–25% as a gate to pass | It is the most load-dependent parameter available and ranks last of four on discrimination. Use the trajectory, and use the valve |
Measuring only at minimum flow | The parameters that outperform the conventional triad all measure the change across a flow reduction. Scan before and after |
Trusting TAPSE after cardiac surgery or lung transplantation | It is normally reduced in those groups and the Red Book says it is not reliable there |
Forgetting the vent is still draining | At low circuit flow a transapical vent can become the dominant drain. The vent belongs in the weaning protocol |
Using natriuretic peptides or troponin kinetics to judge recovery | Their kinetics do not predict cardiac recovery in the first week |
Turning the sweep gas off on a VA circuit | It pumps deoxygenated blood into the aorta. The exception during PCRTO is safe only because the flow is retrograde |
Abandoning weaning after one failed trial | Around a third of eventual successes come from a second or third attempt |
Repeating trials indefinitely | Support beyond 7 days independently predicts weaning failure and accumulates the complications of Chapters 15–17 |
Decannulating without deciding whether you would re-cannulate | If the answer is no, this is a one-way wean and should be named as one before the cannulas come out |
Pulling a large femoral arterial cannula percutaneously by reflex | Open removal is commonly needed; some patients require patch repair or embolectomy |
Treating decannulation as the end of ECMO care | SIRS, deep vein thrombosis, limb surveillance and late arterial stenosis all belong to the period afterwards |
18.12 Key points
- Weaning and surviving are different endpoints. Pooled successful weaning is 45%; pooled in-hospital mortality is 46.6%; one review reports mortality up to 70% after initially successful weaning. Every threshold in this chapter predicts that the circuit can come out.
- Ask the two questions separately: can this circulation work without the machine, and where is this patient going? The second is Chapters 70–75, and it sometimes changes when you ask the first.
- Preconditions first: the cause treated, organs stabilising, FiO₂ at or below 0.6, full anticoagulation, low vasoactive support, and a preload bolus before the trial.
- Three techniques, and the difference that matters is right ventricular loading. Flow reduction leaves the RV partly unloaded; the AV bridge and PCRTO both achieve full loading, and PCRTO adds resistance.
- PCRTO reverses the flow — pump speed under about 1000 RPM gives 0.5–1 L/min retrograde, the pump becomes a brake, circuit oxygen is removed, and the distal perfusion cannula must be disconnected and flushed. Its risk is systemic flow steal.
- Minimum-flow duration is disputed from under 15 minutes to 2 hours, and the disagreement is about circuit state, not time. The rule that survives: never go below the oxygenator's rated minimum flow, and shorten the trial if the circuit already has clot.
- Aortic valve opening is the best-performing single echocardiographic parameter (AUC 0.88), ahead of LVOT VTI, tissue Doppler velocity and ejection fraction. Three chapters of this book now converge on the same observation.
- The conventional triad — VTI at or above 10 cm, LVEF above 20–25%, TDSa at or above 6 cm/s — comes from 51 patients and 20 events. It is a floor, not a target, and it has never been validated as one.
- Ejection fraction is the most load-dependent parameter in a maximally load-distorted patient. Two good studies disagree about it because they measured it differently. Use the trajectory over 48 hours, not the value at minimum flow.
- The parameters that outperform the triad all measure change across a flow reduction, not a value at one flow. Tissue Doppler response, left atrial reservoir strain, isovolumic time. Scan before and after — the delta is the information.
- The right ventricle is where the assessment is weakest: wrong loading conditions, no single parameter, and TAPSE unreliable in post-cardiotomy and post-transplant patients. Use a count of abnormalities, assess serially, and prefer coupling ratios that divide function by the pressure it works against.
- End-tidal CO₂ is a free recovery signal — a rise of 5 mmHg or more across two consecutive 12-hour periods occurred in all successfully weaned patients in one study, and it preceded the haemodynamic changes.
- Cardiac markers do not track recovery. Troponin, NT-proBNP, proadrenomedullin and copeptin kinetics do not predict recovery in the first week.
- A vent distorts the trial. At low circuit flow a transapical vent can drain more than the drainage cannula. Include the vent in the protocol.
- Failing one trial is not failing to wean — a third of eventual successes come after a second or third attempt. But support beyond 7 days independently predicts failure, so repetition has a limit that is a destination decision.
- Decide before decannulating whether you would re-cannulate. If not, this is a one-way wean, and it is an end-of-life decision in physiological clothing.
- Removing a large femoral arterial cannula is a vascular operation, often open, sometimes requiring patch repair or embolectomy, and technical difficulty at explantation predicts late arterial stenosis.
- The sweep-gas-off trial is a VV manoeuvre. On VA it produces the reverse harlequin effect. PCRTO's oxygen removal is safe only because the flow is retrograde.
- Decannulation is not the end of ECMO care. SIRS, deep vein thrombosis, limb surveillance and the mortality in §18.9 all follow it.
- A failed weaning study still generates the data that chooses the next device — above all, whether an LVAD candidate also needs right ventricular support.
[VERIFICATION REQUIRED] — open items in this chapter
- Two reviews were retrieved in full: the 2022 Critical Care systematic review (DOI 10.1186/s13054-022-04249-w) and the 2023 Perfusion practice-oriented review (DOI 10.1177/02676591221115938). A fourth first-author discrepancy affects the first: a bibliographic index returns it under Charbonneau, while the retrieved article named Cavayas as lead author. DOI, journal, year, scope and findings agree. No author list is asserted.
- Everything else in the external set was read as structured abstracts — the 2026 meta-analysis of 37 studies, the founding 51-patient study, the tissue-Doppler, left-atrial-strain, RV-coupling, 3D-RVEF, ejection-time, machine-learning and weaning-failure cohorts, and the three other meta-analyses. Author lists, volumes and page numbers are not asserted except where a retrieved source supplied them.
- Two numerical problems in the Red Book's own weaning chapter are flagged and not corrected: the vasoactive-inotropic score threshold given as under 10 in Table 29-1 and under 30 in Table 29-2's footnote, and the low-dose norepinephrine definition printed as "under 0.05 µg/min" where every other agent in the same list is in µg/kg/min.
- The isolated-RV-failure protocol (Table 29-3) is explicitly stated by the Red Book to be "based on authors' personal experience", and it records that no cut-off value has been defined for RVOT VTI. It is reproduced as a described practice, not as evidence.
- No weaning threshold anywhere in this chapter has randomised support. The 2022 systematic review found most studies at moderate or high risk of bias, that 15 of 47 were conference abstracts without protocols, and that "heterogeneity in methods, timing, and conditions of measurements precluded any meta-analysis." The 2026 meta-analysis describes its own thresholds as "exploratory rather than confirmatory."
- Reported weaning success ranges from 30 to 75% across studies, largely because the definition varies. The commonest is survival after complete removal without further mechanical support or transplant, usually over 30 days — but some studies use 48 hours, which is a materially different question.
- The up to 70% post-weaning mortality figure is quoted from the 2023 review's own framing; the primary source behind it was not retrieved, and the figure should be read as the upper end of a range, not a central estimate.
- The levosimendan meta-analysis referenced by the Red Book was not retrieved, and its results are deliberately not stated. Do not infer a direction from its inclusion. The same applies to the 2020 EACTS/ELSO/STS/AATS post-cardiotomy expert consensus, which the Red Book cites and which this book has not read — a high-value retrieval target for Part XII.
- Specific unverified figures: peak CK-MB under 183 U/L (sensitivity 86%, specificity 71%); skin blood flow at or above 34 perfusion units (AUROC 0.93); LVETc/PAWP above 15.9; 3D RVEF 24.6% and its hazard ratio of 15.86 (3.56–70.73); S′/RVSP above 0.33; LVEF above 33.4%. Several derive from single studies of fewer than 100 patients, and the confidence intervals on the weaning-failure odds ratios in §18.7 are so wide as to be nearly uninformative.
- The resolution offered in §18.4 — that the duration disagreement is about circuit state rather than time — and the resolution in §18.5.3 — that the two LVEF studies differ because of loading conditions and the job asked of the parameter — are this book's reasoning. The underlying facts are cited; the syntheses are not published positions.
- The observation in §18.5.2 that three chapters of this book independently converge on aortic valve opening is likewise this book's, not a published finding.
- Red Book, ISCCM Manual and Taha are cited by chapter; editions, editors, years and page numbers are not confirmed.
Cross-references
- Chapter 10 — VV ECMO Weaning and Decannulation: the same task in the configuration where the sweep-gas-off trial is the definitive test. Worth reading alongside for the contrast
- Chapter 11 — VA ECMO: Indications and Patient Selection: "prognosis is not indication", and the bridge destination declared at cannulation that this chapter finally cashes
- Chapter 12 — VA ECMO Cannulation: what has to be undone, and why open removal is usually needed
- Chapter 13 — VA ECMO Haemodynamics: §13.4 pulsatility rising as flow falls — the weaning signal; §13.5 why the PA catheter's output number stays unusable; §13.9 "target the aortic valve opening, not the number"
- Chapter 15 — LV Distension and LV Unloading: the vent that distorts the trial, and the device that outlasts the circuit
- Chapter 16 — Differential Hypoxaemia: the reverse harlequin effect behind the sweep-gas warning, and conversion to VV as an alternative to weaning
- Chapter 17 — Limb Ischaemia and Vascular Complications: decannulation as a vascular operation, distal embolisation, and late arterial stenosis
- Chapter 25 — Haemodynamic Monitoring and Chapter 26 — Echocardiography During ECMO: how to acquire the measurements this chapter interprets
- Chapters 23–24 — Neurological Outcomes, ECPR Failure and Withdrawal: why the ECPR patient is excluded from this literature
- Chapters 70–73 — Bridge to Decision, Recovery, Transplant and LVAD: where failure to wean leads
- Chapters 74–75 — Futility and Withdrawal: where the one-way wean belongs
- Chapter 91 — Decannulation Checklist: the bedside form of §18.8
References
Retrieved in full
- Parameters associated with successful weaning of veno-arterial extracorporeal membrane oxygenation: a systematic review. Critical Care. 2022;26:375. DOI 10.1186/s13054-022-04249-w. [VERIFICATION REQUIRED] — first author given as Charbonneau by one index and Cavayas by the retrieved article; no author list asserted. 47 studies from 14,578 records; source for the biomarker, haemodynamic, microcirculatory and load-dependence material.
- Hermens JAJ, Meuwese CL, Szymanski MK, et al. Patient-centered weaning from venoarterial extracorporeal membrane oxygenation: a practice-oriented narrative review of literature. Perfusion. 2023;38(7):1349–1359. DOI 10.1177/02676591221115938. Source for the post-weaning mortality argument, the bailout strategy and early identification of high-risk patients.
Syntheses — read as structured abstracts
- Niu M, Feng. Meta-analysis of the predictive value of critical care echocardiography for weaning outcomes in patients with VA-ECMO-assisted cardiogenic shock. Frontiers in Medicine. 2026. DOI 10.3389/fmed.2026.1835564. 37 studies, 3,458 patients; the AUC hierarchy.
- Hsu H, et al. Predictors of successful weaning from veno-arterial extracorporeal membrane oxygenation: a systematic review and meta-analysis. PLOS One. 2025. DOI 10.1371/journal.pone.0310289.
- Burgos L, et al. Multiparameters Associated to Successful Weaning from VA ECMO in Adult Patients with Cardiogenic Shock or Cardiac Arrest: Systematic Review and Meta-Analysis. Annals of Cardiac Anaesthesia. 2023. DOI 10.4103/aca.aca_79_22. Pooled weaning 45%, in-hospital mortality 46.6%.
Cohorts — read as structured abstracts
- Aissaoui N, et al. Predictors of successful extracorporeal membrane oxygenation (ECMO) weaning after assistance for refractory cardiogenic shock. Intensive Care Medicine. 2011. DOI 10.1007/s00134-011-2358-2. The founding 51-patient study and the conventional triad.
- Kim D, et al. Echocardiographic Predictors of Successful Extracorporeal Membrane Oxygenation Weaning After Refractory Cardiogenic Shock. Journal of the American Society of Echocardiography. 2020. DOI 10.1016/j.echo.2020.12.002. Lateral e′ and tricuspid annular S′ response.
- Kim D, et al. Prognostic Implication of RV Coupling to Pulmonary Circulation for Successful Weaning From Extracorporeal Membrane Oxygenation. JACC: Cardiovascular Imaging. 2021;14(8):1523–1531. DOI 10.1016/j.jcmg.2021.02.018.
- Huang K-C, et al. Three-Dimensional Echocardiography-Derived Right Ventricular Ejection Fraction Correlates with Success of Decannulation and Prognosis in Patients Stabilized by Venoarterial Extracorporeal Life Support. Journal of the American Society of Echocardiography. 2017. DOI 10.1016/j.echo.2017.09.004.
- Tavazzi G, et al. Echocardiographic parameters for weaning from extracorporeal membrane oxygenation — the role of longitudinal function and cardiac time intervals. European Heart Journal: Cardiovascular Imaging. 2024. DOI 10.1093/ehjci/jeae274. Position B in Controversy 2.
- Alonso-Fernandez-Gatta M, et al. Echocardiographic Prediction of Successful Weaning From Venoarterial Extracorporeal Membrane Oxygenation. American Journal of Critical Care. 2022. DOI 10.4037/ajcc2022588. Position A in Controversy 2.
- Cusanno A, et al. Predictors of weaning failure in case of VA ECMO implantation. Scientific Reports. 2022. DOI 10.1038/s41598-022-18105-y. Repeat trials and predictors of failure.
- Sawada K, et al. Predicting Parameters for Successful Weaning from Veno-Arterial Extracorporeal Membrane Oxygenation in Cardiogenic Shock. ESC Heart Failure. 2020. DOI 10.1002/ehf2.13097. LVETc/PAWP.
- Shin H, et al. Clinical Significance of Change in Left Atrial Strain During Weaning from Venoarterial Extracorporeal Membrane Oxygenator. Journal of the American Society of Echocardiography. 2025. DOI 10.1016/j.echo.2025.10.012.
Techniques — cited by the Red Book; not retrieved
- Babar ZUD, Sharma AS, Ganushchak YM, et al. An arterio-venous bridge for gradual weaning from adult veno-arterial extracorporeal life support. Perfusion. 2015;30(8):683–688.
- Ling L, Chan KM. Weaning adult patients with cardiogenic shock on veno-arterial extracorporeal membrane oxygenation by pump-controlled retrograde trial off. Perfusion. 2018;33(5):339–345.
- Ju MH, Lim MH, Lee SY, et al. Early experience of pump-controlled retrograde trial off for weaning from veno-arterial extracorporeal membrane oxygenation in adult patients with cardiogenic shock. Perfusion. 2021;36(4):401–406.
- Westrope C, Harvey C, Robinson S, et al. Pump controlled retrograde trial off from VA-ECMO. ASAIO Journal. 2013;59(5):517–519.
- Aissaoui N, El-Banayosy A, Combes A. How to wean a patient from veno-arterial extracorporeal membrane oxygenation. Intensive Care Medicine. 2015;41(5):902–905.
- Fried JA, Masoumi A, Takeda K, et al. How I approach weaning from venoarterial ECMO. Critical Care. 2020;24(1):307.
- Lorusso R, Whitman G, Milojevic M, et al. 2020 EACTS/ELSO/STS/AATS expert consensus on post-cardiotomy extracorporeal life support in adult patients. European Journal of Cardio-Thoracic Surgery. 2021;59(1):12–53. Not retrieved — high-value target.
Textbooks
- ELSO Red Book, 6th edition, Chapter 29 — Weaning and Decannulation in Adult Cardiac Failure: Table 29-1 preconditions, Table 29-2 predictors, Table 29-3 isolated RV failure, the three weaning techniques, weaning a vented ventricle, failure to wean, and the decannulation sequence. Plus the paediatric weaning chapter for trial-off mechanics and the oxygenator minimum-flow rule. [VERIFICATION REQUIRED] — page numbers not confirmed.
- ISCCM Manual of RRT and ECMO in ICU, Chapter 38 (weaning; the cardiac-marker negative finding; the difficult-weaning echo differential). [VERIFICATION REQUIRED].
- Taha AR, Caridi-Scheible M, Leiendecker E, et al. ECMO: A Practical Guide to Management, Chapter 11 — Weaning and Decannulation (the weaning protocol, the end-tidal CO₂ table, the 40–60 minute low-flow period, and the note that IABP during weaning has not been evaluated). [VERIFICATION REQUIRED].
Chapter status
Drafted and audited 11 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 III.
Two controversies were set out rather than smoothed. §18.4 collects five published minimum-flow durations spanning under 15 minutes to 2 hours and argues — as this book's reasoning — that the disagreement is about circuit state rather than elapsed time, with the oxygenator's rated minimum flow as the rule that survives. §18.5.3 sets two well-conducted studies against each other on whether ejection fraction predicts anything, and resolves them through load dependence and the difference between a single value and a trajectory.
The numerical audit found two problems inside the Red Book's own weaning chapter — a vasoactive-inotropic score threshold given as under 10 in one table and under 30 in another, and a norepinephrine dose printed in units inconsistent with every other agent in the same list. Both are flagged and neither is corrected.
The redundancy audit removed echocardiographic technique, which belongs to Chapter 26, and the destination decision, which belongs to Chapters 70–75; both are referenced rather than restated. The sweep-gas-off warning is repeated here deliberately — it is the fifth chapter to carry it, and decannulation is the moment the VV habit is most likely to surface.
The chapter's organising argument — that the entire weaning literature predicts device removal rather than survival — is drawn from the 2023 Perfusion review's own framing and from the arithmetic of the pooled figures, and is labelled as such.