🏁

Chapter 10 — VV ECMO Weaning and Decannulation

🧭

Chapter question: How do I know the native lung has recovered, and how do I get the patient off the circuit safely?

📅

Evidence search date: 6 September 2026. Principal source: ELSO Red Book 6th edition, Chapter 25, Weaning and Decannulation in Adult Respiratory Failure (Malley, Ramanan, Pellegrino, Harvey) — the most complete published treatment of this subject and the backbone of this chapter. Supplemented by the ISCCM Manual (Ch 38) and Taha, ECMO: A Practical Guide to Management.

What this chapter covers — and what it does not

This chapter owns

Deferred to

Recognising native lung recovery; the four steps of liberation; what weans what; ventilator management during weaning; the sweep gas off trial in full; pass and fail criteria; the decannulation procedure; post-decannulation SIRS and venous thromboembolism

The daily down-titration ladder → Chapter 8 Hypoxaemia during a failed trial → Chapter 9 VA weaning and decannulation → Chapter 18 Mobilisation and rehabilitation → Chapter 59 Cannula-site and vascular complications → Chapter 34 Anticoagulation management → Part VII Ventilation strategy in depth → Chapter 53 Neurological outcome and prognostication → Chapter 23 The printable decannulation checklist → Chapter 91

10.1 Weaning begins on the day of cannulation

💡

Clinical pearl — the sentence to keep in mind

Taha and colleagues put it directly: "The only thing worse than not placing a patient on ECMO soon enough is staying on ECMO longer than is required." Their operational conclusion is that the discontinuation plan should begin the moment the patient is placed on extracorporeal life support, and should be discussed daily on rounds.

This is the same discipline as Chapter 8's destination review, seen from the other end. A patient nobody is actively trying to liberate will not be liberated on time.

The risk this guards against is specific. Every additional day on ECMO is another day of anticoagulation, another day of blood–biomaterial contact, another day of immobility and deconditioning, another day of cannula-associated thrombus accumulating in a large vein, and another day of circuit-related risk that Chapters 8 and 9 catalogue. None of that is justified once the native lung can do the work.

10.2 The four steps

The ELSO Red Book frames liberation from VV ECMO as four sequential steps, and this chapter follows that structure:

Recognition of lung recovery (§10.3)
Decreasing the oxygenation and CO₂ removal provided by the circuit as native lung function improves (§10.4–10.5)
Performing a sweep gas off trial (§10.6)
Decannulation following a successful trial (§10.7)

📊

Evidence — how much of this is actually established

The Red Book is candid about the evidence base: no prospective randomised trials comparing weaning methods have been conducted, and the multiple reasonable approaches in use are "based on expert opinion."

What recent evidence does support is standardisation itself. Structured liberation protocols — analogous to the spontaneous awakening and spontaneous breathing trials used for sedation and ventilator liberation, and run either by physicians or by respiratory therapists and perfusionists — have shown reductions in both time to first sweep gas off trial and time to decannulation, without increased adverse events. Strikingly, they also "shown success with liberating patients from VV ECMO safely at higher levels of support than traditionally would have indicated a sweep gas off trial was warranted."

No single protocol has been shown superior to another. The Red Book's conclusion is that institutions should implement a local standardised VV ECMO liberation protocol — the standardisation matters more than which protocol is chosen.

Certainty: low for the specific steps (expert opinion); low to moderate for the benefit of protocolisation (consistent observational signal, no randomised trial).

That last finding deserves emphasis because it is a criticism of unaided clinical judgement. Protocols liberated patients successfully at support levels at which experienced clinicians would not have attempted a trial. The implication is that individual judgement is systematically too conservative about readiness — which costs patients days on the circuit.

10.3 Step 1 — Recognising lung recovery

The useful feature of VV ECMO is that recovery can be assessed while the patient is still on full support, because the native lung is being ventilated throughout at rest settings. Nothing has to be switched off to look.

Domain	Sign of recovery
Respiratory mechanics	Improved static compliance of the respiratory system (the ISCCM manual gives greater than 20 mL/cmH₂O as a marker); reduced airway resistance; larger tidal volume for the same pressure-control settings; increased functional residual capacity on chest radiograph
Gas exchange	Rising PaO₂ or falling PaCO₂ on unchanged rest settings; rising end-tidal CO₂ on rest settings — the single most useful early sign, because it means blood is again passing through ventilated alveoli
Circuit demand	A falling sweep gas flow requirement to maintain normocarbia — Chapter 8's earliest marker of recovery, and usually the first to appear
Imaging	Improved lung aeration on chest radiograph; resolving B-lines on lung ultrasound

Composite of the ISCCM Manual (Ch 38, Table 1) and Taha.

🫀

Physiology — why rising end-tidal CO₂ means recovery

In severe ARDS on lung-rest settings, much of the lung is either unventilated or ventilated but poorly perfused. Exhaled gas therefore carries little CO₂, and end-tidal CO₂ is low — a dead-space picture.

As alveoli are recruited and ventilation–perfusion matching improves, blood once again meets ventilated gas, and CO₂ begins to leave through the native lung rather than through the membrane. End-tidal CO₂ rises and sweep requirement falls, and they are the same event seen from two sides.

This is also why CO₂ clearance recovers before oxygenation (Chapter 8): CO₂ is far more diffusible, so partially recruited lung clears carbon dioxide well before it can oxygenate.

10.3.1 How much native function is enough?

Taha and colleagues report the working figure that ECMO discontinuation can be considered when the native lung is capable of supporting 50–80% of total gas exchange, and cite Mols and colleagues as having weaned patients with ARDS when 80% of total oxygen delivery was supplied by the patient's own lung.

🕳️

Pitfall — these percentages are orientation, not a test

The 50–80% figure is a way of describing a clinical impression, not something routinely measured at the bedside. Partitioning gas exchange between native lung and membrane requires simultaneous circuit and patient measurements that most units do not make.

Treat it as a mental model — is the lung doing most of the work yet? — and let the sweep gas off trial be the actual test. The Mols figure is quoted here second-hand through Taha; the primary study has not been retrieved.

10.4 Step 2 — Down-titrating support: what weans what

Chapter 8 gave the daily down-titration ladder. Here is the part that decides which control to turn:

To wean this	Turn this	Why
Oxygenation support	Extracorporeal blood flow (FsO₂ titration is optional and not required)	Oxygen transfer is limited by how much blood passes the membrane, because haemoglobin leaving the oxygenator is already near-saturated
CO₂ removal	Sweep gas flow	CO₂ clearance is limited by the gas-side gradient, which sweep flow sets

Source: ELSO Red Book 6th ed., Ch 25.

This produces two different weaning trajectories, and which one a patient follows depends on why they were cannulated:

The severely hypoxaemic patient was started on high blood flow. Their weaning is largely a blood flow story, reduced as the native lung takes over oxygenation.
The purely hypercapnic patient was already on low blood flow, and the magnitude of their support depends on sweep. Their weaning is almost entirely a sweep gas flow story.

💡

Clinical pearl — wean the circuit's oxygen before the ventilator's

Both Taha and the ISCCM manual make the same recommendation: reduce FsO₂ (the oxygen fraction in the sweep gas) before reducing ventilator FiO₂ in the early phases, to avoid oxygen toxicity on the recovering native lung.

The logic is that the lung you are trying to heal is the one exposed to ventilator FiO₂; the membrane is not harmed by oxygen. Give the recovering alveoli the lower oxygen concentration, and let the circuit run the higher one.

Note the tension with Chapter 8's down-titration ladder, which puts FdO₂ first as the cost-free reduction. Both are right at different moments: early, when the lung is fragile and support is high, protect the lung by weaning circuit oxygen last. Later, when the patient is stable and heading for a trial, circuit oxygen is the cheapest thing to reduce.

💡

Clinical pearl — target the pH, not the PaCO₂

The Red Book records that many centres judge adequate CO₂ clearance from a normal arterial pH rather than a normal arterial CO₂, because metabolic compensation of hypercapnia is frequently present in patients who required VV ECMO.

A patient with a chronically elevated bicarbonate will have a normal pH at a PaCO₂ that looks abnormal on paper. Weaning them to a "normal" PaCO₂ means over-ventilating them into an alkalosis and misjudging how much support they still need.

10.5 Ventilator management during weaning — the settings go up

This is the step most often got wrong, because it reverses everything Chapters 7 and 8 established. During support, the ventilator is held at ultra-protective rest settings and the circuit does the gas exchange. During weaning, ventilator support must be deliberately increased from ultra-protective back to conventional lung-protective settings — otherwise the patient will fail a trial that only tested an inadequate ventilator.

Parameter	Target during weaning
FiO₂	30–60%
PEEP	At least 5–10 cmH₂O
Respiratory rate	30 breaths/min or fewer
Volume control	Increase tidal volume in increments of 1–2 mL/kg predicted body weight to a maximum of 6–8 mL/kg PBW, with plateau pressure 28 cmH₂O or less
Pressure control or pressure support	Driving pressure increased to no more than 15–20 cmH₂O, delivering tidal volumes in the 6–8 mL/kg PBW range

Source: ELSO Red Book 6th ed., Ch 25, citing the ELSO VV ECMO guideline.

⚠️

Danger — patient self-inflicted lung injury during weaning

Most patients are still mechanically ventilated when weaning begins, but as sedation lightens and support falls, spontaneous effort returns. The Red Book warns explicitly that spontaneously breathing patients must be monitored for respiratory distress and increased work of breathing, because patient self-inflicted lung injury (P-SILI) is possible — whether or not they are intubated.

The mechanism matters: the sweep gas that was suppressing respiratory drive (Chapter 8) is being withdrawn. Vigorous spontaneous effort against a recovering, heterogeneous lung generates large transpulmonary pressure swings and regional overdistension that the ventilator's displayed plateau pressure does not show.

An increase in work of breathing during weaning is a failure criterion, not something to encourage through.

10.6 Step 3 — The sweep gas off trial

🫀

Physiology — why turning off a gas removes all the support

In the sweep gas off trial the gas flow is turned fully off, or the line disconnected. Blood continues to circulate through the circuit at full flow, the pump keeps running, the membrane is still there — and yet, as the Red Book puts it, this "completely eliminates all extracorporeal support because no oxygen is added, or carbon dioxide removed despite blood still moving through the circuit."

The membrane lung is a diffusion device with no gradient once the gas side equilibrates. Blood entering and blood leaving become identical. The circuit is reduced to an inert loop of tubing.

This is a property of VV alone. In VV, the circuit was only ever doing gas exchange, so removing gas exchange removes everything it contributed. This is why the trial is so clean — and why it is so dangerous elsewhere.

⚠️

Danger — the sweep gas off trial is for VV only

The Red Book states it plainly: the simple sweep gas off trial "applies only to VV ECMO and would be dangerous in the V-A or V-VA ECMO configurations."

In VA the circuit is providing circulatory support as well as gas exchange. Turning off the sweep removes the gas exchange while the pump continues to divert cardiac output and impose afterload — so the patient loses oxygenation without being tested on their own circulation, and deoxygenated blood is pumped into the aorta. VA weaning uses flow reduction and a clamp trial instead, with echocardiography (Chapter 18).

10.6.1 Running the trial

Before starting. Restore ventilator settings to the conventional lung-protective targets of §10.5. A trial run on rest settings tests nothing except the rest settings.

Turning the gas off. The ISCCM manual adds a practical warning worth heeding: oxygen can leak around the flow meter even when it is turned off, so clamp the gas tubing — or disconnect it, as the Red Book describes. A trial conducted with a trickle of sweep still running is not a trial.

Confirming the circuit is truly inert. The oxygen remaining in the circuit is usually consumed within about 20 minutes. Confirmation is visual and requires no equipment: the arterial and venous limbs become the same dark red colour — the same colour test used in Chapter 9 to diagnose gas-supply failure, here used deliberately to confirm success.

What does not change.

Blood flow continues. Because the circuit keeps circulating, there is no increased clotting risk during a VV trial-off — one of the genuine advantages of the configuration.
Anticoagulation is left unchanged throughout the trial.
Any infusions running into the circuit are undisturbed, and CRRT connected to the circuit continues without alteration.

Monitoring during the trial. Standard clinical observation, cardiac monitoring and pulse oximetry, plus serial arterial blood gases. The ISCCM manual adds continuous mixed venous oxygen saturation monitoring to assess the adequacy of oxygen delivery, and — for patients on assisted spontaneous modes — close attention to respiratory pattern and mechanics.

💡

Clinical pearl — how long, and why longer is rarely better

Trial durations from 2 to 24 hours are described in practice. But in a study of 192 sweep gas off trials in VV ECMO, no significant changes in blood gas parameters were found after 2 hours from the time the sweep was turned off.

ELSO guidelines recommend a minimum of 2–3 hours, and the Red Book notes that longer trials are rarely necessary. The exception is the marginal patient in whom the team believes pulmonary toilet, ventilator adjustment, sedation change or gentle recruitment may yet tip the balance — there, extending the trial is reasonable. It is unusual to extend beyond 24 hours, because a decision can almost always be reached before then.

The 192-trial study is cited by the Red Book to its own reference and has not been retrieved; the authors and journal are unknown to this chapter.

10.6.2 Pass criteria

Exact cut-offs vary between institutions, but the published thresholds converge:

Domain	Passing
Ventilator support	Remains within lung-protective bounds as defined in §10.5 — this is the criterion that matters most
Oxygenation	SpO₂ 88–92% or above, PaO₂ 70 mmHg or above
Acid–base	pH 7.30 or above
Haemodynamics and effort	No significant new tachycardia, hypertension, hypotension, or increased work of breathing

Source: ELSO Red Book 6th ed., Ch 25, citing the ELSO VV ECMO guideline and several protocol studies.

🕳️

Pitfall — passing on the numbers while failing on the ventilator

A patient can hold an acceptable saturation and pH — and still have failed, if the settings required to achieve them exceed lung-protective limits. Gas exchange achieved at a plateau pressure of 34 cmH₂O is not a pass. The question the trial asks is not "can this patient oxygenate without the circuit?" but "can this patient oxygenate without the circuit and without injurious ventilation?"

The Red Book is explicit that passing "should include a patient maintaining adequate oxygenation and ventilation while not exceeding lung protective levels of ventilator support."

10.6.3 If the trial fails

Restart extracorporeal support and return the ventilator to lung-rest settings. Anticoagulation continues as before. Nothing has been lost except a few hours, and a failed trial is genuinely informative — it usually identifies which of oxygenation or CO₂ clearance is still lacking, which directs the next few days of management. Chapter 9's six-cause differential applies if the deterioration is out of proportion or unexpected.

10.7 Step 4 — Decannulation

10.7.1 Before the procedure

Cardiopulmonary stability, with ventilatory support re-established if it was reduced during the trial.
A blood gas confirming adequate exchange.
Sedation, and paralysis if required.
Blood and blood products available as required.
Anticoagulation held. Current guidance is that it is prudent to hold anticoagulation for at least 30–60 minutes before decannulation; local practice varies with the operator.
For long ECLS runs, formally reassess pulmonary compliance before decannulation, to be certain that safe controlled positive-pressure ventilation is possible if it becomes necessary afterwards.

🕳️

Pitfall — a passed trial is not the same as a patient who can be ventilated

This is the reason for that last item. The trial proves the patient does not need the membrane. It does not prove that, should they deteriorate an hour later, the lung can be safely ventilated by conventional means. In a patient who has been on ECMO for weeks, those are different questions — and the answer to the second one is much easier to obtain before the cannulae come out than after.

10.7.2 The procedure

Decannulation from VV ECMO can generally be performed at the bedside, in a semi-sterile fashion — complete sterility is not possible with cannulae already in place. Experienced providers capable of managing acute bleeding should be present, along with staff trained to manage the circuit.

The Red Book describes the sequence:

If an internal jugular cannula is present, place the patient in Trendelenburg, as for removal of any internal jugular central catheter.
Place sutures but do not tie them — 0 silk, pursestring or mattress, around each cannula where it enters the skin. Extreme care must be taken when suturing near an ECMO cannula.
Remove the securing sutures and devices, with the operators holding the cannulae in place.
Hold pressure over the entry sites with gauze.
Prevent air entrainment. If an internal jugular cannula is present, place the ventilator on a brief inspiratory hold. A spontaneously breathing patient should bear down, hold exhalation, or hum. (The ISCCM manual describes the equivalent as a Valsalva manoeuvre on the ventilator at the moment of removal.)
In a coordinated fashion: clamp the circuit near the cannula at the return limb first, then the drain limb; stop the pump; then each operator withdraws their cannula with a smooth, rapid, single movement. Both cannulae are removed in a coordinated fashion by two operators.
Resume normal respiration, tie the sutures down securely, and hold pressure — either until haemostasis is achieved or for an extended period, by operator preference. The ISCCM manual specifies holding pressure at venous puncture sites for at least 30 minutes.

⚠️

Danger — the three immediate complications

Decannulation is generally well tolerated, but sudden haemodynamic or respiratory decompensation should prompt rapid evaluation for three life-threatening complications:

Pulmonary embolism — a cannula-associated deep vein thrombus dislodged by removal. Given the DVT rates in §10.8, this is not a remote possibility.
Air embolism — entrained during removal, and specifically noted as a risk with multistage drainage cannulae, whose side holes provide multiple entry points. This is what step 5 exists to prevent.
Bleeding — normally controlled with sutures and local pressure.

💡

Clinical pearl — returning the circuit blood when transfusion is refused

There is a published case report of modifying the procedure for an anaemic patient who could not receive blood because of religious restrictions: the drainage limb was clamped first and sterile saline infused into the circuit until the blood in it had been returned to the patient, after which the return limb was clamped and the cannulae removed.

Not usually required, but described as a safe option in exceptional cases — and worth knowing before the situation arises rather than during it.

10.8 After decannulation

Two things happen often enough that a unit should have a standing plan for both.

10.8.1 Post-decannulation SIRS

📊

Evidence — fever after decannulation is usually not infection

Fever, leucocytosis and vasodilation causing tachycardia or vasopressor requirement are described in the 24–48 hours after decannulation. The estimated incidence of fever or SIRS is 50–60% in the 48 hours after decannulation (a separate Red Book chapter gives about 60%). Limited evidence and anecdote suggest it is more common after longer ECMO runs.

These symptoms frequently prompt an infectious workup, and across studies confirmed infection was found in 8–35% of febrile patients — meaning most were not infected. Critically, post-decannulation fever or SIRS in the absence of diagnosed infection was not associated with increased mortality.

There is currently no clear evidence for prophylactic antibiotic use around the time of VV ECMO decannulation.

Certainty: low — limited studies, wide ranges, and the underlying primary sources were not retrieved for this chapter.

The ISCCM manual records a formal definition, attributed to Thangappan and colleagues: the post-decannulation SIRS phenomenon as two of three of — fever above 101.5 °F; leucocytosis (white cell count above 12,000, or a 25% increase from the pre-decannulation baseline); and escalation of vasopressors compared with the pre-decannulation baseline — regardless of infection. Heart rate and respiratory rate were deliberately excluded, because on ECMO they are confounded by inotropes, sedation depth and ventilator settings. It can appear on the first day after decannulation and may continue for almost a week.

🕳️

Pitfall — the reflex antibiotic course

The practical difficulty is that this looks exactly like sepsis in a patient who has every reason to be septic. The evidence does not support prophylactic antibiotics, and it does not support ignoring the fever either — 8–35% of these patients do have infection.

What it supports is culturing, looking hard, and being willing to stop: a patient who is febrile on day two after decannulation with negative cultures and no source is far more likely to be exhibiting a well-described inflammatory phenomenon than an occult infection, and treating them for a week regardless is a decision, not a default.

10.8.2 Venous thromboembolism

⚠️

Danger — most of these patients have a deep vein thrombosis

Cannula-associated venous thrombus is described as "a very frequent complication in adult VV ECMO patients." Multiple studies using CT venography or Doppler surveillance have estimated the prevalence of DVT at over 60% after decannulation — even in patients anticoagulated while on ECLS. One small study of VV ECMO patients with COVID-19 found 100% had evidence of DVT or pulmonary embolism.

Risk factors: femoral cannulation site, larger cannula size (including dual-lumen cannulae), and lower aPTT values on anticoagulation.

Given these rates, routine surveillance of all patients decannulated from VV ECMO is warranted: Doppler ultrasound of at least the cannulated vessels, 24 hours after decannulation, for all patients. Some centres scan all limbs, on the grounds that these patients are critically ill, relatively immobile and at high baseline venous thromboembolism risk.

Note the connection back to Chapter 5: the larger cannula that gave better flow is also the one more likely to leave a thrombus behind.

The ISCCM manual adds that clots found on surveillance should prompt anticoagulation, and that the need for an inferior vena cava filter should be evaluated — though the evidence base for filters in this specific population is not established and no recommendation is made here.

🕳️

Pitfall — new dyspnoea after decannulation has a differential

Worsening dyspnoea in the days after decannulation should not be attributed to deconditioning by default. The differential includes fluid overload and thromboembolism — and, given a DVT prevalence above 60%, the second of those deserves active consideration rather than reassurance.

10.9 Controversies

⚖️

Controversy 1 — Is the oxygen challenge (Cilley) test worth doing?

The question. The Cilley or oxygen challenge test raises ventilator FiO₂ to 100% with no other change; a prompt rise in saturation, and a PaO₂ above 225 mmHg at 15 minutes, is taken as evidence of recruitable, recovering lung. Should it be used to decide who is ready for a sweep gas off trial?

The case for. It is free, takes fifteen minutes, needs no extra equipment, and can be repeated daily. The ISCCM manual recommends performing it daily in patients showing improvement in lung parameters, and Taha lists it among the standard consensus criteria for considering weaning. As a crude screen for "has this lung become recruitable at all?", it has obvious face validity.

The case against. It tests the response of oxygenation to oxygen — which is a shunt measurement, not a measurement of whether the lung can sustain gas exchange unaided. It says nothing about CO₂ clearance, nothing about the work of breathing, and nothing about the ventilator pressures required.

What the evidence actually shows. The ELSO Red Book states that methods other than the sweep gas off trial, specifically sweep gas flow thresholds and the 100% oxygen challenge test, "have been shown to be less reliable" for predicting successful decannulation. Taha cites a single-centre study of 253 patients which concluded that the oxygen challenge test was not a valid predictor of decannulation readiness from VV ECMO. Certainty: low — one single-centre study, quoted second-hand, with no author or journal available to this chapter, plus a guideline-level statement of unreliability.

Where practice actually sits. The test survives in textbooks and in many units as a daily screening habit — a prompt to think about weaning — while the decision rests on the sweep gas off trial. That division of labour is defensible and is what this book recommends.

What would resolve it. A prospective study comparing the oxygen challenge test against sweep gas off trial outcome as the reference standard, with prespecified thresholds. The 253-patient study points this way but has not been read here.

The practical resolution. Use it, if you like it, as a screen that costs nothing. Do not decannulate on it. The sweep gas off trial is the standard of care for predicting successful decannulation, and it is the only test that examines the thing you actually care about.

⚖️

Controversy 2 — Protocolised liberation, or clinician judgement?

The question. Should VV ECMO liberation follow a written local protocol, potentially driven by respiratory therapists or perfusionists, or should readiness and timing remain a physician judgement made case by case?

The case for judgement. ECMO patients are heterogeneous, the population is small, and no protocol has been validated against outcomes. Experienced clinicians integrate information a protocol cannot encode — trajectory, the cause of the lung injury, comorbidity, the plan for what follows.

The case for protocol. The analogy with ventilator liberation is close, and there the history is instructive: protocolised spontaneous breathing trials shortened ventilation without harm, precisely because unaided judgement was systematically too cautious.

What the evidence actually shows. Structured protocols — physician-driven and therapist- or perfusionist-driven alike — have produced reductions in time to first sweep gas off trial and time to decannulation, without increased adverse events, and have liberated patients safely at higher levels of support than would traditionally have prompted a trial. No single protocol has been shown superior to any other. No randomised trial exists; the Red Book states that "clinical efficacy trials are lacking." Certainty: low to moderate — consistent observational signal across several studies, no randomisation, plausible mechanism, and a strong analogous evidence base in ventilator liberation.

Where practice actually sits. The Red Book's recommendation is unambiguous: institutions should implement a local standardised VV ECMO liberation protocol. Note what this does and does not say — it does not name a protocol, because the evidence does not support one over another. The benefit appears to come from standardisation itself, not from any specific set of numbers.

What would resolve it. A randomised comparison of protocolised versus usual-care liberation, with ECMO duration, decannulation failure and adverse events as endpoints. Given the observational consistency and the ventilator-liberation precedent, equipoise for such a trial may already be thin.

The uncomfortable implication. The finding that protocols liberate patients safely at support levels clinicians would not have trialled is a finding about clinicians. It suggests that the intuition "they still need too much support to try" is unreliable, and that the cost of that intuition is paid in extra ECMO days — with their anticoagulation, immobility and thrombosis.

🕳️

Pitfall — a third question with no evidence either way

Most patients are still mechanically ventilated at decannulation. For those who are not, practice genuinely diverges: some centres decannulate awake, spontaneously breathing patients; others routinely reintubate for the procedure, to allow sedation and to reduce the risk of entraining air during cannula removal.

No comparative evidence is available to this chapter. Both are described as accepted practice. The trade-off is explicit — reintubation adds a procedure and its own risks to a patient who had escaped it, while awake decannulation depends on the patient reliably performing a Valsalva or hum at the right moment. The decision should be made on the individual patient's ability to cooperate, and made in advance rather than at the bedside.

10.10 The errors that recur

Error	Correction
Running the trial on lung-rest ventilator settings	Restore conventional lung-protective settings first (§10.5); otherwise the trial tests the ventilator, not the lung
Turning the sweep dial to zero without clamping the gas line	Oxygen leaks around the flow meter even when off — clamp or disconnect
Weaning to a normal PaCO₂	Target pH; metabolic compensation of hypercapnia is common in this population
Calling it a pass because the gases were acceptable	Check what ventilator settings produced them — a pass requires lung-protective settings and acceptable gases
Decannulating on an oxygen challenge test	It is a screen. The sweep gas off trial is the test
Waiting for the patient to look obviously ready	Protocols liberate safely at support levels clinicians would not have trialled — judgement runs conservative
Removing an internal jugular cannula without an inspiratory hold or Valsalva	Air embolism, particularly with multistage drainage cannulae
Treating post-decannulation fever as sepsis by default	50–60% get SIRS; only 8–35% of the febrile ones have confirmed infection; culture, look, and be willing to stop
Not scanning for DVT because the patient was anticoagulated throughout	Over 60% have DVT despite anticoagulation. Doppler the cannulated vessels at 24 hours in every patient

10.11 Key points

The plan to come off begins on the day of cannulation and is reviewed daily. Staying on ECMO longer than required is its own harm.
Liberation has four steps: recognise recovery, down-titrate support, sweep gas off trial, decannulate.
Blood flow weans oxygenation; sweep gas flow weans CO₂ removal. Which one dominates depends on why the patient was cannulated.
Rising end-tidal CO₂ and falling sweep requirement are the same event — the earliest signs of native lung recovery, and they precede improvement in oxygenation.
Wean circuit oxygen (FsO₂) before ventilator FiO₂ early on, to spare the recovering lung from oxygen toxicity.
Target pH, not PaCO₂ — metabolic compensation of hypercapnia is common in patients who needed VV ECMO.
Ventilator support goes up during weaning, from ultra-protective rest settings to conventional lung-protective settings, before any trial is meaningful.
The sweep gas off trial removes all extracorporeal support while blood keeps flowing — and is safe only in VV. Clamp the gas line; confirm by the two limbs turning the same dark red.
Anticoagulation, circuit infusions and in-circuit CRRT all continue unchanged during a VV trial-off, because the circuit keeps circulating.
Two to three hours is enough. Blood gases do not change meaningfully after two hours.
A pass means acceptable gas exchange at lung-protective ventilator settings — not acceptable gas exchange at any cost.
Hold anticoagulation 30–60 minutes before decannulation; remove with an inspiratory hold or Valsalva; clamp return limb then drain limb, stop the pump, withdraw in one smooth movement.
After decannulation, expect SIRS in 50–60% — most of it not infection — and DVT in over 60%, even in patients anticoagulated throughout. Doppler every patient at 24 hours.
Standardise locally. The evidence favours having a protocol over having a particular protocol.

🔍

[VERIFICATION REQUIRED] — open items in this chapter

The ELSO VV ECMO guideline (Tonna JE, Abrams D, Brodie D, et al., ASAIO J 2021;67(6):601–610) is the source, via the Red Book, of the ventilator targets in §10.5, the pass criteria in §10.6.2, the 2–3 hour minimum trial duration and the 30–60 minute anticoagulation hold. Its bibliographic details are transcribed from the Red Book's reference list, not from the article itself, and its DOI and PMID are not stated here because they were not verified — an attempt to confirm them via Europe PMC was blocked and then rate-limited. The guideline itself has not been read.
The study of 192 sweep gas off trials (no blood-gas change after 2 hours) is cited by the Red Book to its own reference; authors, journal and year are unknown to this chapter.
The single-centre study of 253 patients finding the oxygen challenge test invalid is cited by Taha; authors, journal and year unknown.
Mols et al. (weaning at 80% of oxygen delivery from the native lung) and the 50–80% native gas exchange figure are second-hand through Taha; the primary study has not been retrieved.
The SIRS incidence (50–60%), the 8–35% confirmed-infection range, the DVT prevalence above 60%, and the 100% DVT/PE COVID series are all quoted from the Red Book's citations, none of which has been retrieved. Ranges are reproduced as stated.
The Thangappan SIRS definition is quoted as described in the ISCCM manual; the primary paper has not been read.
The compliance threshold of 20 mL/cmH₂O as a recovery marker is from the ISCCM manual and is not an established or validated cut-off.
No randomised evidence exists for any part of this chapter. Every step described is expert opinion or observational; this is the Red Book's own assessment, not a hedge added here.
Numeric weaning criteria appearing in the Red Book's neonatal weaning chapter (including an echocardiographic pulmonary artery pressure criterion) were deliberately excluded from this chapter, as they have no established adult equivalent and importing them would be a category error.

Cross-references

Chapter 5 — VV ECMO Cannulation: cannula size and site, which determine the post-decannulation thrombosis risk of §10.8.2
Chapter 6 — VV ECMO Physiology: what blood flow and sweep gas each control — the basis of §10.4
Chapter 8 — Daily VV ECMO Management: the daily down-titration ladder that leads to a trial, and the destination review that §10.1 mirrors
Chapter 9 — Persistent Hypoxaemia on VV ECMO: the differential when a trial fails unexpectedly, and the colour test used here to confirm circuit inertness
Chapter 18 — VA ECMO Weaning and Decannulation: flow reduction and clamp trials, and why the sweep gas off trial must not be used there
Chapter 23 — Neurological Outcomes and Prognostication
Chapter 34 — Cannula Problems: cannula-associated thrombosis and vascular injury
Chapter 53 — Mechanical Ventilation During ECMO: lung rest, driving pressure, and P-SILI in depth
Chapter 55 — CRRT During ECMO: why in-circuit CRRT continues unaltered through a trial-off
Chapter 59 — Mobilization and Rehabilitation: what happens after the cannulae come out
Part VII — Anticoagulation: the periprocedural hold of §10.7.1
Chapter 91 — Decannulation Checklist: the printable, bedside-ready form of §10.7

References

Malley BE, Ramanan R, Pellegrino V, Harvey C. Weaning and Decannulation in Adult Respiratory Failure. Chapter 25 in: Extracorporeal Life Support: The ELSO Red Book, 6th edition. The four steps of liberation; evidence on protocolised weaning; ventilator management during weaning; sweep gas management and the pH target; the sweep gas off trial, its VV-only applicability, duration and pass criteria; the decannulation procedure; post-decannulation SIRS and venous thromboembolism.
Tonna JE, Abrams D, Brodie D, et al. Management of Adult Patients Supported with Venovenous Extracorporeal Membrane Oxygenation (VV ECMO): Guideline from the Extracorporeal Life Support Organization (ELSO). ASAIO J. 2021;67(6):601–610. [VERIFICATION REQUIRED] — transcribed from the reference list of the source above; the guideline itself has not been read and its DOI and PMID are not stated because they were not verified.
ISCCM Manual of RRT and ECMO in ICU, Chapter 38, Extracorporeal Membrane Oxygenation Weaning. Indian Society of Critical Care Medicine. Parameters of lung recovery; the Cilley test; clamping the gas line; 20-minute circuit oxygen consumption and the colour confirmation; trial-off monitoring; the pre-decannulation checklist; pursestring technique and 30-minute pressure; Valsalva at removal; post-decannulation Doppler; the Thangappan SIRS definition. [VERIFICATION REQUIRED] — edition, editors, year and page numbers not confirmed in this session.
Taha AR, Zaher A, et al. ECMO: A Practical Guide to Management. Discontinuation planning from day one; the 50–80% native gas exchange concept and the Mols observation; the Cilley/oxygen challenge test and the 253-patient study questioning it; weaning sweep FsO₂ before ventilator FiO₂; differing weaning trajectories for hypoxaemic versus hypercapnic patients. [VERIFICATION REQUIRED] — full editor list, publisher, year and page numbers not confirmed in this session.

📅

Chapter status

Drafted and audited 6 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 II — VV ECMO. It contains no fabricated citations, thresholds, doses or guideline statements; every unretrieved item is declared, and one identifier (the ELSO guideline DOI/PMID) is deliberately omitted rather than reconstructed. Where the source literature is expert opinion rather than trial evidence — which is nearly all of it here — the chapter says so.