💨

Chapter 35 — Air and Circuit Emergencies

❓

The question this chapter answers

An ECMO circuit is a suction line joined to a pressure hose, and every breach behaves according to which half it is in. A hole before the pump draws air in; a hole after the pump pushes blood out. Air that enters before the membrane lung is usually caught there; air that gets past it goes to the patient, and on venoarterial support it goes to the arterial tree.

So the bedside question when something goes wrong is not what broke? but which side of the pump is the breach on, which side of the membrane lung is the air on, and does this patient survive the next minute better with the pump running or with it stopped?

This chapter covers air in the circuit, circuit rupture and breach, and the two ways of separating a patient from the circuit in an emergency — emergent and controlled.

🗓️

Evidence search date: 24 September 2026. One source is read as complete full text and is the Verified source new to this chapter: an in vitro study of air entry through a central venous catheter and of what measured volumes of air do to a running circuit. The Japanese national accident analysis read in full for Chapter 34 is used again for its air and circuit-damage events. Primary sources are the ELSO Red Book 6th edition (the circuit chapter, the circuit malfunction and crisis management chapter, the cannulation chapter's airless connection techniques, and the registry complication definitions), ECPR and Resuscitative ECMO (initiation and the circuit catastrophes chapter, including its emergent and controlled separation protocols), ECMO: A Practical Guide to Management (the complications and emergencies algorithm), Comprehensive Healthcare Simulation: ECMO Simulation, and the ISCCM Manual of RRT and ECMO in ICU.

What this chapter covers — and what it does not

This chapter owns	This chapter does not own
Air in the circuit — where it comes from, where it goes, how it is found and how it is removed	The central line as an air source is introduced in Chapter 34 §34.8; this chapter adds the in vitro quantification and does not repeat the procedure rule
Air that has reached the patient — the first minutes only	Neurological injury from air embolism. Chapter 38a owns neurological complications
Circuit rupture and breach, and the temporising measures for each side of the pump	Cannula-site bleeding and cracked cannulas. Chapter 34 §34.8 and Chapter 36 own them
Emergent and controlled separation from the circuit — CLAMP, STOP, RESCUE, and the stepwise version	The planned circuit exchange. Chapter 31 owns the exchange decision and the planned procedure, including the Red Book's pause to identify both limbs of both circuits
The bubble detector — where it sits and what it should be allowed to do	Pump stoppage and the hand crank. Chapter 33 owns them, including retrograde flow on venoarterial support
The emergency cart, the primed backup circuit and training as they bear on air and breach	Chapter 90 owns the emergency circuit checklist as a printable tool

35.1 One circuit, two pressure worlds

🧠

The structural claim of this chapter · The pump divides the circuit into a side that sucks and a side that pushes

A centrifugal pump draws blood from the drainage cannula and pushes it through the membrane lung to the return cannula. Upstream of the pump the pressure is below atmospheric; downstream it is above. The Red Book gives the range at the inlet: subatmospheric pressures down to minus 250 to minus 300 mmHg can be generated (Chapter 34 §34.2). The ECPR text states the consequence for air in one line: "due to the negative pressure in the drainage limb, the venous side of the circuit can be a continual source of air entrainment. Typically, this indicates a leak in the system."

So the same defect has opposite presentations depending on where it is.

A breach on the drainage side draws air in. Nothing leaks out. The first sign may be bubbles, a change in the pump's sound, or a falling flow — not blood.
A breach on the return side pushes blood out. The ECPR text divides its temporising measures exactly this way: occlusion stops "further air entrainment (pre-pump side) or hemorrhage (post-pump side)."

And the membrane lung divides the positive side a second time. The Red Book's terms: air between the membrane lung and the patient is "arterial air"; air before the membrane lung is "venous air." Venous air may be caught by the membrane lung. Arterial air has nothing between it and the patient.

Three consequences, and the chapter is built on them.

Before touching anything, place the problem in one of three zones: before the pump, between pump and membrane lung, or after the membrane lung. The zone decides the danger and the first move.
A drainage-side leak is silent and continuous. It keeps drawing air for as long as the pump runs and the hole is open.
The Red Book's judgement on the whole subject: this complication "is nearly always avoidable." Its prevention list is three items — limit access points on the negative-pressure side, take care with central lines, and handle stopcocks meticulously.

Certainty: high for the pressure physics and the zone definitions, which the Red Book and the ECPR text state concordantly; the three-zone ordering is this book's framing.

🔬

Evidence · How air got into circuits, when a national system recorded it

From the Japanese national accident analysis read in full for Chapter 34 (Hadano 2023, J Artif Organs, DOI 10.1007/s10047-023-01386-z): air in the circuit was 7 percent (12) of 178 reported ECMO accidents. The reported causes:

Cause	Proportion of 12 air events
Incorrect clamp release	33 percent (4)
Incorrect three-way stopcock operation	33 percent (4)
Machine breakage	17 percent (2)

Emergency circuit change was needed in half (6 of 12). Separately, 9 of 26 bleeding events were due to circuit damage, including a fissure in the connection between the arterial cannula and the circuit and cracks in a flow sensor.

The reading this chapter takes. Two thirds of the reported air events were a hand on a clamp or a stopcock. That matches the Red Book's prevention list, which names stopcock technique explicitly. The machine was the cause in one sixth.

Certainty: low for proportions — twelve events, no denominator, voluntary reporting. The direction — air is mostly a handling error — is consistent with every primary text in this chapter.

35.2 Where air comes from

🧮

The sources of air, by the side of the circuit they act on

Source	Side	What the sources say
Connections, three-way stopcocks and pigtails	Drainage (negative)	ECPR and Resuscitative ECMO lists "any point of connection or 3-way connector" first; the Red Book's prevention list names meticulous stopcock technique
An open side port on the cannula	Either	Some cannulas, typically arterial, have a built-in side port which may become unsecured and open to air (ECPR and Resuscitative ECMO)
Damage to the circuit while securing the cannulas	Either	Listed separately by the same text; Chapter 34 §34.8
A central venous catheter near the drainage cannula	Drainage	Chapter 34 §34.8 for the procedure rule; §35.3 for how little it takes
Partial decannulation of a multistage drainage cannula	Drainage	Chapter 34 §34.5 — air with flow still present
Cavitation from excessive negative pressure with chatter	Drainage and pump	ECPR and Resuscitative ECMO: "gas is pulled out of solution resulting in air embolism within the circuit"; the Red Book adds that the micro-bubbles can pass the membrane lung
The membrane lung itself, when gas flow far exceeds blood flow	After the pump	ECPR and Resuscitative ECMO names it as a potential source of bubbles; Chapter 32 owns sweep and the membrane
Infusion pumps	Either	"May inject small bubbles (generally not clinically significant)" (ECPR and Resuscitative ECMO)
Procedures near the venous system	Drainage, via the patient	Yan and colleagues (Perfusion 2020, abstract read) describe three cases: air from a right jugular venous sheath during percutaneous septal repair, from a pulmonary artery catheter in intensive care, and from sewing holes on the atrial wall during surgery
Connection at initiation	Either	The Red Book's airless connection techniques — the bump, the fill, the drip, the stopcock fill — exist to make the wet-to-wet connection without introducing air

Notice what most of the list has in common. Most items are an opening to the atmosphere on the negative-pressure side of the circuit, or in the patient's venous system upstream of it. The patient's great veins are, functionally, part of the drainage limb.

Certainty: moderate — each item is named by at least one primary text; no source ranks them by frequency apart from the twelve Japanese events.

🚩

Pitfall · The patient's own veins are on the negative side

The drainage cannula does not only pull from the circuit's tubing; it pulls from the vena cava. Any opening into the central veins near the drainage cannula — a central line hub, an introducer sheath, a pulmonary artery catheter port, a surgical opening in the atrium — is an opening in the drainage limb.

The consequence the sources draw. Yan and colleagues conclude that more attention should be paid to patients with a communication between the right and left heart, "especially in situations when venous accesses' exposure to air could not be avoided"; their first case was suspected of cerebral air embolism through a transseptal right-to-left shunt (the abstract does not state the configuration). ECPR and Resuscitative ECMO makes the same point from the other direction: lowering the head of the bed may keep air that reached the patient out of the cerebral circulation "if on V-A ECMO, or on V-V ECMO via a patent foramen ovale."

So venovenous support is not immune to arterial air. A patent foramen ovale or an atrial septal defect turns venous air into systemic air.

35.3 How much air, and what it does

🔬

Evidence · Measured volumes of air in a running circuit

Franco D, Krasna N, Novak R, Esposito G, Izzo R, Belohlavek J, Noc M, Goslar T. "Prevention of Air Embolism in Extracorporeal Membrane Oxygenation Systems: An In Vitro Study on Protection of Central Venous Catheter Lumen." Medicina (Kaunas) 2024;60(11):1883. DOI 10.3390/medicina60111883, PMC11596159. Full text read.

The design. A venovenous in vitro model: a Cardiohelp console and HLS set, a 23 Fr 55 cm multistage drainage cannula and a 17 Fr return cannula in latex "vessels", a 40 percent glycerol and water mixture at 36 °C, and a reservoir set to simulate a central venous pressure of 10 cm H₂O. Pump speeds of 1,500, 2,500, 3,500 and 4,500 rpm gave flows of 1.61, 3.15, 4.56 and 5.91 L/min. Every experiment was repeated three times; results are descriptive.

Experiment 1 — a central line beside the drainage cannula. A 7 Fr triple-lumen catheter was placed with its tip 0, 1 or 2 cm from the drainage cannula's distal hole or its proximal side hole, with the hub left open, closed with a three-way stopcock, or fitted with a needle-free valve (clave).

Central line tip beside	Hub open — air entered at	Hub closed by stopcock or clave
Distal hole, tips touching (0 cm)	2,500, 3,500 and 4,500 rpm — not at 1,500	No air at any speed or distance
Distal hole, 1 or 2 cm apart	3,500 and 4,500 rpm only	No air
Proximal side hole, 0, 1 or 2 cm	3,500 and 4,500 rpm at every distance; not at 1,500 or 2,500	No air

Experiment 2 — known volumes of air injected into the drainage cannula.

The characteristic sound of air in the pump occurred with as little as 1 mL, and at every larger volume. It was the first warning sign.
Air passed through the oxygenator and reached the return-side bubble detector with 1 mL at 4,500 rpm, and with 5 mL and 25 mL at 3,500 and 4,500 rpm. At lower speeds, smaller volumes were retained by the membrane lung.
25 mL or 50 mL stopped flow completely at every pump speed tested. Smaller volumes reduced flow briefly; the oxygenator then vented slowly and flow recovered.

The authors' reading, which this book adopts. At lower speed the pump is more likely to airlock; at higher speed air is more likely to be driven through the membrane lung to the patient. And once the pump has stopped from air, "increasing the pump speed risks restarting the pump and forcing air through the membrane lung, so the only way to avoid air embolism in the patient is to clamp the circuit."

Certainty: low to moderate. In vitro, venovenous only, one catheter size, one console, a blood substitute, no respiration and a fixed venous pressure; descriptive with three repeats. The authors themselves expect a vigorous inspiration to make entry through the line far more likely than their model shows. What it establishes firmly is direction and order of magnitude: millilitres are audible, tens of millilitres stop the pump, and a closed hub prevents all of it.

🧠

Physiology · Why the membrane lung is a trap, and why the trap is not reliable

The membrane lung is a large chamber in which blood spreads out and slows down, and gas rises (this book's description of the geometry). So venous air that reaches it tends to collect at its top rather than pass through. The Red Book: air within the membrane lung will usually rise to the highest point and may be aspirated through a pigtail or stopcock attached for this purpose. ECPR and Resuscitative ECMO: air bubbles on the drainage side "may not lead to patient harm, as they may be removed by the membrane lung. This should not be relied upon."

Three things defeat the trap.

Speed. Franco's model shows air passing the membrane at 1 mL at the highest speed tested, and at 5 mL from 3,500 rpm. The faster blood moves through the membrane, the less time a bubble has to rise out of it.
Volume. Larger volumes overwhelm the membrane's capacity to hold them; in the same model 25 mL reached the return detector at 3,500 rpm and above.
Microbubbles. The Red Book notes that micro-bubbles from cavitation "may pass the artificial lung, increasing the risk of clotting and gaseous embolism." Born and colleagues (Artif Organs 2020, abstract read), in an in vitro venoarterial circuit, found microbubble generation highly dependent on negative pressure at the inflow line, identified the centrifugal pump as a main source, and showed that the oxygenator's ability to withhold larger microbubbles is limited.

The practical form: venous air is an emergency on a timer. It is less dangerous than arterial air only for as long as it stays in the membrane lung, and pump speed shortens the timer.

Certainty: moderate — mechanism described concordantly by two primary texts and two in vitro studies; no in vivo measurement of how much venous air reaches patients was retrieved.

⚠️

Danger · A pump that has stopped from air has, for the moment, protected the patient

When enough air reaches a centrifugal pump head, the impeller spins in gas and stops moving blood — ECPR and Resuscitative ECMO calls this pump head airlock, "due to de-priming of the pump resulting in cessation of blood flow." On a console it looks like pump failure (Chapter 33 §33.2).

The instinct is to turn the speed up to get flow back. That is the move that sends the air into the patient. Franco and colleagues: increasing speed risks restarting the pump and forcing air through the membrane lung, and the only way to prevent embolism is to clamp. The Red Book's instruction for air in the pump head is to reduce pump speed to zero before removing the head, with the return limb always clamped first.

So in an airlock the order is fixed: clamp the return limb, speed to zero, then deal with the air. Chapter 33's rule that a stopped pump on venoarterial support is an open arteriovenous shunt applies here too — the clamp serves both purposes.

The Practical Guide's emergency algorithm has the same shape. For massive air on either configuration, it begins CLAMP and STOP the pump, call for help and a backup circuit; for micro air it begins identify the source and location, clamp if air is noted in the arterial limb. Both branches include increase ventilator and inotropic support as a parallel action and consider circuit exchange if unable to quickly remove air.

35.4 The bubble detector: where it sits decides what it can see

🧮

What the sources say about the detector

Every modern system has at least one. The Red Book: "All modern ECLS systems have at least one integrated bubble detector which may be the first indicator of air in the ECLS circuit."
Position is a trade-off. In systems where the sensor can be moved, placing it before the membrane lung alerts the user earlier and may allow more time to intervene; placing it after the membrane lung alerts the user as air is about to enter the patient but avoids alarms for inconsequential air that would stay in the membrane.
Response is configurable. Most systems can be set to stop the pump when air is detected, or to alarm without intervening. §35.10 takes up which to choose.
A detector after the membrane lung cannot see air that the membrane has caught. Franco and colleagues make this the practical point of their study: most injected air was retained by the membrane lung and never reached the return-side detector, so a bubble detector attached to the return cannula is inappropriate for monitoring air in the circuit, and without a drainage-side detector the pump sound "is the only accurate indication of air in the system."
At initiation, a detector can be the reason there is no flow. ECPR and Resuscitative ECMO lists "arterial bubble detector stopped pump" among the three causes of no flow at all after initiation, alongside a clamp still on and pump airlock.

The reading this chapter takes. Two sources, two correct observations, and they point in opposite directions. A post-membrane detector is quiet because it only fires for air that matters; it is also blind to the venous air that will matter next. Whatever the detector's position, the ear is the earliest detector in the room — which is Chapter 33's point about the pump's note, made again for air.

💡

Clinical pearl · Three observations that find air before the detector does

The sound. A new hiss or murmur from the pump head. In Franco's model it appeared with 1 mL.
The top of the membrane lung. Air collects at the highest point. Look there at every circuit check, with a light.
The drainage limb from cannula to pump. A stream of small bubbles, or a line of foam at a connection, marks a leak on the negative side. Follow the bubbles upstream to their origin before trying to remove the air — the Red Book's methodical circuit check, from one end along the entire length, is how the source is found, and finding the source is what stops it recurring.

35.5 The first minute when air is seen

⚠️

Danger · Call, decide whether air is about to reach the patient, and put one person on the return limb

The Red Book's sequence is the most explicit in the sources, and it contains a decision that the shorter protocols leave out.

Call for help as soon as air is detected.
Decide whether a significant amount of air will enter the patient. If it will, clamp the return limb as close to the patient as possible.
If it will not — "in the absence of substantial air entry" — it is generally preferable to maintain circuit flow while the problem is resolved, "as many patients will not tolerate complete cessation of blood flow." If flow is maintained, decrease it to the lowest level tolerated to slow air movement toward the patient.
Assign one person to watch the arterial limb continuously, with a clamp, ready to clamp the moment air moves toward the patient, while others find the source.
Locate the air exactly. A circuit check from one end, methodically along its entire length, both localises the air and identifies where it came from — "a critical step to ensure the problem does not recur."
Remove it (§35.6). If it cannot be removed quickly and the patient is deteriorating, change the circuit.

The short protocol, for when the answer to step 2 is yes. ECPR and Resuscitative ECMO: for a large amount of air, or even small amounts within the arterial section, follow CLAMP, STOP, RESCUE (§35.9) — which "will also stop the movement of air towards the patient." Then lower the head of the bed, which may keep air that did reach the patient out of the cerebral circulation.

Two things run in parallel from the first second, in the emergency protocols of the Practical Guide and ECPR and Resuscitative ECMO: more ventilator support and more vasoactive support, delivered by a team that is not the one working on the circuit (Chapter 33 §33.6).

Certainty: high as expert consensus, concordant across the Red Book, ECPR and Resuscitative ECMO and the Practical Guide's algorithm; no study compares their outcomes.

🧠

Physiology · Why the right first move depends on configuration

Stopping flow does two things at once: it stops air moving toward the patient, and it removes all extracorporeal support. Which of those matters more depends on what the circuit is doing.

Venovenous support. The patient has a heart; stopping the pump removes oxygenation. Air that passes the membrane goes to the right heart and lungs, and to the systemic side only through an intracardiac shunt. The Red Book states the trade directly: "While air entering a patient should never be considered acceptable, patients on VV ECMO may tolerate small amounts of air better than pump stoppage, which may cause rapid desaturation and cardiac arrest."
Venoarterial support. Air that passes the membrane goes into the arterial tree — on femoral support up the aorta toward the coronary and cerebral circulations. And stopping the pump opens the retrograde shunt Chapter 33 describes unless the return limb is clamped. The Red Book singles out centrally cannulated venoarterial support: "even a small amount of air entering the patient can have adverse consequences and an automated pump stoppage in response to air detection may be warranted."

So the same bubble gets a different first move. On venovenous support with air still in the membrane lung, slow the pump, clamp in hand, and find the leak. On venoarterial support with air past the membrane lung, clamp now.

Certainty: high for the mechanism; moderate for the configuration-specific guidance, which is expert judgement from one primary text and not tested.

35.6 Getting the air out

🧮

De-airing techniques, by where the air is

Where the air is	Technique	Source and caveat
In the membrane lung	Aspirate from the highest point through a pigtail or stopcock; or use the membrane's de-airing port to vent passively	Red Book; ECPR and Resuscitative ECMO. Passive de-airing ports "often stop functioning after the membrane has been in use for an extended period" (Red Book)
Small amounts in the tubing	Move it to the membrane lung by gravity — the preferred place to remove air — or withdraw it with a syringe at a Luer-lock access point	Red Book; ECPR and Resuscitative ECMO
A large amount in the tubing	Isolate it between two clamps, cut the tubing, and reconnect with a straight connector using a wet connection	Red Book
In the pump head (airlock)	Speed to zero, return limb clamped, remove the head from the motor and position it so gravity lets the air rise out, then remove it by one of the above	Red Book. Never raise the speed to clear it (§35.3)
Throughout, after flow has been stopped	Push and pull: a 60 mL syringe filled with crystalloid on an emergency priming line or access port pushes air out into a second syringe on a different port	ECPR and Resuscitative ECMO
On venoarterial support	Disconnect the drainage tubing and let native cardiac activity passively fill the circuit, displacing the air	ECPR and Resuscitative ECMO. Requires a heart that is ejecting
Airlock in an integrated console	A backflush technique	Chan and colleagues (ASAIO J 2022, abstract read) report it as safe and effective for centrifugal pump airlock in their system; the technique is not described in the abstract. [VERIFICATION REQUIRED.]

When to stop trying and change the circuit. The Red Book: "In some cases of massive air entrainment, removal in a timely manner may be difficult or impossible. If the problem cannot be resolved rapidly and the patient is deteriorating, an immediate circuit change should be undertaken." ECPR and Resuscitative ECMO adds the physical reason: if the pump was not stopped fast enough and blood has foamed where it mixed with air, "which is difficult to remove," a new primed circuit should be considered.

And before restarting, the source must be closed. ECPR and Resuscitative ECMO: once flow is re-established, the specialist should watch that no additional air is entering the system. A circuit de-aired while the leak is still open will fill again.

Certainty: moderate — consistent expert technique across two primary texts; no comparison of methods exists.

🔬

Evidence · The only clinical series retrieved

Chan KM, Wan WTP, Ling L, So JMC, Wong CHL, Tam SBS. "Management of Circuit Air in Extracorporeal Membrane Oxygenation: A Single Center Experience." ASAIO J 2022;68(3):e39–e43. DOI 10.1097/MAT.0000000000001494. Abstract read.

4 of 116 patients supported between October 2009 and July 2020 developed gas bubbles in the circuit or systemic gas embolism.
In all four the source was located; in three it was on the venous side.
Two had centrifugal pump airlock with cessation of flow.
Flow was re-established by circuit change or by de-airing with a backflush technique.
All four were weaned from ECMO and three survived to hospital discharge. The authors conclude that with effective management, systemic gas embolism could be avoided.

What it adds. A single-centre incidence — about 3.4 percent of patients, the book's arithmetic from 4 of 116 — and confirmation that the source is usually findable and usually venous.

Certainty: very low — four patients, one centre, one console, abstract only.

35.7 When air has reached the patient

⚠️

Danger · Stop the supply first, then treat the patient

The sources give a short and consistent list. None of it works while air is still being delivered, so the clamp comes first.

Clamp the circuit. The ISCCM manual gives this as the first step for both venous and arterial air embolism.
Position. ECPR and Resuscitative ECMO: lower the head of the bed, which may keep air out of the cerebral circulation on venoarterial support, or on venovenous support through a patent foramen ovale. The ISCCM manual: Trendelenburg position for venous air.
Resuscitate. The ISCCM manual gives cardiopulmonary resuscitation for both venous and arterial air embolism once the circuit is clamped, since a circuit full of air and froth cannot pump — cardiac arrest on venoarterial support, profound hypoxia on venovenous support.
Aspirate venous air — the ISCCM manual suggests a pulmonary artery catheter or central line.
For arterial air, neuroprotection. The ISCCM manual: neuroprotective measures such as hypothermia "should be considered as soon as possible." Chapter 38a owns what follows.
Image. The ECMO simulation text's air embolism scenario includes considering imaging for air emboli once the circuit is safe.

What the sources do not give. No source retrieved for this chapter addresses hyperbaric oxygen for air embolism in a patient on ECMO, and none gives a tolerable volume of air in a human on the circuit. The in vitro study cites animal data putting a lethal venous dose at 200 to 300 mL, or 3 to 5 mL/kg, from a secondary reference that was not retrieved. [VERIFICATION REQUIRED.]

Certainty: low to moderate — expert guidance from textbooks, not tested on ECMO patients.

🚩

Pitfall · The case that teaches the whole sequence

The ECMO simulation text builds an air embolism scenario that is worth reading as a real sequence. A child arrives on venoarterial support from another hospital; venous chatter is noted; loose venous cannula purse strings are found and air is seen throughout the circuit, including after the oxygenator; heart rate, blood pressure and saturation begin to fall.

The expected actions, in the order the scenario gives them: clamp the arterial ECMO tubing immediately to prevent air entering the patient; Trendelenburg position; clamp the venous tubing and open the bridge; consider imaging for air emboli; start CPR, ventilate, de-air the circuit, and have the surgical team re-suture the venous cannulation site; re-establish flow.

Every theme of the last two chapters is in it. Chatter is a warning of negative pressure (Chapter 34 §34.2). The leak was at the cannula, on the drainage side (§35.2). The air crossed the membrane (§35.3). The clamp on the return limb came first (§35.5). And the source was closed before flow was restarted (§35.6).

35.8 Circuit rupture and breach

⚠️

Danger · Which side the hole is on decides what comes out of it

ECPR and Resuscitative ECMO names the moments of highest risk: "Patient transport, ambulation, or repositioning are high risk events for circuit damage." Its approach to a damaged component, such as a stopcock, is in two steps.

Temporise. Clamp the pigtail, cover the hole with occlusive tape or bone wax, "or even a sterile gloved finger" — to stop further air entrainment on the pre-pump side, or haemorrhage on the post-pump side.
Replace. If temporising works, change the broken component between one or two clamps on either side of it, using controlled separation (§35.9). If it does not work, emergent separation, and replacement of the damaged component or the whole circuit.

The Red Book adds the large-breach case: "A massive circuit breach will result in pump stoppage" — so the breach may present as pump failure (Chapter 33 §33.2).

The physics behind the temporising measures is §35.1, and this is the book's reading of it. On the drainage side a finger over a hole stops air being drawn in; the pressure is below atmospheric and the finger is held on by it. On the return side a finger is pushing against pump pressure and blood will find its way round it. A return-side breach needs a clamp between it and the pump, or the pump stopped, much sooner.

The Red Book's registry definitions show how this is counted. "Other tubing rupture" is rupture of ECLS tubing; "raceway rupture" is rupture of the raceway tubing in a roller pump; and "air in circuit" is air "requiring circuit intervention or circuit clamping for bubble detector alarm, visualized air, air entry into patient." A bubble-detector alarm that leads to a clamp is a registered complication.

Certainty: moderate — consistent expert description; no frequency data for rupture apart from the Japanese circuit-damage bleeding events (9 of 26).

💡

Clinical pearl · Three cheap ways to prevent a breach

Plastic ties on all circuit connections. The Red Book: "Many centers use plastic ties ('zip ties') on all circuit connections to prevent accidental separation."
The fewest possible access points, above all on the negative-pressure side. Every added device — a haemofilter, a monitoring line — brings connectors that are "a nidus for clot formation and, when on the negative pressure side of the circuit, conduits for air entrainment." The Red Book notes that many successful programmes avoid integrating additional devices altogether, while others do so routinely, and asks each programme to weigh it. Chapter 55 owns integrated renal replacement therapy.
Tubing laid where it cannot be stepped on, kinked, or snagged during transfers and turns — the high-risk moments.

35.9 Two ways off the circuit: emergent and controlled separation

🧮

The two protocols, side by side — from ECPR and Resuscitative ECMO

Emergent separation — "CLAMP, STOP, RESCUE"	Controlled separation — a failing component still giving passable support
CLAMP the arterial and venous cannulas	1. Prepare a sterile field with the components to be exchanged
STOP the flow by decreasing speed to zero	2. Time-out to clarify roles
RESCUE the patient with increased haemodynamic support and ventilator settings	3. Adjust the ventilator and have vasoactive drugs ready
Correct the problem — which may require a circuit change	4. Decrease ECMO flow to about 1.5 L/min
Reinitiate flow with the pre-initiation checklist	5. With no bridge, clamp the return cannula, then the drainage cannula
	6. Speed to zero
	7. Correct the problem
	8. Restart at about 1,500 rpm
	9. Unclamp the drainage cannula, then the return cannula
	10. Return flow to full support

When each is used. Emergent separation is for any catastrophe "when temporizing measures are not sufficient." Controlled separation is for a component beginning to fail but still providing passable support — the text's examples are a membrane lung with a post-membrane oxygen tension under 100 mmHg, or a circuit rupture temporised with bone wax.

Read the order of the clamps in the controlled protocol, because it is not arbitrary. Off: return first, then drainage. On: pump turning at about 1,500 rpm first, then drainage open, then return open. This book's reading of why: clamping the return side first, and opening it last, means that at no moment is the return limb open to the patient while the pump is stopped or turning too slowly to push forward — which is exactly the retrograde-flow window Chapter 33 §33.4 describes. The Practical Guide's decoupling pathway makes the same point: raise the speed to 1,500 and then unclamp.

With a bridge. The same text gives a separate sequence for circuits with an arteriovenous bridge: to stop, clamp after the oxygenator and the shunt, clamp the return and drainage cannulas above the bridge, open the bridge and remove the post-oxygenator clamp so blood keeps circulating through the bridge; to resume, clamp after the oxygenator and shunt, close the bridge, raise the speed to 1,500, unclamp drainage and return, remove the post-oxygenator clamp, and clear the bridge. Chapter 31 §31.2 owns why a bridge must be kept clear.

Certainty: moderate — a single primary text's protocols, consistent with the Red Book and the Practical Guide; untested against alternatives.

💡

Clinical pearl · What has to be in the room before any of this happens

The Red Book's preparation list, which Chapter 33 began and this chapter completes.

A primed backup circuit, readily available — ideally one for each area of the hospital where ECMO patients are treated. Circuits primed with crystalloid in a sterile fashion "can be safely maintained for 30 days or longer without loss of function or microbial contamination." ECPR and Resuscitative ECMO asks every programme to consider one.
A backup pump system — hand crank or second console — in the room and positioned for immediate use (Chapter 33 §33.6).
An emergency supply cart on every unit with everything needed to resolve a circuit emergency, and two tubing clamps on the ECMO cart itself (ECPR and Resuscitative ECMO).
Checklists and protocols for every circuit emergency, reviewed regularly, with individual roles defined in advance.
Training. The Red Book: high-fidelity simulation is more effective than water drills; elective circuit and component exchanges are an opportunity for junior staff to practise an emergency skill under supervision; and ELSO recommends training sessions of 4 to 8 hours every six months as a starting point, covering at minimum circuit ruptures and disruptions, component exchanges, air entrainment (venous and arterial), power or console failure, and inadvertent decannulation.

Chapter 90 turns this into a checklist; the training programme is owed to Chapters 88 to 90.

35.10 Controversy 1 · Should the bubble detector be allowed to stop the pump?

⚖️

The clinical question. Most consoles can be set to stop the pump automatically when the bubble detector fires, or to alarm only. Which should be the default?

Position A: stop the pump, because air in the patient cannot be undone. A detector that only alarms relies on a human seeing, understanding and clamping within seconds — at three in the morning, with the nurse at another bed. Franco's model shows that air is driven through the membrane faster at higher speeds, so the seconds matter most exactly when flow is highest. The Red Book agrees for one group: in centrally cannulated venoarterial support, "even a small amount of air entering the patient can have adverse consequences and an automated pump stoppage in response to air detection may be warranted."

Position B: alarm only, because stopping the pump can kill faster than a bubble. The Red Book states the other side in the same paragraph: patients on venovenous support "may tolerate small amounts of air better than pump stoppage, which may cause rapid desaturation and cardiac arrest." It adds the specific hazard: "A sensor placed before the ML and configured to stop the pump in response to bubble detection may create a dangerous situation in response to a small amount of air that would likely have been trapped in the ML." And on venoarterial support an automatic stop without a clamp opens the retrograde shunt (Chapter 33 §33.4) — a machine can stop a pump, but it cannot clamp a line.

Where the evidence actually sits: there is none. No study retrieved compares automatic stop against alarm-only on any outcome. The Red Book's conclusion is itself the recommendation: "The decision to implement an automated intervention for air deserves thoughtful consideration and each program must weigh risks and benefits before settling on a standardized approach." [VERIFICATION REQUIRED — no outcome comparison retrieved.]

What this book concludes, as a reconciliation of mechanisms.

Detector position and detector action must be chosen together. A pre-membrane detector sees air early and fires for air the membrane would have caught; pairing it with automatic stop gives the Red Book's dangerous combination. A post-membrane detector fires only for air heading to the patient; that is the position where an automatic stop makes most sense.
Configuration decides the default. Central venoarterial: automatic stop at a post-membrane detector is defensible, with the return clamp as the immediate next action. Venovenous: alarm-only is defensible, with the Red Book's sequence (§35.5) as the response.
Whatever is chosen, it is written down, standardised across the unit, and known to every person who may answer the alarm. A console that stops itself surprises people who expected an alarm; the Red Book asks each programme to settle on a standardised approach.

Certainty: very low for any default. The reconciliation rests on the Red Book's mechanisms and one in vitro study; it is not an outcome finding.

35.11 Controversy 2 · Air is seen: stop the pump, or keep it running?

⚖️

The clinical question. Air is visible in the circuit, not yet past the membrane lung. Clamp and stop now, or keep flow going at a reduced rate while the source is found?

Position A: clamp and stop. ECPR and Resuscitative ECMO's protocol is simple and teachable: large amounts of air, "or even small amounts within the arterial section," trigger CLAMP, STOP, RESCUE, which "will also stop the movement of air towards the patient." The Practical Guide's algorithm begins its massive-air pathways the same way. A simple rule is more likely to be followed under stress (this book's argument, not a sourced finding), and a stopped pump cannot drive air through a membrane. Franco's model supports the direction: speed pushes air through; stopping stops it.

Position B: slow down and keep going, with a hand on the clamp. The Red Book: "In the absence of substantial air entry, it is generally preferable to maintain circuit flow while the problem is resolved as many patients will not tolerate complete cessation of blood flow." Flow is reduced to the lowest tolerated level, and one person watches the arterial limb, ready to clamp. The Practical Guide's micro-air pathways do not begin with stopping the pump: they begin with identifying the source and clamping if air is in the arterial limb. A patient dependent on the circuit may be harmed more by the stop than by a bubble that the membrane lung is holding.

Where the evidence actually sits. Two expert protocols that agree on more than they appear to. Both clamp at once when air is past the membrane or large. They differ only for small venous air still in the membrane lung, and there the difference is between "stop" and "slow, watch, and be ready to stop." No study compares them. Chan's four patients were all weaned and three survived, but the abstract does not say which approach was used first.

What this book concludes.

Air past the membrane lung, or a large volume anywhere: clamp the return limb and stop. Both protocols agree.
Small venous air held in the membrane lung, a stable patient, and a person free to stand at the return limb with a clamp: slow the pump and find the leak. This is the Red Book's approach, and it avoids a stop the patient may not survive.
If there is nobody free to stand at the return limb, the second option does not exist. Stop.
In an airlock, the question does not arise. The pump has already stopped; clamp the return limb and do not raise the speed (§35.3).

Certainty: low. The conclusion reconciles two expert protocols using the Red Book's mechanism; it has not been tested.

35.12 A worked case: a hiss after a blood draw

🧮

Worked example · 22:40, venovenous ECMO, and a central line in the neck

The patient. Day 3 of femoro-jugular venovenous ECMO for severe ARDS. Femoral multistage drainage cannula with its tip near the cavo-atrial junction; right internal jugular return cannula. A left internal jugular triple-lumen central line. Speed 3,600 rpm, flow 4.8 L/min, inlet pressure minus 60 mmHg. The patient is lightly sedated and breathing spontaneously on pressure support. The bubble detector is after the membrane lung and set to alarm only.

The event. At 22:40 a nurse draws blood from the central line's distal lumen and is called away before closing it. Two minutes later the ECMO specialist hears a new hissing sound from the pump head. Flow drops to 3.9 L/min and recovers partly. A thin line of bubbles is visible in the drainage tubing, and a small collection of air sits at the top of the membrane lung. The return-side bubble detector is silent. Saturation is unchanged.

The reading, zone by zone.

Zone	Finding	Meaning
Before the pump	Continuous small bubbles	An open source on the negative side, still drawing
Pump head	Hiss; flow dipped	Air is passing through the pump — Franco's first warning sign
Membrane lung	Air collecting at the top	The trap is working, for now
After the membrane lung	No bubbles; detector silent	Nothing has reached the patient yet — and the silent detector is exactly what Franco predicts for air the membrane has caught

What should happen, in order.

Call for help. One person picks up a clamp and stands at the return limb near the patient, watching the tubing after the membrane lung.
Reduce pump speed to the lowest flow the patient tolerates — the Red Book's approach for venous air with no substantial entry to the patient (§35.5, §35.11). Lower speed also makes it less likely that air is driven through the membrane.
Follow the bubbles upstream. They begin at the drainage cannula, so the source is upstream of the circuit, in the patient's veins — and the only open venous access is the central line. Close the lumen. The bubbles stop.
Remove the air: aspirate from the top of the membrane lung through its port (§35.6). Check the pump head and tubing for residual air.
Only then restore flow, and watch that no more air enters.
Record it and fix the process: the lumen was left open. Franco's model found that a closed stopcock or a needle-free valve on the catheter prevented air entry at every speed and distance tested. The spontaneously breathing patient matters too: the authors expected a vigorous inspiration to make entry through an open line far more likely than their model showed.

The counterfactual. Change one detail — the pump note rises and then the flow falls to zero, with the membrane lung full of air. That is an airlock (§35.3). Clamp the return limb, speed to zero, close the source, and de-air before any attempt to restart. Turning the speed up to "get flow back" is the one move that could drive the air into the patient.

The teaching point. The detector did not alarm, and it was not supposed to. The ear found this; the source was found by following the bubbles upstream; and the fix was a closed lumen — Chapter 34 §34.8's central-line rule, measured.

This case is constructed to illustrate the mechanisms described in this chapter. The values are illustrative and are not drawn from a published report.

35.13 What this chapter could not close

🚩

Open questions, recorded rather than concealed

How often air enters patients' circuits, and how often it reaches the patient, is unknown in adults. The in vitro study quotes an incidence of up to 4.6 percent in adults and 4.9 percent in neonates from secondary references that were not retrieved; Chan's single centre gives about 3.4 percent (4 of 116). [VERIFICATION REQUIRED.]
No human data give a tolerable or lethal volume of air on ECMO. The 200 to 300 mL (3 to 5 mL/kg) figure is animal data quoted secondhand.
No study compares automatic pump stoppage against alarm-only for bubble detection (§35.10), or stop-first against slow-and-watch (§35.11).
The in vitro volumes are from one console, one cannula pair, a blood substitute, and no breathing. They set the order of magnitude, not a threshold.
No source retrieved addresses hyperbaric oxygen for arterial air embolism in a patient on ECMO, or whether and how such a patient could be treated.
The backflush de-airing technique is reported as effective in one centre's abstract without a description. [VERIFICATION REQUIRED.]
No frequency data exist for circuit rupture in modern centrifugal circuits beyond the Japanese circuit-damage bleeding events.
The 30-day sterility of a primed backup circuit rests on four studies cited by the Red Book that were not retrieved here (Deptula 2021, Weinberg 2015, Tan 2018, Naso 2013). The claim is the Red Book's.

Corrections and additions owed to other chapters

Chapter 31 (Circuit Thrombosis) is owed the Red Book's warning that every added access point is a nidus for clot, and that the exchange triggered by air belongs in its exchange-decision framework as an emergency trigger distinct from its planned ones.
Chapter 32 (Oxygenator Failure) is owed the membrane lung as an air trap that fails with speed and volume, and the observation that passive de-airing ports stop working after extended use.
Chapter 33 (Pump Failure) is owed airlock as a cause of a stopped pump, with the rule that the speed is never raised to clear it, and the clamp order in controlled separation as the procedural answer to its retrograde-flow window.
Chapter 34 (Cannula Problems) is owed Franco's measurement of its central-line rule: a closed stopcock or needle-free valve prevented air entry at every speed and distance; an open lumen drew air at 2,500 rpm with tips touching and at 3,500 rpm or more otherwise.
Chapter 38a (Neurological Complications) is owed arterial air embolism — its sources, the head-down position, the suggestion of hypothermia in the ISCCM manual, and the absence of any hyperbaric guidance.
Chapter 55 (AKI and CRRT) is owed the Red Book's warning about negative-side connectors for integrated devices, and a 2026 Critical Care narrative review's statement (DOI 10.1186/s13054-026-06033-6, abstract read) that integrated configurations require careful pressure management to avoid alarms, haemolysis and air embolism.
Chapter 60 (Transport) is owed transport, ambulation and repositioning as the high-risk moments for circuit damage.
Chapters 88 to 90 are owed the emergent and controlled separation protocols, the bridge sequences, the preparation list, and ELSO's training recommendation.

Carried forward and still open: Chapter 33's unmeasured retrograde-flow interval; Chapter 34's missing cannula-days denominator; Chapter 32's unretrieved source for the 150 mL/min oxygen transfer threshold.

35.14 Evidence and certainty summary

Claim	Best evidence	Certainty
The drainage side is below atmospheric pressure, so breaches there draw air in; the return side is above, so breaches there push blood out	ELSO Red Book; ECPR and Resuscitative ECMO	High — physics
Air before the membrane lung is venous air and may be caught there; air after it is arterial air with nothing between it and the patient	ELSO Red Book definitions; ECPR and Resuscitative ECMO	High for the definitions; moderate for the membrane as a trap
Circuit air is nearly always avoidable; two thirds of reported air events were clamp or stopcock errors	Red Book; Hadano 2023, 12 events. Full text read	Moderate for the direction; low for proportions
An open central-line lumen near the drainage cannula draws air at higher pump speeds; a closed stopcock or needle-free valve prevents it	Franco 2024, in vitro. Full text read	Low to moderate — in vitro, one set-up
The pump's sound changes with as little as 1 mL of air; 25 mL stops flow at any speed; air passes the membrane more readily at higher speeds	Franco 2024	Low to moderate — order of magnitude, not thresholds
A return-side bubble detector is blind to air held in the membrane lung	Franco 2024; consistent with the Red Book's description of detector positions	Moderate
Microbubbles depend on negative inflow pressure; the centrifugal pump is a main source; the oxygenator's ability to hold them is limited	Born 2020, in vitro. Abstract read; Red Book on cavitation	Low
In an airlock, raising the speed may drive air through the membrane lung; clamp the return limb and set the speed to zero	Franco 2024; Red Book pump-head de-airing instruction	Moderate
For air with no substantial entry to the patient, maintaining reduced flow with a person ready to clamp is generally preferable to stopping	ELSO Red Book	Moderate as expert guidance; untested
For large air or any arterial air: CLAMP, STOP, RESCUE, and lower the head of the bed	ECPR and Resuscitative ECMO; Practical Guide algorithm	High as consensus; untested
Venovenous patients may tolerate small amounts of air better than a pump stop; central venoarterial patients may warrant automatic stoppage	ELSO Red Book	Moderate — expert judgement
Air can be removed from the membrane's highest point, moved there by gravity, isolated and cut out, or pushed out with syringes; foam means circuit change	Red Book; ECPR and Resuscitative ECMO	Moderate — consistent technique
4 of 116 patients had circuit air in one centre; the source was always found, usually venous; all were weaned	Chan 2022. Abstract read	Very low
After air reaches the patient: clamp, head down, CPR if needed, aspirate venous air, consider neuroprotection for arterial air	ISCCM manual; ECPR and Resuscitative ECMO; ECMO simulation text	Low — textbook guidance, not tested on ECMO
A drainage-side breach can be temporised by occlusion; a return-side breach needs a clamp or a stop; a massive breach presents as pump stoppage	ECPR and Resuscitative ECMO; Red Book	Moderate
Controlled separation: flow to about 1.5 L/min, clamp return then drainage, speed zero, fix, restart at about 1,500 rpm, open drainage then return	ECPR and Resuscitative ECMO	Moderate as protocol; the retrograde-flow rationale is this book's reading
A sterile crystalloid-primed backup circuit can be kept for 30 days or longer	ELSO Red Book, citing four studies not retrieved	Moderate as quoted
Training of 4 to 8 hours every six months, including venous and arterial air, as a starting point; simulation beats water drills	ELSO Red Book	Moderate as recommendation

35.15 Key points

The pump divides the circuit into a side that sucks and a side that pushes. A drainage-side hole draws air in; a return-side hole pushes blood out.
The membrane lung divides the positive side again. Air before it is venous air and may be caught; air after it is arterial air and goes to the patient.
Place every problem in one of three zones before touching anything: before the pump, between pump and membrane, after the membrane.
The patient's own central veins are part of the drainage limb. Any open line, sheath or surgical opening near the drainage cannula is a hole in the circuit.
A patent foramen ovale turns venovenous air into arterial air.
Circuit air is nearly always avoidable. In reported events, two thirds were a clamp or a stopcock handled wrongly.
A closed stopcock or needle-free valve on the central line prevented air entry at every speed and distance tested; an open lumen drew air.
Millilitres are audible and tens of millilitres stop the pump. The hiss from the pump head is often the first sign.
The membrane lung is a trap on a timer. Higher speed and larger volumes push air through it.
A detector after the membrane lung cannot see air the membrane has caught. The ear and a light on the top of the membrane can.
A pump stopped by air has, for the moment, protected the patient. Never raise the speed to clear an airlock. Clamp the return limb, speed to zero, then de-air.
When air is seen: call, decide whether it will reach the patient, and put one person on the return limb with a clamp.
Air past the membrane, or a large volume anywhere: CLAMP, STOP, RESCUE, and lower the head of the bed.
Small venous air held in the membrane, a stable patient, and someone free to hold the clamp: slow the pump and find the leak. If nobody is free to hold the clamp, stop.
Find the source by following the bubbles upstream, and close it before restarting. A circuit de-aired with the leak still open fills again.
Foam, or air that cannot be removed quickly in a deteriorating patient, means a circuit change.
Controlled separation clamps the return first and opens it last, with the pump already turning at about 1,500 rpm — which keeps the retrograde-flow window closed.
A drainage-side breach can be held with a gloved finger or bone wax; a return-side breach needs a clamp or a stop.
Whether the bubble detector may stop the pump is a unit decision made with its position in mind, written down, and known to everyone.
A primed backup circuit, a backup pump, an emergency cart, and training every six months are what make every section of this chapter possible.

📌

Chapter answer: Air behaves according to where it is. Before the pump the circuit is below atmospheric pressure, so any opening — a stopcock, a loose connection, a central line near the drainage cannula, a partly withdrawn cannula — draws air in silently for as long as the pump runs; after the pump any opening pushes blood out. Venous air usually collects at the top of the membrane lung, but faster flow and larger volumes carry it through, and air past the membrane has nothing between it and the patient. So the first moves are to call for help, decide whether air is about to reach the patient, and put one person at the return limb with a clamp. If air is past the membrane, or there is a lot of it, clamp and stop, lower the head of the bed, and support the patient with a second team. If a small amount is held in the membrane and the patient needs the circuit, slow the pump, follow the bubbles upstream to the source, close it, and only then remove the air. A pump that has stopped from air must never be sped up to get flow back. And because most reported air events are a clamp or a stopcock handled wrongly, the best treatment is a closed system, a primed spare circuit, and a team that has practised all of this.