The question this chapter answers
Registries remember bleeding. Incident reports remember cannulas. When a national patient-safety system was searched for every ECMO accident over twelve years, the two commonest events were not haemorrhage or clot but a cannula in the wrong place and a cannula that came out — together nearly half of all reports, and the malposition events were fatal in almost half their cases.
So the bedside question is not is the cannula in? but where is it now, what is holding it there, which of its failures is happening, and does this patient survive the next five minutes without it?
This chapter is about the cannula after insertion: why a fixed cannula moves, how a drainage problem is told apart from a return problem, how to correct a migrated cannula, what happens in the seconds after one comes out, and what the evidence says about holding it in.
Evidence search date: 24 September 2026. Two sources are read as complete full text and are the Verified sources in this chapter: a Japanese national analysis of ECMO accident reports, and an international survey of cannula and circuit management from 45 countries. Primary sources are the ELSO Red Book 6th edition (the circuit chapter, the circuit malfunction and crisis management chapter, the bedside ultrasound chapter, the diagnostic procedures chapter and the paediatric respiratory weaning chapter), ECPR and Resuscitative ECMO (initiation, securement, the circuit check and the circuit catastrophes chapter), ECMO: A Practical Guide to Management (cannulation steps, nursing cannula care, and the complications and emergencies chapter), Comprehensive Healthcare Simulation: ECMO Simulation (the proning and mobilisation scenarios), ECMO in the Adult Patient 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 |
The cannula after insertion — migration, malposition found during the run, kinking, cracks, side ports and connections | Insertion itself. Chapter 5 (venovenous), Chapter 12 (venoarterial) and Chapter 21 (ECPR) own cannulation, including wire control and immediate confirmation. §34.9 uses the insertion-injury data only to show what malposition does |
Telling a drainage-cannula problem from a return-cannula problem with the pressures the console already shows — the part of Chapter 33's six-item differential that sits upstream and downstream of the pump | Imaging the cannula. Chapter 26 §26.5 owns it: why the radiograph cannot localise the tip, why position drifts, and the four presentations of the malposition syndrome. Chapter 26 owns seeing it; this chapter owns fixing it |
Correcting a migrated cannula, and the decision to leave one alone | Persistent hypoxaemia and recirculation. Chapter 9 owns hypoxaemia on venovenous support and the chatter it describes; Chapter 6 owns recirculation physiology |
Accidental decannulation — its physiology, its causes and the first minute | The air. Chapter 35 owns air in the circuit and de-airing; Chapter 90 owns the emergency circuit checklist |
Securement as a clinical intervention, and the evidence on how to do it | Cannula-site bleeding as a haemostatic problem (Chapter 36), cannula-site infection (Chapter 58), limb ischaemia (Chapter 17), and thrombosis after removal (Chapter 10) |
The cannula as a mechanical object that can crack, and the central line inserted near it | Thrombus on or in the cannula as a thrombotic diagnosis. Chapter 31 owns it; this chapter owns what it does to drainage |
34.1 The component nobody can replace in a minute
Every other part of the circuit can be exchanged while the patient waits on a clamp: a pump head in seconds, a membrane lung in minutes, a whole circuit in a few minutes with a practised team. The cannula is the one component that cannot be exchanged without re-entering a large vessel — and it is the only component that is attached to something that moves.
The structural claim of this chapter · The cannula is the circuit's only moving part that nobody set in motion
A cannula is fixed at the skin and free at the tip. Everything between those two points belongs to the patient's anatomy, and the anatomy changes shape: the diaphragm rises as lung volume falls, the head turns, the hip flexes, the patient is proned and supinated, washed, transported, woken. Chapter 26 §26.5 set out why: "The cannula did not move; the patient did."
Three consequences follow, and the chapter is built on them.
- A cannula position is a measurement with a time stamp, not a property of the cannula. The confirmation at insertion is a baseline. Every later problem in this chapter begins with the position having changed since somebody last looked.
- The two ends fail differently. The tip fails by drifting — slowly, silently, and announced by a flow or oxygenation problem somewhere else in the circuit. The skin end fails by letting go — suddenly, loudly, and announced by blood or air.
- Because the cannula cannot be exchanged quickly, prevention outranks treatment by more than it does anywhere else in Part VI. A clotted membrane is an inconvenience with a primed spare. A cannula out of a venoarterial patient with no native output is a cardiac arrest with a groin wound.
Certainty: high for the mechanism, which the Red Book, Chapter 26's sources and every emergency text in this chapter describe concordantly; the ordering of consequences is this book's framing.
Evidence · What a national accident-reporting system found when it looked
Hadano H, Kamio T, Fukaguchi K, Sato M, Tsunano Y, Koyama H. "Analysis of adverse events related to extracorporeal membrane oxygenation from a nationwide database of patient-safety accidents in Japan." J Artif Organs 27(1):15–22, published online February 2023. DOI 10.1007/s10047-023-01386-z, PMC9933024. Full text read.
The design is the point. Registries and systematic reviews record complications as clinical outcomes, and the authors note that almost all such reports come from high-volume centres. This study instead searched the Japan Council for Quality Health Care's accident-reporting database — mandatory for tertiary teaching hospitals, voluntary for others, about 20 percent of Japanese hospitals — for every ECMO-related event from January 2010 to December 2021, and read the free-text narratives. 178 ECMO-related accidents remained after exclusions.
Event | Proportion of 178 reports |
Cannula malposition (extravascular cannulation or the wrong vessel) | 28 percent (50) |
Accidental decannulation (removal requiring reinsertion or position adjustment) | 19 percent (34) |
Bleeding | 15 percent (26) |
Air in the circuit | 7 percent (12) |
Thromboembolism | 5 percent (9) |
Limb ischaemia; gas supply; power supply | 4 percent each |
Foreign body remnants | 3 percent |
Of all 178 accidents, 23 percent resulted in death and a further 26 percent had a high potential for residual disability. Of the 50 malposition events, 44 percent (22) were classified as resulting in death.
The reading this chapter takes. Cannula events made up 47 percent of what was reported — more than bleeding, air and clot combined. The authors draw the conclusion themselves: historically bleeding is the most feared and frequent adverse event, "however, data from the accident-reporting system used in this study suggest that cannula malposition and decannulation are more common adverse events."
Certainty: low for any frequency. There is no denominator — the authors state that an incidence rate cannot be calculated — reporting depends on individuals, the system is designed around errors, and the authors expect bleeding to be under-reported in it. What it establishes firmly is the shape of harm when things go wrong at the cannula: common among reported accidents, and — for malposition — fatal in 44 percent, a proportion exceeded only by thromboembolism (5 of 9) among the categories for which the authors report deaths.
Clinical pearl · Why the two data sources disagree, and why both are right
A registry counts what happened to patients; an accident system counts what went wrong in care. Bleeding is frequent and usually managed; a cannula event is less frequent and is, by its nature, something that went wrong. So the registry answers what should I expect? and the accident system answers what will kill someone if we are careless? Chapter 36 is owed the first question. This chapter is about the second.
34.2 Upstream or downstream: using the pressures the console already shows
Chapter 33 §33.2 listed six reasons the flow falls at an unchanged speed, and handed two of them to this chapter entire: the patient's preload, and the drainage. Chapter 9 owns the pressure-localisation rule in the setting of hypoxaemia. What this chapter adds is the discrimination the cannula itself requires — which cannula, and which end of it.
The pressure signatures, read from the Red Book's own table
The Red Book tabulates how a pressure-generating pump responds to each kind of change. For a falling flow at an unchanged pump speed, it gives three distinct patterns:
Cause, as the Red Book names it | Pump inlet pressure | Pump outlet pressure | Post-artificial-lung pressure | Where the problem is |
Preload decreased — "CVP usually or position of cannula" | Falls (more negative) | Falls | Falls | Upstream: the patient's volume, or the drainage cannula |
Resistance building in the artificial lung | Rises | Rises | Falls | In the membrane — Chapter 32 |
Afterload increased — "patient MAP or cannula resistance" | Rises | Rises | Rises | Downstream: the patient's arterial pressure, or the return cannula |
Note what the Red Book puts in each of the two cannula rows. It does not separate the patient from the cannula — both rows name a patient variable and a cannula variable together. The pressure pattern tells you which side of the circuit; it cannot tell you whether the patient or the plastic on that side has changed. That second discrimination is always made at the bedside: by looking at the cannula, by the depth mark, and by imaging (Chapter 26).
The practical rule that follows. A more negative inlet pressure with a falling flow means something upstream is starving the pump — hypovolaemia, bleeding, tension in the chest or abdomen, tamponade, a kink, or a drainage cannula against a wall or in the wrong compartment. A rising post-lung pressure with a falling flow and a flat pressure drop means something downstream is resisting it — vasoconstriction, a kink, a return cannula too small or malpositioned. ECPR and Resuscitative ECMO gives the same two lists under "venous drainage insufficiency (very negative venous drainage pressure)" and "arterial return obstruction (very high arterial return pressure)", and adds the item that belongs on neither: an extravascular cannula.
Certainty: high for the pressure logic, which is the Red Book's; the observation that each row pairs a patient variable with a cannula variable is this book's reading of the table.
Danger · What a drainage cannula does when the pump pulls harder than the vein can supply
The Red Book describes the sequence precisely. At the pump inlet, subatmospheric pressures down to minus 250 to minus 300 mmHg can be generated, risking cavitation: micro-bubbles that add shear stress and haemolysis, that implode in the positive-pressure zone at the pump exit with serious blood cell damage, and that may pass the membrane lung and raise the risk of clotting and gaseous embolism.
And at the cannula itself: "high suction forces can result in blood vessel collapse around the cannula, causing the drainage tubing to rhythmically move in a process referred to as 'chattering' or 'chugging'." The consequences it lists are all cannula consequences — occluded cannula holes, reduced net drainage, damage to the vessel intima, extreme negative pressure pulses of minus 600 to minus 650 mmHg, and haemolysis from high shear at the cannula tip.
Read that list as a cannula chapter would. Chatter is not only a symptom of a drainage problem; it is a cause of injury at the cannula tip while it continues. The vein is being repeatedly sucked onto the side holes. Chapter 9 owns chatter as a cause of hypoxaemia; this chapter's point is that every minute of it is also a minute of intimal trauma, which is why Chapter 33 §33.2's rule — never raise the speed to silence a low-flow alarm — is, at the cannula, a rule about the vessel wall.
The ISCCM manual gives the design response: drainage cannulas are longer, wider and multiperforated along a larger proportion of their length, features that reduce the pressure drop and limit chattering, avoiding a significant collapse of the wall of the right atrium or the inferior vena cava.
Clinical pearl · The two cannula findings on the console that need no transducer
- Colour. ECPR and Resuscitative ECMO's circuit check begins by looking at the colour difference between the drainage limb (dark) and the return limb (bright). Turned around, the same observation is a cannula test: a drainage limb that brightens on venovenous support means returned blood is being drawn straight back — recirculation, which is usually a cannula-position problem announcing itself in the tubing (Chapter 6, Chapter 9). The ECMO simulation text teaches the laboratory form: a rising circuit venous saturation with a falling patient saturation after a position change means cannula malposition until shown otherwise.
- Movement. A drainage line that twitches rhythmically is chatter, visible before the inlet pressure is read. A drainage line that is still and a flow that is low is not chatter — it is a kink, a clot, a clamp, or a pump.
34.3 Migration: finding it before the flow does
Chapter 26 §26.5 established that positions drift and that the radiograph cannot see it. The practical question for this chapter is earlier and cheaper: what is the first sign, and who is placed to see it?
Physiology · Two kinds of movement, and only one of them shows at the skin
Movement at the skin. The cannula slides in or out through the insertion site. Sutures loosen, a dressing is changed, a hip is flexed, a patient pulls. This shows as a change in the depth mark, and it is the kind of movement that ends in decannulation (§34.5).
Movement of the patient around a still cannula. The skin mark is unchanged, but the anatomy has moved: the diaphragm has risen with loss of lung volume and carried the right atrium cranially relative to a femoral cannula; the head has turned and changed the geometry of a jugular one. The depth mark cannot see this. It is Chapter 26's drift, and it shows only as a flow, recirculation or oxygenation change — or on an echocardiogram.
The Red Book states the consequence of either in its diagnostic-procedures chapter: "Migration of cannulas can also occur during ECLS, potentially compromising flow or causing injury to surrounding tissue, including the right ventricle, right atrium, or inferior vena cava."
So the depth mark is necessary and not sufficient. It catches the dangerous kind of movement early and misses the insidious kind entirely. An unchanged depth mark does not clear a cannula whose flow has changed.
Clinical pearl · The depth mark, done properly, is the cheapest monitor in the book
Every primary source asks for it, and they agree on the essentials.
- Mark and record at insertion. ECMO: A Practical Guide to Management: the insertion point to the skin should be marked according to institution policy at the end of cannulation.
- Check it often. The same text, in its nursing chapter: cannula site and depth should be assessed and documented at minimum every hour. The Red Book: depth of insertion and integrity of sutures and other securing devices should be noted and recorded at regular intervals.
- Check it at the moments that move cannulas. ECPR and Resuscitative ECMO's circuit check, performed at initiation, during any circuit emergency and at the start of every shift, includes noting the depth of cannula insertion, "as cannulas can become malpositioned with patient movement," and — with a flashlight — inspecting the drainage cannula for white fibrin strands and darker clots, which commonly develop over time (Chapter 31 owns what they mean).
The rule this book draws from the three: a depth that has changed since the last check is an event, recorded with a time and the reason, and followed by a flow review. A depth that has changed with no reason recorded is a securement failure in progress.
Pitfall · Where migrated cannulas go, as the published cases record it
The sources and case reports describe a small, recurring set of destinations. Each has its own presentation, and none of them presents as "cannula moved."
Where it went | How it presented | Source |
Dual-lumen bicaval cannula into a hepatic vein | The authors list inadequate flow, hypoxia and structural injury as what malposition causes; the abstract does not give each case's presentation | Tanaka and colleagues, J Card Surg 2015, two cases; Muhammad and colleagues, Perfusion 2023, one case in a 17-year-old |
Dual-lumen bicaval cannula into the right ventricle | Not described in the abstract; the Red Book lists the right ventricle among the structures a migrating cannula can injure | Tanaka and colleagues, second case |
Reinjection port drifting to face the liver, with the cannula unmoved at the skin | Inefficient oxygenation and rising liver enzymes after loss of lung volume | Red Book bedside ultrasound chapter; Chapter 26 §26.5 |
Femoral drainage cannula tip into an accessory hepatic vein on venoarterial support | Nothing at all — found on day 8 by ultrasound, with no disturbance of venous return | Winiszewski and colleagues, J Extra Corpor Technol 2018 |
The last row is the important one. A malposition that produces no flow change will be found only by someone who looks for it. Chapter 26's rule — re-image after any change in ventilation strategy, after proning, and whenever flow falls — is the only mechanism by which it is found at all.
Certainty: very low for frequency — these are case reports; moderate for the list of destinations, which recurs across independent sources.
34.4 Correcting a migrated cannula
The Red Book's instruction is short: "malposition of the cannula should be corrected expeditiously using fluoroscopy, echocardiography, and/or guidewires." The case literature supplies the techniques, and one thing the instruction leaves out: whether a particular malposition should be corrected at all (§34.11).
The techniques that have been reported, and what each requires
Situation | Technique reported | What it required |
Cannula tip too deep, reinjection jet misdirected (the Red Book's hepatic reinjection case) | Withdrawal under transoesophageal guidance until the reinjection is seen in the right atrium | An operator on the probe, colour Doppler, and a second person moving the cannula |
Dual-lumen cannula in a hepatic vein | Repositioning under echocardiographic guidance without a guidewire (Tanaka); or retracting the cannula, reinserting the wire and introducer together, then manipulating under serial radiographs and transthoracic echocardiography (Muhammad) | In Muhammad's account, confirming the wire in the infrahepatic inferior vena cava is the step the authors identify as ensuring safe position |
Dual-lumen cannula in the right ventricle | A guidewire passed from the femoral vein under fluoroscopy — without antegrade wire placement into the cannula, without discontinuing ECMO, and without bleeding (Tanaka) | Fluoroscopy, a second venous access, and an operator with interventional skills |
Three principles are common to every account, and they are the transferable part.
- Correction is done under real-time imaging, never by the depth mark alone. The depth mark shows how far the cannula moved at the skin; it does not show where the tip is going.
- A wire is the safety device whenever the tip must pass a branch point. The Practical Guide's insertion guidance makes the same point for dual-lumen cannulas: the wire is advanced through the intrahepatic inferior vena cava to the level of the iliac bifurcation specifically "to ensure inadvertent cannulation of the hepatic vein" is avoided. A repositioning is a partial reinsertion and inherits the insertion rules (Chapter 5).
- Flow does not have to stop. Tanaka's title is the point: How to Adjust Without Interruption of Flow. Reduced flow during manipulation is reasonable; separation from support is usually unnecessary.
Certainty: very low — three case reports and one textbook case. The principles are moderate, because they are the insertion principles of Chapters 5 and 26 applied again.
Pitfall · The two things a repositioning must be followed by
- Re-securing, and a new depth mark. The old mark now records a position that no longer exists. A repositioned cannula secured with the sutures that failed to hold the old position will migrate again.
- The Red Book's post-procedure views: after any cannulation, the absence of a new pericardial or pleural effusion should be ascertained, and the size of any pre-existing effusion reassessed. A repositioning manipulates a large-bore cannula near the right atrium in an anticoagulated patient, and deserves the same look. This is Chapter 26's cheapest safety step, and it applies here with the same force.
And the embolic risk Chapter 26 recorded: manipulating a large-bore cannula carries a risk of dislodging thrombus on it. The flashlight inspection for fibrin and clot along the drainage cannula (§34.3) is worth doing before, not after, a repositioning. No source retrieved quantifies that risk. [VERIFICATION REQUIRED.]
34.5 Accidental decannulation: what comes out determines what happens
Danger · The two cannulas fail in opposite directions
The Red Book defines the event and states both consequences in two sentences. Inadvertent decannulation "occurs when a cannula is displaced from the vessel, either partially or completely. If the return cannula is involved there will be rapid exsanguination while displacement of the drainage cannula may result in air embolism or loss of flow as blood in the pump head is replaced by air. Either scenario will result in blood loss from the cannula site."
ECPR and Resuscitative ECMO states the same asymmetry in six words each: venous drainage decannulation "will result in air entrainment into the circuit," and arterial return decannulation "will result in hemorrhage."
Read the physics and the difference in tempo appears.
- The return cannula is on the positive side. The pump keeps pushing. Whatever the set flow is, that is the rate at which blood now leaves the patient — through the circuit and out of the displaced cannula, and through the hole in the vessel. At 4 litres per minute the pump alone moves the whole of an adult's circulating volume in not much more than a minute. This is arithmetic from the set flow, not a measured figure; no source retrieved reports the blood loss in a real event.
- The drainage cannula is on the negative side. The pump keeps pulling. If the side holes are outside the vessel, it pulls air — through the skin tract or the open holes — into the drainage line and toward the pump head. The Red Book's two outcomes follow directly: the air reaches the patient, or it fills the pump head and the pump deprimes and stops moving blood. A multistage drainage cannula has holes along a long length, so a partial withdrawal can expose the proximal holes while the tip is still in the vein — which is why the Red Book's definition includes partially.
The partial case is the one that deceives. A partly withdrawn drainage cannula can still carry some flow; the first sign may be a stream of bubbles in the drainage line or a pump alarm, not blood on the sheet. Chapter 35 owns the air; this chapter's point is where it came from.
Certainty: high for the mechanism, stated concordantly by three primary texts; no source quantifies the tempo of either event.
Evidence · How often, and how it happens
How often. ECMO: A Practical Guide to Management puts accidental decannulation at approximately 1.3 percent of patients, calling it among the most fatal complications on support. The Japanese accident analysis attributes the same figure to Kim and colleagues, and notes that that study could not determine the causes. The Red Book's characterisation is "among the most uncommon but most dangerous ECLS related emergencies." Certainty: low — one quoted figure, no denominator or definition given in either text that quotes it. [VERIFICATION REQUIRED — the Kim primary report was not retrieved.]
How it happens — the only two sources that asked.
Japanese national accident reports (Hadano 2023) — 34 decannulations | International survey of 45 countries (Bull 2019) — 71 line events described by 61 centres |
All 34 were complete removals requiring reinsertion | Events were malposition, dislodgement or decannulation leading to an adverse patient outcome, reported by 34 percent of respondents over five years |
The arterial cannula in 85 percent (29 of 34) | Inadequate or ineffective securement — 28 percent (20) |
During another procedure, such as surgery or coronary intervention — 41 percent (14 of 34) | During insertion or manipulation of the cannula — 14 percent (10) |
Insufficient fixing — reported as 44 percent with a numerator of 23 of 34 (the published numerator and percentage do not agree) | The patient removed the cannula — 13 percent (9) |
A blinded insertion position — the site out of view — 41 percent (14 of 34) | During turning or bathing — 11 percent (8) |
Inadequate communication — 21 percent (7 of 34) | During transport 4.2 percent; cannula material failure 2.8 percent; ambulation 1.4 percent; not stated 25 percent |
The Japanese authors' own account of the mechanism: decannulation occurred when doctors performed another procedure and when nurses changed the patient's position, with attention distracted from the cannulation site. The Practical Guide's list of commonest causes matches: an improperly secured cannula, inadvertent removal of dressings or sutures over time, and migration during care — bathing, turning or transport.
The reading this chapter takes. Two unrelated data sources, one built from incident narratives and one from centre directors, name the same three things: securement that did not hold, a cannula nobody was watching, and a moment when the patient was being moved or worked on. None of the three is a device failure, and material failure accounts for under 3 percent of the survey's events. Accidental decannulation is overwhelmingly a process failure.
Certainty: moderate for that conclusion, because the two sources are independent and concordant; low for every proportion, because neither has a patient denominator and both depend on voluntary reporting.
Clinical pearl · The four moments the data keep pointing to, and the one rule for each
- Another procedure on an ECMO patient (a catheter laboratory, an operating table, a line insertion): one named person's only job is the cannulas. The Japanese finding that 41 percent occurred during another procedure is the strongest single signal in either dataset.
- Turning, washing and proning: cannulas visible, not under linen — ECPR and Resuscitative ECMO's instruction — and one person holding them. The Red Book's proning procedure assigns one person at the head to hold the endotracheal tube and the jugular cannula, and one at the feet guiding the procedure while holding the cannulas and avoiding any tension, with at least six experienced staff.
- Transport: Chapter 60. The cannulas are the first thing checked on arrival, not the last.
- An agitated or waking patient: the Practical Guide notes that confused or combative patients may require chemical or mechanical restraints, and suggests a knee immobiliser for awake patients who tend to bend their legs. Chapter 57 owns sedation; the cannula's claim on that decision is real.
34.6 The first minute after a cannula comes out
Danger · Clamp, stop, press — and a second team for the patient
The three primary sources give the same sequence in slightly different words.
- Call for help and clamp the circuit near the patient. Red Book: "The immediate reaction should be to call for help and immediately clamp the circuit near the patient." ECPR and Resuscitative ECMO: immediately clamp the return and drainage tubing and stop ECMO flow. The Practical Guide: ensure the existing cannulas are clamped and ECMO support is ceased. The clamp near the patient stops the return side exsanguinating the patient and stops the drainage side entraining more air.
- Pressure on the vessel. ECPR and Resuscitative ECMO: after flow has stopped, pressure needs to be applied at the exit site and the anticoagulant infusion held. The Practical Guide adds the rest of the vessel-side work: holding pressure, performing primary vascular repair, transfusion and resuscitation, until new access can be established.
- Support the patient, with a separate team. Red Book: increased ventilator support, blood product transfusion, vasopressor and inotropic administration "should be initiated immediately by a team distinct from the one troubleshooting the ECLS circuit." This is Chapter 33 §33.6's separation-of-roles requirement, arriving at its most urgent use.
- Keep the circuit alive. ECPR and Resuscitative ECMO: if a bridge is present, it can be used to maintain circuit patency while a new cannula is prepared. Chapter 35 owns de-airing a circuit that has entrained air.
- Prepare to replace the cannula — a new appropriately sized cannula, and, in ECPR and Resuscitative ECMO's words, "cutdown skills may be beneficial in this scenario."
Two things the sequence deliberately does not include. It does not include pushing the cannula back in, and it does not include restarting flow through a cannula whose position is unknown. None of the sources describes re-advancing a displaced cannula blind: every repositioning in §34.4 was done under real-time imaging, and most over a wire. The book reads the silence as deliberate and does not recommend it. [VERIFICATION REQUIRED — no source retrieved addresses it explicitly.]
Physiology · The decision that follows the clamp: does this patient need to go back on at all?
The Red Book's paediatric weaning chapter gives the decision explicitly for unplanned removal of a venous cannula: "Priorities in this situation are hemostasis and respiratory support." Then two scenarios are possible.
- Ventilation and oxygenation are adequate at acceptable ventilator settings and FiO₂: mechanical ventilation continues and is weaned as possible, using a lung-protective strategy. The patient has, in effect, been decannulated early, and the event becomes an unplanned trial off (Chapter 10).
- Even with high ventilation and FiO₂ the patient is not adequately supported: ECLS should be re-established expeditiously, while supporting the patient as well as possible, with separate teams responsible for each task if possible.
The same logic divides the configurations, and it is the reason the tempo differs.
- On venovenous support the patient has a heart. The ventilator is the backup, and it buys time. The question is whether the lungs can carry the load for as long as reinsertion takes.
- On venoarterial support in a patient with little or no native output, there is no backup. The heart that needed the circuit an hour ago cannot carry the circulation now. Accidental decannulation in that patient is a cardiac arrest from the moment the clamp goes on, and the Japanese data say it is usually the arterial cannula that comes out. Chapter 21's ECPR cannulation skills, and Chapter 90's checklist, are the treatment.
Certainty: high for the Red Book's two scenarios as expert guidance; the configuration contrast is this book's reasoning from Chapter 13's physiology, and no study compares outcomes of decannulation by configuration.
34.7 Securement is a treatment, and it has a technique
What the sources ask for, side by side
Element | What the source says | Source |
When | Securing sutures should be placed at the time of cannulation | Red Book, circuit malfunction chapter |
Where the sutures go | Positioned to prevent both migration of the cannula and bowing of the cannula away from the skin between the skin entry site and the securing sutures | Red Book |
How many points | At least four securement points with thick suture down the visible cannula and circuit tubing is typical; centres using adhesive devices generally use at least two per cannula. The Practical Guide: secured and fixed to the skin in several points, and regardless of method, the cannula must be secured at multiple points | ECPR and Resuscitative ECMO; ECMO: A Practical Guide to Management |
At the skin entry | The cutaneous insertion point tied closely around the cannula with separated stitches and/or a cutaneous purse-string, to avoid external bleeding | ECMO: A Practical Guide to Management, cannulation chapter |
Maintenance | Sutures can break or loosen over time and should be inspected regularly and replaced as needed. Additional adhesive devices can secure tubing to the patient | Red Book |
Connections | Many centres use plastic ties on all circuit connections to prevent accidental separation | Red Book |
Dressing | Chlorhexidine-impregnated dressings unless contraindicated, changed at least every 7 days and immediately if integrity is compromised; the cannula and tubing can cause pressure injury, so tubing is repositioned and a barrier dressing used | ECMO: A Practical Guide to Management, nursing chapter (Chapter 58 owns the infection argument) |
The overriding rule | "Whatever the approach to cannula securement, it should be standardized." ECPR and Resuscitative ECMO: "No high-quality data exist suggesting one method is superior. Whatever approach is used, it should remain consistent throughout each institution." | Red Book; ECPR and Resuscitative ECMO |
The Red Book's "bowing" instruction is the one most often missed and it is pure mechanics. A loop of cannula standing off the skin between the entry site and the first anchor is a lever. Any pull on the tubing is transmitted through the loop to the vessel instead of being taken by the anchor. Anchoring close to the entry site, with the cannula lying flat, converts a lever into a fixed point.
Certainty: moderate for every element as expert consensus, concordant across four texts; none of it has been tested against outcome, which §34.10 takes up.
Evidence · What 45 countries actually do
Bull T, Corley A, Lye I, Spooner AJ, Fraser JF. "Cannula and circuit management in peripheral extracorporeal membrane oxygenation: An international survey of 45 countries." PLoS One 2019;14(12):e0227248. DOI 10.1371/journal.pone.0227248, PMC6936833. Full text read.
Directors and coordinators of 396 ELSO-member centres in 51 countries were invited; 391 responses from 45 countries were received (43 percent response rate), with some double enrolment acknowledged.
- Suturing at the insertion site: 93 percent. Regional variation was large — 96 percent in the Americas and Europe against 68 percent in Australia and New Zealand.
- The primary securement along the line: direct skin suture 45 percent, securing to the bed or another fixed object 19 percent, a sutureless securement device 16 percent, adhesive tape or bandage 10 percent. In Australia and New Zealand the sutureless device was the commonest method at 42 percent.
- Fixation points along a femoral line: two in 48 percent, more than two in 26 percent, one in 21 percent.
- A cannula or line malposition, dislodgement or accidental decannulation leading to an adverse patient outcome in the previous five years: 34 percent of respondents.
- The respondents' own verdict: 78 percent said an evidence-based guideline for ECMO line management would be beneficial. The authors note that the securement products in use included several not designed or marketed for ECMO — grip-style devices, Foley catheter holders, colostomy ring patches.
The reading this chapter takes. One fixation point on a femoral ECMO line is reported by one centre in five. Every text in §34.7 asks for more than one. That gap between guidance and practice is the most actionable single finding in the survey.
Certainty: low to moderate. Self-report, recall bias, no patient denominator, possible double enrolment, and a respondent pool dominated by the United States. It describes practice reliably and outcome not at all.
34.8 The cannula as a mechanical object, and the line inserted next to it
Pitfall · Five mechanical cannula failures that present as something else
- A crack in the cannula presenting as a bleeding insertion site. Dugal and colleagues (Am Surg 2025, abstract read) describe an adolescent on venovenous support whose cannula site oozed for 13 days despite dressing changes and attempted suture repair; inspection then found a crack on the return side of the bicaval cannula. It was sealed with a skin adhesive rather than exchanging the cannula, the site became haemostatic, and support continued for 49 more days to recovery. The transferable lesson is not the adhesive — a single case — but the sequence: persistent oozing from a cannula site that does not respond to local measures should prompt inspection of the cannula itself, not only the skin. The Japanese analysis found the circuit counterpart: of 26 bleeding events, 9 were due to circuit damage, including a fissure in the connection between the arterial cannula and the circuit.
- A kink presenting as a flow problem. Both ECPR and Resuscitative ECMO's drainage and return lists include tube kinking; the Red Book's venoarterial monitoring table lists relieving a cannula kink among the treatments for low flow. A kink produces the pressure signature of the side it is on (§34.2), which is why it must be looked for along the whole length before the patient is blamed.
- An open side port presenting as air. ECPR and Resuscitative ECMO: "Some cannulae (typically intended for arterial insertion) have a built-in side-port which may have become unsecured and open to air" — and it lists damage to the circuit itself, typically when securing the cannulae as another source. The act of securing the cannula can puncture it.
- Reversed connections presenting as a patient who does not improve. ECPR and Resuscitative ECMO warns that the drainage limb must be in line with the drainage cannula and the return limb with the return cannula, that it has numerous examples of the limbs being connected wrongly, that the problem is difficult to correct once the tubing is fully hubbed, and that double-checking is prudent. Chapters 5, 12 and 21 own the check; the consequence surfaces in this chapter's territory.
- Fibrin on the drainage cannula presenting as a slowly falling flow. §34.3's flashlight inspection; Chapter 31 owns the thrombus.
Danger · The central line inserted near a drainage cannula
Two hazards, and both come from the drainage cannula's negative pressure.
- Guidewire aspiration. ECPR and Resuscitative ECMO: "Guidewire aspiration into the ECMO circuit is a rare, but catastrophic complication. The location of the ECMO venous drainage cannula and negative circuit pressure contribute to this risk. A high drainage cannula tip and increased ECMO pump speed have been associated with guidewire aspiration." The mitigations it lists: constant firm guidewire handling throughout, minimising guidewire insertion depth, and reducing pump speed.
- Air entrainment through the line. The same text: the negative pressure in the drainage cannula can overcome the resistance of one-way valves on the central line or sheath, entraining air; reducing pump speed and minimising open connections reduce the risk. The Red Book adds the staffing rule: "The person responsible for managing the ECLS pump should be aware and in the immediate area when central lines are being placed", and those drawing blood or giving drugs through central lines near a drainage cannula should take precautions against exposing the line to air.
The practical form: a central line on ECMO is a two-person procedure, one of whom is at the console with a hand on the speed. It is also exactly the kind of "other procedure" during which the Japanese data found 41 percent of decannulations happened (§34.5).
Certainty: moderate — expert description of a rare event with a clear mechanism; no incidence reported.
34.9 Malposition at insertion: what the injury data say about the cannula
Chapter 5, Chapter 12 and Chapter 21 own insertion. This section records only what the malposition data establish about the cannula as a cause of harm, because it bears on what a low flow at initiation means.
Evidence · Fifty malposition events, read from the full text
From the Japanese national analysis (Hadano 2023):
- Configuration: extracorporeal cardiopulmonary resuscitation 46 percent (23), venoarterial 28 percent (14), venovenous 26 percent (13). 74 percent of patients were over 60.
- Where the injury was: the retroperitoneum or intraperitoneal structures in 63 percent (27) — in every one of these the cannula had been inserted via the femoral vessels. With right internal jugular insertion, the commonest single insertion site for malposition (20 percent, 10), mediastinal or thoracic structures were injured.
- Guidance: 38 percent (19) had neither fluoroscopic nor ultrasound guidance. Where guidance was used it was one modality alone — a combined approach was not used in any case — and some malpositions occurred despite fluoroscopy confirming the guidewire position.
- What followed: surgical treatment 54 percent (27), trans-arterial embolisation 18 percent (9), and 26 percent (13) died before any treatment.
What this means at the bedside of a new patient. ECPR and Resuscitative ECMO places "extravascular cannula" first in its list of causes for persistently low flow despite rising pump speed at initiation. The data above explain why it is first: it is common among reported accidents, it is concentrated in the crash cannulations where imaging is least available, and a quarter of these patients did not live long enough to be treated. The book's reading: a femoral cannulation whose flow will not rise, with a very negative inlet pressure and a falling haemoglobin, should be treated as an extravascular or retroperitoneal cannula until imaging says otherwise.
Certainty: low for proportions (no denominator, voluntary reporting, a 12-year window in which equipment changed); moderate for the direction, which matches the vascular-injury literature the authors cite. Chapter 5, Chapter 12 and Chapter 21 are owed the guidance finding in full.
34.10 Controversy 1 · Sutures, or sutureless devices?
The clinical question. Almost every centre sutures its cannulas at the skin, and many also suture along the line. A minority use adhesive sutureless securement devices instead. Is one safer?
Position A: suture, because a suture holds. Suturing is what the Red Book, the Practical Guide and ECPR and Resuscitative ECMO all describe first. A suture through skin resists a pull that an adhesive may not, and many of the adhesive products in use were never designed for ECMO lines — Bull and colleagues found centres using products not designed or marketed for ECMO at all, and warned that some methods may be risky outside manufacturers' recommendations. The survey itself identified inadequate securement as the leading named cause of cannula events, which argues for more fixation, not less.
Position B: sutureless, because the data point the other way — and sutures have their own costs. ECPR and Resuscitative ECMO gives the theoretical advantage of adhesive devices: holding the cannula without further skin punctures, to reduce sources of bleeding and infection. And the only comparison in the literature runs against suturing: in Bull's survey, centres that "always" sutured at the insertion site reported a cannula adverse event in 35 percent (95 percent CI 30 to 41), against 0 percent (95 percent CI 0 to 28) in centres that "never" sutured (chi-square 4.40, p equal to 0.04).
Where the evidence actually sits.
- The comparison did not meet the survey's own prespecified significance level of p under 0.01, set by Bonferroni adjustment for multiple testing. The authors call it counterintuitive and of clinical significance, but not statistically significant.
- The "never suture" group is very small. A 95 percent upper bound of 28 percent on a zero proportion corresponds to roughly a dozen centres — this is the book's arithmetic from the published interval, not a figure the paper gives. Zero events in about a dozen centres is compatible with an event rate close to the suturing group's.
- The authors' own leading explanation is behavioural: centres that do not suture may be more vigilant because they believe their method is riskier. They also note reporting and recall bias.
- The outcome has no patient denominator. It is whether a centre recalls any event in five years. A high-volume centre is more likely to recall one simply because it has more patients — and the same survey found adult centres doing more than 30 cases a year reported events in 56 percent against 22 percent for centres doing 1 to 6 (p under 0.001). This is the book's reading of why the volume finding and the suturing finding should both be read as exposure effects before they are read as practice effects.
- No randomised or prospective comparison of securement methods was retrieved. ECPR and Resuscitative ECMO states it directly: no high-quality data exist suggesting one method is superior.
What this book concludes.
- Do not abandon suturing on the basis of this survey. The signal is non-significant by the authors' own threshold, rests on about a dozen centres, and has an unmeasured exposure problem.
- Do take from it the thing it measured well: securement fails in one centre in three over five years, and inadequate securement is the commonest named reason. Whichever method is used, the number of fixation points, the absence of bowing at the entry site, and the regular inspection and replacement of sutures or devices are what the evidence and the texts agree on.
- Standardise. Every source says so, and it is the only securement recommendation that is unanimous.
What would resolve it. A prospective multicentre registry of securement method with cannula-days as the denominator and dislodgement, migration, site bleeding and site infection as outcomes. The survey's respondents asked for a guideline; a guideline without that denominator would be consensus again.
Certainty: low in both directions. The comparison that exists is hypothesis-generating; the recommendation to standardise is expert consensus.
34.11 Controversy 2 · A malpositioned cannula that is working: move it or leave it?
The clinical question. Imaging finds a cannula in the wrong place — a tip in a hepatic vein, a port facing the wrong way — but the flow is adequate and the patient is stable. Does it have to be corrected?
Position A: correct it, and promptly. The Red Book's instruction is unconditional: malposition "should be corrected expeditiously using fluoroscopy, echocardiography, and/or guidewires," because migration can compromise flow or injure the right ventricle, right atrium or inferior vena cava. A cannula in a hepatic vein is a large-bore rigid tube in a thin-walled vein inside the liver; the next patient movement may turn a silent malposition into a laceration. And in Tanaka's account the repositioning from the right ventricle was done without stopping flow and without bleeding (§34.4).
Position B: leave it, when moving it is the greater hazard. Winiszewski and colleagues (J Extra Corpor Technol 2018, abstract read) found a femoral venous drainage cannula in an accessory hepatic vein on day 8 of venoarterial support, with no disturbance of venous return. "Moving or replacing the cannula was considered a high-risk maneuver potentially resulting in hepatic laceration with hemoperitoneum." Because drainage was adequate, repositioning was deferred until day 10, when the underlying defect was repaired and the patient was weaned from ECMO. The authors' general conclusion still urges awareness "to correct it and prevent insufficient venous drainage or traumatic complications" — they deferred, they did not dismiss. The record is indexed with the term "Fatal Outcome"; the abstract does not describe the death or relate it to the cannula, and the full text was not retrieved.
Where the evidence actually sits: case reports on both sides, and no comparison. No study compares correction against observation for an asymptomatic malposition. The risks being traded — laceration or embolism during manipulation, against injury or flow loss if left — have never been quantified. [VERIFICATION REQUIRED.]
What this book concludes, as a framework rather than a rule.
- A malposition that is producing a problem — low flow, chatter, recirculation, hypoxaemia, rising liver enzymes — is corrected. This is the Red Book's instruction and nobody disputes it.
- A malposition that is producing no problem is a risk-balance decision, and three questions settle most of them. How long is support expected to last? A cannula due out in two days is a different proposition from one needed for three weeks. What structure is the tip in, and what does injury to it cost? A tip against the free wall of the right ventricle is not the same as a tip in a hepatic vein. What will correction require here — withdrawal under echocardiography at the bedside, or fluoroscopy, a second venous access and an interventional operator?
- If it is left, it is a monitored decision, not an omission: documented, re-imaged at a stated interval, and re-imaged at once if flow or drainage pressure changes.
Certainty: very low. This is a reasoned framework from two opposing case reports and one textbook instruction; it is not a finding.
34.12 A worked case: bubbles after a turn
Worked example · Day 5, femoro-jugular venovenous ECMO, and a pressure-area turn
The patient. Day 5 of venovenous ECMO for severe ARDS. Femoral multistage drainage cannula, right internal jugular return. Speed 3,300 revolutions per minute, flow 4.6 L/min, inlet pressure minus 55 mmHg, sweep 5 L/min. Lung-rest ventilation. Depth mark at the groin recorded at 46 cm at the 06:00 check.
The event. At 07:20 a turn for pressure care is performed by two nurses; the femoral cannula is under a sheet. Thirty seconds after the turn, the ECMO specialist sees a stream of small bubbles in the drainage line and the flow falls to 3.0 L/min. The inlet pressure reads minus 20 mmHg. There is fresh blood on the dressing.
The readings, and what each excludes.
Reading | Value | What it means |
Depth mark | 40 cm — 6 cm out since 06:00 | The cannula has moved at the skin. This is decannulation in progress, not drift |
Air in the drainage line | Continuous small bubbles | Side holes are outside the vessel and entraining air — the Red Book's drainage-side failure, partial form |
Inlet pressure | Less negative (minus 55 to minus 20) with a falling flow | Not a hypovolaemia or wall-suction pattern. The pump is drawing on a leak, not a collapsed vein |
Impeller and sound | Turning; note changing | Air is reaching the pump head (Chapter 33 §33.5, Chapter 35) |
The reading. A multistage drainage cannula withdrawn 6 cm has exposed its most proximal side holes outside the vein while the tip remains inside. The circuit still carries some flow; it is also drawing air. This is §34.5's partial decannulation — the form that deceives because there is still flow.
What should happen, in order.
- Call for help. Clamp near the patient, drainage and return, and stop the pump. Air must not reach the membrane or the return cannula.
- A separate team takes the patient: ventilator to full support — FiO₂ 1.0 and the settings the team would use without a circuit — per the Red Book's separation-of-roles rule.
- Pressure over the insertion site. Do not push the cannula back in — the book's reasoning, since no source addresses it directly: its external segment has been outside the sterile field, its side holes are in the tissue tract, and none of the reported repositioning techniques was done blind (§34.4, §34.6).
- Chapter 35 owns the circuit: de-air, or exchange.
- Then the Red Book's decision (§34.6): is the patient adequately supported on the ventilator? On day 5 of severe ARDS, probably not — reinsertion under imaging, over a wire, with the circuit kept patent in the meantime.
- Write down the time, the depth before and after, and what the patient was doing. This is precisely the event the Japanese and survey data record as turning or bathing, and a unit learns nothing from an event it does not describe.
The counterfactual that makes the case worth reading. Change one detail — the depth mark is unchanged at 46 cm, there is no air, and the inlet pressure has fallen to minus 110 mmHg with chatter. That is not decannulation. It is a drainage problem at an unchanged cannula — hypovolaemia, a kink from the turn, or drift of the tip relative to a moved patient — and it belongs to §34.2 and Chapter 26: reduce the speed, look along the line, give volume if the patient is dry, and image. The depth mark, read before anything is touched, is what separates the two.
The teaching point. The one observation that would have caught this during the turn, rather than after it, was the depth mark — and it was under a sheet. ECPR and Resuscitative ECMO's instruction exists for exactly this: when repositioning patients, cannulas should be well visualised and not covered by linens.
This case is constructed to illustrate the mechanisms described in this chapter. The values are illustrative and are not drawn from a published report.
34.13 What this chapter could not close
Open questions, recorded rather than concealed
- No study has a cannula-days denominator for dislodgement or migration. The Japanese analysis states an incidence cannot be calculated; the survey counts centres recalling any event. Every rate in this chapter is a proportion of reports, not of patients. [VERIFICATION REQUIRED.]
- The 1.3 percent accidental decannulation figure traces to one study (Kim and colleagues) that was not retrieved, is quoted without definition or denominator by the texts that use it, and — on the Japanese authors' account — did not determine causes.
- No comparative evidence on securement method exists. The only comparison is a non-significant survey association resting on about a dozen centres (§34.10).
- No study compares correcting against observing an asymptomatic malposition (§34.11), and the risks traded — laceration or embolism during manipulation against injury or flow loss if left — have never been quantified.
- The tempo of either decannulation is unmeasured. The blood loss after a return-cannula dislodgement and the air volume after a drainage-cannula dislodgement are given here as arithmetic and mechanism; no source reports them from real events.
- Nobody has reported how often a depth-mark change precedes a clinical event, which is the number that would say whether hourly depth checks prevent anything or merely document it.
- Whether re-advancing a partly displaced cannula is ever acceptable is unaddressed in every source retrieved. The book's advice against it is reasoning, not evidence.
- The Japanese finding that 41 percent of decannulations occurred during another procedure has no counterpart elsewhere. If it replicates, it is the single most modifiable risk in this chapter.
Corrections and additions owed to other chapters
- Chapter 5, Chapter 12 and Chapter 21 (Cannulation) are owed §34.9 in full: 38 percent of reported malpositions had no imaging guidance, combined ultrasound and fluoroscopy was used in none, femoral insertion accounted for all retroperitoneal and intraperitoneal injuries, 26 percent died before treatment — and some malpositions occurred despite fluoroscopic confirmation of the wire. Also ECPR and Resuscitative ECMO's warning on reversed limb connections, and the Red Book's instruction that securing sutures are placed at the time of cannulation and positioned to prevent bowing.
- Chapter 9 (Persistent Hypoxaemia) is owed the observation that each cannula row in the Red Book's pressure table pairs a patient variable with a cannula variable, so the pressure pattern localises the side but never separates patient from plastic.
- Chapter 10 and Chapter 18 (Decannulation) are owed the Red Book's two-scenario decision after unplanned venous decannulation — an accidental decannulation is an unplanned trial off.
- Chapter 26 (Echocardiography) is owed the accessory hepatic vein case in which a malposition produced no flow change at all and was found only because someone imaged — the strongest argument yet for its re-imaging rule.
- Chapter 31 (Circuit Thrombosis) is owed ECPR and Resuscitative ECMO's flashlight inspection of the drainage cannula for fibrin strands and clot at every shift check.
- Chapter 35 (Air and Circuit Emergencies) is owed the partial drainage-cannula decannulation as a source of air with flow still present, the open arterial-cannula side port, and circuit damage caused by the act of securing.
- Chapter 36 (Bleeding) is owed the cracked cannula presenting as 13 days of site oozing, and the Japanese finding that 9 of 26 bleeding events were circuit damage rather than coagulopathy.
- Chapter 57 (Sedation) and Chapter 59 (Mobilisation) are owed the patient-removed cannula (13 percent of the survey's events) and turning or bathing (11 percent), and the Practical Guide's knee immobiliser and restraint guidance.
- Chapter 58 (Infection) is owed the survey's dressing, antisepsis and circuit-access data, which this chapter did not use.
- Chapter 60 (Transport) is owed the transport share of events and the rule that cannulas are checked first on arrival.
- Chapter 87, Chapter 88 and Chapter 92 are owed the depth mark as a charted value with a time, and any change as an event with a reason.
- Chapter 89 and Chapter 90 are owed §34.6's sequence and §34.5's four high-risk moments.
- Chapters 88 to 90 are owed the central-line rule: two people, one at the console with a hand on the speed.
Carried forward and still open: Chapter 33's unmeasured retrograde-flow interval on venoarterial support; Chapter 32's unretrieved primary source for the 150 mL/min oxygen transfer threshold; and the absence of any venoarterial-specific circuit-complication dataset — now partly relieved, since the Japanese malposition and decannulation data are predominantly venoarterial and ECPR.
34.14 Evidence and certainty summary
Claim | Best evidence | Certainty |
Cannula malposition (28 percent) and accidental decannulation (19 percent) were the two commonest ECMO accidents in a national reporting system — more than bleeding, air and clot combined | Hadano and colleagues, Japan Council for Quality Health Care database, 2010–2021, 178 events. Full text read | Low as frequencies — no denominator, reporting-dependent; moderate that cannula events dominate reported serious harm |
44 percent of reported malposition events resulted in death | Same | Low — reporter-classified harm |
The pressure pattern localises a falling flow to the drainage side, the membrane, or the return side, but each cannula row pairs a patient variable with a cannula variable | ELSO Red Book pump-response table; ECPR and Resuscitative ECMO's two lists | High for the localisation; the pairing observation is this book's reading |
Chatter is sustained intimal trauma at the drainage cannula tip, with negative pressure pulses to minus 600 to minus 650 mmHg | ELSO Red Book, circuit chapter | Moderate — mechanistic description, not measured outcome |
Migration can injure the right ventricle, right atrium or inferior vena cava, and should be corrected expeditiously under imaging and/or with guidewires | ELSO Red Book, diagnostic procedures chapter | Moderate as expert guidance |
Migrated cannulas can be repositioned without stopping flow, under echocardiography or with a femoral wire under fluoroscopy | Tanaka 2015, two cases; Muhammad 2023, one case; Red Book TEE case. Abstracts read | Very low — case reports |
A malposition can produce no flow disturbance at all | Winiszewski 2018, accessory hepatic vein on venoarterial support. Abstract read | Very low — single case, but decisive as an existence proof |
Return-cannula dislodgement exsanguinates; drainage-cannula dislodgement entrains air or deprimes the pump; either can be partial | ELSO Red Book; ECPR and Resuscitative ECMO; Hadano | High for the mechanism; no source quantifies the tempo |
Accidental decannulation occurs in about 1.3 percent of patients | ECMO: A Practical Guide to Management, citing a reference attributed to Kim and colleagues | Low — primary report not retrieved, no definition given |
In reported decannulations, the arterial cannula came out in 85 percent, and 41 percent occurred during another procedure | Hadano, 34 events. Full text read | Low — small numbers, voluntary reporting; the during-another-procedure finding is unreplicated |
Inadequate securement is the commonest named cause of cannula events, followed by manipulation, patient removal and turning or bathing; material failure is under 3 percent | Bull and colleagues, 71 events from 61 centres; concordant with Hadano and the Practical Guide. Full text read | Moderate that the causes are process failures — two independent concordant sources; low for each proportion |
After decannulation: call for help, clamp near the patient, stop flow, press on the site, hold anticoagulation, support the patient with a separate team, keep the circuit patent, prepare to reinsert | Red Book; ECPR and Resuscitative ECMO; Practical Guide — concordant | High as expert consensus; untested |
After unplanned venous decannulation, the decision is whether the ventilator can support the patient; if not, re-establish ECLS expeditiously | ELSO Red Book, paediatric weaning chapter | Moderate — expert guidance; the configuration contrast is this book's reasoning |
Secure at cannulation, at multiple points, without bowing at the entry site; inspect and replace; record depth regularly; standardise | Red Book; ECPR and Resuscitative ECMO; Practical Guide | Moderate as consensus; no outcome data |
One centre in five uses a single fixation point on a femoral line; one in three reports a cannula adverse event in five years | Bull, 391 responses, 45 countries | Low to moderate — self-report |
Centres that always suture reported more cannula events than centres that never suture | Bull: 35 percent against 0 percent, p equal to 0.04, not significant at the prespecified 0.01 | Very low — about a dozen never-suture centres, no denominator, exposure confounding |
Guidewire aspiration into the circuit during central line insertion is associated with a high drainage cannula tip and high pump speed | ECPR and Resuscitative ECMO | Low — rare event, expert description |
A cracked cannula can present as persistent site oozing | Dugal 2025, one case. Abstract read; Hadano's circuit-damage bleeding events | Very low as a frequency; moderate as a diagnostic prompt |
Reported malpositions at insertion lacked imaging in 38 percent and combined imaging in all; femoral insertion accounted for every retroperitoneal injury | Hadano, 50 events. Full text read | Low for proportions; moderate for direction |
34.15 Key points
- Registries remember bleeding; incident reports remember cannulas. In a national accident system, malposition and decannulation were nearly half of all reported ECMO accidents, and malposition was fatal in 44 percent.
- The cannula is the only component that cannot be exchanged in a minute, and the only one attached to something that moves. Prevention outranks treatment here more than anywhere else in Part VI.
- A cannula position is a measurement with a time stamp. The confirmation at insertion is a baseline.
- The two ends fail differently. The tip drifts — slowly, silently, announced as a flow or oxygenation problem. The skin end lets go — suddenly, announced by blood or air.
- A more negative inlet pressure with a falling flow is upstream; a rising post-lung pressure with a flat pressure drop is downstream. The pattern localises the side — it never separates the patient from the cannula.
- Chatter is not only a symptom. Every minute of it is the vein being sucked onto the side holes. Never raise the speed to silence it.
- The depth mark catches movement at the skin and misses drift of the patient around a still cannula. An unchanged mark does not clear a cannula whose flow has changed.
- Record the depth with a time; a change is an event with a reason. A change with no reason is a securement failure in progress.
- A malposition can produce no flow change at all. Only re-imaging finds it.
- Correct a migrated cannula under real-time imaging, over a wire whenever the tip must pass a branch point, and usually without stopping flow. Then re-secure it and make a new depth mark.
- Return-cannula dislodgement exsanguinates at pump flow; drainage-cannula dislodgement entrains air or deprimes the pump. A partly withdrawn multistage drainage cannula can do it while still carrying flow.
- Accidental decannulation is a process failure far more than a device failure: securement that did not hold, a cannula nobody was watching, a moment when the patient was being moved or worked on.
- Four moments recur across the data and the texts: another procedure, turning or washing, transport, and an agitated patient. One named person holds the cannulas at each.
- The first minute: call, clamp near the patient, stop, press, hold anticoagulation, support the patient with a separate team, keep the circuit patent, prepare to reinsert. Do not push it back in.
- Then one question: can the ventilator carry this patient? On venovenous support it often buys time; on venoarterial support with no native output, decannulation is a cardiac arrest.
- Secure at cannulation, at several points, with no bow between skin and anchor; inspect and replace; standardise. One centre in five uses a single fixation point on a femoral line.
- The only comparison of sutures against sutureless devices is a non-significant survey association resting on about a dozen centres. It does not justify changing method; it does justify taking securement seriously.
- Persistent oozing that does not respond to local measures should prompt inspection of the cannula itself. It may be cracked.
- A central line on ECMO is a two-person procedure, one of them at the console with a hand on the speed, because the drainage cannula can aspirate the wire and entrain air.
- A femoral cannulation whose flow will not rise, with a very negative inlet pressure and a falling haemoglobin, should be treated as an extravascular cannula until imaging says otherwise.
Chapter answer: The cannula is the one piece of the circuit that nobody can replace in a minute and the one piece attached to a body that keeps changing shape — which is why, when a national safety system collected every ECMO accident for twelve years, the commonest entries were a cannula in the wrong place and a cannula that came out. The tip fails slowly: the patient moves around it, a port drifts to face the liver, drainage falls, recirculation rises, and the circuit reports a problem somewhere else; the pressure pattern tells you which side of the pump to look on, and only the bedside — the depth mark, the line itself, and imaging — tells you whether the fault is the patient or the plastic. The skin end fails suddenly: a return cannula that comes out empties the patient at the pump's flow rate, and a drainage cannula that comes out draws air — sometimes while still carrying flow. The first minute is the same every time — call, clamp near the patient, stop, press, support with a second team, and do not push it back in — and the next decision is whether the ventilator can hold this patient while a new cannula goes in. Almost none of it is device failure. It is securement that did not hold, a cannula under a sheet, and a moment when everyone was looking at something else; and the cheapest monitor against all of it is a depth mark that somebody reads, writes down, and treats as an event when it changes.