🩸

Chapter 31 — Circuit Thrombosis

❓

The question this chapter answers

Every extracorporeal circuit begins clotting the moment blood enters it, and every circuit that runs long enough contains clot. So the bedside question is never is there clot in this circuit? — the answer is yes. The question is which clot matters, what is it doing to this patient, and what should be done about it.

This chapter is about thrombus as an object inside the circuit: where it forms, why it forms there, how it announces itself, what it does when it is not found, and when the circuit has to be replaced.

🗓️

Evidence search date: 23 September 2026. The anchor source is read as complete full text and is the only Verified source in this chapter. Primary sources are the ELSO Red Book 6th edition (the VV complications chapter and the circuit malfunction and crisis management chapter), ECMO in the Adult Patient (circuit monitoring and pressure measurement), ECPR and Resuscitative ECMO (the circuit check and the five circuit catastrophes), the ISCCM manual (technical complications, acute versus planned exchange), and ECMO: A Practical Guide to Management.

What this chapter covers — and what it does not

This chapter owns	This chapter does not own
Thrombus wherever it forms in the circuit — the sites, the mechanism, the natural history, and the surveillance system that finds it	Why blood clots on a foreign surface at all. Chapter 39 owns ECMO coagulation physiology and this chapter uses it without re-deriving it
Transmembrane pressure as a thrombus signal, and its normalisation to blood flow	The oxygenator as a gas exchanger. Chapter 32 owns membrane lung dysfunction — post-membrane gas interpretation, transfer calculations, plasma leak and the non-thrombotic causes
The system exchange decision and the exchange itself — indications, timing, the procedure, and whether a planned exchange is safe	The specific triggers that belong to other findings: Chapter 32 owns the gas-exchange trigger and Chapter 37 owns the haemolysis trigger. Each chapter owns its own trigger; this chapter owns how the decision is made and how the exchange is done
Device-related coagulopathy — the circuit as a cause of bleeding, and the finding that it reverses on exchange	Chapter 36 owns bleeding as a clinical problem and its management; Chapter 46 owns transfusion
Pump head thrombosis as a thrombotic event	Chapter 33 owns pump failure — mechanical, electrical and drive failure, and the hand crank
Clot at and around the cannula as a circuit finding	Chapter 34 owns cannula problems — position, malfunction, vascular injury; Chapter 17 owns limb ischaemia
Nothing about anticoagulation targets	Chapter 42 owns anticoagulation monitoring and Appendix 6 owns the algorithms. This chapter states what clot does, not what dose to give

31.1 The circuit is clotting from the moment it is primed

Blood meeting a non-endothelial surface activates the contact system, adsorbs plasma proteins within seconds, and begins depositing fibrin. Chapter 39 owns that physiology. What belongs here is its consequence, which is simpler and more useful: thrombus in an ECMO circuit is not an adverse event, it is the expected state. The clinically meaningful distinction is between clot that is being tolerated and clot that is now doing something.

🧠

The structural claim of this chapter · A clot that matters declares itself, and it does so in one of five languages

A circuit cannot be opened and examined. Everything known about the clot inside it is inferred from five independent channels, and each one detects a different kind of clot in a different place.

The eye — visible fibrin and clot on torch inspection, and the sound the pump makes.
The pressure — the pressure drop across the membrane lung, read as a ratio to blood flow.
The gas — what the blood leaving the membrane lung looks like.
The haemolysis panel — plasma free haemoglobin and lactate dehydrogenase.
The coagulation panel — D-dimer, fibrinogen and the platelet count.

The five do not substitute for one another. In the anchor cohort, pump head thrombosis was found by free haemoglobin and was invisible to the pressure channel; acute oxygenator thrombosis was found by the pressure channel and was missed by the coagulation channel in a third of cases; device-induced hyperfibrinolysis was found by the coagulation channel in circuits whose gas exchange was only minimally affected. A unit that watches one channel will be surprised by the other four.

And the honest limit, which §31.6 develops: watching all five converts about half of the clot emergencies into planned procedures. It does not abolish them.

🔬

Evidence · The anchor cohort, and what 83 system exchanges were actually for

Lubnow M, Philipp A, Foltan M, Bull Enger T, Lunz D, Bein T, Haneya A, Schmid C, Riegger G, Müller T, Lehle K. "Technical Complications during Veno-Venous Extracorporeal Membrane Oxygenation and Their Relevance Predicting a System-Exchange — Retrospective Analysis of 265 Cases." PLoS One 2014;9(12):e112316. DOI 10.1371/journal.pone.0112316, PMC4251903. Read as complete full text. Retrospective analysis of prospectively collected data, single centre (Regensburg), January 2009 to December 2013.

The cohort. 265 consecutive adults on venovenous ECMO. Age 50 (36–60); SOFA 11.0 (8.0–14.3); Murray lung injury score 3.33 (3.33–3.67). Indication was primary lung failure in 68 percent. Cumulative support 3,411 days; median 9 (6–15) days per patient. Before cannulation, 46 patients had a platelet count under 80/nL and 67 already showed signs of disseminated intravascular coagulation.

Eighty-three patients (31 percent) required at least one system exchange. Thirty-seven were acute and 46 were elective. The reasons, which sum exactly to 83:

Reason	n	Acute or elective	Which channel found it
Mechanical failure (pump 5, membrane lung 2, cannula 1)	8	Acute	The eye, and the ear
Pump head thrombosis	13	Acute	Haemolysis panel (and pump sound)
Acute oxygenator thrombosis	16	Acute	Pressure (dpMO with falling flow)
Worsening gas transfer	10	Elective	Gas
Device-related coagulation disorder (local fibrinolysis 24, device-induced hyperfibrinolysis 8)	32	Elective	Coagulation panel
Suspected circuit infection	4	Elective	Clinical, plus inflammatory markers

Two numbers in that table deserve to be read twice. Mechanical failure was 8 of 83 — under 10 percent. Nearly everything else was clot. And the single largest category, 32 of 83, was not a circuit that stopped working but a circuit that had started to injure the patient's coagulation system (§31.5).

Certainty: moderate to high for the categories and their relative frequency — prospectively collected data, a single high-volume centre, six different ECMO systems, and definitions stated in advance. Low for the absolute exchange rate of 31 percent, which depends on that centre's thresholds, its anticoagulation protocol and its case mix.

💡

Clinical pearl · The first channel is free, takes thirty seconds, and is the one that gets skipped

Every primary source in this book independently describes the same physical examination of the circuit, and ECPR and Resuscitative ECMO gives it the name it deserves: the circuit check is a standardised physical examination, and it should happen at initiation, at the start of every shift, and during any circuit emergency.

What to look at, in order.

Colour. Dark in the drainage limb, bright in the return limb. Both dark means the sweep gas or the membrane lung, not the patient. Both bright means the bridge is open, both cannulae are on the same side of the circulation, or oxygen extraction has failed.
A torch, along the whole circuit. ECMO in the Adult Patient is explicit that thrombus is usually noticed first by a nurse doing hourly checks of the tubing and oxygenator with a pen torch. ECPR and Resuscitative ECMO says where to point it: white fibrin strands and darker clots along the drainage cannula, and inside the membrane lung, where they commonly form in the corners and on the inflow side.
The sound. A change in the pump's note, or new vibration, is a pump head finding and not a nuisance. The ISCCM manual adds a modern and genuinely useful trick: a smartphone slow-motion recording can be used to visualise the thrombus in the pump head.
Every connection, stopcock and pigtail. Hand-tightened, never clamp-tightened; stopcocks off. A bridge, if present, should be free of blood and filled with crystalloid, because a bridge with blood standing in it is a clot waiting to be released.

The circuit check finds the two things no laboratory test can find in time: the clot you can see, and the mechanical failure that is about to happen.

31.2 Where clot forms, and why there

🧠

Physiology · Three conditions, and the places in a circuit where all three are met

Thrombus needs a surface, a stimulus and time. In an extracorporeal circuit the surface is everywhere and the stimulus is continuous, so what varies from one part of the circuit to another is residence time and flow character. Clot forms where blood slows, where it recirculates, and where shear is high enough to activate platelets but the geometry then lets them settle.

The sites, and the reason each one is a site.

The membrane lung. Tens of thousands of hollow fibres present an enormous surface at low local velocity, and the flow path has corners. This is why torch inspection is directed at the corners and the inflow side — those are where velocity is lowest relative to the bulk flow.
The pump head. The opposite problem: the highest shear in the circuit, and a rotating geometry with a zone behind the impeller where flow is slow. Shear activates the platelets; the slow zone lets them deposit. A clot here does not obstruct first — it grinds first, which is why it is found by haemolysis rather than by pressure.
Connectors, stopcocks, pigtails and sampling ports. Every junction is a step change in diameter and a small stagnant volume.
A bridge with blood in it. A bridge is a dead end by design.
The cannula tips and the length of cannula inside the vessel, where fibrin strands are visible on the drainage cannula and where surrounding stasis is greatest.
Anywhere the flow has been allowed to fall. ECMO in the Adult Patient states the rule directly: flow through any tubing should not fall below 1.5 L/min, that centre's own practice being at least 2 and usually not below 2.5 L/min, because low blood flow leads to blood stagnation and thrombus development.

The bedside consequence is a rule that recurs through Part VI: anything that lowers flow raises clot risk, and that includes the things done for good reasons — weaning trials, a trial off, a clamped bridge, a reduction made to treat a number. Chapter 18 and Chapter 26 own the weaning decision; this chapter owns the warning that the weaning trial is the highest-risk period for circuit thrombus in a whole run.

⚠️

Danger · Four situations in which a circuit clots much faster than it should

Inadequate anticoagulation at initiation. ECPR and Resuscitative ECMO names this first: a catastrophic circuit clot can occur during ECMO initiation if the patient is not given adequate anticoagulation.
Protamine given while a heparin-coated circuit is running. The same source names it explicitly as a cause of catastrophic circuit clot. This is a drug-interaction hazard with a device, and it is not on most ward drug charts. If protamine is being considered in a patient on ECMO — after cardiac surgery, or to reverse heparin for a bleed — the circuit is part of the decision. Chapter 42 owns the reversal decision; what belongs here is that the circuit must be named in it.
A trial off, or any period of reduced flow. The Red Book states it for weaning: "The main risk during a trial off ECLS is circuit thrombus formation and thromboembolism to the patient," particularly with central cannulation, and instructs that flow should not be reduced below the lowest flow rate established for the circuit's oxygenator and that systemic anticoagulation should be adjusted pre-emptively. It notes that an open bridge regulated by an adjustable thumb clamp, or a pump-controlled retrograde trial-off, both avoid the intermittent clamping that raises thrombus risk.
Activated factor VII. In the anchor cohort, 2 of the 4 patients who received activated coagulation factor seven while on ECMO developed an acute oxygenator thrombosis requiring a system exchange. Four patients is a very small number and the observation is uncontrolled. It is recorded here because the direction is mechanistically obvious, the consequence is a circuit emergency, and a 50 percent event rate in even four patients should change the conversation before the drug is given. [Certainty: very low as a rate; the association is hypothesis-generating.]

31.3 Pump head thrombosis

🔬

Evidence · What pump head thrombosis actually looked like, in 13 patients

From the anchor cohort. Clots were documented within every single pump head after removal — the diagnosis was confirmed in all 13.

Plasma free haemoglobin rose approximately 10-fold before the exchange (p under 0.001) and fell approximately 6-fold after it (p under 0.001). Patients without a system exchange, at comparable support times, had significantly less free haemoglobin.
In 6 of 13, the rise was already visible the day before: free haemoglobin 55 (49–60) mg/dL two days before, rising to 143 (117–266) mg/dL one day before (p equal to 0.031).
Which means that in 7 of 13, it was not. The warning existed in fewer than half.

The clinical signature the sources agree on, from the ISCCM manual and the anchor paper together: a sudden change in the sound of the pump head, with vibration in the circuit, and progressive pump-induced haemolysis — rising free haemoglobin and lactate dehydrogenase — and in some, a falling platelet count.

Certainty: high that pump head thrombosis presents as haemolysis rather than as obstruction; moderate for the magnitude of the free haemoglobin rise; moderate that roughly half give a day of warning.

⚠️

Danger · A rising free haemoglobin has five other explanations that must be excluded first, and the anchor paper lists them

This is the most useful single list in the source, and it is the bridge between this chapter and Chapter 29. Before a rising plasma free haemoglobin is called pump head thrombosis, exclude:

Powerful aspiration of the blood sample — the sampling artefact Chapter 29 identified as commoner than the real thing.
Excessive suction from cannula malposition, usually with high negative inflow pressures and an inadequately high pump speed relative to the blood flow achieved.
Continuous renal replacement therapy — which Chapter 29 recorded as independently associated with haemolysis when connected to the ECMO circuit.
Thrombosis elsewhere.
Larger haematomas.

Two of these five are the same finding read two ways, and that matters. A high negative inlet pressure with a pump speed disproportionate to the flow achieved is both a cause of haemolysis and a sign of a circuit or drainage problem. ECMO in the Adult Patient puts a number on it: pre-pump pressure ideally around minus 60 mmHg, and a value more negative than minus 100 mmHg will increase blood trauma and free haemoglobin release.

So the order of the bedside enquiry is: repeat the sample; look at the inlet pressure and the speed-to-flow relationship; look at the cannula; look at the renal replacement connection; and only then call it the pump. Chapter 29 owns the assay. Chapter 34 owns the cannula. Chapter 37 owns haemolysis as a clinical problem.

31.4 Acute oxygenator thrombosis

🧮

The measurement that identifies it, and why the raw number will not do

The pressure drop across the membrane lung is dpMO = pressure at the membrane lung inlet minus pressure at its outlet. It rises when the membrane lung's resistance rises, which is what a clot does.

But resistance is not the only thing that determines the pressure drop — flow does too. A circuit running at 5 L/min through a clean membrane will show a higher dpMO than the same membrane at 3 L/min. So a raw dpMO is uninterpretable without the flow that produced it, and a dpMO that falls while the flow is falling can conceal a rising resistance completely.

The anchor study's solution is the one to adopt: normalise the pressure drop to blood flow and follow the ratio. In its 16 patients with acute oxygenator thrombosis:

dpMO divided by blood flow rose 2.6-fold (2.1 to 4.0) before the exchange, p under 0.001.
It fell 3.2-fold (2.0 to 5.2) after the exchange, p under 0.001.
In every other acutely exchanged patient, the ratio did not rise. Only one electively exchanged patient showed a rise (3.1-fold, exchanged for suspected infection), which the authors read as an exchange that pre-empted a thrombosis that was already forming.

That is an unusually specific signal, and it is the reason this chapter asks for the ratio rather than the number. It is also exactly the general rule Chapter 30 §30.2 arrived at from the other direction: a set value and its measured partner belong on the same line, and the informative quantity is the relationship between them.

The clinical picture the ratio accompanies: an acute rise in dpMO followed by a fall in blood flow at unchanged pump speed. That second half is the part that gets misread, because a falling flow invites the reflex of turning the speed up.

🔬

Evidence · What else was happening around those 16 thromboses, and what was not

D-dimer was elevated before the event in 11 of 16, at 31 (23–35) mg/dL by the local assay, declining to 13 (6–31) within two days of exchange (p equal to 0.016). The other 5 of 16 had a sudden acute oxygenator thrombosis that routine D-dimer testing missed entirely.
The platelet count showed only a trend — 161 (76–300)/nL two days before, 145 (74–286)/nL on the day, not significant — improving after exchange in 6 of 11.
In 8 of 11, gas transfer had already deteriorated before the thrombosis: a 53 percent (45–59) fall in the oxygen tension at the membrane lung outlet, with carbon dioxide at the outlet of 40 (32–48) mmHg despite gas flow rates of 10 L/min or more.

The reading that matters. Three channels were pointing at this membrane lung before it occluded — pressure, coagulation and gas — and in about a third of cases none of them gave notice. That is the quantitative basis for §31.6.

A units caution, exactly of the kind Chapter 30 §30.9 asks for. D-dimer here is reported in mg/dL, which is not the unit most laboratories use; D-dimer is commonly reported in mg/L or µg/mL, and in fibrinogen-equivalent units that differ again. The numbers 31 and 13 are therefore not transferable to another laboratory as numbers. What transfers is the pattern — a several-fold rise before, and a fall after exchange. Reproduced with the unit as printed and not converted.

31.5 When the clot in the circuit is why the patient is bleeding

This is the most important inversion in the chapter, and the least expected. The largest single category of system exchange in the anchor cohort — 32 of 83, larger than pump head thrombosis and acute oxygenator thrombosis combined — was not a circuit that had stopped working. It was a circuit that had started to consume the patient's coagulation system.

🔬

Evidence · Two device-related coagulation disorders, and one of them reverses when the circuit is changed

The anchor cohort separated 32 elective exchanges into two patterns with different laboratory signatures.

Pattern 1 — device-induced hyperfibrinolysis (n equal to 8).

Fibrinogen fell significantly below 200 mg/dL, with D-dimer rising, over the four days before the exchange, against patients without exchange at comparable support times. The platelet count fell without reaching significance.
These patients had a diffuse, otherwise unaccountable bleeding tendency.
After the exchange, coagulation normalised and the bleeding diathesis was reduced. The authors' own conclusion: "a device-related hyperfibrinolysis was reversible by a system-exchange."
And the finding that makes this dangerous to miss: the gas exchange of these membrane lungs was only minimally affected. The circuit was working. It was also bleeding the patient.

Pattern 2 — local fibrinolysis within the membrane lung (n equal to 24).

A significant rise in D-dimer and a fall in platelet count over the three to four days before exchange — but fibrinogen remained unchanged.
Forty percent of these patients nevertheless developed bleeding complications.
In 7, gas transfer had worsened (a 40 percent fall in outlet oxygen tension in 3; outlet carbon dioxide 38 to 43 mmHg despite gas flows of 6 to 10 L/min). In 5, the pressure drop rose moderately, 1.8-fold (1.4 to 2.2), p equal to 0.001. In 1, free haemoglobin was slightly raised.
Exchange reversed the D-dimer and platelet changes, improved gas transfer, reduced free haemoglobin and normalised the pressure drop (a 2.0-fold [1.6–2.4] fall, p equal to 0.018).

Certainty: moderate that a device-related coagulopathy exists and improves after exchange — the before-and-after design is uncontrolled but the reversal is consistent, the comparison group is stated, and the mechanism is plausible. Low for any threshold, because the study offers none and the D-dimer units are local.

⚠️

Danger · The bleeding ECMO patient whose problem is the circuit

The reflex when an ECMO patient bleeds is to look at the heparin, then at the platelets, then at the products. This chapter adds a fourth item to that list and it is frequently the right one.

The pattern to recognise: a patient some days into a run, bleeding diffusely from everywhere and nowhere in particular, with a rising D-dimer, a falling platelet count, and — in the more severe form — a falling fibrinogen with no disseminated intravascular coagulation, no thrombosis, no pulmonary embolism, no trauma and no surgery to explain it.

The diagnosis is presumptive until the exchange is done. Both the anchor paper and the ISCCM manual say this explicitly: device-induced coagulopathy is a diagnosis of exclusion that is confirmed only by the markers improving after the system is exchanged. That is uncomfortable, and it is the honest position.

Three practical consequences.

Fibrinogen is the discriminating test between the two patterns, and a normal fibrinogen does not exclude the problem — 24 of the 32 had a normal fibrinogen and 40 percent of them bled.
Do not require the circuit to be failing before you suspect it. In the hyperfibrinolysis group the gas exchange was nearly normal.
A patient who keeps bleeding through escalating product support, with a rising D-dimer and no other cause, has a treatable cause that no amount of product will fix. Chapter 36 owns the bleeding decision and Chapter 46 owns the transfusion; what belongs here is the finding that the circuit itself belongs on that differential, and near the top of it.

🚩

Pitfall · Attributing a falling platelet count and a rising D-dimer to the illness

Every critically ill patient on ECMO has reasons for thrombocytopenia and a raised D-dimer, and Chapter 29 listed most of them. That is exactly what makes device-related coagulopathy easy to miss: the abnormality is unremarkable in isolation and only the trajectory identifies it.

What separates it from the noise, in the anchor data, is the time course: a change developing over three to four days, in a patient who is not otherwise deteriorating, with no new clinical event to account for it. A D-dimer that has been high since cannulation is the illness. A D-dimer that begins climbing on day 8 in a stable patient is the circuit until proved otherwise.

And there is a second, quieter reason to look: the circuit can be infected. Four patients (5 percent) were exchanged for suspected circuit infection. In one, mycobacteria were identified microscopically in the membrane lung after exchange; another was colonised with Enterococcus species, confirmed by long-term ex-vivo incubation. Two were exchanged on clinical grounds alone, both having developed a diffuse unaccountable bleeding diathesis after 15 to 18 days of support that ceased after the exchange. In one patient the noradrenaline requirement had risen five-fold before the exchange, to 0.35 µg/kg/min, and fell by a factor of 15 after the membrane lung was replaced.

Chapter 58 owns infection on ECMO. What belongs here is that an unexplained septic picture with no focus, and an unexplained bleeding diathesis, can both be the same object — and that object can be removed.

31.6 Controversy 1 · Does surveillance actually prevent anything?

⚖️

The clinical question. Daily D-dimer, fibrinogen, platelets, free haemoglobin, lactate dehydrogenase, pressure drop and post-membrane gases, on every ECMO patient, forever. Does this surveillance panel convert circuit emergencies into planned procedures, or does it simply generate abnormal numbers in patients who were going to be fine?

Position A: yes, and the anchor study was designed to show it. Its own conclusion is unambiguous: "The majority of ECMO system-exchanges could be predicted by regular inspection of the complete ECMO circuit, evaluation of gas exchange, pressure drop across the MO and laboratory parameters (DD, FG, platelets, LDH, frHb). These parameters should be monitored in the daily routine to reduce the risk of unexpected ECMO failure." The supporting arithmetic is real: progressive clot formation with worsening gas transfer was identified early in 42 patients (51 percent), which is what made those exchanges elective rather than emergent. And 46 of the 83 exchanges were elective — a majority.

Position B: read the acute events, and the panel's performance is about a coin toss. The same paper says, of the urgent exchanges for acute clot, that "nearly 50% could be foreseen" by the panel. Taking the two acute clot categories apart:

Pump head thrombosis: a warning rise in free haemoglobin on the preceding day in 6 of 13.
Acute oxygenator thrombosis: D-dimer elevated beforehand in 11 of 16, but 5 of 16 were "a sudden AOT missed by routine testing of DD."

So of 29 acute clot events, roughly half arrived with no notice from a panel that was being drawn daily. And the panel's abnormal values are non-specific: Chapter 29 established that a rising D-dimer, a falling platelet count and a rising free haemoglobin each have several ordinary explanations in this population.

Where the evidence actually sits. The two positions are answering different questions and both answers are correct.

For progressive, slow clot — which is most clot — surveillance works. Half the exchanges in this cohort were identified early enough to be planned, and a planned circuit change is a different procedure from an emergency one.
For acute clot — pump head thrombosis and acute oxygenator thrombosis — surveillance halves the surprise and no more. About half of acute circuit thrombosis is, on current monitoring, unpredictable.

What this book concludes. Run the panel, and run it for the trend rather than the value — this is Chapter 30's rule and it applies here exactly. But do not let a normal panel lower the index of suspicion, because in this cohort a normal panel the day before was compatible with a pump head full of clot the next morning. Chapter 28 reached the identical structure for the brain: these are rule-in tests, and none of them reassures.

What would resolve it. Prospective evaluation of a defined surveillance panel with prespecified thresholds, reporting sensitivity and specificity for acute circuit thrombosis with the decision window stated, against a comparator of clinical surveillance alone — and reporting the false-positive exchanges, which no study currently does.

Certainty: moderate that surveillance shifts exchanges from acute to elective; moderate that roughly half of acute clot events are not foreseen; and none at all for any specific threshold in any parameter.

31.7 Controversy 2 · When should a circuit be changed, and is changing it safe?

⚖️

The clinical question. There is visible clot in the membrane lung. The pressure drop has crept up. Gas exchange is adequate. The patient is stable. Should this circuit be changed now, or watched?

The starting position, which both sides accept: there is no rule. The ELSO Red Book states it plainly — "There are no standard and agreed indications to change components or a circuit at regular intervals during VV ECMO" — and lists what makes it unanswerable in general: the level of systemic anticoagulation, the duration of support, the configuration, the make of components, and the patient's pathophysiology.

Position A: change late. Exchange when the circuit is failing, not when it is imperfect. ECPR and Resuscitative ECMO gives the conservative rule in one sentence: "The ultimate decision to exchange a circuit component should be made when clot or fibrin results in compromised circuit flow." The arguments are good ones. Visible clot is near-universal in a circuit that has run for a week and most of it never does anything. A circuit exchange means a period off support, a fresh uncoated interface, new consumables and a procedure performed in a room rather than a theatre. And the Red Book is explicit that D-dimer surveillance offers no threshold: progressive clot formation "can be monitored by serial measurement of D-dimers in the plasma, although no cutoff value is indicative of a need to change the circuit."

Position B: change early, because the cohort that changed early was not harmed. The anchor study is the strongest available evidence here, and its outcome data are the part most often overlooked.

Patients requiring at least one exchange had longer support (20 [14–27] versus 8 [6–10] days), longer intubation (30 [20–43] versus 16 [11–26] days), longer intensive care stay (34 [24–46] versus 21 [14–34] days) and more renal failure requiring renal replacement during ECMO (28 versus 15 percent, p equal to 0.028) — all of which describe a sicker, longer-running group rather than harm from the exchange.
They received a median of 2 (2–3) membrane lungs, with a range extending to five.
And there was no difference in mortality. Successful weaning 69 versus 79 percent; died on ECMO 31 versus 21 percent; p equal to 0.113.
Comparing acute against elective exchanges within that group, blood product use and outcomes did not differ significantly, and mortality was not increased (died on ECMO 17 versus 29 percent, p equal to 0.320).

Where the evidence actually sits. The mortality comparison is not randomised and it is not powered, and a p value of 0.113 with a 10-point absolute difference is not evidence of equivalence. What it does support is the narrower and still useful claim that a programme which exchanged circuits in a third of its patients, sometimes up to five times, did not visibly pay for it in survival. That removes the strongest argument against a low threshold, which was the fear that the procedure itself is dangerous.

What this book concludes.

Exchange when the circuit is doing something: obstructing flow, failing to exchange gas, haemolysing, or consuming coagulation factors. Those four are the indications, and each is owned by the chapter that detects it.
Visible clot alone is not an indication. It is a reason to look harder at the other four channels and to raise the frequency of looking.
Prefer the planned exchange. The whole purpose of surveillance is to move the procedure from 03:00 to a weekday morning with the team assembled, and in this cohort the acute group had a shorter working life of the first membrane lung (7.0 [5.0–10.3] versus 9.5 [7.0–13.0] days, p equal to 0.023) — that is, the circuits that failed acutely failed early, and were not the ones being watched for weeks.
A circuit that is causing a coagulopathy should be changed even if it is exchanging gas perfectly (§31.5). This is the clearest departure from Position A and it rests on the strongest evidence in the chapter.

What would resolve it. A randomised comparison of a protocolised early-exchange strategy against symptom-driven exchange, with prespecified triggers, reporting circuit-related emergencies, transfusion, bleeding and survival. None exists.

Certainty: high that no threshold exists; moderate that early elective exchange is not harmful; low for any specific trigger; none for a routine time-based exchange interval, which no source recommends.

31.8 The system exchange itself

💡

Clinical pearl · What is actually exchanged, and how

What. In the anchor centre, "in case of technical complications the whole ECMO circuit (tubing, MO, blood pump), except for the cannulas, was exchanged." That is worth stating because the alternative — changing the membrane lung alone by cutting either side of it — is also common practice and is described by ECMO in the Adult Patient as "a relatively straightforward procedure involving cutting the connecting tubing," always performed under aseptic conditions by perfusionists or ECMO specialists. Both are legitimate. The choice turns on where the clot is and on whether the pump head has been implicated, and a pump head thrombosis is not treated by changing the oxygenator.

How, from the Red Book's crisis management chapter.

Lay out and prime the new circuit.
Pause and identify the drainage and return limbs of both the existing and the new circuit, and confirm the planned connections. This is the step that prevents the catastrophic error.
Clamp the existing circuit in two places per limb, about 15 cm apart, and cut between the clamps.
Connect the prepositioned ends of the new circuit and resume flow.
Perform a full circuit check and re-establish the previous settings on the new circuit.

The principle underneath the procedure, from ECPR and Resuscitative ECMO: a component can be changed with one or two clamps either side of it if the situation is temporisable — a controlled separation. If it is not temporisable, the emergent separation applies, and that is CLAMP, STOP, RESCUE. Chapter 35 and Chapter 90 own the emergency; this chapter owns the planned procedure.

Two things to have decided before the clamps go on. How the patient will be supported during the minute or two without ECMO — which for a venoarterial patient on full support is an entirely different problem from a venovenous patient with some native lung. And who is doing what: the exchange is a team procedure and the pause in step 2 needs somebody whose job is to call it.

🚩

Pitfall · Exchanging the circuit and not looking at what came out

The old circuit is the only direct evidence about the clot that will ever be available, and it is usually in a bin within ten minutes.

Three things are worth doing before it is discarded, and none takes long.

Open or transilluminate the membrane lung and record where the clot was — inflow side, corners, distributed, or a single large thrombus. Over a programme this builds the only local data that exist on the circuit's actual failure mode.
Photograph it. Chapter 30's argument applies exactly: the observation exists only if it is recorded, and a photograph in the notes is a record that survives.
If circuit infection was the suspicion, send it. In the anchor cohort, one membrane lung yielded mycobacteria microscopically and another grew Enterococcus on long-term ex-vivo incubation — findings that only exist because somebody cultured the device rather than only the patient.

And record the indication for the exchange in words, which is Chapter 30 §30.7's rule: confounding by indication is unfixable when the indication was never written down, and the circuit-exchange decision is one that a later reviewer will certainly want to understand.

31.9 A worked case: the circuit that was working perfectly

🧮

Worked example · Day 9 of venovenous ECMO, and the wrong three questions

The patient. Day 9 of venovenous ECMO for influenza pneumonitis. Blood flow 4.2 L/min at 3,300 revolutions per minute, unchanged for four days. Sweep 4 L/min. Arterial saturation 94 percent. Ventilator settings unchanged and lung-protective. The patient is not deteriorating.

The problem. For 48 hours there has been oozing from the cannulation sites, the oral cavity, the endotracheal tube and every puncture site. Two units of red cells have been given in 24 hours. The team has reduced the heparin twice, and the bleeding has not changed.

The panel, with the columns Chapter 30 asks for.

Value	Today (day 9)	Day 7	Day 5	Day 1
D-dimer (local units)	34	26	12	9
Fibrinogen (mg/dL)	118	171	245	410
Platelets (/nL)	71	96	130	188
Plasma free haemoglobin (mg/dL)	18	16	14	12
dpMO / blood flow (ratio, arbitrary units)	1.1	1.1	1.0	1.0
Post-membrane oxygen tension (mmHg)	385	390	402	410
Post-membrane carbon dioxide (mmHg)	31	30	31	30

The three questions the round asked, and why each was the wrong one.

"Is the heparin too high?" It has already been halved twice with no effect, and the bleeding began while the anticoagulation was unchanged. Anticoagulation does not consume fibrinogen.
"Is this disseminated intravascular coagulation?" It is the right shape, and it must be excluded — but there is no sepsis, no new organ failure, no trigger, and the patient is otherwise stable and improving. Chapter 39 owns the distinction.
"Should we give more product?" Two units of red cells, and the natural next step is fibrinogen concentrate and platelets. This would treat the consequence of the process while leaving the process running.

The question nobody asked, and the four rows that answer it. The pressure ratio is flat. The post-membrane oxygen is barely changed. The post-membrane carbon dioxide is unchanged. The free haemoglobin is unchanged. By every measure of function, this circuit is working. And that is precisely the picture §31.5 describes: in the device-induced hyperfibrinolysis group, "gas exchange of respective MOs was only minimally affected."

What the coagulation rows say. Fibrinogen has fallen from 410 to 118 mg/dL — below the 200 mg/dL the anchor study used as its marker — with D-dimer rising nearly four-fold and platelets halving, over roughly four days, in a patient with no other explanation. That is the described pattern, on the described timescale.

What should happen.

Exclude the alternatives properly — sepsis, a new thrombosis or pulmonary embolus, occult bleeding, a drug — because the diagnosis is one of exclusion and stays presumptive until the exchange (Chapter 39, Chapter 36).
Examine the circuit with a torch and record what is seen, including the corners and inflow side of the membrane lung.
Discuss a system exchange with the perfusion team on the basis that the circuit is the most likely cause of a reversible coagulopathy — not on the basis that the circuit is failing, because it is not.
Give product to cover the exchange, not instead of it.
Record the indication in words, and examine the explanted circuit (§31.8).

The teaching point. Every functional measurement of this circuit was normal, and the circuit was the problem. A unit that defines "circuit failure" as "the circuit has stopped working" will not find this patient, and this patient was the single largest category of system exchange in the best cohort available.

This case is constructed to illustrate the pattern described in the anchor study. The values are illustrative and are not drawn from a published report.

31.10 What this chapter could not close

🚩

Open questions, recorded rather than concealed

There is no agreed indication for changing a circuit and no threshold in any parameter. The Red Book says so for the indications; the same source says so specifically for D-dimer — no cut-off is indicative of a need to change the circuit. Every number in this chapter is a description of what happened before an exchange in one cohort, not a trigger.
About half of acute circuit thrombosis was not foreseen by a daily panel in the best study available (§31.6), and no monitoring strategy has been shown to do better.
The anchor cohort is a single centre, venovenous only, and 2009 to 2013. Six ECMO systems were used, with different pumps and coatings but an identical polymethylpentene gas-exchange membrane throughout — so the study cannot say anything about membrane material, and its exchange rate reflects that centre's thresholds. No comparable venoarterial dataset was retrieved, which matters because embolism from a venoarterial circuit goes to the brain rather than to the lung.
The consequences of circuit thrombus for the patient — embolic stroke on venoarterial support, pulmonary embolism on venovenous — were not quantified by any source retrieved for this chapter. The literature measures the circuit, not what leaves it. This is a real gap and it is where the subject matters most.
Activated factor VII and acute oxygenator thrombosis: 2 of 4 patients. Four patients. Recorded as a signal and nothing more.
No evidence was retrieved on whether a circuit should ever be changed pre-emptively by time, and no source recommends it.
The frequency of circuit infection is unknown. Four of 83 exchanges here, with no routine infection surveillance performed — so this is a detection rate in the exact sense Chapter 30 §30.5 defines, and the true frequency is unknown in both directions.

Corrections and additions owed to other chapters

Chapter 36 (Bleeding) is owed §31.5 entire: the circuit belongs near the top of the differential for a diffusely bleeding ECMO patient, the distinguishing pattern is a rising D-dimer with a falling platelet count over three to four days, fibrinogen discriminates the severe form but a normal fibrinogen does not exclude it, and the treatment is a system exchange rather than more product.
Chapter 42 (Anticoagulation Monitoring) is owed the protamine hazard — protamine given while a heparin-coated circuit is running is a named cause of catastrophic circuit clot — and the anchor centre's protocol, including that it started ECMO without heparin for up to seven days in trauma with severe fractures or in bleeding diathesis, relying on the circuit coating.
Chapter 18 and Chapter 26 (weaning) are owed the Red Book's statement that the trial off is the period of highest circuit-thrombus risk, that flow must not fall below the oxygenator's established minimum, that anticoagulation should be adjusted pre-emptively, and that an open bridge with a thumb clamp or a pump-controlled retrograde trial-off avoids the intermittent clamping.
Chapter 32 (Oxygenator Failure) is owed the gas-transfer definitions used here as an exchange trigger, and the normalisation of the pressure drop to blood flow, which is this chapter's measurement and Chapter 32's diagnostic.
Chapter 37 (Haemolysis) is owed the five-item differential for a rising free haemoglobin (§31.3) and the fact that pump head thrombosis presents as haemolysis and not as obstruction.
Chapter 58 (Infection) is owed the finding that the circuit itself can be the septic focus, and that culturing the explanted device is what proves it.
Chapter 88 and Chapter 92 are owed the circuit check as a named, timed, recorded item.

Carried forward and still open: the pre-cannulation laboratory panel, still unowned (Chapter 29); the four numeric anticoagulation disagreements awaiting Chapter 42 (Chapter 29); and the alpha-stat versus pH-stat gap, for which no ECMO-specific evidence exists.

31.11 Evidence and certainty summary

Claim	Best evidence	Certainty
About a third of venovenous ECMO patients require at least one system exchange; roughly 45 percent of exchanges are acute and 55 percent elective	Lubnow and colleagues, 265 patients, 83 exchanged (31 percent), 37 acute and 46 elective. Full text read	Moderate for the proportions; low for the absolute rate, which is centre-specific
Mechanical failure accounts for under 10 percent of exchanges; the rest is clot or its consequences	Same cohort: 8 of 83	Moderate to high
Pump head thrombosis presents as haemolysis, not obstruction — free haemoglobin rose about 10-fold before exchange and fell about 6-fold after, with clot confirmed in every pump head	Same cohort, n equal to 13, p under 0.001 both directions	Moderate to high for the mechanism; moderate for the magnitude
Acute oxygenator thrombosis is identified by the pressure drop normalised to blood flow, which rose 2.6-fold before and fell 3.2-fold after exchange, and did not rise in any other acutely exchanged patient	Same cohort, n equal to 16, p under 0.001 both directions	Moderate to high, and unusually specific for an ECMO measurement
Roughly half of acute circuit thromboses give no warning on a daily surveillance panel	Same cohort: free haemoglobin warning in 6 of 13 pump head thromboses; D-dimer missed 5 of 16 acute oxygenator thromboses; the authors state "nearly 50 percent could be foreseen"	Moderate
A device-related coagulopathy exists, is the largest single reason for exchange, and reverses when the circuit is changed	Same cohort, n equal to 32 (hyperfibrinolysis 8, local fibrinolysis 24); the authors state that device-related hyperfibrinolysis "was reversible by a system-exchange"	Moderate — uncontrolled before-and-after with a stated comparison group, consistent reversal, plausible mechanism
It can occur in a circuit whose gas exchange is essentially normal, and a normal fibrinogen does not exclude it	Same cohort: gas exchange "only minimally affected" in the hyperfibrinolysis group; fibrinogen unchanged in the 24 with local fibrinolysis, 40 percent of whom bled	Moderate — the most clinically consequential finding in the chapter
Repeated system exchange, up to five membrane lungs, was not associated with increased mortality	Same cohort: weaning 69 versus 79 percent, death on ECMO 31 versus 21 percent, p equal to 0.113; acute versus elective p equal to 0.320	Low — not randomised, not powered, and a non-significant 10-point difference is not equivalence. It removes the strongest argument against a low threshold; it does not establish safety
There is no agreed indication or interval for changing a circuit, and no D-dimer cut-off indicates the need	ELSO Red Book, stated in its own words	High as a statement about the state of the evidence
The trial off ECMO is the period of highest circuit-thrombus risk, and flow must not fall below the oxygenator's established minimum	ELSO Red Book, weaning chapter	Moderate — expert practice with a clear mechanism
Flow below about 1.5 L/min in any tubing promotes stagnation and thrombus	ECMO in the Adult Patient, stated as a rule with that centre's own stricter practice	Moderate — expert practice, mechanistically sound, no outcome evidence
Protamine given while a heparin-coated circuit is running is a cause of catastrophic circuit clot	ECPR and Resuscitative ECMO, stated as a named cause	Moderate — expert statement, mechanistically obvious, no quantification
The circuit can be the septic focus	Same cohort, 4 of 83: mycobacteria identified microscopically in one membrane lung, Enterococcus cultured from another after ex-vivo incubation; noradrenaline fell 15-fold in one patient after exchange	Low for frequency (no routine surveillance was performed); high that it happens
Activated factor VII may precipitate acute oxygenator thrombosis	Same cohort: 2 of 4 exposed patients	Very low — four patients, uncontrolled. Recorded as a signal

31.12 Key points

Every circuit contains clot. The question is never whether there is clot but which clot matters and what it is doing.
A clot that matters declares itself in one of five languages — the eye, the pressure, the gas, the haemolysis panel and the coagulation panel — and the five do not substitute for one another. A unit watching one will be surprised by the other four.
Mechanical failure is under 10 percent of system exchanges. Almost everything else is clot or what clot does to the patient.
The circuit check is a physical examination, it takes thirty seconds, and it finds the two things no laboratory test finds in time: the clot you can see and the mechanical failure about to happen. Torch along the whole circuit; the membrane lung clots in the corners and on the inflow side.
Pump head thrombosis grinds before it obstructs. It presents as rising free haemoglobin with a change in the pump's sound, not as falling flow — and clot was found in every pump head removed.
Before calling a rising free haemoglobin pump head thrombosis, exclude five things: sampling aspiration, excessive suction from cannula malposition, circuit-connected renal replacement, thrombosis elsewhere, and large haematomas. Inlet pressure more negative than about minus 100 mmHg is itself a cause.
Acute oxygenator thrombosis is a rising pressure drop with a falling flow at unchanged pump speed — and the reflex to turn the speed up is the wrong move.
Read the pressure drop as a ratio to blood flow, not as a number. The ratio rose 2.6-fold before exchange and fell 3.2-fold after, and it did not rise in any other acutely exchanged patient.
Anything that lowers flow raises clot risk, including the things done for good reasons. The trial off ECMO is the highest-risk period in a whole run; flow must not fall below the oxygenator's established minimum and anticoagulation should be adjusted before, not after.
Protamine and a running heparin-coated circuit is a named cause of catastrophic clot. Put the circuit in the reversal conversation.
The largest single reason for changing a circuit is not that it stopped working. It is that it started consuming the patient's coagulation system.
A diffusely bleeding ECMO patient with a rising D-dimer and a falling platelet count over three to four days, and no other cause, may need a new circuit rather than more product — and the gas exchange of those circuits was only minimally affected.
Fibrinogen distinguishes the severe form, but a normal fibrinogen does not exclude the problem: 24 of 32 had a normal fibrinogen and 40 percent of them bled.
Device-related coagulopathy is presumptive until the exchange. It is confirmed by the markers improving afterwards, and that is the honest position rather than a weakness in the argument.
Surveillance converts about half of circuit emergencies into planned procedures. It does not abolish them. Roughly half of acute clot events arrived with no notice from a panel drawn daily. A normal panel is not reassurance — the same conclusion Chapter 28 reached about the brain.
There is no threshold in any parameter, and no agreed interval. The Red Book says so, and says it specifically of D-dimer.
Exchange when the circuit is doing something — obstructing, failing to exchange gas, haemolysing, or consuming coagulation factors. Visible clot alone is not an indication.
Repeated exchange, up to five membrane lungs, was not associated with higher mortality. That is not proof of safety, but it removes the main argument for waiting.
Before the clamps go on, pause and identify both limbs of both circuits. Clamp twice per limb about 15 cm apart and cut between the clamps.
Look at what came out. Record where the clot was, photograph it, culture it if infection was the question, and write down why the circuit was changed.

📌

Chapter answer: The circuit is clotting from the moment it is primed, so nothing is learned by asking whether clot is present. What is learned comes from asking what the clot is doing — and it can be doing one of four things: obstructing flow, ruining gas exchange, grinding red cells, or consuming the patient's coagulation system. The last of these is the largest category and the easiest to miss, because the circuit doing it is still working. Five channels watch for all four, none of them reassures when normal, and together they turn about half of the emergencies into planned procedures. That is a worthwhile gain and it is not a guarantee, so the circuit is examined by hand and by eye every shift, the pressure drop is read against the flow that produced it, and the circuit is changed when it is doing something — not when it merely looks imperfect, and not only when it has already failed.