Chapter question: Every laboratory test at the bedside was validated by changing one thing and watching the number move. An ECMO patient changes everything at once — haemoglobin, platelets, fibrinogen, factor levels, haematocrit, temperature, heparin and free haemoglobin, all in the same twelve hours. What survives that, what does not, and how should the panel be read?
Evidence search date: 22 September 2026.
Primary sources: ELSO Red Book 6th edition — the adult cardiac chapter (Table 28-2, the laboratory frequency schedule adapted from McMichael; Table 28-3, transfusion thresholds; the consumptive-coagulopathy panel; factor XIII and acquired von Willebrand syndrome) and the neonatal chapter (Table 9-1, the haematologic laboratory schedule; what the activated clotting time is affected by). ISCCM Manual of RRT and ECMO in ICU — the coagulation monitoring table with its targets, advantages and disadvantages, and the unit's own paediatric monitoring protocol. ECMO in the Adult Patient (Core Critical Care, 2017) — the haematology chapter: assay interference by free haemoglobin, the blood film, thrombocytopenia, platelet increments, red cell antibodies and the factor XI case. ECMO: A Practical Guide to Management (Taha, Caridi-Scheible, Leiendecker) — the haemolysis marker set, the two mechanisms of free-haemoglobin organ injury, and the coagulation test comparison table.
External evidence: an in vitro and in vivo study of four haemostasis assays against isolated and simultaneous changes in heparin and coagulation factors; an adult venoarterial cohort of low-level haemolysis and stroke; a paediatric cohort of haemolysis and acute kidney injury; a single-centre venovenous haemolysis analysis; a 215-patient study of free haemoglobin before and after cannulation; a pump comparison; a carboxyhaemoglobin study; two studies and two conference abstracts on acquired von Willebrand syndrome and platelet receptor shedding; a point-of-care free-haemoglobin method comparison; and two haematology reviews.
Every external reference was read as an indexed abstract or full-text passage. None was retrieved as a complete article. Digital object identifiers are reproduced from the search record and have not been independently resolved. Every database row seeded from this chapter is marked Verified = No.
What this chapter covers — and what it does not
This is the fifth chapter of Part V and the most tightly hemmed in, because four other chapters own pieces of the laboratory. The line this chapter holds is between the test and the thing the test is about. Chapter 37 owns haemolysis as a complication to be prevented and treated; this chapter owns the measurement of haemolysis, and what that measurement does to every other measurement. Chapter 42 owns anticoagulation monitoring as a therapeutic decision; this chapter owns the assays as instruments, and what they are actually correlated with.
This chapter owns | Owned elsewhere — cite, do not re-argue |
What the circuit does to a sample before the laboratory sees it — sampling site, dilution, heparin contamination, timing | Blood gas sampling sites and their four questions — Chapter 27 |
The haemolysis indices as measurements: plasma free haemoglobin, lactate dehydrogenase, haptoglobin, bilirubin, phosphate, the blood film, carboxyhaemoglobin — and the specific confounder attached to each | Haemolysis as a complication — its causes, its prevention, the circuit change — Chapter 37; the carboxyhaemoglobin line on the co-oximeter — Chapter 27 |
Free haemoglobin as an assay interferent — the tests it breaks and the direction it breaks them in | The renal consequences of free haemoglobin — Chapter 55; the neurological consequences — Chapter 28 |
The full blood count on a circuit, and the blood film | Transfusion thresholds and strategy — Chapter 46 |
Acquired von Willebrand syndrome and platelet receptor shedding as laboratory diagnoses — which test makes the diagnosis and which test cannot | The coagulation physiology that produces them — Chapter 39; bleeding as a complication — Chapter 36 |
What each coagulation assay is actually correlated with, measured directly | Choosing and titrating an anticoagulant against those assays — Chapter 42, Appendix 6 |
Assembling a laboratory schedule, and the cost of the panel itself | Integrating the whole dataset and the defect classes of interpretation — Chapter 30 |
29.1 One variable at a time, in a patient where everything changes at once
This is the organising idea, and unusually for this book it has been measured directly rather than argued.
Evidence · The study that should change how the coagulation panel is read
What was done. Four assays — prothrombin time, partial thromboplastin time, activated clotting time and anti-factor Xa heparin activity — were tested against isolated changes in heparin and coagulation factor levels, and then against simultaneous changes, using both in vitro samples and in vivo samples from patients on extracorporeal life support.
What was found with one variable changing at a time.
- Anti-Xa correlated with heparin at r equal to 0.97 and with antithrombin at r equal to 0.98, and was unaffected by other parameters.
- Prothrombin time correlated with coagulation factors at r equal to 0.88 and was minimally affected by heparin.
- Activated clotting time correlated with coagulation factors at r equal to 0.99 — a near-perfect correlation — but was insensitive to heparin below 0.5 U/mL, and was affected by factor XII and by platelets.
- Partial thromboplastin time correlated with heparin and with coagulation factors when each was varied individually.
What was found when multiple parameters changed, in vitro and in vivo.
- The activated clotting time correlated with neither heparin nor coagulation factors.
- The partial thromboplastin time had low correlation with either.
- Anti-Xa and prothrombin time retained their specificity.
The authors' conclusion, reproduced because it is unusually blunt for a laboratory paper: the activated clotting time is "too insensitive to heparin, sensitive to too many other factors, and too imprecise to be useful for monitoring hemostasis during ECLS."
Certainty: moderate to high for the correlation structure, which is a measurement rather than an inference and was reproduced in vitro and in vivo. Low for the practice recommendation, which no outcome study supports.
Source: Saifee NH, Brogan TV, McMullan DM, Yalon L, Matthews DC, Burke CR, Chandler WL. ASAIO J 2020;66(2):230–237. DOI 10.1097/MAT.0000000000000993.
Physiology · Why a good assay becomes a bad measurement
The finding above is not a statement about laboratory quality. Every one of those assays performed exactly as designed. What failed was the inference from the assay to the patient, and the reason is worth stating in general form because it applies to everything in this chapter.
A clotting assay is a race. A reagent starts a reaction and the instrument reports when a clot forms. That time depends on everything in the tube at once — the concentration of each factor, fibrinogen, platelets, the haematocrit that determines how much plasma there is per millilitre of sample, the temperature, and any drug that inhibits any step. A validation study varies one of those and shows a tight correlation. That correlation is real, and it is conditional.
An ECMO patient violates the condition continuously. In a single twelve-hour period the same patient can have a falling haematocrit, a falling platelet count, a falling fibrinogen, consumed antithrombin, a changing heparin infusion, a rising free haemoglobin, a temperature set by a heat exchanger and two units of red cells and a unit of plasma delivered. Every one of those moves the clotting time, and the assay reports a single number that cannot distinguish them.
The consequence is a rule, not a caveat: a test whose result depends on several variables is uninterpretable in a patient in whom several variables are moving, no matter how good the assay is. The tests that survive are the ones that are specific to a single quantity — anti-Xa for heparin, prothrombin time for factor levels — and the way to survive with the others is to measure more than one and read them against each other.
Clinical pearl · The one-line version, for the ward round
"This number moved. Which of the six things that move it, moved?"
It is worth asking out loud, because the reflex answer on ECMO is always the heparin, and the heparin is frequently not the answer. A rising activated clotting time in a patient whose platelets have halved and whose fibrinogen is 1.1 g/L is describing the platelets and the fibrinogen at least as much as the infusion, and turning the heparin down will not correct it.
This chapter's sections are, in effect, six answers to that question for six panels.
Pitfall · Treating the reference interval as if it applied
Every reference interval on the report was derived from a population of healthy volunteers who were not anaemic, not transfused, not anticoagulated, not febrile, not in multi-organ failure, and not connected to a pump. The interval is a statement about them, not a statement about this patient.
That is true in any intensive care unit. What makes it sharper on ECMO is that the deviations are not random — they are systematic, they are in a known direction, and they are produced by the treatment. A haematocrit of 0.28 is not noise; it is the expected consequence of circuit priming and haemodilution. A raised lactate dehydrogenase is not an incidental finding; it is what a pump does.
So the useful comparison on ECMO is almost never the reference interval. It is this patient's value yesterday — which is the same rule Chapter 25 reached for haemodynamics, Chapter 26 for ejection fraction and Chapter 28 for cerebral monitors. Part V has now arrived at it from five directions.
29.2 What happens to the sample before the laboratory sees it
Chapter 27 established that a blood gas is uninterpretable without its sampling site. The same is true of the laboratory panel, for different reasons, and the errors are more easily made because a full blood count does not obviously belong to a site the way a gas does.
Pre-analytical problem | What it does to the result | What to do |
Heparin contamination from a circuit sampling port | Prolongs every heparin-sensitive clotting time, sometimes dramatically. A coagulation sample drawn through a heparinised line is not a coagulation sample | Draw coagulation studies from a peripheral or dedicated arterial line where possible; if the circuit must be used, discard an adequate volume first and record that the sample was circuit-drawn |
Haemodilution from the circuit prime and from fluid | Lowers haematocrit, platelet count, fibrinogen and every factor level together, in proportion. A proportional fall across the whole panel is dilution until proven otherwise | Read the panel as a pattern. An isolated fall in one component is a finding; a fall in all of them at once is arithmetic |
Mechanical haemolysis during the draw itself | Raises free haemoglobin, potassium and lactate dehydrogenase in the tube, not in the patient. Indistinguishable on the report from the haemolysis this chapter is trying to detect | This is the single most important artefact in §29.3. A first abnormal free haemoglobin should be repeated, from a different site, before it is believed |
Iatrogenic blood loss from the panel itself | The schedules in §29.12 draw blood many times a day. Over a week that is a measurable contribution to the anaemia the panel is being used to monitor | Use paediatric tubes where the assay allows; audit the daily phlebotomy volume as a number, because no one does |
Timing relative to a heparin change | An anti-Xa drawn before a dose change has equilibrated is describing the previous dose | The Red Book schedule states it: anti-Xa 4 hours after each heparin change |
Timing relative to transfusion | A haemoglobin drawn during a transfusion is a number in transit. Haptoglobin falls after red cell transfusion independently of haemolysis (§29.3) | Record what was transfused in the preceding hours alongside the panel |
Danger · A single abnormal free haemoglobin has an artefactual explanation that is commoner than the real one
Every part of this chapter that follows depends on the free haemoglobin being real. A difficult draw through a small needle, excessive suction, or a long transit produces exactly the same result as pump-induced haemolysis, and in most laboratories the two are not distinguishable on the report.
Before a rising free haemoglobin triggers a circuit change — which is what Chapter 37 will have it trigger — repeat it, from a different site, and look at the concordant markers. Real intravascular haemolysis on a circuit moves a set of numbers together: free haemoglobin up, lactate dehydrogenase up, unconjugated bilirubin up, phosphate up, haptoglobin down, and frequently the urine or the dialysis effluent visibly discoloured. A high free haemoglobin with a normal lactate dehydrogenase, a normal haptoglobin and clear urine is a sampling artefact until proven otherwise.
Source for the concordant marker set: ECMO: A Practical Guide to Management, the venovenous physiology chapter, which lists plasma free haemoglobin, lactate dehydrogenase, unconjugated bilirubin and phosphate rising with haptoglobin falling, and recommends trending them rather than reading single values.
29.3 The haemolysis indices, and the confounder attached to each
Haemolysis is the laboratory signature of extracorporeal support, and it is measured with five or six tests that are quoted as a set. Read individually, every one of them is confounded by something an ECMO patient definitionally has.
Marker | What it measures | Its confounder in this population | How to use it |
Plasma free haemoglobin | Haemoglobin released into plasma. The most specific of the set, and the one on which the ELSO definition rests | Sampling artefact (§29.2). Also elevated by the resuscitation that preceded cannulation, before the circuit had touched the patient (§29.3 evidence box) | Daily, per the Red Book schedule. Repeat a first abnormal value before acting. It is the marker with the outcome data behind it (§29.4) |
Lactate dehydrogenase | An intracellular enzyme released by lysed red cells | Released by many kinds of injured tissue, not only red cells — so it rises in the shock, the infarction, the hepatic congestion and the muscle injury that this patient also has. And it was already grossly elevated before cannulation (§29.3 evidence box) | As a trend against its own baseline, never as an absolute. A rising LDH with a flat free haemoglobin is usually not the circuit |
Haptoglobin | Falls as it binds free haemoglobin | Two independent confounders pulling opposite ways. It is an acute-phase reactant, so inflammation raises it and can conceal consumption. And it falls after red cell transfusion independently of haemolysis, from the small quantity of free haemoglobin in each unit | Useful mainly when it is unmeasurably low with concordant markers. A mid-range haptoglobin in an inflamed, transfused patient carries almost no information |
Unconjugated bilirubin | The breakdown product of released haem | Hepatic dysfunction and congestion are common on venoarterial support; conjugated and unconjugated fractions must be separated or the number means nothing | Order the split bilirubin, not the total |
Phosphate | Released from lysed cells | Renal replacement therapy removes it; refeeding lowers it; it is replaced by protocol in most units | The most overlooked member of the set, and worth reading precisely because it moves for different reasons than the others — a concordant rise adds information |
Carboxyhaemoglobin | Endogenous carbon monoxide from haem catabolism, read off the co-oximeter with the blood gas | Smoking, inhaled exposure, and the small number of patients who run high without haemolysis (§29.3 evidence box) | Chapter 27 owns the line on the report. Its value here is speed: it comes back with the gas, not with the send-away |
The blood film | Red cell fragmentation, and much else | Requires a haematologist to look at it, which is the reason it is rarely requested | Underused. See the pearl below |
Evidence · Two findings that reverse the usual attribution
First: the free haemoglobin was high before the circuit started, and the circuit lowered it. A retrospective analysis of 215 adult patients with cardiac failure on extracorporeal life support, divided into three groups — cannulated during ongoing cardiopulmonary resuscitation (n equal to 110), after resuscitation with return of circulation and sustained shock (n equal to 45), and in severe shock without preceding resuscitation (n equal to 60) — measured free haemoglobin, lactate and pH daily before and during support.
- Resuscitation itself produced a pronounced rise in pre-cannulation free haemoglobin: 318 (138 to 586) versus 212 (107 to 439) versus 79 (53 to 232) mg/L across the three groups, p under 0.001.
- Within 24 hours on support, free haemoglobin declined significantly. The authors' conclusion: implantation did not aggravate haemolysis but instead decreased it within 24 hours.
- Pre-cannulation free haemoglobin had no prognostic value on multivariable analysis. Only a low pre-cannulation lactate predicted successful weaning (p under 0.0001) and hospital discharge (p equal to 0.0028).
Source: Lehle K, Lubnow M, Philipp A, Foltan M, Zeman F, Zausig Y, Lunz D, Schmid C, Müller T. Eur J Heart Fail 2017;19(S2):110–116. DOI 10.1002/ejhf.854.
Second: the lactate dehydrogenase was grossly abnormal before the circuit started. A comparison of two centrifugal pumps in 80 adult ECMO patients found no difference between them in daily plasma haemoglobin (5.7 plus or minus 3.6 versus 5.7 plus or minus 4.2 mg/dL), lactate or lactate dehydrogenase (2656 plus or minus 1607 versus 2689 plus or minus 1726) — but recorded pre-ECMO lactate dehydrogenase levels of 4004 plus or minus 3583 and 3604 plus or minus 3354. That is, the pre-cannulation value was higher than anything measured on support, in both arms.
Source: Palanzo DA, El-Banayosy A, Stephenson E, Brehm C, Kunselman A, Pae WE. Artif Organs 2013;37(9):E162–E166. DOI 10.1111/aor.12158.
What the two together mean. The circuit is routinely blamed for haemolysis that the illness and the resuscitation caused. Both markers were abnormal before cannulation, in the direction that is attributed to the pump, and in the larger study the free haemoglobin improved once support began.
The bedside consequence is a single instruction: measure the panel BEFORE the pump starts. Without a pre-cannulation value, every subsequent number is being read against a reference interval built from healthy people rather than against the only comparison that could settle the question.
Certainty: moderate. Both are retrospective and single-centre, and neither was designed to test attribution. But they point the same way, and the direction is the opposite of the clinical assumption.
Clinical pearl · Ask for the blood film, and ask a haematologist to look at it
ECMO in the Adult Patient devotes a section to what the film shows, and it is the least-ordered informative test in this chapter. Its own example, which it labels "the sign of dreadfulness", is a single field containing pseudo-Pelger-Huët neutrophils, prominent basophilic inclusion bodies in red cells, acanthocytosis suggesting hepatic failure, and a nucleated red cell beside a mature neutrophil. The same source notes that neutrophilia, monocytosis and lymphopenia are common, especially during septic episodes, and that the white cells are often affected by the drugs being given.
The film is the one test in this chapter that can distinguish mechanical fragmentation from the other causes of a falling haemoglobin, and it is the one that most often changes the differential rather than confirming it.
The same chapter makes a practical point worth carrying: a bone marrow biopsy can be performed safely on a thrombocytopenic patient, even on ECMO. It is not commonly needed, and it is not contraindicated.
Pitfall · Attributing all haemolysis to the circuit and stopping there
ECMO in the Adult Patient states it in one line: "Severe intravascular haemolysis should be investigated, and causes other than ECMO should be suspected." The differential that does not get considered because the pump is present includes transfusion reaction, drug-induced haemolysis, thrombotic microangiopathy, sepsis-associated haemolysis, glucose-6-phosphate dehydrogenase deficiency unmasked by an oxidant drug, and an underlying haemoglobinopathy.
The same source adds a second and subtler harm: "Significant haemolysis may lead to anaemia followed by inappropriate investigations if not readily recognized" — that is, the anaemia gets worked up as blood loss, and a patient on full anticoagulation is then scanned, scoped and possibly transfused for a bleed that is not occurring.
29.4 Controversy 1 · Is the ELSO threshold of 50 mg/dL a safety margin or a definition?
The clinical question. A patient on venoarterial support has a plasma free haemoglobin of 30 mg/dL on day 2. Every reference in the unit defines significant haemolysis as a value over 50 mg/dL. Is this patient's result normal?
Position A: 50 mg/dL is the operative threshold, and below it there is nothing to act on. It is the figure the Extracorporeal Life Support Organization uses, it is reproduced across the field, and the practical arguments behind it are real: a value below it does not distinguish itself from sampling artefact, and there is no intervention of demonstrated benefit at that level. A method-comparison study of a point-of-care free haemoglobin device is instructive here — the device showed within-run imprecision under 3 percent and acceptable concordance with laboratory methods from 50 mg/dL upward, which is to say that agreement between methods below the threshold was not established. Source: Baud B, Dupuy AM, Zozor S, Badiou S, Bargnoux AS, Mathieu O, Cristol JP. Bioanalysis 2023;16(2):65–74. DOI 10.4155/bio-2023-0116.
Position B: the strongest adult outcome signal in this literature comes from values below the threshold. A single-centre retrospective review of all adult venoarterial ECMO patients over almost six years defined low-level haemolysis as a 48-hour free haemoglobin of 11 to 50 mg/dL — that is, entirely within the range the field calls normal. Of 150 patients (mean age 55 plus or minus 14; 33 percent women), 62 (41 percent) met that definition.
- Non-haemorrhagic stroke occurred in 20 of 62 (32 percent) with low-level haemolysis versus 4 of 88 (5 percent) without: adjusted hazard ratio 7.6 (95 percent confidence interval 2.2 to 25.9, p equal to 0.001).
- The relationship was graded: free haemoglobin 26 to 50 mg/dL gave a hazard ratio of 11.3 (3.6 to 35.1), and 11 to 25 mg/dL gave 4.4 (1.36 to 14.85), each against no low-level haemolysis.
- A mechanism was measured alongside: patients with low-level haemolysis had a two-fold greater rise in mean platelet volume after cannulation (0.98 plus or minus 1.1 versus 0.49 plus or minus 0.96 fL, p equal to 0.03), consistent with platelet activation.
- Patients with non-haemorrhagic stroke had operative mortality of 83 percent versus 45 percent (adjusted hazard ratio 3.1, 1.8 to 5.3, p under 0.001).
Source: Saeed O, Jakobleff WA, Forest SJ, Chinnadurai T, Mellas N, Rangasamy S, Xia Y, Madan S, Acharya P, Algodi M, Patel SR, Shin J, Vukelic S, Sims DB, Reyes Gil M, Billett HH, Kizer JR, Goldstein DJ, Jorde UP. Ann Thorac Surg 2019;108(3):756–763. DOI 10.1016/j.athoracsur.2019.03.030.
Where the evidence actually sits. The two positions are not really about the same thing, and separating them resolves most of the argument. 50 mg/dL is a definition of an event for registry and comparison purposes. It was never established as a level below which free haemoglobin is biologically inert, and the Saeed data are the clearest adult evidence that it is not. Supporting data from the paediatric side point the same way: in a 501-child cohort, an elevated free haemoglobin — described by the authors as the most specific of the haemolysis markers — carried an adjusted hazard ratio of 2.9 (1.4 to 5.6) for a composite of acute kidney injury or renal replacement therapy, rising to 6.6 (3.3 to 13.2) for stage 2 or worse injury. Source: Strong AE, Zee J, Fulchiero R, Kilbaugh TJ, Connelly J, Makeneni S, Campos D, Laskin BL, Denburg MR. Kidney360 2023;4(11):1536–1544. DOI 10.34067/KID.0000000000000253.
What this book concludes. Treat 50 mg/dL as the threshold for calling an event and not as the threshold for paying attention. A free haemoglobin that is rising within the 10 to 50 mg/dL band, with concordant markers, is a finding — it identified a group with a seven-fold adjusted hazard of stroke in the only adult cohort that looked. It does not by itself justify a circuit change, which is Chapter 37's decision and carries its own risk; it justifies looking for the cause, tightening the surveillance, and reviewing the anticoagulation.
The unresolved part is the direction of causation, and it is not trivial. Low-level haemolysis may be producing the strokes, or it may be marking a circuit that is generating thrombus which is producing the strokes. The interventional question — whether acting earlier changes anything — has not been asked.
What would resolve it. A prospective multicentre cohort with protocolised free haemoglobin measurement by a method validated below 50 mg/dL, adjudicated neurological and renal outcomes, and circuit-thrombus assessment recorded alongside — so that haemolysis as a cause can be separated from haemolysis as a marker.
Certainty: moderate that sub-threshold haemolysis marks risk; very low that it causes it; and there is no evidence at all about acting on it.
Danger · The threshold and the assay were not validated over the same range
This is the sharpest practical consequence of the controversy above, and it applies to any unit adopting point-of-care free haemoglobin measurement.
The interesting range is 11 to 50 mg/dL. The point-of-care method's concordance with the laboratory was established from 50 mg/dL upward. A device that agrees well at 200 mg/dL and has not been shown to agree at 25 mg/dL cannot be used to act on the Saeed finding, however attractive the turnaround time.
Before a unit builds a protocol around sub-threshold haemolysis, it must know the analytical performance of its own assay in that band. That is a question for the laboratory, not the perfusion team, and it should be asked in writing.
29.5 Haemolysis is also an interferent: the tests it breaks
This is the second structural claim of the chapter, and it is the reason the haemolysis section comes before everything else rather than after it.
Danger · The patient with the most abnormal laboratory results is the patient whose results are least trustworthy
ECMO in the Adult Patient states the mechanism in one sentence: "High levels of free haemoglobin can interfere with the optical instruments used in the laboratory to measure enzyme levels or coagulation tests."
Why it happens. A great many laboratory methods are optical. A clotting analyser detects clot formation by a change in light transmission or scatter; a biochemistry analyser reads absorbance at a defined wavelength. Free haemoglobin is intensely coloured and absorbs strongly across much of the visible spectrum. A plasma sample that is pink or red is optically different from the plasma the method was calibrated against, and the instrument cannot tell the difference between light absorbed by haemoglobin and light absorbed by the reaction it is trying to measure.
Two named examples from this book's own sources.
- The ISCCM manual lists, as a disadvantage of the anti-factor Xa assay, that free haemoglobin and bilirubin affect it.
- ECMO: A Practical Guide to Management lists, as a disadvantage of anti-Xa, that it "can be modified by hyperlipidaemia and hyperbilirubinaemia."
The compounding is the point. Haemolysis raises free haemoglobin and unconjugated bilirubin and lactate dehydrogenase. So the patient who is haemolysing is simultaneously the patient in whom the assay used to titrate the anticoagulation is least reliable, and the patient in whom a wrong anticoagulation decision is most dangerous. Chapter 28 met the same structure with neuron-specific enolase; this is its general form.
What to do about it. Ask the laboratory for the haemolysis index on the sample, which most modern analysers report automatically and most clinicians never see. It is the instrument's own statement about how much it trusts the result. A result flagged as haemolysed should be repeated, not interpreted — and if it cannot be repeated cleanly, the decision should be made on a test that does not share the interference.
Clinical pearl · Which laboratory question to ask, and of whom
Three questions are worth asking of the laboratory once, at the start of a programme, and writing the answers into the protocol. None requires a research study and all three change interpretation.
- "Does your analyser report a haemolysis index, and can it be added to our ECMO panel?" If yes, it should appear on every result in this chapter.
- "Which of our assays are optical, and what is their stated free-haemoglobin interference limit?" Manufacturers publish these. The limits are frequently lower than the values an ECMO patient reaches.
- "What is the analytical performance of our free haemoglobin method between 10 and 50 mg/dL?" (§29.4.)
A unit that cannot answer these is not measuring haemolysis; it is generating numbers about it.
29.6 The full blood count on a circuit
Component | What the circuit does to it | How to read it |
Haemoglobin and haematocrit | Fall at cannulation from the prime, then fall further from phlebotomy, haemolysis and bleeding. In one perioperative series haematocrit fell from 0.41 plus or minus 0.03 before support to 0.28 plus or minus 0.03 after it (p under 0.0001) | The Red Book declines to give a threshold: transfusion should be based on active bleeding or evidence of reduced systemic or regional oxygen delivery "rather than a fixed hemoglobin level", and it states that insufficient evidence exists to define a safe lower threshold in venoarterial ECMO. Chapter 46 owns the decision; Chapter 27 owns haemoglobin as an oxygen-delivery lever |
Platelet count | Falls in almost everyone. In the same perioperative series, 337 plus or minus 147 before support to 224 plus or minus 77 after (p equal to 0.0001), with short runs | A count is not a function (§29.7). The ISCCM table says so explicitly: "No proven threshold. Platelet function does not reflect by its count." ECMO in the Adult Patient reports that clinicians' stated minimum safe count ranges from 20 to 150 × 10⁹/L, and notes this variation is not peculiar to ECMO |
White cell count and differential | Neutrophilia, monocytosis and lymphopenia are common, especially during septic episodes; the differential is also affected by the drugs being given | Trend, not value (§29.10) |
Mean platelet volume | Rises with platelet activation. In the low-level haemolysis cohort, patients with sub-threshold free haemoglobin had a two-fold greater rise after cannulation | Already on the report, and almost never read. It costs nothing to look at |
Red cell antibodies | Repeated transfusion immunises. ECMO in the Adult Patient: up to 4 percent develop red cell antibodies, so compatibility testing must be repeated, usually every 72 hours | Put the 72-hour re-crossmatch on the weekly checklist. It is a logistics failure that presents as an emergency |
Clinical pearl · Measure the platelet increment, not just the count
A platelet transfusion on ECMO is usually given, and the result usually checked the next morning by which time it means nothing. ECMO in the Adult Patient gives the measurement that does mean something: a satisfactory increment is a rise of at least 10 × 10⁹/L measured 60 minutes after the infusion.
Inconsistent increments between transfusions should raise HLA immunisation. Platelets express minimal ABO antigen but do express HLA class I, and immunisation can follow pregnancy or red cell transfusion — so it can be present in a patient who has never received platelets before. HLA-matched platelets can be sourced.
The same chapter notes that arbitrary transfusion thresholds are often unachievable in patients with multifactorial platelet consumption, and that empirical approaches are sometimes adopted — transfusion every 12 hours, immunoglobulin, or HLA-matched units even with a negative antibody screen. Chapter 46 owns whether to do this. What belongs here is the measurement that tells you whether it worked.
And one reassuring fact worth telling families and juniors: the majority of thrombocytopenic patients recover their platelet count when ECMO is discontinued.
Pitfall · Working up the thrombocytopenia as though the circuit explained it
The circuit consumes platelets, so a falling count is attributed to it and the workup stops. ECMO in the Adult Patient lists what gets missed: acute or acute-on-chronic immune thrombocytopenia (which justifies prednisolone and immunoglobulin); drug-induced thrombocytopenia, which can be profound and should be excluded; post-transfusion or post-transplant thrombocytopenia, which it notes is often markedly more haemorrhagic than immune thrombocytopenia; thrombotic thrombocytopenic purpura, characterised by fever, microangiopathic haemolysis, thrombocytopenia and renal and central nervous system features; and sepsis or haemophagocytic syndrome, which may present with extremes of pancytopenia.
Heparin-induced thrombocytopenia belongs on that list for adults; the Red Book notes it is extremely rare in neonates.
The discriminating test is frequently the blood film (§29.3), and the discriminating question is whether the fall is isolated or part of a proportional fall across the panel (§29.2).
29.7 Primary haemostasis: the diagnosis the routine panel cannot make
Acquired von Willebrand syndrome is the commonest laboratory abnormality on ECMO that no routine test detects. The Red Book's adult cardiac chapter names it and factor XIII deficiency in the same sentence as "common", and then moves on — which is roughly how both are handled at the bedside.
Physiology · Why shear removes the largest multimers first
Von Willebrand factor circulates as multimers of widely varying size, and haemostatic potency rises steeply with multimer size — the high-molecular-weight multimers do most of the work of tethering platelets to injured subendothelium under flow.
Multimer size is regulated by shear itself. Under high shear the multimer unfolds, exposing a cleavage site for the protease ADAMTS13, which cuts it. In an intact circulation this is a regulatory mechanism. In a circuit running at pathological shear, it becomes a continuous stripping of exactly the multimers that matter. A 2024 review of acquired bleeding disorders places extracorporeal support alongside aortic stenosis and left ventricular assist devices in this mechanism — "increased proteolysis through shear stress followed by ADAMTS13" — and associates it with frequent and recurrent mucosal bleeding. Source: Tiede A, Susen S, Lisman T. Haemophilia 2024;30(S3):29–38. DOI 10.1111/hae.14995.
The laboratory consequence is the whole problem: the quantity of von Willebrand factor can be normal or high while its function is severely impaired, because what has been lost is a size distribution, not a concentration. An antigen level cannot see a size distribution.
Danger · The screening test in general use has roughly half the sensitivity of the diagnostic test
The conventional screen for acquired von Willebrand syndrome is the ratio of von Willebrand factor activity to antigen, on the reasoning that function falls while quantity does not. Two datasets say that ratio misses about half of the cases.
A paediatric ECMO series compared the ratio against direct multimer analysis by immunofixation electrophoresis, with the diagnosis adjudicated independently by two specialists from the gel. Of 28 specimens from children on support, 24 were classified as acquired von Willebrand syndrome on the gel. All 24 had a high-molecular-weight multimer fraction below 40 percent — but only 46 percent of them had an activity-to-antigen ratio below 0.7. Among controls, 92 percent had a normal multimer fraction and 89 percent a normal ratio. A cut-off of multimers below 40 percent gave sensitivity 100 percent, specificity 75 percent and overall accuracy 96 percent. The authors' opening sentence is the finding: "It is known that VWF activity to antigen ratio cannot be a screening test for AVWS."
Source: an abstract in the International Journal of Laboratory Hematology 2024;46(S2):3–121 supplement. DOI 10.1111/ijlh.14396. Abstract only, paediatric, single centre. [VERIFICATION REQUIRED: whether this series has been published in full.]
The bedside consequence. A normal activity-to-antigen ratio does not exclude acquired von Willebrand syndrome on ECMO. If the question is being asked because a patient is bleeding, the test that answers it is multimer analysis, and it must be requested by name. Most laboratories do not perform it routinely and some send it away, which is the real reason the ratio is used.
Evidence · How common, how severe, and how quickly it reverses
Prevalence in adults. A single-centre study of mechanical circulatory support measured von Willebrand factor and platelet receptors in 21 continuous-flow ventricular assist device patients, 20 ECMO patients, 12 heart failure patients and 7 with aortic stenosis. Loss of high-molecular-weight multimers, consistent with acquired von Willebrand syndrome, was found in 14 of 20 ECMO patients and 18 of 19 device patients. Platelet receptor shedding was demonstrated in parallel: elevated soluble glycoprotein VI in plasma, and significantly reduced surface glycoprotein Ibα and glycoprotein VI compared with healthy donors.
The control group is the part to notice. Platelet receptor levels were also significantly reduced in the heart failure patients — who had no device. So an unknown share of what is attributed to the circuit belongs to the disease that made the circuit necessary, which is the same correction §29.3 made for free haemoglobin and lactate dehydrogenase.
Source: Lukito P, Wong A, Jing J, Arthur JF, Marasco SF, Murphy DA, Bergin PJ, Shaw JA, Collecutt M, Andrews RK, Gardiner EE, Davis AK. J Thromb Haemost 2016;14(11):2253–2260. DOI 10.1111/jth.13497.
Time course, in children. A prospective observational study sampled 23 children before ECMO, daily to day 14, and 12 to 24 hours after decannulation — 191 samples. Eighteen of 23 (78 percent) had at least one collagen-binding-to-antigen ratio below 0.6; eight (35 percent) below 0.4; three (13 percent) below 0.2. The ratio fell over the course of support and normalised 12 to 24 hours after decannulation. Multimer patterns, available in three patients, showed loss of high-molecular-weight multimers during the first 3 days.
Source: Drop J, Wildschut E, de Hoog M, de Maat M, van Ommen CH. Abstract, Res Pract Thromb Haemost 2021;5(S2). DOI 10.1002/rth2.12591. Abstract only, paediatric, 23 patients. The same abstract records that bleeding complications occur in 50 to 70 percent of paediatric ECMO patients and that the aetiology is poorly understood.
Certainty: moderate that the syndrome is present in most patients on support and reverses rapidly after decannulation; low for any specific prevalence figure; very low for its contribution to any individual bleeding episode.
Pitfall · A bedside aggregometry result that named the wrong abnormality
A study of 20 patients undergoing lung transplantation with short-term perioperative ECMO measured multiple-electrode aggregometry alongside formal von Willebrand testing, before support and after explantation. The aggregometry suggested severe von Willebrand deficiency — the ristocetin channel fell from a mean area under the curve of 104.7 plus or minus 44.2 to 39.6 plus or minus 24.6, p under 0.0001.
The formal von Willebrand assays said the opposite. Antigen rose from 136.4 to 181.8 percent (p equal to 0.0068), ristocetin cofactor from 129.4 to 168.8 percent (p equal to 0.0088), collagen binding from 117.5 to 160.5 percent (p equal to 0.01). The ratios did not change. The high-molecular-weight multimer index fell only from 98.4 to 94.8 percent — statistically significant at p equal to 0.0015, and described by the authors as not clinically significant. There was no significant correlation between the ristocetin aggregometry channel and any of the laboratory von Willebrand tests.
Two lessons, and they point in different directions.
- A ristocetin-based point-of-care channel can collapse for reasons other than von Willebrand factor — in this study the authors concluded it was detecting acquired platelet dysfunction, which the adenosine diphosphate channel also showed (71.4 to 45.7, p equal to 0.0004). The channel is named after its reagent, not after the abnormality it detects.
- Short perioperative support did not produce clinically significant multimer loss. That is a genuine qualification on §29.7's prevalence figures, which come from longer runs — and it fits the paediatric time course, in which the ratio fell progressively over days.
Source: Garaj M, Durila M, Vajter J, Solcova M, Marecek F, Hrachovinová I. Artif Organs 2021;46(5):899–907. DOI 10.1111/aor.14142. Twenty patients, lung transplantation, short-run support — the population is not a general ECMO population.
29.8 The coagulation assays as instruments
Chapter 42 owns the anticoagulation decision and Appendix 6 owns the algorithms. What belongs here is narrower: what each assay is correlated with, which is §29.1's finding applied test by test, and where this book's own sources disagree about the target.
Assay | Correlates with (measured) | Confounded by | Targets quoted in this book's primary sources |
Anti-factor Xa | Heparin, r equal to 0.97; antithrombin, r equal to 0.98. Unaffected by other parameters | Optical interference from free haemoglobin, bilirubin and lipaemia (§29.5). Requires calibration; slower than point-of-care | 0.3 to 0.7 IU/mL (Red Book neonatal Table 9-1; ISCCM) versus 0.2 to 0.3 units/mL (Practical Guide, for venovenous support). These do not overlap at the lower bound and the discrepancy is unexplained |
Prothrombin time and international normalised ratio | Coagulation factors, r equal to 0.88. Minimally affected by heparin | Warfarin, liver disease, dilution | INR under 1.5 in a bleeding patient, under 3 if not bleeding (Red Book Table 28-3); 1.2 to 1.5 (ISCCM paediatric protocol) |
Activated partial thromboplastin time | Heparin and coagulation factors individually; low correlation with either when several parameters change | Low factor levels, low fibrinogen, von Willebrand disease, and — per the Practical Guide table — a raised acute-phase response | 1.5 to 2 times normal (ISCCM) versus 40 to 60 seconds (Practical Guide). The ISCCM entry adds the sound advice to use a ratio rather than an absolute, to avoid inter-laboratory variation |
Activated clotting time | Coagulation factors, r equal to 0.99. Insensitive to heparin below 0.5 U/mL. Not correlated with heparin OR factors when several parameters change | Platelet count, factor deficiency, hypofibrinogenaemia, haemodilution, hypothermia — all five listed by the Red Book itself, and all five present on ECMO | 160 to 200 seconds and 180 to 200 seconds — both appear in the ISCCM manual, in the text and in the table of the same document |
Fibrinogen | Fibrinogen | An acute-phase reactant — the ISCCM table says so — so consumption can be masked by inflammation | Over 1.5 g/L bleeding or pre-procedure, over 1 g/L otherwise (Red Book Table 28-3); over 100 mg/dL, or over 150 if bleeding or surgery planned (Red Book Table 9-1). The ISCCM table prints "Fibrinogen (mg/L) 100–150", which as written is 10 to 15 mg/dL and is implausible; the intended unit is almost certainly mg/dL. Reproduced with the discrepancy flagged, not silently corrected |
Antithrombin | Antithrombin activity | Consumed by heparin therapy itself; neonates have physiologically low levels | Over 50 to 80 percent (Red Book Table 9-1, with the note to consider replacement if on maximum heparin with an out-of-range anti-Xa) versus 80 to 120 percent (ISCCM). Again non-overlapping at the lower bound |
D-dimer | Fibrin turnover | Elevated by almost everything in a critically ill patient | Under 0.5 µg/mL (ISCCM), which describes almost no ECMO patient. Its stated value is different: the ISCCM manual reports it as having prognostic value for oxygenator failure, and cites a 13-patient study using a rise as a very early precursor. Chapter 32 owns oxygenator failure |
Viscoelastic testing | Global clot formation and lysis | Poor sensitivity and specificity for adjusting heparin, per both the ISCCM table and the Practical Guide | No numeric target given by either source |
Danger · Four unresolved numeric disagreements between this book's own primary sources
None of these is a trivial rounding difference, and a unit cannot follow both members of any pair.
- Anti-Xa target: 0.3 to 0.7 IU/mL versus 0.2 to 0.3 units/mL. The ranges do not overlap except at a single point. A unit targeting the lower range will run a patient at a heparin level the other source calls subtherapeutic.
- Activated clotting time target: 160 to 200 versus 180 to 200 seconds — within a single document.
- Antithrombin target: over 50 to 80 percent versus 80 to 120 percent. The first is a floor for replacement; the second reads as a therapeutic range. They are probably answering different questions, but neither source says so.
- The fibrinogen unit error above.
This book records all four and asserts none. Chapter 42 must resolve them or state that they are irreconcilable, and Appendix 6 cannot be written until it does. What matters clinically is the rule underneath: write down which source your protocol follows, and cite it on the protocol itself. An anticoagulation protocol without a citation is the commonest way these discrepancies become bedside errors.
Clinical pearl · A discordance between two assays is a finding, not a nuisance
ECMO in the Adult Patient gives a case that makes the argument better than any principle. A patient developed multiple thrombi in the cardiac chambers despite what was thought to be adequate anticoagulation. The clue was in the panel: anti-Xa levels were disproportionately low in relation to the activated partial thromboplastin time ratio. Further testing revealed a factor XI deficiency. The chapter's own conclusion: multimodal monitoring would have detected this anomaly earlier.
The generalisable rule: the two assays measure different things (§29.8's table), so when they disagree, the disagreement localises the abnormality. A prolonged aPTT with a low anti-Xa is not a heparin problem — it is a factor or inhibitor problem, and it needs a haematologist rather than a dose change.
The same source recommends what follows from this: a standard decision tree at the bedside, including when to call haematology. That is a cheap intervention and almost no unit has one.
29.9 Biochemistry, renal and hepatic tests
Test | What the circuit does to it | Reading rule |
Creatinine | Diluted at cannulation like everything else, so the first post-cannulation value underestimates renal injury. Muscle mass falls over a long run, lowering production | The pre-cannulation creatinine is the baseline that matters — the paediatric haemolysis cohort used exactly that, taking the value closest to ECMO start as baseline. Chapter 55 owns the injury; this chapter owns the caveat that a normal creatinine on day 1 of support is not evidence of a normal kidney |
Potassium | Rises with genuine haemolysis and with sampling haemolysis, identically (§29.2) | A high potassium with a haemolysed sample flag is repeated, not treated, unless the electrocardiogram says otherwise |
Phosphate | Released by lysed cells; removed by renal replacement therapy; falls with refeeding | See §29.3 — its value is that it is pushed by different forces than the other haemolysis markers, so concordance is informative |
Bilirubin | Unconjugated rises with haemolysis; conjugated rises with hepatic congestion, sepsis and parenteral nutrition. The NeoReviews review lists direct hyperbilirubinaemia among the morbidities increased by haemolysis | Always split. A rising total bilirubin in a patient on venoarterial support has at least three explanations that the total cannot separate |
Lactate | Chapter 25 and Chapter 27 own it as a perfusion and delivery instrument | The one laboratory value in this chapter with demonstrated prognostic weight before cannulation: in the 215-patient cohort, a low pre-cannulation lactate was the only marker predicting successful weaning and hospital discharge, while free haemoglobin predicted neither |
Albumin and total protein | Fall from dilution, capillary leak and the inflammatory response to the circuit | Relevant to drug binding (Chapter 47) and to the interpretation of corrected calcium. Not a nutritional marker here — Chapter 56 |
29.10 Infection markers on a circuit
Pitfall · Reading an inflammatory marker as though the circuit were not there
The extracorporeal circuit is a large, foreign, non-endothelial surface. Exposure to it activates complement and the contact system and triggers an inflammatory cascade that the nutrition literature describes as resembling the systemic inflammatory response syndrome, with neutrophil activation and vascular endothelial infiltration. The Red Book adds that extracorporeal support itself contributes to endothelial dysfunction and exacerbates the inflammatory response.
So the patient has a permanent, treatment-generated inflammatory stimulus, and every marker used to diagnose infection is a marker of inflammation.
- C-reactive protein and fibrinogen are acute-phase reactants and will be raised in a patient with no infection at all. Fibrinogen's role as a consumption marker (§29.8) is degraded by the same mechanism.
- Haptoglobin is an acute-phase reactant too, which is why §29.3 treats a mid-range value as uninformative.
- The white cell count and differential are disturbed by the circuit and by the drugs.
- Temperature is set by the heat exchanger and is therefore not available as a sign at all — the most underappreciated consequence of the circuit for infection diagnosis, and one Chapter 58 must carry.
What is left is the trend and the culture. A single value on any of these is close to uninterpretable; a sustained change in one is worth investigating. Chapter 58 owns the diagnosis and the antimicrobial decision; this chapter owns only the statement that the usual screening panel has lost most of its specificity.
This is recorded as a gap rather than a solution: no source retrieved for this chapter offers an ECMO-specific threshold or strategy for any infection marker. [VERIFICATION REQUIRED: procalcitonin and C-reactive protein performance for infection on extracorporeal support.]
29.11 Controversy 2 · Does running at a lower flow reduce haemolysis?
The clinical question. A venovenous patient has a rising free haemoglobin. The circuit looks intact and there is no visible clot. Should the blood flow be reduced?
Position A: yes — lower flow means less shear, and less shear means less haemolysis. This is the mechanical intuition and it is stated as advice in one of this book's primary sources. ECMO in the Adult Patient writes that "ECMO-related haemolysis can be decreased by appropriate circuit management and aiming for the lowest possible blood flow." The mechanism is not controversial: haemolysis is produced by shear, negative pressure and turbulence, and the NeoReviews review lists the risk factors as small cannula size, high negative inlet pressure in the pump head, and thrombosis in the pump head and oxygenator.
Position B: no — in the only adult cohort that measured it, lower flow was associated with MORE haemolysis. A retrospective single-centre analysis of 35 consecutive adult venovenous ECMO patients over almost six years took daily plasma free haemoglobin and haptoglobin as the outcome and examined them against patient characteristics, daily laboratory findings and circuit variables.
- A reduction in ECMO blood flow of 1 L/min was associated with a daily haptoglobin consumption of 93.371 mg/dL (p equal to 0.001) — that is, lower flow tracked with more haemolysis, not less.
- Membrane oxygenator dead space was associated with haptoglobin reduction (B equal to minus 215.307, p equal to 0.004).
- Continuous renal replacement therapy connected to the ECMO circuit was associated with both higher daily free haemoglobin and lower haptoglobin.
Source: Costantini S, Belliato M, Ferrari F, Gazzaniga G, Ravasi M, Manera M, De Piero ME, Curcelli A, Cardinale A, Lorusso R. Perfusion 2022;38(3):609–621. DOI 10.1177/02676591211073768.
Where the evidence actually sits. The two positions are probably both right about different things, and the resolution is mechanically plausible rather than demonstrated. At a fixed circuit, raising the pump speed raises shear. But lowering the flow in a circuit that is developing a problem does not undo the problem — and a falling flow is frequently a consequence of the thing causing the haemolysis (a developing clot, a rising resistance, an inadequately draining cannula, negative inlet pressure from a hypovolaemic patient) rather than a remedy for it. On that reading, the Costantini association is confounded by indication in the most ordinary way: flow was lowered because something was wrong, and the something that was wrong was haemolysing the blood.
The dead-space finding supports this: it points at the oxygenator as a site, not at the pump speed as a setting.
What this book concludes. Do not lower the blood flow as a treatment for haemolysis. Look for the cause — inlet pressure, cannula position and adequacy of drainage, visible clot, transmembrane gradient, and any circuit-connected renal replacement therapy — and address that. Lowering flow to treat a number risks under-supporting the patient while leaving the mechanism in place. Where flow is genuinely excessive for the patient's needs, reducing it is reasonable on general grounds; that is a different decision from treating a laboratory value.
And record the CRRT finding, because it is actionable and easy to miss: a circuit-connected renal replacement circuit is itself a haemolysis source. Chapter 55 owns the connection decision and should carry this.
What would resolve it. A prospective study with flow, inlet pressure, transmembrane pressure and free haemoglobin recorded together at high frequency, analysed with flow as a time-varying exposure and with circuit events adjudicated — so that "flow was lowered" and "the circuit was failing" can be separated.
Certainty: very low in both directions. Thirty-five patients, retrospective, and one primary source stating the opposite as advice.
Evidence · Two things that do not explain haemolysis, and one that does
The pump brand does not. Eighty adult ECMO patients received one of two centrifugal pumps in otherwise identical circuits — one costing 20 to 30 times more than the other. Over the first 10 days there was no significant difference in daily plasma haemoglobin (5.7 plus or minus 3.6 versus 5.7 plus or minus 4.2 mg/dL), lactate (2.8 versus 3.0) or lactate dehydrogenase (2656 versus 2689), and no difference in days of support, age, body surface area, aetiology or configuration. The authors' conclusion: "The difference in cost of the devices does not correlate with the performance and outcomes." Source: Palanzo and colleagues, DOI 10.1111/aor.12158.
The pump TYPE does, but the effect is age-dependent. The ISCCM manual's comparison table gives the mechanism — a roller pump is occlusive and flow is set by stroke volume and revolutions, with higher chances of mechanical trauma to blood; a centrifugal pump is non-occlusive with flow determined by preload and afterload, and less haemolysis. But the NeoReviews review reports the opposite direction in the young: Extracorporeal Life Support Organization data indicate a significant increase in intravascular haemolysis in neonatal and paediatric patients receiving ECMO when the centrifugal pump is used, compared with its use in adults. The stated reasons are elsewhere in the same review — small cannula size, high negative inlet pressure in the pump head, and thrombosis in the pump head and oxygenator — and the Red Book adds that neonatal red cells are more susceptible to mechanical stress and that neonates have higher haemoglobin concentration and blood viscosity.
So the honest summary is that the haemolysis-generating variable is not the pump's name or its price but the pressure and geometry it is asked to work against — which is why §29.11 sends the clinician to the inlet pressure rather than to the flow dial.
Source for the paediatric direction: Garg M. NeoReviews 2020;21(11):e728–e740. DOI 10.1542/neo.21-11-e728.
Clinical pearl · What haemolysis costs the patient, in outcomes rather than in numbers
Worth having to hand, because the panel is easy to treat as bookkeeping.
- Renal. In 501 children, elevated free haemoglobin carried an adjusted hazard ratio of 2.9 (1.4 to 5.6) for acute kidney injury or renal replacement therapy, and 6.6 (3.3 to 13.2) for stage 2 or worse. Elevated lactate dehydrogenase gave 3.1 and 2.8 respectively. The two mechanisms named in the Practical Guide are direct renal tubular toxicity and nitric oxide depletion.
- Neurological. In 150 adults on venoarterial support, sub-threshold haemolysis carried an adjusted hazard ratio of 7.6 (2.2 to 25.9) for non-haemorrhagic stroke (§29.4).
- Haematological. Free haemoglobin scavenges nitric oxide, producing endothelial dysfunction and platelet activation — the mechanism behind the mean platelet volume finding in §29.6 and behind the thrombosis seen in other haemolytic states.
- And a calibrating counterpoint: the NeoReviews review states that haemolysis increases renal injury, direct hyperbilirubinaemia and thrombosis without an increase in mortality, and describes haemolysis as "an unavoidable side effect of current ECMO technology" for which there are no approved treatments or treatment guidelines in the neonatal population.
So the panel identifies real, organ-specific harm, and there is no treatment for the finding itself. What it changes is surveillance, the threshold for inspecting the circuit, and the anticoagulation review — not a prescription.
29.12 Building a laboratory schedule
Two primary sources give an explicit schedule and they broadly agree. The Red Book's adult cardiac chapter reproduces one adapted from McMichael; its neonatal chapter gives a second; the ISCCM manual prints a unit's own paediatric protocol. None is evidence-based and none claims to be.
Test | Red Book adult (Table 28-2) | Red Book neonatal (Table 9-1) | Comment |
Activated clotting time | Every 1 to 2 hours — hourly until the heparin dose has been unchanged for 6 hours, then 2-hourly | — | The most frequent test in the schedule is the one §29.1 found least interpretable. See the note below |
Activated partial thromboplastin time | Every 6 to 12 hours | — | Use a ratio, not an absolute (ISCCM) |
Anti-factor Xa | Every 6 to 12 hours, or 4 hours after each heparin change | Every 6 to 12 hours, target 0.3 to 0.7 u/mL | The 4-hour rule is the part most often omitted |
Platelets | Every 6 to 12 hours | Every 6 to 12 hours, target over 50,000 to 100,000/µL | |
Haematocrit | — (within the full blood count) | Every 6 to 12 hours, target over 35 to 40 percent | The neonatal target is far above any adult threshold in Table 28-3 |
International normalised ratio | Every 12 to 24 hours | — | |
Fibrinogen | Every 12 to 24 hours | Every 12 to 24 hours | Targets in §29.8 |
Full blood count | Every 12 to 24 hours | — | |
Antithrombin | Daily, or as needed | Daily, target over 50 to 80 percent | |
Plasma free haemoglobin | Daily | — | The only haemolysis marker with a stated frequency in either schedule |
Thromboelastometry | Daily, or as needed for bleeding or thrombotic complications | — |
Clinical pearl · A defensible panel, built from what this chapter found
The schedules above are the published starting point. What follows is this book's synthesis of the additions and subtractions this chapter's evidence supports — offered to be adapted and audited, not as a guideline.
Before the pump starts, if the clock allows — the highest-value item in this chapter
- Full blood count, coagulation screen, fibrinogen, lactate dehydrogenase, bilirubin split, haptoglobin, plasma free haemoglobin and lactate. §29.3 showed that free haemoglobin and lactate dehydrogenase are frequently abnormal before cannulation, and that without this sample every later value is read against the wrong comparison.
- Group and antibody screen, with the 72-hour repeat scheduled from here.
Additions to the standard schedule
- The haemolysis index on every sample, if the analyser reports it (§29.5).
- The split bilirubin rather than the total, and phosphate, read with the haemolysis set (§29.3).
- Mean platelet volume — already on the count, and free (§29.6).
- A platelet increment at 60 minutes after any platelet transfusion (§29.6).
- A blood film with a haematologist's report at the first unexplained fall in haemoglobin or platelets (§29.3).
- Multimer analysis, requested by name, if acquired von Willebrand syndrome is being asked about in a bleeding patient — the activity-to-antigen ratio will not answer it (§29.7).
Subtractions worth considering
- The activated clotting time, in units that have an anti-Xa assay available in a workable turnaround. This is the one genuinely contentious recommendation in the chapter and §29.12's danger box states the case against it.
- Routine daily D-dimer, unless it is being used for the specific purpose the ISCCM manual describes — oxygenator surveillance, which Chapter 32 owns.
And one measurement nobody makes
- The daily phlebotomy volume. The schedule above generates something in the order of a dozen draws a day. Over a week that is a transfusion. It should be counted at least once, in every programme, and the panel pruned against the count.
Danger · The most frequent test in the schedule is the least interpretable, and this book will not tell you to stop doing it
The position has to be stated carefully, because the evidence and the practice point in opposite directions and the stakes are high.
What the evidence says. Measured directly, the activated clotting time correlates with coagulation factors at r equal to 0.99, is insensitive to heparin below 0.5 U/mL, is affected by factor XII and platelets, and — when several parameters change at once, as they always do on ECMO — correlates with neither heparin nor coagulation factors. The Red Book's own neonatal chapter lists five things that affect it, all five of which are present in every ECMO patient. The primary authors call it "too insensitive to heparin, sensitive to too many other factors, and too imprecise to be useful."
What practice says. It is recommended hourly to 2-hourly by the adult schedule, it is the bedside test with a result in minutes, and it is what most units titrate on. A survey quoted in the ISCCM manual found about 42 percent of centres using aPTT or ACT, about 11 percent anti-factor Xa, and about 9 percent viscoelastic testing or a combination.
What this book concludes, and the reasoning matters more than the conclusion. No outcome study shows that abandoning the activated clotting time improves anything, and a unit that removes its fastest bedside test without a workable replacement will be slower to detect a catastrophic over- or under-anticoagulation, which is a real harm. So this is not a recommendation to stop. It is a recommendation to stop treating it as a measurement of heparin effect.
The practical form: never act on an isolated activated clotting time. Pair it with an anti-Xa and a prothrombin time, and when they disagree, believe the specific test over the global one (§29.8). And when the activated clotting time moves without a heparin change, look at the platelets, the fibrinogen, the haematocrit and the temperature before touching the infusion — because r equal to 0.99 with coagulation factors means it is probably telling you about them.
Chapter 42 owns the final position. This chapter's contribution is the measurement that Chapter 42 has to answer to.
29.13 A worked case: the number that was about a different organ
Worked example · Reading a panel as a pattern rather than a list
The patient. A 61-year-old woman, day 4 of peripheral venoarterial ECMO for postcardiotomy shock. Circuit flow 3.4 L/min. Continuous renal replacement therapy connected to the ECMO circuit since day 2. Heparin infusion unchanged for 14 hours.
The panel, this morning.
Test | Today | Day 3 | Pre-cannulation |
Haemoglobin | 7.8 g/dL | 8.6 | 11.2 |
Platelets | 68 × 10⁹/L | 94 | 212 |
Fibrinogen | 1.9 g/L | 2.1 | 4.4 |
Activated clotting time | 226 s | 178 | — |
Anti-factor Xa | 0.34 IU/mL | 0.36 | — |
Plasma free haemoglobin | 38 mg/dL | 21 | 14 |
Lactate dehydrogenase | 780 U/L | 690 | 1240 |
Haptoglobin | 0.4 g/L (low) | 0.9 | 2.1 |
Bilirubin, unconjugated | 34 µmol/L | 21 | 9 |
Phosphate | 1.55 mmol/L | 1.10 | 1.05 |
Mean platelet volume | 11.4 fL | 10.2 | 9.1 |
The question on the ward round. The activated clotting time has risen by 48 seconds with no heparin change. The reflex is to reduce the infusion.
Working it through.
- The anti-Xa has not moved (0.36 to 0.34). §29.8: anti-Xa correlates with heparin at r equal to 0.97 and is unaffected by other parameters. The heparin effect has not changed, so the activated clotting time is not describing the heparin.
- What has moved is everything the activated clotting time is actually correlated with. Platelets 94 to 68, fibrinogen 2.1 to 1.9, haematocrit falling with the haemoglobin — and §29.1 measured the activated clotting time's correlation with coagulation factors at r equal to 0.99. Reducing the heparin would be treating the platelets and the fibrinogen with a dose change, and would leave the patient under-anticoagulated on a circuit.
- Now read the haemolysis set as a set, not as a free haemoglobin. Free haemoglobin 21 to 38, lactate dehydrogenase up, unconjugated bilirubin up, phosphate up, haptoglobin down, and mean platelet volume up. That is six markers moving concordantly in the direction of real intravascular haemolysis, which is what §29.2's danger box asks for before believing a free haemoglobin. This is not a sampling artefact.
- The free haemoglobin is 38 mg/dL, which is "normal" by the ELSO definition and squarely inside the band with an adjusted hazard ratio of 11.3 for non-haemorrhagic stroke in the Saeed cohort (§29.4). The number that looks normal is the one that matters.
- The pre-cannulation column does the work that no reference interval could. Lactate dehydrogenase is 780 today against 1240 before the pump started — so the absolute value, which looks alarming, is in fact the lowest it has been, and the LDH is not adding information here. Free haemoglobin is 38 against 14, and haptoglobin 0.4 against 2.1 — those two have genuinely changed. Without the baseline, all three would have been read the same way.
- The cause is probably on the circuit, and there is a specific candidate in the chart. Renal replacement therapy connected to the ECMO circuit was independently associated with higher free haemoglobin and lower haptoglobin (§29.11). That is not proof, but it is the cheapest thing to examine.
What should happen.
- Do not change the heparin. Anti-Xa is in range and unchanged.
- Inspect the circuit — inlet pressure, transmembrane gradient, visible clot, cannula position and drainage — and review the renal replacement connection with the perfusion and renal teams (Chapter 55 owns the connection; Chapter 32 owns the oxygenator assessment; Chapter 37 owns the circuit-change decision).
- Repeat the free haemoglobin and add a blood film.
- Raise the neurological surveillance, because this is the exact laboratory picture associated with subsequent non-haemorrhagic stroke (Chapter 28 §28.14).
- Address the falling fibrinogen and platelets on their own merits (Chapter 46), not as a way of correcting a clotting time.
The teaching point. Two numbers on this panel were abnormal and uninformative — the activated clotting time and the lactate dehydrogenase. One number was normal and was the most important on the page. Reading the panel as a list of values against reference intervals gets both of those backwards; reading it as a pattern against the patient's own baseline gets both right.
This case is constructed to illustrate the reasoning and the operating characteristics cited in §§29.1 to 29.11. It is not drawn from a published report, and the values are illustrative.
29.14 What this chapter could not close
Open questions, recorded rather than concealed
- Four numeric targets disagree between this book's own primary sources (§29.8): anti-Xa 0.3 to 0.7 IU/mL versus 0.2 to 0.3 units/mL; activated clotting time 160 to 200 versus 180 to 200 seconds within one document; antithrombin over 50 to 80 percent versus 80 to 120 percent; and the ISCCM fibrinogen unit, printed as mg/L where mg/dL is meant. Chapter 42 must resolve or formally declare these irreconcilable before Appendix 6 can be written.
- The analytical performance of free haemoglobin assays below 50 mg/dL is not established, and that is exactly the range in which the strongest adult outcome signal sits (§29.4). The one method-comparison study retrieved reports concordance from 50 mg/dL upward.
- Whether sub-threshold haemolysis causes the strokes it marks, and whether acting on it changes anything, has not been studied at all.
- No ECMO-specific evidence was retrieved on any infection marker — procalcitonin, C-reactive protein or otherwise — despite the circuit generating a permanent inflammatory stimulus (§29.10). Recorded as a finding, not an omission.
- Two of this chapter's von Willebrand sources are conference abstracts (the multimer cut-off study and the paediatric time-course study), both paediatric and single-centre. The adult prevalence rests on 20 ECMO patients.
- The flow-and-haemolysis relationship is supported in opposite directions by a primary textbook and a 35-patient cohort (§29.11), with no study designed to separate them.
- The Red Book adult and neonatal schedules are reproduced without an evidence base, and neither source claims one. The frequencies in §29.12 are convention.
- Haemolysis has no treatment. The NeoReviews review states there are no approved treatments or treatment guidelines for ECMO-associated haemolysis in the neonatal population, and none was retrieved for adults either.
Corrections and additions owed to other chapters, arising from this chapter's audit
- Chapter 14 (Initial VA ECMO Management) and the venovenous initiation chapters should carry the pre-cannulation laboratory panel as a named step. It is the single highest-value item in this chapter and no chapter currently owns it. This is a small scope gap, and unlike Chapter 28's it is closable by an addition rather than a new chapter.
- Chapter 55 (AKI and CRRT) is owed the finding that a renal replacement circuit connected to the ECMO circuit is itself associated with haemolysis (§29.11).
- Chapter 88 (ECMO Morning Round Template) is owed the panel and the pattern-reading rule from §29.13, and the daily phlebotomy volume as a number to record.
- Chapter 42 is owed §29.1 and §29.8 entire, and the position on the activated clotting time that this chapter declines to finalise.
Carried forward from earlier chapters and still open: a home for extracorporeal carbon dioxide removal (Chapter 27); a home for neurological complications as complications (Chapter 28); the alpha-stat versus pH-stat gap, for which no ECMO-specific evidence exists; and the Hunt near-infrared spectroscopy cohort.
29.15 Evidence and certainty summary
Claim | Best evidence | Certainty |
Anti-Xa is specific for heparin (r equal to 0.97) and unaffected by other parameters; prothrombin time is specific for factor levels (r equal to 0.88) | Saifee and colleagues, in vitro and in vivo, isolated and simultaneous parameter changes | Moderate to high — a direct measurement, reproduced in two settings |
The activated clotting time correlates with coagulation factors at r equal to 0.99, is insensitive to heparin below 0.5 U/mL, and correlates with neither when several parameters change | Same study; the five confounders are listed independently by the ELSO Red Book | Moderate to high for the measurement; low for what to do about it |
Free haemoglobin and lactate dehydrogenase are frequently abnormal BEFORE cannulation, and free haemoglobin falls in the first 24 hours of support | Lehle and colleagues, 215 patients; Palanzo and colleagues, 80 patients (pre-ECMO LDH 3604 to 4004) | Moderate — two retrospective single-centre studies pointing the same way |
Pre-cannulation lactate predicts weaning and discharge; pre-cannulation free haemoglobin predicts neither | Lehle and colleagues, multivariable analysis | Moderate |
Sub-threshold haemolysis (free haemoglobin 11 to 50 mg/dL) is associated with non-haemorrhagic stroke on venoarterial support, with a graded relationship | Saeed and colleagues, 150 adults; aHR 7.6 overall, 11.3 for 26 to 50 mg/dL, 4.4 for 11 to 25 mg/dL | Moderate for the association; very low for causation; none for acting on it |
Elevated free haemoglobin is associated with acute kidney injury | Strong and colleagues, 501 children; aHR 2.9, rising to 6.6 for stage 2 or worse | Moderate, paediatric |
Free haemoglobin interferes optically with enzyme and coagulation assays, including anti-Xa | ECMO in the Adult Patient; ISCCM manual; Practical Guide, all stating it independently | High as an analytical fact; low for the magnitude in any given assay |
Haptoglobin is confounded in two directions — an acute-phase reactant, and falls after red cell transfusion independently of haemolysis | Haematology review of transfusion-related haemolysis markers; haemopexin review | Moderate to high |
Lactate dehydrogenase is released by many injured tissues, not only red cells | Same review | High |
Acquired von Willebrand syndrome is present in most patients on support and reverses within 12 to 24 hours of decannulation | Lukito and colleagues, 14 of 20 adults; Drop and colleagues, 78 percent of 23 children, normalising after decannulation (abstract) | Moderate for presence; low for prevalence; low for the reversal timing |
The von Willebrand activity-to-antigen ratio is not an adequate screening test; multimer analysis is required | Paediatric abstract: multimers under 40 percent gave sensitivity 100 percent and specificity 75 percent, where the ratio identified only 46 percent of gel-positive cases | Low — abstract only, paediatric, 28 specimens — but the direction is supported by the mechanism, since a ratio cannot see a size distribution |
Platelet receptor shedding occurs on support — and also in heart failure without a device | Lukito and colleagues, with a cardiac control group | Moderate — the control group is the strength of this study |
Short perioperative ECMO does not produce clinically significant multimer loss | Garaj and colleagues, 20 lung transplant patients; multimer index 98.4 to 94.8 percent | Low — small, specific population |
A ristocetin-based aggregometry channel can collapse without any von Willebrand abnormality | Same study: ristocetin area under the curve 104.7 to 39.6 while all three von Willebrand assays rose | Low for frequency; high as a demonstration that the failure mode exists |
Lowering blood flow reduces haemolysis | Asserted by ECMO in the Adult Patient; the opposite association found by Costantini and colleagues in 35 adults | Very low in both directions — probably confounded by indication |
Circuit-connected renal replacement therapy is associated with haemolysis | Costantini and colleagues, 35 adults | Low — but actionable and easily overlooked |
Pump brand does not affect haemolysis; pump type and the pressures it works against do | Palanzo and colleagues (no difference between two centrifugal pumps differing 20 to 30-fold in cost); ISCCM comparison table; NeoReviews for the paediatric reversal | Moderate for the brand finding; low for the age-dependent direction |
Up to 4 percent of repeatedly transfused patients develop red cell antibodies, requiring 72-hourly compatibility testing | ECMO in the Adult Patient | Moderate — expert practice, and a logistics rule rather than a clinical claim |
A satisfactory platelet increment is a rise of at least 10 × 10⁹/L at 60 minutes | ECMO in the Adult Patient | Moderate — standard haematology practice, not ECMO-specific |
There is no proven platelet threshold on ECMO, and clinicians' stated minima range from 20 to 150 × 10⁹/L | ISCCM table; ECMO in the Adult Patient | High that the variation exists; none for any particular threshold |
There is no approved treatment for ECMO-associated haemolysis | NeoReviews review, stated for the neonatal population; nothing retrieved for adults | Moderate to high |
29.16 Key points
- Every assay in this chapter was validated by changing one variable at a time. An ECMO patient changes several at once, always. Measured directly: anti-Xa correlates with heparin at r equal to 0.97 and prothrombin time with factor levels at r equal to 0.88, and both hold; the activated clotting time and the partial thromboplastin time lose their correlation entirely when several parameters move together.
- The useful comparison on ECMO is almost never the reference interval — it is this patient's value yesterday, and ideally before the pump started. Part V has now reached that rule from five directions.
- Take a full laboratory panel BEFORE cannulation. Free haemoglobin and lactate dehydrogenase are frequently abnormal beforehand, and in one 215-patient cohort free haemoglobin fell in the first 24 hours of support. Without the baseline, the circuit is blamed for what the illness and the resuscitation caused. No chapter currently owns this step.
- Repeat a first abnormal free haemoglobin before acting on it. Sampling haemolysis and circuit haemolysis are indistinguishable on the report. Real haemolysis moves six numbers together — free haemoglobin, lactate dehydrogenase, unconjugated bilirubin and phosphate up; haptoglobin down; frequently the urine discoloured.
- Each haemolysis marker has a confounder that an ECMO patient definitionally has. Lactate dehydrogenase comes from any injured tissue. Haptoglobin is an acute-phase reactant and also falls after red cell transfusion. Phosphate is the most overlooked and the most independently informative.
- 50 mg/dL is a definition of an event, not a threshold below which nothing is happening. In the only adult cohort that looked, free haemoglobin of 11 to 50 mg/dL carried an adjusted hazard ratio of 7.6 for non-haemorrhagic stroke, graded with concentration. It justifies looking for a cause, not a circuit change.
- Free haemoglobin is an assay interferent as well as a finding. It breaks optical methods — including anti-Xa. The patient who is haemolysing is the patient in whom the test used to titrate the anticoagulation is least reliable. Ask for the haemolysis index; it is already being generated.
- A platelet count is not platelet function, and there is no proven threshold. Measure the increment at 60 minutes instead, and treat inconsistent increments as possible HLA immunisation.
- Do not let the circuit close the thrombocytopenia workup. Immune, drug-induced, post-transfusion, microangiopathic and haemophagocytic causes all present identically, and the blood film is the test most likely to change the differential.
- Acquired von Willebrand syndrome is present in most patients on support, and the ratio everyone screens with misses about half of it. Von Willebrand antigen rises on ECMO while function falls, because what is lost is a size distribution. If the question matters, request multimer analysis by name.
- The abnormality reverses fast — within 12 to 24 hours of decannulation in the paediatric time-course data — which is worth knowing when planning a procedure.
- Four numeric targets in this book's own primary sources disagree, including two ranges that do not overlap and one unit that is wrong as printed. Write down which source your protocol follows and cite it on the protocol.
- When two coagulation assays disagree, the disagreement localises the abnormality. A prolonged aPTT with a low anti-Xa is a factor or inhibitor problem, not a dose problem — in one reported case, factor XI deficiency found only after intracardiac thrombus had formed.
- Never act on an isolated activated clotting time. When it moves without a heparin change, look at platelets, fibrinogen, haematocrit and temperature first.
- Do not lower the blood flow to treat a haemolysis number. The association between low flow and haemolysis is most easily read as confounding by indication: flow falls because something is wrong, and that something is haemolysing the blood. Look at inlet pressure, drainage, transmembrane gradient and any circuit-connected renal replacement therapy instead.
- The pump's price does not predict haemolysis; two centrifugal pumps differing 20 to 30-fold in cost performed identically across 80 patients.
- Count the daily phlebotomy volume once. The published schedules generate roughly a dozen draws a day, and over a week that is a transfusion.
Chapter answer: What survives is the specific: anti-Xa for heparin, prothrombin time for factor levels, free haemoglobin for haemolysis, the multimer gel for von Willebrand, the film for everything the counter cannot see. What does not survive is the global — the activated clotting time, the partial thromboplastin time, the total bilirubin, the haptoglobin in an inflamed and transfused patient — because a single number cannot separate six moving causes. And the panel should be read as a pattern against the patient's own pre-cannulation values, because the reference interval on the report describes a person who was never here.