📈

Chapter 25 — Haemodynamic Monitoring

🧭

Chapter question: Every haemodynamic instrument at the bedside was designed, validated and given its normal range in a circulation with one pump, in series, producing a pulse. The patient in front of you has two pumps in parallel, or a shunt in series, and may have no pulse at all. The monitors still display numbers. What do those numbers now mean — and which of them has quietly stopped meaning anything?

📅

Evidence search date: 15 September 2026.

Primary sources: ECMO: A Practical Guide to Management (Taha, Caridi-Scheible, Leiendecker) — Chapter 9, Table 9.1, the monitoring-modality table. ECPR and Resuscitative ECMO (Shinar and Badulak) — Chapter 7, Table 1, monitoring the venoarterial patient. ISCCM Manual of RRT and ECMO in ICU — Chapter 33, Monitoring during Extracorporeal Membrane Oxygenation, including the ELSO circuit-pressure monitoring recommendation. ELSO Red Book 6th edition.

External evidence: the cardiac-output-on-ECMO literature (the modified Fick principle and its published exchange, the shunt-fraction method, modified thermodilution in a porcine model and in vitro), the haemodynamic phenotyping study of 58 pigs and 128 patients, a porcine venoarterial-against-ECMELLA comparison, a pressure-volume simulation study, a systematic review of sublingual microcirculation on venoarterial support, and a general non-invasive cardiac-output validation cohort.

Every external reference was read as a structured abstract or as an indexed passage. None was retrieved in full text. 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

Part V is about measurement. This chapter opens it, and it is about the instruments of circulatory measurement specifically: what each one physically measures once a pump has been placed in the circulation, and how far that has drifted from what its label says.

This chapter owns	Owned elsewhere — cite, do not re-argue
What the arterial waveform, pulse pressure, filling pressures, venous saturation, lactate and cardiac-output devices measure on a circuit	Why the circulation behaves as it does on venoarterial support — afterload, series-versus-parallel flow, the mixing point — Chapter 13
The pulse-pressure target problem, and why low pulse pressure has more than one cause	Left ventricular distension: its mechanism, its detection by imaging, and the decision to unload — Chapter 15
Measuring or estimating native cardiac output while the pump is running	Echocardiography as a modality — views, measurements, the turndown study — Chapter 26
Venous saturation as a measure of perfusion adequacy, and which sample is the true mixed venous one	Arterial blood gases, oxygenation targets and gas-exchange interpretation — Chapter 27; differential hypoxaemia — Chapter 16
Lactate as a haemodynamic signal, including the wash-out trap	Lactate as a laboratory value, and the haemolysis panel — Chapter 29
The microcirculation, and its divergence from the macrocirculation	Cerebral oximetry and cerebral flow — Chapter 28
Circuit pressures as haemodynamic information — what a drainage pressure says about the patient	Circuit pressures as device diagnostics — delta P, oxygenator failure, cannula problems — Chapter 32, Chapter 34, Chapter 38
Assembling a monitoring set, and the errors of interpretation	What to do with the numbers — fluids, vasopressors, inotropes, flow changes — Chapter 54; integration, trends and scores as a class — Chapter 30

The boundary with Chapter 54 is the one that matters most. This chapter ends at the interpretation. A number, correctly understood, is where this chapter stops and Chapter 54 begins — and the commonest error in ECMO haemodynamics is not a wrong treatment but a correct treatment applied to a misread instrument.

25.1 Every instrument was calibrated in a circulation that no longer exists

This is the organising idea, and it is mechanical rather than clinical.

🧠

Physiology · Three circulatory architectures, one set of monitors

Every haemodynamic instrument in routine use assumes one pump, in series, generating pulsatile flow through a lung and then a body. Thermodilution assumes the indicator injected upstream all passes the thermistor downstream. Pulse pressure assumes stroke volume is the only source of arterial pressure oscillation. Mixed venous saturation assumes that all systemic venous return mixes in the pulmonary artery. Pulse contour analysis assumes an arterial pressure wave generated by a ventricle ejecting into a closed vascular tree.

Venoarterial ECMO replaces that architecture with two pumps in parallel. The native heart ejects antegrade; the circuit delivers retrograde; the two meet at a mixing point whose position moves with the ratio of the two flows (Chapter 13, Chapter 16). Venous return is partly diverted before it reaches the right heart. Every one of the four assumptions above is now false.

Venovenous ECMO leaves one pump but inserts a shunt in series with it. Blood drains from the vena cava, is oxygenated outside the body, and is returned to the same venous compartment; whatever fraction of venous return bypasses the circuit is a true right-to-left shunt. The heart is intact and the assumptions about pressure hold — but the assumptions about oxygen do not, and §25.7 shows that this converts an oxygenation measurement into a flow measurement.

The instruments do not announce any of this. A pulmonary artery catheter on venoarterial support will display a cardiac output. It is a number, generated correctly by a working device, and it is not the patient's cardiac output.

Instrument	What it measures with a native circulation	What it measures on venoarterial ECMO
Mean arterial pressure	The driving pressure for organ perfusion, generated by the heart	The sum of two pumps' output against the vascular resistance. Still the perfusion pressure, still useful — and it says nothing whatever about the heart
Pulse pressure	Stroke volume, broadly, in a given vascular tree	Whether the native ventricle is ejecting at all — and, as §25.4 shows, not only that
Thermodilution cardiac output	Cardiac output	An artefact. The indicator is partly drained into the circuit before it reaches the thermistor, and pulmonary blood flow is not systemic blood flow. This book's own primary source states the problem plainly: readings are inaccurate because of redirection of pulmonary blood flow
Pulmonary artery pressures	Right ventricular afterload and left-sided filling	Transpulmonary flow and left-sided filling combined. A low pulmonary artery pressure may mean a well-drained right heart or a failing one — the two are indistinguishable from the number (§25.4, §25.6)
Central venous pressure	Right-sided filling pressure	Right-sided filling minus whatever the drainage cannula has removed. The circuit is now a determinant of the number
Mixed venous saturation from a pulmonary artery catheter	Global oxygen delivery against consumption	The saturation of the fraction of venous return that escaped the drainage cannula. The true mixed venous sample is now in the circuit (§25.7)
Pulse contour cardiac output	Stroke volume from the arterial waveform	Unreliable by construction: the waveform is generated by two sources, one of them non-pulsatile
Dynamic preload indices (pulse pressure variation, stroke volume variation)	Fluid responsiveness in a ventilated, sinus-rhythm patient	Uninterpretable. They require a pulsatile arterial signal driven solely by a ventricle whose preload is being modulated by the ventilator — and on a circuit, preload is being modulated by a pump
Lactate	Anaerobic metabolism, broadly	Still anaerobic metabolism — plus a wash-out artefact on initiation, plus altered clearance (§25.8)

💡

Clinical pearl · The three questions a haemodynamic monitor on ECMO can actually answer

Rather than asking what each device shows, ask which of three separate questions it bears on. They have different instruments, different time courses, and they can move in opposite directions.

Is total systemic perfusion adequate? Mean arterial pressure, total flow (circuit flow plus native output), venous saturation, lactate trend, urine output, microcirculation. The pump can fix this one within minutes.
Is the native heart doing anything? Pulse pressure and the arterial waveform, pulmonary artery pulsatility, echocardiography (Chapter 26). The pump cannot fix this one at all, and is partly responsible for the answer.
Is the circuit delivering what it claims? Flow probe distal to all shunts, drainage and pre- and post-oxygenator pressures, recirculation. The pump is the subject of this question, not the instrument for it.

Most bedside confusion comes from answering one question with another's instrument — most often from reading a good mean arterial pressure as evidence that the heart is recovering, which is question 1 answering question 2.

25.2 The arterial line: where it goes, and what the waveform is for

The arterial line is the single most informative instrument in this chapter, and its usefulness depends almost entirely on a decision made at insertion.

🚨

Danger · On peripheral venoarterial support, the arterial line goes in the right arm — and a line elsewhere answers a different question

This book's primary sources agree without qualification. One states that the arterial line "should be placed in the right arm of all VA ECMO patients." Another instructs that during peripheral venoarterial support "an arterial line should be secured on the right arm and blood gases' analysis should be taken from the same side," because this provides information about oxygenation at the level of the coronary, innominate and left common carotid arteries.

The reason is Chapter 16's, and it is worth restating in one sentence: the right radial artery is the last territory the circuit reaches, so it is the first to reveal that the native, poorly oxygenated circulation has taken over the upper body. A femoral or left radial line is bathed in circuit blood and will look reassuring while the brain and heart are hypoxic.

Two consequences that are routinely missed.

A left-sided or femoral line is not a substitute with a caveat. It is a different measurement. It reports the circuit's perfusion, not the watershed.
When a right radial line cannot be sited, the pulse oximeter probe goes on the right hand — a substitution both primary sources make explicitly — while accepting that a non-pulsatile patient may give no plethysmographic trace at all.

Chapter 16 owns the physiology and the management of differential hypoxaemia. This section owns only the consequence for where the catheter goes.

⚠️

Pitfall · Pulse oximetry is less accurate on a circuit than off it, and the direction of the error is known

One primary source states that pulse oximetry is typically accurate to within 2 to 3% of the true arterial saturation, but that in patients on ECMO the discrepancy is 3 to 7.2% higher and more inaccurate. The proposed mechanism is twofold: shear stress in the circuit causes haemolysis, which generates endogenous carbon monoxide and therefore carboxyhaemoglobin, which a conventional two-wavelength oximeter reads as oxyhaemoglobin; and reduced pulsatility degrades the signal the device depends on.

The same source's practical instruction follows directly: frequent measured arterial saturations are recommended even when the pulse oximeter reading looks normal.

Two corollaries. First, the error is in the reassuring direction — the oximeter reads high. Second, the mechanism links this to Chapter 37: a circuit that is haemolysing more will mislead the oximeter more, so the reading becomes least trustworthy exactly when the circuit is deteriorating.

💡

Clinical pearl · Read the arterial waveform's shape, not only its two numbers

A monitor reduces the waveform to a systolic, a diastolic and a mean. On a circuit the shape carries information the three numbers discard.

Is there a dicrotic notch? Its presence is the most specific bedside sign that the aortic valve is opening and closing — that is, that the ventricle is ejecting into the aorta rather than merely generating pressure. Its disappearance is an earlier and more specific signal than a falling pulse pressure, because a small pulse pressure can be produced by a circuit artefact and a dicrotic notch cannot.
Is the pulsatility regular, and does it track the electrocardiogram? Pulsatile oscillation that does not follow the electrocardiogram is not ejection; it is the circuit, the ventilator, or a damped line.
Is the trace damped? A femoral arterial line in a leg with a distal perfusion cannula, or a line in a vessel partly obstructed by the arterial cannula, can flatten the waveform for reasons that have nothing to do with the heart. Before concluding that the ventricle has stopped ejecting, confirm that the line itself is faithful — a fast-flush test, and a look at whether the trace changed at a moment when something was done to the line rather than to the patient.

A flat trace with no dicrotic notch, confirmed on a well-damped right radial line that tracks the electrocardiogram, is a finding. A flat trace on an unverified line is a plumbing problem until proved otherwise.

25.3 Pulse pressure: the most-used number in ECMO, and the one with no agreed target

Pulse pressure is quoted on every round, written into every weaning protocol and used to trigger unloading decisions. It has no agreed threshold, and this book's own primary sources do not overlap.

⚖️

Controversy 1 — What pulse pressure should a venoarterial ECMO patient have?

The problem, stated exactly. Two of the primary textbooks used throughout this book give targets that do not overlap.

One monitoring table gives a pulse pressure of 10 to 20 mmHg, with the accompanying statements that pulsatility indicates native left ventricular ejection, that inotropes may be necessary to maintain pulsatility, and that a lack of pulsatility indicates the need for a venting strategy.
The other gives a goal pulse pressure of over 25 mmHg, in a table of monitoring modalities and their normal values.

Ten to twenty against over twenty-five. A patient at 22 mmHg is simultaneously above one target and below the other, and neither source cites evidence for its number.

The case for a low threshold (roughly 10 mmHg). What matters physiologically is not the magnitude of ejection but its presence: an aortic valve that opens each beat prevents stasis in the left ventricle and aortic root, and therefore prevents the thrombus that is the feared consequence. On this reading the target is binary — some ejection, every beat — and 10 mmHg is simply a pragmatic floor for detecting it. A porcine comparison supports the plausibility of the range: after coronary embolisation on venoarterial support, pulse pressure fell to 11.1 (SD 3.3) mmHg, and in the group also supported with a microaxial pump to 5.6 (SD 1.1) mmHg (P = 0.064), with four of seven and seven of eight animals respectively reaching a pre-specified target of 10 mmHg or less. That target was chosen to create a low-pulsatility state, which tells us where investigators consider the floor to lie.

The case for a higher threshold (over 25 mmHg). A pulse pressure at 10 mmHg is close to the noise floor of a damped line, is compatible with an aortic valve that opens only intermittently, and gives no margin. A higher target forces attention to inotropy and to unloading earlier, when the options are still reversible.

What is not in dispute, and is more useful than either number. Both sources agree that a wider pulse pressure does not by itself indicate that the ventricle has recovered, and that a flattened one indicates a low-flow state through the native heart with a risk of stasis and thrombosis.

Where this book lands.

Do not treat a pulse-pressure number as a target in its own right. It is one of the three questions in §25.1, and the instrument answers it only qualitatively.
Treat the presence of ejection as the binary that matters, confirmed by a dicrotic notch on a verified waveform and, when it matters, by echocardiography (Chapter 26).
Record the trend, not the value. A pulse pressure falling from 30 to 15 mmHg at unchanged circuit flow is a real event; a pulse pressure of 15 mmHg on a single reading is not a diagnosis.
Do not let a number trigger an unloading decision on its own — §25.4 is the reason, and the decision itself is Chapter 15's.

What would settle it. A study relating pulse pressure, by band, to aortic root stasis on imaging and to subsequent thrombosis. None was identified. The thresholds in both textbooks appear to be conventions.

[VERIFICATION REQUIRED] — both targets are quoted from textbook tables, and neither table cites a source for its figure.

🧠

Physiology · Why pulse pressure falls on venoarterial support even when the heart is unchanged

Three mechanisms act simultaneously from the moment flow starts, and none of them is a change in contractility.

Preload falls. The drainage cannula removes venous return before it reaches the right heart, so less blood reaches the left ventricle and stroke volume falls by the Frank-Starling mechanism.
Afterload rises. Retrograde flow into the aorta raises the pressure the ventricle must exceed to open the aortic valve. This is Chapter 13's central claim and is not re-argued here.
The arterial tree is already pressurised. A commentary on the pressure-volume literature makes the point cleanly: ECMO simply increases arterial blood flow; in the absence of a change in systemic vascular resistance, this amounts to an increase in arterial volume and pressure. The ventricle is therefore ejecting into a fuller, stiffer system.

The clinical consequence is precise and is the reason §25.4 exists: a falling pulse pressure after a flow increase is the expected mechanical result of that flow increase. It is evidence about the circuit's effect on the heart, not evidence that the heart has deteriorated.

The same commentary makes the converse point about weaning. When flow is reduced and cardiac output rises, that rise is not a direct effect of the flow reduction but a compensatory one — mediated by sympathetic tone, stressed blood volume, or drugs. A heart that "improves" when flow comes down has not necessarily improved; it has been allowed to fill.

25.4 A low pulse pressure has more than one cause, and the commonest is not the one assumed

This is the most immediately practice-changing section in the chapter.

The standard bedside inference runs: pulse pressure is low, therefore the ventricle is not emptying, therefore the ventricle is distending, therefore it should be unloaded. The largest study to measure rather than model the haemodynamic response to venoarterial ECMO found that this chain is correct in a minority of patients.

📊

Evidence · Moderate certainty · Three haemodynamic phenotypes, measured in animals and in patients

Simulation models predict that venoarterial ECMO uniformly produces left ventricular distension in cardiogenic shock. This study tested that prediction against measurement.

The animal arm. Left ventricular end-diastolic pressure and arterial pulse pressure were measured in 58 pigs with cardiogenic shock at two levels of ECMO flow. Cluster analysis identified three phenotypes, whose centroids for the change in end-diastolic pressure and the change in pulse pressure were: Cluster 1, +4.99 and −1.85 mmHg; Cluster 2, +0.02 and −5.57 mmHg; Cluster 3, +1.31 and +15.5 mmHg.

The clinical arm. Pulmonary artery diastolic pressure, pulmonary artery pulse pressure and arterial pulse pressure were collected within two hours of initiation in 128 patients with cardiogenic shock. Three phenotypes again:

"Pulsatile" — 26%, with the highest arterial pulse pressure.
"Low pulsatility" — 52%, with low arterial and low pulmonary artery pulse pressure.
"Left ventricular distension" — 22%, with an elevated pulmonary artery diastolic pressure.

Acute myocardial infarction was more common in the distension phenotype. The low-pulsatility phenotype had a larger right ventricular diameter and more tricuspid regurgitation.

The authors' interpretation of the distension phenotype: a higher pulmonary artery diastolic pressure occurred alongside higher tricuspid annular plane systolic excursion, a smaller right ventricular diameter and a higher pulmonary artery pulse pressure — that is, better right ventricular function and higher transpulmonary flow, in a ventricle that was not emptying. In the low-pulsatility phenotype the small arterial pulse pressure was not accompanied by distension, probably because reduced transpulmonary flow from poorer right ventricular function meant the left ventricle was not being filled in the first place.

🚨

Danger · Low pulse pressure most often means the right ventricle is failing, not that the left is distended

Read the phenotype distribution again. Of the three quarters of patients with a low arterial pulse pressure, roughly two thirds had no distension at all. In that group the low pulse pressure was a downstream consequence of a right ventricle that could not deliver blood across the lungs.

The two situations demand opposite actions, and the arterial pulse pressure cannot distinguish them.

Low pulse pressure with a high pulmonary artery diastolic pressure and a preserved pulmonary artery pulse pressure — blood is arriving at a left ventricle that cannot empty it. This is distension. Unloading is the question (Chapter 15).
Low pulse pressure with a low pulmonary artery diastolic pressure and a low pulmonary artery pulse pressure — blood is not arriving. Unloading the left ventricle here removes preload from a ventricle that is already underfilled, and does nothing about the failing right ventricle that caused the problem.

The discriminating measurement is on the pulmonary artery side, and on venoarterial ECMO it is usually not being made. One cardiogenic-shock centre reporting detailed left ventricular haemodynamics states plainly that pulmonary artery catheters are not routinely inserted at their institution during treatment with venoarterial ECMO.

This is the strongest argument in this book for a pulmonary artery catheter on venoarterial support — and it is an argument for the pulsatility and diastolic pressure, not for the thermodilution cardiac output, which §25.5 shows is invalid.

💡

Clinical pearl · What this book's own primary source already said, and what nobody drew from it

One of the monitoring tables quoted in Controversy 1 contains this line about pulmonary artery pressures on venoarterial support: "Low PA pressures and low PA pulsatility indicate a well-drained right heart and/or poor RV function." And, separately: "Rising PA diastolic pressure may indicate LV overdistension and risk of pulmonary oedema." Both statements are correct, both are in the same table, and the table then offers "no recommendation for absolute numbers."

Those two lines contain the whole of the phenotype finding, five years earlier, without the inference being drawn. The practical form is a three-line bedside read, and it costs nothing if a pulmonary artery catheter is already in:

Arterial pulse pressure low, pulmonary artery diastolic pressure rising → distension physiology.
Arterial pulse pressure low, pulmonary artery pulse pressure also low, pulmonary artery diastolic pressure not rising → the right ventricle, or over-drainage.
Arterial pulse pressure preserved → the ventricle is ejecting; whatever else is wrong, this is not it.

And the confounder that must be excluded first: over-drainage. A drainage cannula taking too much venous return produces low pulmonary artery pressures, low pulmonary artery pulsatility and a low arterial pulse pressure — an exact mimic of right ventricular failure, produced by the machine and reversible by turning a dial. §25.6 covers how to tell them apart.

⚠️

Pitfall · Loss of pulse pressure predicts failure to wean, and the paper that shows it also shows why single numbers mislead

In a study of patients after extracorporeal cardiopulmonary resuscitation, loss of pulse pressure was independently associated with failure to wean from support, with an odds ratio of 0.039 (95% CI 0.006 to 0.275, P below 0.01) — alongside age, duration of resuscitation and ECMO complications. The direction is unsurprising and the association is strong.

The same multivariable table also reports an odds ratio of 339.0 (95% CI 1.850 to 62,124) for arterial pH after cannulation, and an odds ratio of 221,457 (95% CI 2.385 to over twenty billion) for pH on the first intensive care day, both labelled statistically significant. Confidence intervals spanning ten orders of magnitude are the signature of separation or near-collinearity in a small logistic model, not of an enormous effect.

Both results come from the same model. The pulse-pressure finding is plausible and is used in this chapter; the pH estimates are printed here uncorrected as a reminder that the same regression can produce one credible coefficient and two meaningless ones, and that a reader who accepts the table's conclusions wholesale accepts both. Chapter 30 owns this failure mode as a class.

25.5 Cardiac output on a circuit

Two numbers are wanted and they are not the same: total systemic flow, which determines perfusion, and native cardiac output, which determines whether the patient can leave the machine.

Quantity	How to get it	Why it matters
Circuit flow	The flow probe — sited distal to all circuit shunts, as ELSO's circuit-monitoring recommendation specifies, otherwise it reports what left the pump rather than what reached the patient	The one flow in the system that is directly measured. Everything else in this section is derived
Total systemic flow	Circuit flow plus native cardiac output, which is why the second term cannot simply be ignored	Determines oxygen delivery and therefore perfusion adequacy (§25.7)
Native cardiac output	Not directly measurable at the bedside. Estimated by echocardiography (Chapter 26), by a modified Fick principle, or by modified thermodilution — all discussed below	The weaning question, the distension question, and the recovery question all reduce to this number

🚨

Danger · Thermodilution cardiac output is invalid on venoarterial ECMO, and the monitor will not say so

This book's own primary source states it in a monitoring table, in the row for the pulmonary artery catheter: "PAC readings are inaccurate due to redirection of pulmonary blood flow from VA ECMO flow." The same table, two columns earlier, lists cardiac output and cardiac index among the things the catheter measures, with normal values.

The mechanism has three parts, and each alone would be enough.

The indicator is stolen. A cold bolus injected into the right atrium or vena cava is partly drained into the ECMO circuit before it reaches the pulmonary artery thermistor. The area under the thermodilution curve is therefore too small, and the calculated output too high.
Pulmonary blood flow is not systemic blood flow. Even a perfectly measured transpulmonary flow is only the native contribution; the circuit's contribution never passes the thermistor at all.
The baseline temperature is set by the circuit. A heat exchanger holding the blood at a fixed temperature alters both the baseline and its stability, and thermodilution is a measurement of a temperature transient against a baseline.

The practical rule: on venoarterial support, do not report a thermodilution cardiac output. If a pulmonary artery catheter is in place, use it for the pressures and the pulsatility (§25.4) and for the true venous saturation — and delete the cardiac output from the chart. A number that is wrong and plausible is more dangerous than no number.

⚖️

Controversy 2 — Can native cardiac output be measured at the bedside during extracorporeal support?

Three approaches exist, they have been argued about in print, and none is yet routine.

Approach A — the modified Fick principle. Derived from mass balance, this relates blood flow through the native lung to circuit flow through the ratio of the oxygen content differences, or equivalently the gas transfers, across the two gas exchangers. Its proponents state that it estimates native cardiac output with acceptable precision and accuracy, that it was derived theoretically, tested in a pilot study, elucidated in a bench study and then assessed in 16 animals across varying dead-space and shunt conditions with more than 1,500 blood gas analyses, and that — unlike the alternatives — it works while extracorporeal gas exchange is continuing, raising the possibility of continuous monitoring at all stages of therapy.

Approach B — the shunt-fraction method. In central right-atrium-to-aorta support, turning off the sweep gas and the oxygen converts the circuit into a pure right-to-left shunt, after which the familiar shunt equation yields the ratio of pulmonary to systemic flow, and native cardiac output follows from the known circuit flow. In the published case, weaning central flow to 800 mL/min with sweep off gave a pulmonary-to-systemic ratio of 0.8 and a calculated native output of 3.2 L/min; up-titrating inotropes at the same flow raised the ratio to 0.84 and the output to 4.1 L/min; further weaning to 400 mL/min gave 0.9 and 4.0 L/min.

The published exchange between them is instructive rather than merely territorial. The Fick authors argue that the shunt method is a special case of their own general solution, restricted to the state of zero sweep gas; they concede that maximising the content difference in that way may improve accuracy; and they then raise the objection that matters most clinically. Both methods are valid only if the inlet and outlet conditions of both the circuit and the native lung are perfectly mixed — that is, only if there is no differential hypoxia. Their own data show that venous differential hypoxia is as common and as important as the arterial form, although far less often recognised, and that left atrial saturation cannot be assumed to be 100% even in healthy lungs, let alone in shunt states — an assumption the shunt method uses to avoid direct sampling.

Approach C — modified thermodilution. Injecting into the circuit and measuring in both circuit limbs and the pulmonary artery reconstructs the partition of the bolus and therefore the flows. In an in-vitro venovenous simulation it measured recirculation fraction accurately and precisely (bias −5.4%, limits of agreement −18.6 to 7.9% by area under the curve; −5.9%, −18.8 to 7.0% flow-based) — but delivered a cardiac output that was, in the authors' own words, accurate but imprecise: bias 0.56 L/min with limits of agreement of −2.27 to 3.4 L/min, across simulated cardiac outputs of 2.5 to 3.5 L/min. A related porcine study applied the technique to right heart function.

Where this book lands.

The modified Fick principle is the most generalisable of the three and the only one that works with gas exchange running. It is also the most demanding: it requires simultaneous sampling at four sites and an assumption of complete mixing that is frequently false.
The shunt-fraction method is the most executable at a bedside and is a reasonable way to obtain a single number at a weaning decision point on central support. It should not be used as a continuous monitor, and its assumption of a fully saturated pulmonary vein should be stated whenever the result is quoted.
Modified thermodilution's cardiac output is not yet usable. Limits of agreement of roughly ±3 L/min around a cardiac output of 3 L/min describe a method that cannot distinguish a good heart from a failing one. Accuracy without precision is not a measurement, and this is the clearest example in Part V of a technique whose headline bias figure is far more reassuring than its interval. Its recirculation measurement, by contrast, looks genuinely useful (Chapter 9).
Whichever is used, differential hypoxia invalidates it, which places Chapter 16 upstream of every number in this section.

What would settle it. A head-to-head comparison of all three against a reference flow measurement, in patients rather than animals or bench models, across a range of native outputs. None was identified.

[VERIFICATION REQUIRED] — every study here is known through an abstract or an indexed passage; no full text was retrieved and no equation in the sources has been independently checked.

💡

Clinical pearl · The cross-check that costs nothing

Whatever derived number is produced, check it against an independent estimate before acting on it. In a published venovenous case, a cardiac output derived from the shunt equation as 8.7 L/min was checked the same day against echocardiography — an aortic area of 4.95 cm², a velocity-time integral of 20.69 cm and a heart rate of 89 gave 9.07 L/min. Two methods, two sets of assumptions, agreement within about 4%.

Agreement of that kind is not proof, but disagreement is informative. A derived output that differs substantially from the echocardiographic estimate usually means one of the assumptions has failed — most often incomplete mixing, an unrecognised recirculation fraction, or a sampling site that is not where it is believed to be.

Chapter 26 owns the echocardiographic measurement. This section owns the instruction to make two measurements and compare them.

25.6 Filling pressures, and the pressures the circuit generates

On ECMO there are two sets of pressures: the ones measured in the patient and the ones measured in the tubing. The second set contains haemodynamic information about the patient that the first set cannot provide, and it is routinely read only as device diagnostics.

Pressure	What a high value suggests	What a low value suggests
Central venous pressure	Inadequate drainage of the right heart, pneumothorax, tamponade, high airway pressures, or abdominal compartment syndrome	Hypovolaemia or vasoplegia — and a risk of inadequate venous drainage to the pump
Pulmonary artery diastolic pressure	Left ventricular overdistension and a risk of pulmonary oedema (§25.4)	Reduced transpulmonary flow — from good drainage, from right ventricular failure, or both
Pulmonary artery pulse pressure	Preserved right ventricular output across the lungs	A well-drained right heart and/or poor right ventricular function — the ambiguity §25.4 turns on
Drainage (access) line pressure	Less negative than expected: the cannula is well filled, or flow is low	Excessively negative: the cannula is not being filled. Hypovolaemia, a malpositioned or undersized cannula, tamponade, or a pump speed the venous return cannot support
Pre- and post-oxygenator pressures	A rising gradient across the membrane points at the membrane. A rise in both points at an obstruction after it	Chapter 32 and Chapter 34 own these as device diagnostics. They appear here only because the drainage pressure is also a volume signal

💡

Clinical pearl · The drainage line is a volume monitor, and it is the only continuous one you have

The ISCCM monitoring chapter makes the point explicitly: continuous negative pressure monitoring may be helpful in the assessment of intravascular volume status and proper venous catheter positioning. ELSO's circuit-monitoring recommendation covers the access line, the pre-oxygenator port and the post-oxygenator port.

The physiology is straightforward. A centrifugal pump is preload sensitive and afterload dependent: at a fixed revolution rate it will deliver less flow if less blood arrives, and it will generate progressively more negative pressure trying. Chatter, suction events and a drainage pressure becoming steadily more negative at unchanged pump speed are a hypovolaemia alarm that fires before the arterial pressure moves — because the arterial pressure is being defended by the pump itself.

The same source names the harm: excessive negative pressure on the drainage cannula may traumatise the surrounding cardiac and vascular endothelium and cause an outpouring of gaseous emboli from tubing cavitation.

Two cautions. First, this is a trend instrument — the absolute value depends on cannula size, length, position and pump speed, so each patient is their own control. Second, it is not specific to volume: tamponade, a kinked cannula and a cannula that has migrated all produce the same signal, and the distinction is echocardiographic (Chapter 26).

⚠️

Pitfall · The primary sources decline to give absolute numbers, and that is the correct answer

The venoarterial monitoring table quoted throughout this chapter gives, for central venous pressure and for pulmonary artery pressures, the same entry: "No recommendation for absolute numbers." It is tempting to read this as a gap. It is not.

A central venous pressure on venoarterial ECMO is right-sided filling pressure minus whatever the drainage cannula has removed, and the second term is set by pump speed, cannula size and position. A value of 8 mmHg means something different at 2 L/min than at 5 L/min, in the same patient, ten minutes apart. There is no number because the number is a function of a machine setting.

The instruction that follows is to record the pressures with the flow that produced them. "Central venous pressure 8" is not a datum. "Central venous pressure 8 at a circuit flow of 4.2 L/min, drainage pressure −60 mmHg" is.

25.7 Venous saturation: which sample is the mixed venous sample

Venous saturation is the most useful single measure of whether perfusion is adequate, and on a circuit it is drawn from the wrong place more often than any other sample in this chapter.

🧠

Physiology · On venoarterial ECMO the true mixed venous sample is in the tubing

Mixed venous saturation means the saturation of all systemic venous return, after complete mixing, before any oxygen is added. In a native circulation that point is the pulmonary artery, which is why the pulmonary artery catheter exists.

On venoarterial support, most of the systemic venous return is drained into the circuit before it reaches the pulmonary artery. The pulmonary artery sample therefore represents only the fraction that escaped the cannula — a residue, not a mixture. The blood that did enter the circuit is sampled, fully mixed and before any gas exchange, at the pre-oxygenator (premembrane) port.

One of this book's primary sources states the conclusion directly: "Venous saturation on VA ECMO is a true reflection of venous oxygen content. This can be measured from either a central line or premembrane blood gases."

The practical rule: on venoarterial support, the premembrane saturation is the mixed venous saturation. A central venous sample is an acceptable surrogate with the usual caveats about upper-body-only drainage; a pulmonary artery sample is neither, and is the least representative of the three.

The targets the primary sources give are conventional and consistent: a mixed venous saturation above 60%, or a central venous saturation above 70%, with the note that low central venous saturations correlate with increased mortality on venoarterial ECMO. A persistently low value should prompt optimisation of oxygen delivery and a search for raised consumption — shivering, fever, agitation, seizure, or high work of breathing.

🚨

Danger · On venovenous ECMO, arterial saturation is a cardiac output monitor — and this inverts the usual reading

This is the most counter-intuitive consequence in the chapter, and it is worth following through a published case.

In a patient on venovenous support with a non-functioning lung, whatever fraction of venous return bypasses the circuit is a true right-to-left shunt. The shunt fraction can be estimated from saturations alone, and total cardiac output then follows from the circuit flow, which is known exactly.

Day 11, 15:00. Arterial saturation 74%, venous saturation 45%, circuit flow 6 L/min. Shunt fraction 0.47. Total cardiac output 11.3 L/min, of which 5.3 L/min bypassed the circuit.
Day 11, 21:00, after cooling from 36.8 to 35.9 °C. Arterial saturation 89%, venous saturation 63%, circuit flow unchanged at 6 L/min. Shunt fraction 0.30. Total cardiac output 8.6 L/min, bypassing flow 2.6 L/min.
Day 13. Arterial saturation 81%, venous saturation 46%, circuit flow 5.64 L/min. Shunt fraction 0.35, total cardiac output 8.7 L/min — against an echocardiographic estimate of 9.07 L/min.

Nothing was done to the machine. The hypoxaemia was corrected by lowering the cardiac output. A hyperdynamic septic circulation was outrunning a fixed circuit flow; cooling reduced both consumption and output, the shunt fraction fell, and arterial saturation rose by 15 points.

The inversion to carry to the bedside: on venovenous support, a falling arterial saturation at unchanged circuit flow, settings and recirculation is most often a rising cardiac output. Increasing the sweep gas will not fix it — the membrane is already fully saturating what reaches it. The lever is the patient's oxygen consumption and output, and Chapter 9 owns the full differential for persistent hypoxaemia.

⚠️

Pitfall · The two shunt equations in the same published table do not give the same answer

The case above reports the shunt fraction two ways at each time point: from oxygen contents, and from saturations alone with dissolved oxygen ignored.

At the first time point: 0.52 by content, 0.47 by saturation.
After cooling: 0.38 by content, 0.30 by saturation.

The derived cardiac outputs differ accordingly, and the case uses the saturation-based values. A five- to eight-point difference in shunt fraction is not trivial when the fraction sits in the denominator — at the second time point, using the content-based figure would give a total cardiac output of about 9.7 rather than 8.6 L/min.

Printed here uncorrected. The lesson is not that one method is wrong but that a derived flow inherits the uncertainty of every term used to derive it, and that the haemoglobin concentration — 7.5 g/dL in this case — is exactly the sort of term whose effect the saturation-only simplification discards.

25.8 Lactate, and the wash-out trap

⚠️

Pitfall · Lactate rises when flow is restored, and the rise is not a failure

The venoarterial monitoring table used throughout this chapter gives lactate as a trend and adds the warning that matters: "May see an initial rise in lactate due to tissue wash-out; may also be affected by liver and kidney injury."

The mechanism is the same one that makes reperfusion an intervention in Chapter 19. Tissue beds that were underperfused have been producing lactate and not exporting it. When flow is restored, that accumulated lactate is washed into the central circulation, and the measured concentration rises although production has fallen. The patient is improving and the number is worsening.

Three consequences.

Do not escalate flow on the basis of a lactate rise in the first hour or two after initiation, particularly after a prolonged low-flow state. Read the venous saturation and the drainage pressure instead.
The peak matters less than the slope after the peak. A lactate that rises and then falls steadily is the expected trajectory. A lactate that rises and plateaus is the concerning one.
Clearance is not constant. Hepatic hypoperfusion, and renal replacement therapy where lactate-containing or lactate-free fluids are used, both alter the number independently of production. Chapter 29 owns lactate as a laboratory value and Chapter 55 owns the renal-replacement interaction; this section owns only the warning not to read the early rise as failure.

💡

Clinical pearl · Two perfusion signs that survive the loss of a pulse, and one that does not

When pulsatility is minimal, much of the standard peripheral examination stops working — a capillary refill assessed on a non-pulsatile limb, a peripheral pulse that is not there to palpate. Two signs survive and one commonly misleads.

Skin mottling and capillary refill still carry information. One primary source records that these correlate with increasing morbidity and mortality specifically in ECMO patients — while stating, in the same passage, that physical examination alone will not identify inadequate perfusion or define a resuscitation target. Both are true: the signs are informative when abnormal and unreliable when normal.
Urine output survives, and is one of the few genuinely continuous perfusion signals on a circuit that no machine setting directly manufactures.
The limb with the arterial cannula does not count. Colour, warmth and capillary refill in that leg are dominated by the cannula and the distal perfusion catheter (Chapter 17), not by systemic perfusion. Assess the perfusion of the patient on the other side.

25.9 The macrocirculation is not the microcirculation

Everything above this point measures pressure and flow in large vessels. The purpose of pressure and flow in large vessels is to perfuse small ones, and the two have been measured together on venoarterial support.

📊

Evidence · Low certainty · What happens to the microcirculation when the pump starts

A systematic review searched four databases to PRISMA standards (registered CRD42021243930) and screened 1,215 records, of which 11 studies reported sublingual microcirculatory perfusion in adult patients on venoarterial support. Cardiogenic shock was the indication in eight.

Three studies reported increased perfused small vessel density, proportion of perfused vessels and microvascular flow index at 24 hours after initiation compared with before it. Microcirculatory perfusion then stabilised — it did not continue to improve.

Four studies out of four found higher perfused small vessel density and a higher proportion of perfused vessels in survivors than in non-survivors.

Eleven studies, heterogeneous, with no standard technique or threshold — but the direction is consistent, and it is the direction that matters here.

🧠

Physiology · The two circulations recover on different clocks, and only one of them is on the monitor

Turn the pump on and the mean arterial pressure moves within seconds. The microcirculation takes about a day, improves to a plateau, and then stops — and where that plateau sits separates the patients who live from the patients who do not, in every study that has looked.

This is the same dissociation that runs through modern shock physiology, arriving in a setting where it is unusually stark, because the macrocirculatory variable has been taken over by a machine. A mean arterial pressure of 70 mmHg on venoarterial ECMO is a statement about the pump and the vascular resistance. It is not a statement about whether capillaries are perfused, and on a circuit it is even less of one than usual.

The bedside form: a patient whose pressure and flow targets are met, whose lactate is not falling, and whose venous saturation is adequate, is describing a microcirculatory problem — and the macrocirculatory instruments will keep reporting success while it continues.

⚖️

Controversy 3 — Should the microcirculation be measured, and should it be a target?

The case for measuring it. It is the variable the whole apparatus exists to serve; it dissociates from the macrocirculation precisely in the patients who do badly; and it separates survivors from non-survivors in every included study. If a unit is going to chase a perfusion target at all, this is the one with the most direct claim to be the right one.

The case against, which is currently stronger. Four out of four studies showing an association in survivors is not evidence that acting on the measurement helps. The systematic review rests on 11 heterogeneous studies with no standardised acquisition, no agreed threshold and no intervention arm. And the one study that tested the measurement against a decision found nothing: in 30 paediatric patients undergoing weaning trials, 19 stable and 11 unstable after weaning, macrocirculatory and microcirculatory indices showed no differences between the groups. A measurement that does not discriminate at the decision point is not yet a monitor.

Where this book lands.

Do not make the microcirculation a target. There is no threshold to target and no evidence that targeting it changes anything.
Do use it as an explanation. When the macrocirculatory numbers are met and the patient is not improving, the dissociation above is the likeliest reason, and knowing that stops the futile escalation of a flow that is already adequate.
Treat the clinical surrogates as the practical version — mottling, capillary refill away from the cannulated limb, urine output, and the lactate slope after its early peak (§25.8).
Watch this space rather than buying a camera. The plateau finding, if it holds, would be an important early signal that further macrocirculatory escalation has no yield.

What would settle it. A trial in which microcirculatory measurement guides a management decision, with a clinical endpoint. None was identified in adults.

[VERIFICATION REQUIRED] — both the systematic review and the paediatric weaning study are known through abstracts.

25.10 The trade-off nobody can escape

Every haemodynamic decision on venoarterial support is a negotiation between pressure and unloading, and a simulation study puts numbers on it.

📊

Evidence · Low certainty · Flow buys pressure and costs volume

In a pressure-volume simulation with severe left ventricular failure, a normal right ventricle and a mean circulatory filling pressure of 7 mmHg:

At a circuit flow of 2 L/min with a systemic vascular resistance of 1,000 dynes·s·cm⁻⁵, mean arterial pressure was 68.9 mmHg.
Raising the flow to 4 L/min at the same resistance shifted the pressure-volume loop rightward: left ventricular end-diastolic volume rose from 150 to 180 mL.
Reducing the resistance to 800 at that higher flow gave an end-diastolic volume of 160 mL — still greater than at 2 L/min.
End-diastolic volume could be brought below the 2 L/min value only by reducing resistance to 600, which dropped mean arterial pressure to 61.1 mmHg.

A simulation, with the fixed relationships a simulation requires. But the structure it describes is not in doubt.

💡

Clinical pearl · You cannot increase flow and unload at the same time without paying in pressure

The three levers interact, and the monitoring set has to show all three at once or the trade-off is invisible:

More circuit flow buys mean arterial pressure and total oxygen delivery, and costs left ventricular volume — it raises afterload and distends.
Less vasoconstriction buys left ventricular volume back, and costs mean arterial pressure.
More inotropy buys ejection and pulse pressure, and costs myocardial oxygen consumption — which is the thing an infarcted ventricle has least of.

The practical instruction: whenever a flow change is made, record the pulse pressure, the pulmonary artery diastolic pressure if available, and the vasopressor dose before and after. A flow increase that raised the mean arterial pressure and halved the pulse pressure has not improved the patient; it has traded one variable for another, and the note should say so.

This chapter measures the trade-off. Chapter 15 owns the unloading decision and Chapter 54 owns the fluid and vasoactive strategy.

⚠️

Pitfall · Restoring flow is not restoring regional perfusion, and a porcine dataset shows the gap

In a porcine cardiogenic shock model, carotid blood flow fell from a baseline of 276 mL/min to 121 mL/min in shock. After four hours of venoarterial support it had recovered to 239 mL/min — better, and not back to baseline. In the arm also supported with a microaxial pump the figures were 253, 68 and 213 mL/min.

Over the same interval mean arterial pressure was restored essentially to baseline (70 to 42 to 65 mmHg in the venoarterial arm), and central venous pressure returned to its starting value.

The macrocirculatory numbers said the shock was treated. Carotid flow said roughly 13% of it was not. A single small animal study, and it is recorded here for its structure rather than its numbers: a restored mean arterial pressure is compatible with a regional flow deficit, and the monitor that shows the pressure is not the monitor that would show the deficit.

25.11 Building a haemodynamic monitoring set

The instruments are not interchangeable and they do not answer the same question. This is how they assemble.

flowchart TD
    A["Patient on ECMO<br/>haemodynamic question arises"] --> B{"Which question<br/>is being asked?"}
    B -->|"Is systemic perfusion<br/>adequate?"| C["MAP + total flow<br/>PREMEMBRANE venous saturation<br/>lactate SLOPE not value<br/>urine output, mottling off the cannulated limb"]
    B -->|"Is the native heart<br/>doing anything?"| D["Arterial waveform on a VERIFIED<br/>RIGHT RADIAL line<br/>Dicrotic notch present?<br/>Pulse pressure TREND<br/>Echo — Chapter 26"]
    B -->|"Is the circuit delivering<br/>what it claims?"| E["Flow probe DISTAL to all shunts<br/>Drainage pressure trend<br/>Pre- and post-oxygenator pressures<br/>Recirculation — Chapter 9"]
    C --> F{"Targets met but<br/>patient not improving?"}
    F -->|Yes| G["Suspect the MICROcirculation<br/>Do NOT escalate flow further<br/>Explanation, not a target"]
    D --> H{"Pulse pressure low?"}
    H -->|Yes| I{"Check the PULMONARY<br/>ARTERY side first"}
    I -->|"PA diastolic RISING<br/>PA pulsatility preserved"| J["LV DISTENSION physiology<br/>Unloading question — Chapter 15"]
    I -->|"PA diastolic LOW<br/>PA pulsatility LOW"| K["RV failure OR over-drainage<br/>Unloading the LV would HARM<br/>Check drainage pressure first"]
    I -->|"No PA catheter"| L["The discriminating measurement<br/>is not being made<br/>Echo, and say so in the note"]
    E --> M{"Drainage pressure<br/>increasingly negative at<br/>unchanged pump speed?"}
    M -->|Yes| N["Volume, tamponade, cannula position<br/>This fires BEFORE the arterial pressure moves"]

Step	What to do	Why it is in this order
1. Site the arterial line in the right arm	Right radial or right brachial on all peripheral venoarterial patients; the pulse oximeter probe on the right hand if the line cannot be sited there	Every other site reports the circuit's perfusion rather than the watershed, and the decision cannot be revisited later without a new procedure (§25.2)
2. Verify the line before believing the waveform	Fast-flush test; confirm the trace tracks the electrocardiogram; look for a dicrotic notch	A flat trace on an unverified line is a plumbing problem, and the commonest cause of a spurious "loss of ejection" (§25.2)
3. Name which of the three questions you are asking	Perfusion adequacy, native cardiac function, or circuit performance	Most bedside confusion is one question answered with another's instrument (§25.1)
4. Delete the thermodilution cardiac output	Keep the pulmonary artery catheter for pressures, pulsatility and the true venous saturation. Do not chart its cardiac output on venoarterial support	The indicator is stolen by the drainage cannula and pulmonary flow is not systemic flow. A wrong plausible number is worse than none (§25.5)
5. Draw the mixed venous sample from the premembrane port	Central venous as a surrogate; never the pulmonary artery on venoarterial support	The pulmonary artery sample is the residue that escaped the cannula, not a mixture (§25.7)
6. Read pulse pressure as a trend and a binary	Is there ejection, and is it changing? Not: is the number above a threshold	Two primary sources give non-overlapping targets and neither cites evidence (§25.3)
7. Before attributing a low pulse pressure to distension, look at the pulmonary artery side	Rising diastolic pressure with preserved pulsatility means distension. Low diastolic pressure with low pulsatility means the right ventricle or over-drainage	The commonest phenotype is low pulsatility without distension, and unloading it removes preload from an underfilled ventricle (§25.4)
8. Record every pressure with the flow that produced it	"Central venous pressure 8 at 4.2 L/min, drainage pressure −60 mmHg" — not "central venous pressure 8"	On a circuit the filling pressures are partly a function of a machine setting, which is why the primary sources decline to give absolute numbers (§25.6)
9. Read the lactate slope after its peak	Expect an early rise from wash-out; the concerning pattern is a rise that plateaus	Escalating flow for a wash-out rise treats a number rather than a patient (§25.8)
10. Record the trade-off whenever flow changes	Pulse pressure, pulmonary artery diastolic pressure and vasopressor dose, before and after	Flow buys pressure and costs ventricular volume; if all three are not recorded the trade is invisible (§25.10)

💡

Clinical pearl · The one-line handover that carries the haemodynamic state

A haemodynamic handover on ECMO that gives a blood pressure and a flow has given almost nothing. The minimum informative version names the three questions:

"Flow 4.2, mean pressure 68 on noradrenaline 0.15. Premembrane venous saturation 68%, lactate 3.1 and falling since the 06:00 peak of 5.4. Pulse pressure 18 on a right radial line with a dicrotic notch, down from 26 yesterday at the same flow. Pulmonary artery diastolic 24 and rising, pulmonary artery pulsatility preserved. Drainage pressure minus 55, stable."

That sentence states perfusion adequacy, native cardiac function and circuit performance separately; it gives the pressures with the flow that produced them; it distinguishes a falling pulse pressure at constant flow from one that followed a flow change; and it contains the two measurements that separate distension from right ventricular failure. Chapter 92 owns the handover template. This is what it has to carry for this chapter.

25.12 The errors that recur

Error	Why it is made	What to do instead
Charting a thermodilution cardiac output on venoarterial support	The catheter is in, the monitor displays a number, and the number is plausible	The indicator is partly drained into the circuit and pulmonary flow is not systemic flow. This book's own primary source says the readings are inaccurate in the same table that lists them. Keep the pressures; delete the output (§25.5)
Drawing the mixed venous sample from the pulmonary artery	That is where mixed venous blood is, in every other patient	On venoarterial support most venous return never gets there. The premembrane port is the mixed venous sample (§25.7)
Siting the arterial line on the left or in the femoral artery	It is easier, and the cannulated groin is already exposed	Right radial or right brachial. A left-sided line measures the circuit's perfusion, not the watershed, and will look reassuring during upper-body hypoxia (§25.2)
Concluding that the ventricle has stopped ejecting from a flat trace	The waveform is flat and the conclusion is available	Verify the line first — fast flush, does it track the electrocardiogram, is there a dicrotic notch. A damped femoral line in a leg with a distal perfusion cannula flattens for mechanical reasons (§25.2)
Unloading the left ventricle because the pulse pressure is low	Low pulse pressure and distension are taught as the same thing, and simulations predict that they are	In 128 measured patients the commonest phenotype was low pulsatility without distension, associated with a larger right ventricle and tricuspid regurgitation. Check the pulmonary artery diastolic pressure and pulsatility before concluding (§25.4)
Escalating flow because lactate rose in the first hours	Rising lactate means worsening shock, everywhere else	Expect a wash-out rise when a low-flow state is reperfused. Read the venous saturation and the drainage pressure, and judge lactate on its slope after the peak (§25.8)
Reading a good mean arterial pressure as evidence of cardiac recovery	The number is the one everyone watches, and it improved	On venoarterial support the mean arterial pressure is a statement about the pump and the vascular resistance. It is question 1 answering question 2 (§25.1)
Using pulse pressure variation or stroke volume variation	The monitor offers them and they are familiar	They require an arterial waveform generated solely by a ventricle whose preload the ventilator is modulating. On a circuit, preload is being modulated by a pump. Uninterpretable, not merely less accurate (§25.1)
Trusting a normal pulse oximeter reading	It is continuous, non-invasive and looks fine	On ECMO the discrepancy from measured saturation runs 3 to 7.2% higher, plausibly from haemolysis-derived carbon monoxide plus reduced pulsatility — and it is least trustworthy when the circuit is worst. Measure the saturation (§25.2)
Recording a central venous pressure without the flow	It is one number in one box on the chart	On a circuit the filling pressure is right-sided filling minus what the cannula removed, and the second term is a machine setting. Record both, and accept that there is no absolute target (§25.6)
Escalating flow when every macrocirculatory target is already met	The patient is not improving and flow is the available lever	The microcirculation improves for about 24 hours and then plateaus, and its state separates survivors. Met targets with no improvement is a microcirculatory description, not an argument for more flow (§25.9)

25.13 Key points

Every haemodynamic instrument at the bedside assumes one pump, in series, generating a pulse. Venoarterial ECMO creates two pumps in parallel; venovenous ECMO inserts a shunt in series. The instruments go on producing numbers and none of them announces that its assumptions have failed (§25.1).
Ask which of three questions you are asking before choosing an instrument. Is systemic perfusion adequate; is the native heart doing anything; is the circuit delivering what it claims. They have different instruments and different time courses, and most bedside confusion is one question answered with another's tool — above all, reading a good mean arterial pressure as evidence that the heart is recovering (§25.1).
The arterial line goes in the right arm on peripheral venoarterial support, because the right radial territory is the last the circuit reaches and the first to reveal upper-body hypoxia. A line anywhere else measures the circuit's perfusion. And the pulse oximeter reads 3 to 7.2% high on a circuit, plausibly from haemolysis-derived carbon monoxide plus reduced pulsatility — least trustworthy when the circuit is worst (§25.2).
Pulse pressure has no agreed target. Two of this book's primary sources give 10 to 20 mmHg and over 25 mmHg, non-overlapping, neither citing evidence. Treat the presence of ejection as the binary that matters, confirmed by a dicrotic notch on a verified waveform, and read the trend at constant flow rather than the value (§25.3).
A falling pulse pressure after a flow increase is the expected mechanical consequence of that flow increase — preload falls, afterload rises, and the arterial tree is already pressurised. It is evidence about the circuit's effect on the heart, not evidence that the heart has deteriorated. The converse also holds: a heart that looks better when flow comes down has been allowed to fill, not necessarily recovered (§25.3).
Low pulse pressure most often does not mean left ventricular distension. In 58 pigs and 128 patients, three phenotypes emerged: pulsatile 26%, low pulsatility 52%, distension 22%. The commonest was low pulsatility without distension, with a larger right ventricle and more tricuspid regurgitation. The discriminating measurement is on the pulmonary artery side, and it is usually not being made (§25.4).
Thermodilution cardiac output is invalid on venoarterial support — the indicator is partly drained into the circuit, pulmonary blood flow is not systemic blood flow, and the heat exchanger sets the baseline. Keep the catheter for pressures, pulsatility and the true venous saturation; delete the cardiac output from the chart (§25.5).
Native cardiac output can be estimated three ways and none is routine. The modified Fick principle is the most general and the only one that works with gas exchange running; the shunt-fraction method is the most executable at the bedside; modified thermodilution measures recirculation well but gives a cardiac output with limits of agreement of roughly ±3 L/min — accuracy without precision is not a measurement. All three are invalidated by differential hypoxia, which on the venous side is as common as on the arterial side and far less often recognised (§25.5).
On venoarterial support the true mixed venous sample is the premembrane port, not the pulmonary artery. On venovenous support the inversion is larger still: arterial saturation is a cardiac output monitor, and in the published case a 15-point rise in arterial saturation was produced by cooling the patient — that is, by lowering cardiac output — with no change to the machine (§25.7).
The macrocirculation is restored in seconds and the microcirculation over about a day, after which it plateaus — and where that plateau sits separated survivors from non-survivors in every included study. Met macrocirculatory targets with no clinical improvement is a microcirculatory description, not an argument for more flow. Do not make it a target: the one study that tested it at a decision point found no difference between stable and unstable weaning groups (§25.9).

Cross-references

Inherited from earlier chapters

Chapter 13 — VA ECMO Haemodynamics. The afterload argument, the parallel-circulation architecture and the mixing point. Chapter 13 explains why the circulation behaves this way; this chapter explains what the instruments therefore read.
Chapter 15 — LV Distension and LV Unloading. §25.4 supplies the discrimination that should precede the unloading decision; the decision itself is Chapter 15's and is not re-argued here.
Chapter 16 — Differential Hypoxaemia. Upstream of this whole chapter: it dictates the arterial line site (§25.2), and venous differential hypoxia invalidates every derived cardiac output in §25.5.
Chapter 6 — VV ECMO Physiology, and Chapter 9 — Persistent Hypoxaemia on VV ECMO. The shunt physiology that §25.7 converts into a flow measurement, and the full differential for the falling arterial saturation it describes.
Chapter 17 — Limb Ischaemia. The reason the cannulated limb is excluded from the peripheral perfusion examination (§25.8).
Chapter 19 — ECPR Principles. Reperfusion as an intervention, which is the mechanism behind the lactate wash-out rise (§25.8).
Chapter 24 — ECPR Failure and Withdrawal. Its §24.1 recorded that refractory shock is the earliest mode of death and the one the circuit is best at masking. This chapter is how that masking is detected.

Handed forward

Chapter 26 — Echocardiography During ECMO. The independent cross-check that §25.5 requires, the discrimination between hypovolaemia, tamponade and cannula malposition that §25.6 defers, and the imaging assessment of distension.
Chapter 27 — Blood Gas and Oxygenation Monitoring. Gas-exchange interpretation, oxygenation targets, and the sampling discipline that §25.7's venous saturation depends on.
Chapter 28 — Cerebral Monitoring. Regional perfusion, and the carotid-flow deficit that §25.10 records as persisting after the pressure was restored.
Chapter 29 — Laboratory Monitoring. Lactate as a laboratory value and the haemolysis panel that §25.2's oximetry error implicates.
Chapter 30 — ECMO Data Interpretation. Trends, derived indices and the class of failure that §25.4's implausible confidence intervals belong to.
Chapter 32, Chapter 34 and Chapter 38 — circuit complications. The device-diagnostic reading of the pressures that §25.6 borrows only for their volume information.
Chapter 54 — Fluid and Haemodynamic Management. Everything this chapter declines to do. This chapter ends at a correctly interpreted number.
Chapter 18 — VA ECMO Weaning, and Chapter 68 — Paediatric Considerations. The paediatric microcirculation-at-weaning study in §25.9 is held there.
Chapter 92 — ECMO Handover Template. The three-question handover sentence in §25.11.

References

⚠️

How to read this list

Every external reference in this chapter was read as a structured abstract or as an indexed full-text passage. No article was retrieved and read in full. Digital object identifiers are reproduced as they appeared in the search record and have not been independently resolved; treat them as leads rather than as verified citations. Numerical values are those printed in the source passage. Every database row seeded from this chapter carries Verified = No.

The textbook material in this chapter is quoted from the project's primary sources and is the most directly actionable content here — and it is also the source of the chapter's central contradiction (§25.3).

Cardiac output and flow measurement on a circuit

Bachmann KF, Haenggi M, Jakob SM, Takala J, Gattinoni L, Berger D. Comment on: "A novel 'shunt fraction' method to derive native cardiac output during liberation from central VA ECMO" by Lim HS. ESC Heart Failure. 2024;11(4):2464–2466. DOI 10.1002/ehf2.14804. The published exchange that frames Controversy 2, and the source of the statement that venous differential hypoxia is as common and as important as the arterial form.
Lim HS. A novel "shunt fraction" method to derive native cardiac output during liberation from central VA ECMO. ESC Heart Failure. 2023;11(1):570–573. DOI 10.1002/ehf2.14441. The sweep-off shunt method and the three worked calculations quoted in §25.5.
Lim H. The physiology of extracorporeal membrane oxygenation: the Fick principle. Perfusion. 2023;38:236–244. DOI 10.1177/02676591211055971.
Bachmann KF, Berger D, Moller PW. Interactions between extracorporeal support and the cardiopulmonary system. Frontiers in Physiology. 2023;14:1231016. DOI 10.3389/fphys.2023.1231016.
Bachmann KF, Haenggi M, Jakob SM, Takala J, Gattinoni L, Berger D. Gas exchange calculation may estimate changes in pulmonary blood flow during venoarterial extracorporeal membrane oxygenation in a porcine model. American Journal of Physiology: Lung Cellular and Molecular Physiology. 2020;318:L1211–L1221. DOI 10.1152/ajplung.00167.2019. The original derivation of the modified Fick principle.
Bachmann KF, Vasireddy R, Heinisch PP, Jenni H, Vogt A, Berger D. Estimating cardiac output based on gas exchange during venoarterial extracorporeal membrane oxygenation in a simulation study using paediatric oxygenators. Scientific Reports. 2021;11:11528. DOI 10.1038/s41598-021-90747-w.
Berger DC, Zwicker L, Nettelbeck K, Casoni D, Heinisch PP, Jenni H, et al. Integral assessment of gas exchange during venoarterial ECMO: accuracy and precision of a modified Fick principle in a porcine model. American Journal of Physiology: Lung Cellular and Molecular Physiology. 2023;324:L102–L113. DOI 10.1152/ajplung.00045.2022. 16 animals, over 1,500 blood gas analyses.
Bachmann KF, Zwicker L, Nettelbeck K, Casoni D, Heinisch PP, Jenni H, et al. Assessment of right heart function during extracorporeal therapy by modified thermodilution in a porcine model. Anesthesiology. 2020;133:879–891. DOI 10.1097/ALN.0000000000003443.
Modified thermodilution for simultaneous cardiac output and recirculation assessment in veno-venous extracorporeal membrane oxygenation: a prospective diagnostic accuracy study. Anesthesiology. 2024. DOI 10.1097/ALN.0000000000004895. In vitro. Recirculation measured well; cardiac output "accurate but imprecise," bias 0.56 L/min with limits of agreement −2.27 to 3.4 L/min.
Stadlen R, Singhal AK, Reed RM, Hasday JD, Bates ML, Schmidt GA, Eberlein M. Management of two circulations in a COVID-19 patient with secondary superinfection. Physiological Reports. 2023;11. DOI 10.14814/phy2.15602. The venovenous worked example in §25.7, including the echocardiographic cross-check and the two discordant shunt equations flagged in the chapter status.

Pulse pressure, phenotypes and the pressure-volume trade-off

Lim HS, Vondrakova D, Micek M, Ostadal P. Phenotyping hemodynamic response to veno-arterial extracorporeal membrane oxygenation in cardiogenic shock. Physiological Reports. 2026;14(11). DOI 10.14814/phy2.70961. 58 pigs and 128 patients; the three phenotypes that govern §25.4 and the most practice-changing reference in this chapter.
Lee SI, Lim YS, Park C, Choi WS, Choi CH. Importance of pulse pressure after extracorporeal cardiopulmonary resuscitation. Journal of Cardiac Surgery. 2021;36(8):2743–2750. DOI 10.1111/jocs.15614. Loss of pulse pressure independently associated with failure to wean, odds ratio 0.039 (0.006–0.275). The same model's pH estimates are flagged in the chapter status and are not used.
Kalra R, Alexy T, Bartos JA, Prisco AR, Kosmopoulos M, Maharaj VR, et al. Left ventricular haemodynamics with venoarterial extracorporeal membrane oxygenation. Catheterization and Cardiovascular Interventions. 2024. DOI 10.1002/ccd.30951. The estimated pressure-volume loop protocol, and the statement that pulmonary artery catheters are not routinely inserted during venoarterial support at that centre.
Dickstein ML. Direct versus indirect effects of extracorporeal membrane oxygenation. Catheterization and Cardiovascular Interventions. 2024;103(7):1173. DOI 10.1002/ccd.31034. The source of the arterial-volume argument in §25.3 and of the claim that improvement on weaning is compensatory rather than direct.
Wu EL, Boesch J, Nonaka H, Djelovic C, Wickramarachchi A, Benitez J, Fraser JF. Comparison of left ventricular pressure-volume loops with veno-arterial extracorporeal membrane oxygenation. Artificial Organs. 2026. DOI 10.1111/aor.70163. The simulation quantifying the flow-against-volume trade-off in §25.10.
Frederiksen PH, Linde L, Gregers E, Udesen NL, Helgestad OK, Banke A, et al. ESC Heart Failure. 2024. DOI 10.1002/ehf2.14780. Porcine venoarterial against ECMELLA; the pulse-pressure and carotid-flow figures in §25.3 and §25.10.

Perfusion adequacy and the microcirculation

Volleman C, Raasveld SJ, Jamaludin FS, Vlaar AP, van den Brom CE. Microcirculatory perfusion disturbances during veno-arterial extracorporeal membrane oxygenation: a systematic review. Microcirculation. 2024. DOI 10.1111/micc.12891. PROSPERO CRD42021243930. 1,215 records screened, 11 studies; improvement to 24 hours then a plateau, and higher perfused vessel density in survivors in four of four studies.
Suc V, Starck J, Levy Y, Soreze Y, Rambaud J, Léger P. Predictive value of microcirculation for pediatric extracorporeal membrane oxygenation weaning test: a monocentric study. Artificial Organs. 2024. DOI 10.1111/aor.14754. 30 children; macrocirculatory and microcirculatory indices showed no difference between stable and unstable weaning groups. Held largely for Chapter 68 and Chapter 18.
Ameloot K, Van De Vijver K, Broch O, Van Regenmortel N, De laet I, Schoonheydt K, et al. Nexfin noninvasive continuous haemodynamic monitoring: validation against continuous pulse contour and intermittent transpulmonary thermodilution derived cardiac output in critically ill patients. Critical Care Research and Practice. 2013. DOI 10.1155/2013/519080. Not an ECMO study. Included for calibration: 36% error against thermodilution overall, and R² of 0.01 in the low-cardiac-output subgroup — non-invasive output estimation failing where it is most needed.
Rossetti M, Capuano P, Mamone G, Abdulaziz S, Ghisulal P, Panarello G, et al. Diagnostic imaging challenges during extracorporeal membrane oxygenation. Artificial Organs. 2026;50(5):647–659. DOI 10.1111/aor.70092. Transoesophageal echocardiography for watershed identification, and the practical instruction to reduce circuit flow — to no lower than 1.5 L/min — to move the mixing zone. Held largely for Chapter 26.
Reddan T, Venugopal PS, Powell J, Mattke AC. Ultrasonographic assessment of aortic flow characteristics in a paediatric patient with sepsis treated with extracorporeal life support: defining the mixing zone. Australasian Journal of Ultrasound in Medicine. 2020;23(4):255–263. DOI 10.1002/ajum.12206. Held for Chapter 16 and Chapter 26.

Textbooks — the primary sources for this chapter

Taha AR, Caridi-Scheible M, Leiendecker E, et al. ECMO: A Practical Guide to Management, Chapter 9, Table 9.1 — the monitoring-modality table: pulse oximetry accuracy on ECMO, the right-arm arterial line, the goal mean arterial pressure over 65 mmHg and goal pulse pressure over 25 mmHg, the pulmonary artery catheter row including "PAC readings are inaccurate due to redirection of pulmonary blood flow from VA ECMO flow", the in-line central venous saturation row, and the capillary refill and mottling material. [VERIFICATION REQUIRED] — page numbers not confirmed.
Shinar Z, Badulak J. ECPR and Resuscitative ECMO, Chapter 7, Table 1 — Monitoring the V-A ECMO patient: central venous pressure and pulmonary artery pressure trends with "no recommendation for absolute numbers", mixed venous saturation over 60%, lactate as a trend with the wash-out warning, and pulse pressure 10 to 20 mmHg. [VERIFICATION REQUIRED].
ISCCM Manual of RRT and ECMO in ICU, Chapter 33 — Monitoring during Extracorporeal Membrane Oxygenation: the ELSO circuit-pressure monitoring recommendation (access line, pre- and post-oxygenator), the flow probe distal to all shunts, negative drainage pressure as a volume signal, the transmembrane pressure patterns, the right-arm sampling instruction, the statement that venous saturation on venoarterial ECMO can be measured from a central line or premembrane blood gases, and the oxygen delivery to consumption ratio of about three. [VERIFICATION REQUIRED].
ELSO Red Book, 6th edition — consulted for monitoring standards; not separately quoted in this chapter. [VERIFICATION REQUIRED].

📅

Chapter status

Drafted and audited 15 September 2026. Ten-pass quality control completed: clinical, physiology, evidence, citation, numerical, safety, contradiction, redundancy, bedside utility and literature-currency passes. This chapter opens Part V.

A research-tooling change is recorded here for the first time. The structured-abstract search service used for Chapters 19 to 24 exhausted its monthly quota during this chapter's research. Evidence was gathered instead through a biomedical index and a full-text passage-retrieval service, which returned indexed passages with resolvable identifiers rather than structured abstracts. The provenance is therefore better than in Chapters 19 to 24 — passages are quoted from the article body rather than its abstract — but no full text was read end to end, and identifiers remain unresolved. The Verified = No convention is unchanged.

The organising insight is architectural: every haemodynamic instrument at the bedside assumes one pump, in series, generating a pulse. Venoarterial ECMO creates two pumps in parallel; venovenous ECMO inserts a shunt in series. The instruments continue to produce numbers, and none of them announces that its assumptions have failed. The chapter is organised around the three questions those instruments can answer — perfusion adequacy, native cardiac function, circuit performance — on the claim that most bedside confusion is one question answered with another's tool.

The most practice-changing finding is that a low pulse pressure most often does not indicate left ventricular distension. Simulation predicts uniform distension; measurement in 58 pigs and 128 patients found three phenotypes, of which the commonest — 52% — was low pulsatility without distension, accompanied by a larger right ventricle and tricuspid regurgitation. The discriminating measurement is the pulmonary artery diastolic pressure and pulsatility, and it is frequently not being made. Unloading that phenotype removes preload from a ventricle that is already underfilled.

Three controversies were set out rather than smoothed. §25.3 asks what pulse pressure a venoarterial patient should have and finds this book's own primary sources in direct conflict. §25.5 asks whether native cardiac output can be measured at the bedside and records a published exchange between two research groups, landing on the modified Fick principle as most general and the shunt-fraction method as most executable. §25.9 asks whether the microcirculation should be a target and declines, on the grounds that no threshold exists and the one study testing it at a decision point found nothing.

The numerical and contradiction audits flagged six problems and corrected none. (1) Two primary textbooks give non-overlapping pulse-pressure targets — 10 to 20 mmHg and over 25 mmHg — and neither cites evidence, so a patient at 22 mmHg is simultaneously above one target and below the other. (2) A multivariable model reports odds ratios of 339 (1.850–62,124) and 221,457 (2.385–over twenty billion) for arterial pH, labelled significant; the same model's pulse-pressure estimate is credible and is used, its pH estimates are printed and are not. (3) A published case computes the shunt fraction two ways in the same table — 0.52 against 0.47, and 0.38 against 0.30 — and the derived cardiac output differs by more than a litre per minute depending on which is used. (4) Modified thermodilution reports a cardiac output "accurate but imprecise," with limits of agreement of roughly ±3 L/min around an output of 3 L/min — a headline bias figure far more reassuring than its interval. (5) A monitoring table lists cardiac output and cardiac index among a pulmonary artery catheter's measurements and states in the same row that its readings are inaccurate on venoarterial support. (6) A non-invasive cardiac output device reports a 36% error overall and an R² of 0.01, not significant, in the low-output subgroup — failing precisely where the measurement is needed.

The redundancy audit removed the parallel-circulation physiology, which belongs to Chapter 13; the unloading decision, which belongs to Chapter 15; differential hypoxaemia, which belongs to Chapter 16; echocardiographic technique, which belongs to Chapter 26; gas-exchange interpretation, which belongs to Chapter 27; lactate as a laboratory value, which belongs to Chapter 29; circuit pressures as device diagnostics, which belong to Chapters 32, 34 and 38; and every therapeutic response, which belongs to Chapter 54. This chapter ends at a correctly interpreted number.