Extracorporeal Circulatory and Respiratory Support Physiology

An extracorporeal circuit adds a new pathway for blood, gas exchange, and mechanical energy. Its effect depends on where blood is drained, where it returns, and how it interacts with the native heart and lungs. A satisfactory circuit display therefore cannot establish adequate regional oxygen delivery, favorable ventricular loading, or recovery of the organs being supported.

79.1 Begin with the Blood Path Rather Than the Device Name

Extracorporeal membrane oxygenation, or ECMO, commonly combines venous drainage, a pump, a membrane gas exchanger, and a return pathway. The membrane can add oxygen and remove carbon dioxide, while the pump supplies energy to move blood through the circuit. These functions must be distinguished from the route by which that blood reaches tissues.

In venovenous support, abbreviated VV, blood is returned to the venous circulation. It still passes through the right heart, pulmonary circulation, and left heart before reaching systemic arteries. The circuit provides gas exchange but does not directly replace systemic cardiac pumping. Gas correction and changes in ventilation can nevertheless alter ventricular load and improve secondary circulatory dysfunction.

In venoarterial support, abbreviated VA, drained venous blood returns to an artery. The circuit can contribute directly to systemic flow while bypassing part of the right-heart and pulmonary pathway. Native ventricular ejection and extracorporeal arterial return may then coexist, carrying blood with different oxygen contents and reaching different arterial territories.

Hybrid arrangements divide return between arterial and venous sites or use additional drainage pathways. Their physiological effects depend on the actual branch flows and anatomical connections. A name alone cannot specify how much support reaches each pathway or how much blood recirculates.

The distinction also clarifies the meaning of cardiac output. During VV support, circuit flow is not an additional systemic output to be added to native cardiac output. Much of the same blood traverses both pathways. In a simple VA arrangement, native forward ejection and circuit arterial return can contribute in parallel, although shunts, valve regurgitation, and additional devices complicate summation.

The initial description should therefore identify drainage and return sites, measured flows, remaining native paths, and gas-exchanger function. The same pump flow can have different circulatory consequences after a change in configuration, native cardiac performance, or lung function.

79.2 Oxygenator Transfer Is a Content Difference Multiplied by Flow

Oxygen transfer across the membrane depends on blood flow, inlet oxygen content, hemoglobin availability, gas-side conditions, and the membrane's transfer properties. A high outlet oxygen tension demonstrates one aspect of gas-exchanger performance, but oxygen tension alone does not state how much oxygen was added per minute.

For appropriately paired samples under sufficiently stable conditions:

QE is circuit blood flow in L/min. Cpost and Cpre are postmembrane and premembrane oxygen contents in mL O₂/dL. The factor 10 converts liters to deciliters, producing transfer in mL O₂/min. Contents include hemoglobin-bound and dissolved oxygen, using the actual measured blood conditions.

Assume a constructed circuit flow of 3.5 L/min, inlet content 14 mL/dL, and outlet content 20 mL/dL. Net membrane transfer is 3.5 × 6 × 10 = 210 mL O₂/min. The outlet content flow is 700 mL/min, but most of that oxygen was already present before the blood entered the membrane. Calling all 700 newly supplied oxygen would count preexisting oxygen again.

Increasing blood flow can increase total transfer even if the content increment per deciliter becomes smaller as membrane transit time shortens. Conversely, once outlet hemoglobin is nearly saturated, increasing gas oxygen concentration mainly affects the smaller dissolved component. The result depends on hemoglobin, inlet content, membrane performance, and the operating flow.

Schmidt and colleagues altered circuit conditions in ten adults with severe ARDS receiving femoro-jugular VV support. Circuit blood flow and gas oxygen fraction influenced systemic oxygenation, while sweep flow strongly influenced carbon dioxide removal. Their physiological observations explain major determinants without defining a universal flow-to-output ratio or a transfusion target. 1

Transfer across the membrane still differs from delivery to organs. Native gas exchange, recirculation, systemic flow, mixing, and demand determine what happens after oxygenated blood leaves the circuit.

79.3 VV Oxygenation Depends on Effective Flow and Recirculation

In VV support, some returned oxygenated blood may be drained back into the circuit before traversing the systemic circulation. This recirculation uses pump capacity without exposing the same fraction of fresh systemic venous blood to the gas exchanger. Its magnitude depends on cannula geometry, return-jet direction, drainage conditions, cardiac flow, and the patient's position and anatomy.

A useful flow-accounting convention is:

Here QE is total circuit flow in L/min and r is the recirculated fraction, a dimensionless number between zero and one. If a constructed circuit runs at 4 L/min with r = 0.25, effective flow is 4 × 0.75 = 3 L/min. This defines the flow not immediately recycled through the circuit; it is not a direct measurement of systemic oxygen delivery.

Fisser and colleagues used ultrasound dilution to study recirculation in two VV configurations in a nonrandomized two-center investigation. Different displayed flows could produce similar effective flows because recirculation differed. Configuration and center-related factors were not randomized, so the findings do not prove one arrangement universally superior or establish a cannula-position recipe. 2

Effective flow relative to native cardiac output helps determine how much venous blood is treated before it reaches the arterial circulation. If native output and metabolic demand rise without a comparable change in effective circuit support, arterial saturation may fall. That saturation change does not automatically mean systemic oxygen delivery fell, because delivery also depends on total forward flow and hemoglobin content.

Recirculation estimates themselves require a valid method. A very oxygenated drainage sample can suggest recycling, but also depends on systemic extraction and where venous streams mix. A saturation-based calculation that assumes an inaccessible true venous reference may be misleading.

The net content-difference measurement across the oxygenator already reflects its inlet conditions. Applying an additional arbitrary recirculation correction to that measured oxygen-transfer flux can double-correct the same phenomenon. Effective-flow accounting and measured gas transfer answer related but different questions.

79.4 Sweep Gas Changes Carbon Dioxide Removal Through a Separate Pathway

Carbon dioxide crosses the membrane according to blood-side delivery and the gas-side gradient. Sweep gas carries it away and maintains that gradient. Increasing sweep can therefore increase removal without the same change in oxygen delivery produced by increasing blood flow.

This separation is conditional. Membrane area, blood flow, inlet carbon dioxide content, blood chemistry, and gas-side conditions limit transfer. At sufficiently low blood flow, the amount delivered to the membrane becomes important. Sweep is not an unlimited independent controller, and arterial carbon dioxide also reflects native alveolar ventilation, production, stores, and time.

Extracorporeal carbon dioxide removal can permit changes in native ventilation. Lower ventilatory exposure may reduce some mechanical loads, but lower assistance or changed respiratory drive can also alter effort, recruitment, and circulation. Gas removal itself does not guarantee that the resulting ventilatory strategy benefits the patient.

Large changes in carbon dioxide additionally affect cerebral vascular tone and acid–base conditions. Cavayas and colleagues associated a larger early carbon dioxide decrease with neurological complications in a retrospective registry cohort receiving extracorporeal support for respiratory failure. The observation does not establish a causal correction-rate limit or one safe sweep-adjustment rule. 3

REST tested lower tidal-volume ventilation facilitated by extracorporeal carbon dioxide removal against standard low tidal-volume care in acute hypoxemic respiratory failure. The trial stopped early for futility and feasibility, did not significantly reduce 90-day mortality, and reported more serious adverse events with the extracorporeal strategy. This concerns the complete tested strategy and population, rather than proving that carbon dioxide removal never has physiological value. 4

Systems providing relatively low-flow carbon dioxide removal should not be assumed to provide equivalent oxygenation or circulatory support to another ECMO configuration. The required transport depends on blood content and flow, while the clinical value depends on what native support is changed and what complications are introduced.

79.5 Drainage Sets a Boundary on the Flow the Pump Can Deliver

A centrifugal pump's rotational speed is not a blood-flow measurement. Flow depends on the pressure difference the pump must generate and the resistance and availability of its inlet and outlet pathways. At the same speed, changes in vascular pressure, tubing resistance, cannula position, or venous drainage can change actual flow.

Venous drainage is particularly vulnerable because the vessel can collapse when the pressure within it becomes too low relative to surrounding pressure. Increasing suction can then narrow the very pathway needed to supply the pump. Intermittent collapse can produce unstable flow rather than a useful sustained increase.

Low drainage flow can reflect inadequate effective venous filling, but also a cannula against the vessel wall, altered position, obstruction, increased thoracic or abdominal pressure, or other limitations. It should not automatically be interpreted as a total-body volume deficit. Added volume may increase congestion without correcting the relevant anatomical or pressure problem.

The pressure reference matters. An inlet pressure recorded relative to atmosphere is not the same as the pressure distending the local vein. Patient position, transducer level, surrounding pressure, and the exact measurement location influence its interpretation. A fixed suction number cannot represent the same vascular condition in every patient.

Return-side resistance also matters. Arterial pressure, cannula dimensions, tubing, and membrane resistance contribute to the pressure head required for a given flow. A change in pressure across a component may identify altered resistance, but comparisons require comparable flow, viscosity, temperature, and measurement conditions.

These constraints connect circuit performance to patient physiology. Hypovolemia, vasodilation, impaired venous return, edema, and external pressure may each alter drainage differently. The response should be understood through actual flow, pressure behavior, position, cardiac filling, and congestion rather than increasing speed or volume in response to one display.

The pump can move only the blood its inlet supplies through the load presented at its outlet. Treating flow as a fixed command obscures both limitations and can expose blood and vessels to greater mechanical stress without increasing useful support.

79.6 VA Support Can Improve Systemic Pressure While Increasing Left Ventricular Load

VA drainage diverts part of venous return away from the right heart and lungs, reducing some native filling. Arterial return simultaneously supplies systemic flow and changes the pressure against which the left ventricle ejects. These effects can move in opposite directions.

Peripheral arterial return can raise aortic pressure and impede native ejection when left ventricular contractile reserve is poor. More volume can remain at end systole, while continued pulmonary, bronchial, or other venous return still supplies the left heart. Reduced right-heart flow does not mean the left ventricle receives no blood.

If inflow continues while ejection is inadequate, ventricular and atrial pressures can rise, pulmonary congestion can worsen, and blood can stagnate in poorly emptied chambers or the aortic root. Valve disease modifies these paths. Aortic regurgitation, for example, can return arterial blood toward the left ventricle rather than permit every unit of circuit return to represent useful forward delivery.

Ostadal and colleagues studied VA flow changes in five swine with experimentally induced cardiogenic shock. Greater circuit flow improved arterial pressure while increasing end-systolic volume and ventricular stroke work and reducing native ejection. End-diastolic pressure and volume did not significantly change in that experiment. It demonstrates altered load, rather than proving that every increase immediately produces clinical ventricular distension. 5

In a prospective flow-variation study of 80 stabilized patients receiving femoro-femoral VA support, Saura and colleagues found wedge pressure usually remained stable or fell despite increased left ventricular afterload. Reduced right-heart filling could counterbalance the arterial load. The human observations reinforce that ventricular loading and pulmonary congestion cannot be predicted from pump flow alone. 6

An apparently stronger arterial pressure therefore does not establish favorable myocardial conditions. Native pulse pressure and aortic-valve opening provide clues, but their relation to forward flow depends on loading, arterial properties, rhythm, regurgitation, and any additional device.

Conversely, simply reducing circuit flow to encourage ejection may sacrifice systemic perfusion in a patient who still requires support. The competing demands are sufficient organ flow and tolerable ventricular loading. They cannot be resolved through one universal arterial pressure, pump flow, or requirement for a particular pulse pattern.

79.7 Arterial Return Creates Regional Mixing Rather Than One Uniform Blood Gas

During peripheral VA support, retrograde circuit return meets blood ejected from the left ventricle. The location and pattern of their interaction depend on relative flows, arterial anatomy, pulsatility, and time within the cycle. It is not a fixed mixing point that can be inferred from one pump setting.

If native lungs remain poorly exchanging while cardiac ejection improves, proximal arteries may receive relatively deoxygenated native blood while more distal territories receive well-oxygenated circuit blood. Thus, improved ventricular ejection can expose differential hypoxemia rather than guarantee improved oxygen delivery to every organ.

Venous drainage also influences the pattern. Hou and colleagues created differential oxygenation in a sheep respiratory-failure model receiving femoral VA support and altered it by changing drainage and return pathways. Their observations show that differential venous streams, not only arterial return location, can sustain the separation. The animal preparation does not define a universal configuration change. 7

Cai and colleagues subsequently measured reversible changes in upper-body oxygenation when drainage location was altered in seventeen adults receiving femoral VA support. This small physiological study supports a human pathway effect; its short-term oxygenation findings do not demonstrate a survival advantage or establish a bedside cannula-repositioning protocol. 8

Sampling must match the question. A femoral arterial gas may not represent proximal cerebral or coronary inflow. A right radial sample can help assess an upper-body territory, but it cannot certify the oxygen content reaching every coronary or cerebral branch under all mixing conditions.

For a constructed whole-flow accounting example, assume circuit arterial return of 4 L/min at 20 mL O₂/dL and native forward flow of 2 L/min at 12 mL/dL. Summed oxygen-content flow is 1040 mL/min. Dividing by the total flow in dL/min gives 1040 ÷ (6 × 10) ≈ 17.3 mL/dL as a hypothetical average content. That average does not mean every territory receives 17.3: spatially segregated streams could expose different organs to markedly different contents.

79.8 Unloading Changes the Pathway but Introduces Additional Tradeoffs

Left-heart decompression can reduce filling pressure or remove blood that an inadequately ejecting ventricle cannot clear. Different methods act at different boundaries. Draining the left atrium reduces inflow to the ventricle, whereas a ventricular-to-arterial pump directly transfers blood from the ventricle into the arterial circulation.

Reducing chamber pressure and pulmonary congestion can improve the mechanical environment, but each additional pathway introduces new flow dependencies, access requirements, blood contact, and possible injury. Excess drainage can also reduce chamber filling and produce suction-related limitations. Methods cannot be treated as interchangeable simply because all are described as unloading.

The balance includes myocardial work, coronary conditions, aortic-valve behavior, stasis, lung congestion, systemic perfusion, bleeding, hemolysis, and vascular complications. A better ventricular dimension or lower filling pressure is physiological evidence, not by itself proof of a favorable outcome strategy.

EARLY-UNLOAD randomized 116 patients receiving VA support for cardiogenic shock to routine early transseptal left atrial unloading or a conventional approach permitting rescue unloading. The early strategy did not reduce 30-day mortality, although pulmonary congestion resolved sooner. Substantial rescue use in the comparator and the specific unloading method are essential to interpreting the result. 9

A 2026 report by Moussa and colleagues compared selective biatrial drainage with a ventricular-to-arterial pump in 57 patients with left ventricular congestion during VA support. Lower bleeding and transfusion requirements were associated with the atrial approach, without a demonstrated difference in the primary ventilation-free-day endpoint or mortality. This consecutive-patient comparison was not a randomized allocation between unloading methods and cannot establish universal superiority. 10

These findings leave the physiological reason for addressing important congestion intact while limiting claims about routine timing and method. The question is which boundary is causing the problem, how changing it affects the whole circulation, and whether the relevant outcome evidence supports that particular strategy and population.

79.9 Flow and Saturation Measurements Need Configuration-Specific Interpretation

The circuit's flow sensor measures flow through its own pathway. It does not automatically measure native pulmonary flow, native ventricular ejection, or total effective systemic perfusion. An additional device or regurgitant path can make these quantities diverge further.

Indicator-based output measurements also depend on where the indicator travels. Drainage can remove part of an injected indicator before it reaches the intended detector, and recirculation can return it through a different path. The assumptions behind a standard dilution curve may then be violated.

Linden and colleagues compared pulmonary arterial and transpulmonary thermodilution with three-dimensional echocardiography in eighteen VV-supported patients. Substantial differences were consistent with an extracorporeal influence on thermodilution estimates, while the comparison itself does not make echocardiography an error-free reference or provide one correction factor. 11

Modified methods require their own validation. Berger and colleagues studied adapted thermodilution using an in vitro circulation with controlled circuit flow, recirculation, and pulmonary shunt. Their work showed useful recirculation assessment but imprecision in calculated output. Despite its clinical-sounding study title, this preparation was not a patient validation study. 12

Fick calculations require equivalent care. Native pulmonary oxygen uptake alone may underestimate total oxygen uptake when the membrane also contributes. Conversely, adding oxygen-content flow rather than net membrane transfer counts oxygen already present in blood. Samples must represent the actual inflow and outflow of the compartment being analyzed under sufficiently stable conditions.

Venous saturation is particularly location dependent. A drainage sample, superior vena cava sample, and pulmonary arterial sample need not represent the same mixture during extracorporeal support. Their trends can be useful when location and configuration remain defined, but a familiar label cannot restore the ordinary mixing assumptions.

Pressure, flow, echo geometry, gas content, circuit function, and organ perfusion should consequently be used as complementary observations. Disagreement can reveal altered routing or measurement assumptions rather than justify choosing whichever number appears reassuring.

79.10 Supported Performance Must Be Separated from Native Recovery

A normal arterial gas on substantial membrane support does not show that native lungs have recovered. Likewise, satisfactory arterial pressure during VA support does not establish that the native ventricles can sustain the circulation without that pathway. Support changes the very loads under which organ performance is observed.

Lung recovery concerns effective gas exchange and tolerable mechanics and effort under the actual remaining respiratory burden. Reducing extracorporeal gas removal changes carbon dioxide, drive, native ventilation, and potentially pulmonary vascular load. A gas change therefore requires assessment of the altered conditions rather than one universal interpretation.

Cardiac assessment similarly requires loading context. Reducing VA flow can increase native venous return while lowering the arterial opposition to ejection. A change in ejection fraction or forward stroke volume during that transition reflects both altered support and native capacity. It is not a pure isolated contractility measurement.

Aissaoui and colleagues associated clinical and Doppler findings during flow-reduction assessments with subsequent successful removal in a selected cardiogenic-shock cohort. Their observations support evaluating performance under reduced support, while selection and removal criteria limit treating the reported thresholds as universal independent predictors. 13

Sustained recovery also differs from brief tolerance. Reserve, right-heart behavior, lung function, rhythm, metabolic demand, and organ congestion can change after support is reduced. The time course and wider clinical condition therefore matter alongside a short observation.

The circuit itself can contribute competing injury through blood–surface interaction, thrombosis, bleeding, hemolysis, and vascular access effects. These processes can coexist with successful gas exchange or pressure support. Their presence does not identify a single circuit cause, but it prevents successful device performance from being equated with recovery of the patient.

Chapters 67 and 77 develop coagulation and multiorgan consequences. Within extracorporeal physiology, the essential distinction is that organ replacement, organ unloading, device-associated burden, and recovery are separate processes that must be followed together.

79.11 Outcome Trials Test Complete Strategies in Defined Populations

Physiological capability explains how support may help, but cannot establish whom it benefits overall. Trials compare timing, selection, cointerventions, rescue pathways, and complication exposure together. A result from one configuration and disease cannot be transferred automatically to every extracorporeal circuit.

EOLIA randomized 249 patients with very severe ARDS to immediate VV support or continued conventional care with rescue crossover available. Its 60-day mortality difference favored immediate support numerically but did not reach conventional statistical significance, and substantial crossover occurred. More bleeding requiring transfusion accompanied the extracorporeal strategy. This is not an uncomplicated comparison of support with a group permanently denied rescue. 14

The result also does not demonstrate equivalence or absence of a clinically relevant effect. Statistical uncertainty, the comparator, selection, expertise, and the entire lung-protective strategy remain part of its interpretation. It should neither be replaced by an impressive blood gas nor generalized to VA support for circulatory failure.

ECLS-SHOCK tested early extracorporeal circulatory support in infarct-related cardiogenic shock with planned early revascularization. Among 420 randomized patients, 417 entered the reported final analyses. Thirty-day mortality was not reduced, while moderate or severe bleeding and peripheral vascular complications requiring intervention were more frequent. The tested population and strategy do not include every cause of refractory shock or mechanical obstruction. 15

These trials address different problems and cannot be combined into a single claim that extracorporeal support always works or never works. Gas-exchange replacement and arterial circulatory assistance have different immediate effects, but both expose the patient to additional pathways and harms.

Registry experience and physiological response remain valuable for characterizing mechanisms and generating hypotheses. They are also vulnerable to selection, timing, confounding by severity, and changes in practice. A survival association should not silently become proof of efficacy, and a negative trial should retain its actual comparator and uncertainty rather than become a universal prohibition.

79.12 Integrate Circuit Transfer, Native Flow, and Regional Delivery

Return to the constructed VV example: circuit flow is 3.5 L/min and pre- and postmembrane contents are 14 and 20 mL O₂/dL. Net membrane transfer is 210 mL/min. This says how much oxygen the membrane adds; it does not state native cardiac output or the content reaching each systemic artery.

Suppose measured forward systemic flow is 4 L/min and arterial content 17 mL/dL. Global delivery is 4 × 17 × 10 = 680 mL O₂/min. In a later constructed state, flow is 6 L/min and arterial content 15 mL/dL, giving 900 mL/min. Lower arterial content coexists with higher delivery because flow increased. Whether that delivery meets demand, and how it is distributed, still requires separate assessment.

The two states do not assume unchanged membrane transfer, oxygen consumption, native exchange, or venous content. They illustrate why an isolated saturation trend cannot determine the direction of delivery. Hemoglobin, gas content, measurement reliability, and the clinical time course must be retained.

Now consider VA support with a satisfactory postmembrane gas, rising arterial pressure, little native ejection, and increasing pulmonary congestion. Gas-exchanger success has not resolved the question of left-heart loading. If native ejection later increases while lung exchange remains poor, proximal differential oxygenation can become a new limitation despite apparent cardiac improvement.

The integrated assessment therefore follows actual blood paths, effective circuit flow, transfer across the membrane, native forward flow, regional arterial content, chamber geometry, congestion, respiratory load, and organ response. No single measurement supplies all of them, and an additional intervention can improve one while worsening another.

Support is physiologically successful when it supplies the intended function at a tolerable cost while the cause and recovery trajectory remain understandable. A circuit should be interpreted as part of the patient's circulation and gas-exchange system, rather than as a separate machine whose reassuring numbers certify the condition of the whole patient.

References

1. Schmidt M, Tachon G, Devilliers C, et al. Blood oxygenation and decarboxylation determinants during venovenous ECMO for respiratory failure in adults. Intensive Care Med. 2013;39:838–846. Original clinical physiology study.

2. Fisser C, Palmér O, Sallisalmi M, et al. Recirculation in single lumen cannula venovenous extracorporeal membrane oxygenation: a non-randomized bi-centric trial. Front Med. 2022;9:973240. Original study, with published figure-label correction.

3. Cavayas YA, Munshi L, Del Sorbo L, Fan E. The early change in PaCO₂ after extracorporeal membrane oxygenation initiation is associated with neurological complications. Am J Respir Crit Care Med. 2020;201:1525–1535. Original registry cohort.

4. McNamee JJ, Gillies MA, Barrett NA, et al. Effect of lower tidal volume ventilation facilitated by extracorporeal carbon dioxide removal vs standard care ventilation on 90-day mortality in patients with acute hypoxemic respiratory failure: the REST randomized clinical trial. JAMA. 2021;326:1013–1023. Original trial, with published figure corrections.

5. Ostadal P, Mlcek M, Kruger A, et al. Increasing venoarterial extracorporeal membrane oxygenation flow negatively affects left ventricular performance in a porcine model of cardiogenic shock. J Transl Med. 2015;13:266. Original animal experiment.

6. Saura O, Hékimian G, Del Marmol G, et al. Effect of ECMO flow variations on pulmonary capillary wedge pressure in patients with cardiogenic shock. J Am Coll Cardiol. 2025;86:768–778. Original human physiology study.

7. Hou X, Yang X, Du Z, et al. Superior vena cava drainage improves upper body oxygenation during veno-arterial extracorporeal membrane oxygenation in sheep. Crit Care. 2015;19:68. Original animal experiment.

8. Cai T, Li C, Xu B, et al. Drainage from superior vena cava improves upper body oxygenation in patients on femoral veno-arterial extracorporeal membrane oxygenation. Front Cardiovasc Med. 2022;8:807663. Original human physiology study.

9. Kim MC, Lim Y, Lee SH, et al. Early left ventricular unloading or conventional approach after venoarterial extracorporeal membrane oxygenation: the EARLY-UNLOAD randomized clinical trial. Circulation. 2023;148:1570–1581. Original randomized trial.

10. Moussa MD, Roux J, Jungling M, et al. Impella versus selective biatrial canulation for left ventricular unloading during extracorporeal membrane oxygenation. Cardiovasc Ther. 2026;2026:3669575. Original comparative clinical report.

11. Linden K, Schmandt M, Muders T, et al. Estimation of cardiac output under veno-venous extracorporeal membrane oxygenation: comparing thermodilution methods to 3D echocardiography. ASAIO J. 2025;71:75–81. Original clinical comparison.

12. Berger D, Stanger EJ, Jenni H, et al. Modified thermodilution for simultaneous cardiac output and recirculation assessment in veno-venous extracorporeal membrane oxygenation: a prospective diagnostic accuracy study. Anesthesiology. 2024;140:1002–1015. Original in vitro validation study.

13. Aissaoui N, Luyt CE, Leprince P, et al. Predictors of successful extracorporeal membrane oxygenation (ECMO) weaning after assistance for refractory cardiogenic shock. Intensive Care Med. 2011;37:1738–1745. Original clinical physiology cohort.

14. Combes A, Hajage D, Capellier G, et al. Extracorporeal membrane oxygenation for severe acute respiratory distress syndrome. N Engl J Med. 2018;378:1965–1975. Original EOLIA randomized trial.

15. Thiele H, Zeymer U, Akin I, et al. Extracorporeal life support in infarct-related cardiogenic shock. N Engl J Med. 2023;389:1286–1297. Original ECLS-SHOCK randomized trial.