Part I — Foundations · Chapter 2
Evidence search date: 6 September 2026. Physiological relationships in this chapter are mechanistic and largely uncontested; the quantitative claims that are not pure arithmetic are cited and graded individually.
Clinical Question
Before any ECMO number can be interpreted, what must already be understood about how oxygen and carbon dioxide are carried, how much of each the body uses, and how blood flow and shunt determine what reaches the tissues?
Why This Matters
Chapter 1 established the framing: ECMO is a parallel circuit, and every measured value is a mixture. This chapter supplies the quantities that make the mixture calculable.
The pay-off is diagnostic. A patient desaturates on VV ECMO. The circuit is unchanged, the oxygenator is clean, the cannula has not moved. Nothing about the machine explains it — and yet the physiology explains it completely, because the patient became febrile and agitated three hours earlier. Without the relationships in this chapter, that patient gets a circuit exchange. With them, they get paracetamol and sedation.
Everything here is standard cardiopulmonary physiology. It is restated because ECMO makes each relationship quantitatively load-bearing in a way that ordinary intensive care does not.
Content, Tension, Delivery, Consumption
Four distinct things, routinely spoken of as "oxygenation".
Quantity | Symbol | Units | What it tells you | Typical adult value |
Oxygen tension | PaO₂ | mmHg | The driving pressure for diffusion. Almost none of the oxygen in blood. | 80–100 breathing air |
Oxygen content | CaO₂ | mL O₂/dL | How much oxygen the blood is actually carrying. Dominated by haemoglobin. | ~20 mL/dL at Hb 15, SaO₂ 100% |
Oxygen delivery | DO₂ | mL/min | Content × flow. What the tissues are offered. | ~1000 mL/min (~550 mL/min/m²) |
Oxygen consumption | VO₂ | mL/min | What the tissues actually take. Set by metabolism, not by supply — until supply fails. | ~200–250 mL/min (~120–130 mL/min/m²) |
Oxygen extraction ratio | O₂ER | fraction | The reserve remaining. The single most informative derived number in shock. | 0.25–0.30 |
Physiology
At a PaO₂ of 100 mmHg, dissolved oxygen contributes 0.3 mL/dL of roughly 20 mL/dL total — about 1.5%. This is why a falling PaO₂ above the shoulder of the dissociation curve is nearly irrelevant to delivery, and why a rising PaO₂ in a patient who is already saturated buys nothing at all.
The corollary matters more on ECMO than anywhere else: post-oxygenator PO₂ of 400 mmHg is not four times better than 200 mmHg. Both fully saturate the haemoglobin. The extra tension is dissolved oxygen and oxidative stress, not delivery.
The oxygen content equation itself was introduced in Chapter 1 and is not repeated here; see also Appendix 1 and Appendix 2.
The Fick Relationship
Everything that follows in this chapter is a rearrangement of one statement: oxygen consumed equals oxygen delivered minus oxygen returned.
Variables and units
- — oxygen consumption, mL/min
- — cardiac output, L/min (on ECMO, this must mean total systemic flow: native output plus circuit flow returned to the arterial side)
- , — arterial and mixed venous oxygen content, mL/dL
- 10 — converts dL to L
Assumptions. Steady state; no significant intracardiac shunt; mixed venous blood genuinely mixed — an assumption that ECMO frequently violates, as below.
Rearranged, the same statement gives the extraction ratio:
The approximation on the right holds whenever dissolved oxygen is negligible, which on ECMO it almost always is.
Worked example
An adult on VA ECMO. Total systemic flow 5 L/min. Hb 10 g/dL. SaO₂ 98%, PaO₂ 90 mmHg. SvO₂ 70%, PvO₂ 40 mmHg.
Step | Calculation | Result |
CaO₂ | (1.34 × 10 × 0.98) + (0.003 × 90) | 13.40 mL/dL |
CvO₂ | (1.34 × 10 × 0.70) + (0.003 × 40) | 9.50 mL/dL |
Arteriovenous difference | 13.40 − 9.50 | 3.90 mL/dL |
VO₂ | 3.90 × 5 × 10 | 195 mL/min |
DO₂ | 13.40 × 5 × 10 | 670 mL/min |
O₂ER | 195 ÷ 670 | 29% |
DO₂ : VO₂ | 670 ÷ 195 | 3.4 : 1 |
Now change one thing that has nothing to do with the circuit. The patient becomes febrile and agitated; VO₂ rises to 300 mL/min. Flow, haemoglobin and SaO₂ are unchanged.
Step | Calculation | Result |
Required a–v difference | 300 ÷ (5 × 10) | 6.00 mL/dL |
New CvO₂ | 13.40 − 6.00 | 7.40 mL/dL |
New SvO₂ | (7.40 − 0.12) ÷ 13.4 | 54% |
New DO₂ : VO₂ | 670 ÷ 300 | 2.2 : 1 |
Clinical interpretation. The SvO₂ has fallen 16 points and the delivery-to-consumption ratio has moved from comfortable to the edge of supply dependence — with no change in any ECMO setting, no circuit problem, and no new organ failure. The correct intervention is antipyresis and sedation, not more flow.
Clinical Pearl
On ECMO, VO₂ is a variable you can treat. Fever, shivering, agitation, seizures, work of breathing and the thermogenic response to rewarming all raise it, and each is reversible within minutes to hours. Lowering consumption is often faster, safer and cheaper than raising delivery.
The Fick Equation as a Diagnostic Tree
When SvO₂ falls or lactate rises, exactly one of four terms has changed. Ask them in this order.
This tree recurs, elaborated, in Chapters 9, 25 and 30.
The DO₂–VO₂ Relationship and the Critical Threshold
Across a wide range, VO₂ is independent of DO₂: as delivery falls, extraction rises to compensate and consumption is preserved. Below a critical point, extraction can rise no further, consumption becomes supply-dependent, and anaerobic metabolism begins.
Evidence — where the critical threshold actually sits
Ronco and colleagues measured VO₂ by indirect calorimetry and DO₂ by the Fick principle in 18 critically ill adults (9 septic, 9 non-septic) during discontinuation of life support, and identified the critical oxygen delivery from the biphasic DO₂–VO₂ relationship. Critical DO₂ was 3.8 ± 1.5 mL/min/kg in septic and 4.5 ± 1.3 mL/min/kg in non-septic patients — not significantly different, and considerably lower than earlier estimates. The authors concluded that sepsis does not alter the critical oxygen delivery, and questioned interventions aimed at driving DO₂ to supranormal levels. (Ronco JJ, et al. JAMA 1993)
Certainty: moderate, with important limits. Eighteen patients, measured during withdrawal of life support — a state of low and falling metabolic rate that is not the ICU patient in front of you. The threshold is a population estimate, not a bedside target, and individual thresholds vary with temperature, sedation and disease.
A practical framing used throughout the extracorporeal literature: DO₂ normally exceeds VO₂ by roughly five-fold; aerobic metabolism is maintained comfortably down to about 3:1; and below roughly 2:1 anaerobic metabolism and lactate accumulation supervene (ELSO Red Book, 6th ed.). In the worked example above, the febrile patient reached 2.2:1 — which is why that example matters.
Pitfall — chasing supranormal delivery
The existence of a critical threshold does not imply that more delivery above it is better. Driving DO₂ to supranormal targets has repeatedly failed to improve outcomes and, on ECMO, is bought with higher flows, more transfusion and more circuit exposure — all of which carry their own harm. Treat the threshold as a floor to stay above, not a target to maximise.
Mixed Venous Saturation — and Why ECMO Breaks It
SvO₂ is the integrated report on the adequacy of delivery relative to demand. Four things lower it, and they are exactly the four terms of the Fick equation: falling SaO₂, falling haemoglobin, falling cardiac output, rising VO₂.
On ECMO, however, the measurement itself becomes treacherous.
Danger — the venous sample on VV ECMO may be fiction
Blood drawn from the drainage limb of a VV circuit is contaminated by recirculated, already-oxygenated blood. Recirculation makes that sample look reassuringly high while true tissue extraction is worsening. A "good" pre-oxygenator saturation is therefore compatible with a patient in oxygen debt.
The same trap applies to pulmonary artery catheter sampling on VV ECMO: the catheter sits downstream of the return cannula, so it measures the mixture, not true mixed venous blood.
On VA ECMO, blood in the pulmonary artery reflects only the small native fraction, not systemic extraction. Interpret venous saturations on ECMO by knowing the sampling site relative to the cannulae — always. See Chapter 27.
Central venous saturation (ScvO₂) from a superior vena cava catheter is not the same as true mixed venous saturation — it omits the inferior vena cava and coronary sinus contributions — but it tracks trends usefully when the sampling site is stable and known. Absolute equivalence should not be assumed.
Carbon Dioxide Is a Different Animal
Oxygen and carbon dioxide are not symmetrical, and almost every practical asymmetry in ECMO follows from this.
Oxygen | Carbon dioxide | |
Carried mainly as | Bound to haemoglobin (~98%) | Bicarbonate (~70%), carbamino compounds (~20%), dissolved (~10%) |
Content–tension relationship | Sigmoid, saturable — plateaus above ~90% | Near-linear across the physiological range, effectively unsaturable |
Solubility in plasma | 0.003 mL/dL/mmHg | Roughly 20 times greater |
Limiting factor in a membrane lung | Blood flow and haemoglobin presented to the membrane | Sweep gas flow — the analogue of minute ventilation |
Practical consequence | Full support needs high blood flow | Full CO₂ clearance is achievable at low blood flow — which is what makes ECCO₂R possible |
Physiology — why the smaller gradient wins
Across a membrane lung the oxygen gradient is enormous (roughly 600 mmHg in the gas phase against ~40 mmHg in venous blood) while the CO₂ gradient is small (about 45 mmHg against 0). Despite this, at a sweep-to-blood flow ratio of 1:1 the quantities of O₂ and CO₂ exchanged are approximately equal, because CO₂ is far more soluble and diffusible and its content curve is not saturable. Raising the sweep-to-blood ratio to 4:1 or 8:1 multiplies CO₂ removal while doing nothing whatever for oxygenation.
This is the mechanistic basis for the rule stated in Chapter 1: sweep governs CO₂, blood flow and FdO₂ govern oxygen. (ELSO Red Book, 6th ed.; ECMO: A Practical Guide to Management)
Because CO₂ clearance is so much easier than oxygenation, the ELSO Red Book's position is that circuit management is driven by oxygenation, with sweep simply titrated to the desired PaCO₂ — and that there is exactly one circumstance in which CO₂ clearance fails before oxygenation does: water vapour accumulating on the gas side of the membrane. That single exception is a diagnostic gift, and it is developed in Chapter 32.
Carbon dioxide production (VCO₂) relates to VO₂ through the respiratory quotient, typically about 0.8 on mixed substrate — so an adult consuming 250 mL/min of oxygen produces roughly 200 mL/min of CO₂, all of which must leave through the native lung, the membrane lung, or both.
Danger — how fast you correct hypercapnia matters
A membrane lung can drop a PaCO₂ of 100 mmHg to 40 mmHg within minutes. Cerebral vasoconstriction, a rapid CSF pH shift and a sudden fall in cerebral blood flow follow. Rapid CO₂ correction at the initiation of ECMO is a plausible contributor to early neurological injury.
Start with a low sweep and reduce PaCO₂ gradually, particularly when hypercapnia has been prolonged and metabolic compensation is established. Chapters 7, 22 and 28 return to this.
Shunt, Dead Space and Why FiO₂ Stops Working
Gas exchange fails in two fundamentally different ways, and distinguishing them determines whether ECMO is even the right answer.
where is end-capillary oxygen content (assumed fully equilibrated with alveolar gas), and is the fraction of cardiac output that reaches the systemic circulation without being oxygenated.
Mechanism | What is wrong | Response to raising FiO₂ | Effect on PaCO₂ | ECMO relevance |
Shunt | Perfusion without ventilation — consolidation, collapse, ARDS | Little or none once shunt exceeds roughly 30% | Usually preserved until very severe | The dominant lesion in the VV ECMO population |
Dead space | Ventilation without perfusion — PE, low output, high airway pressure | Little effect on CO₂; oxygenation often preserved | Rises | Suggests a circulatory or obstructive lesion; ask whether VA rather than VV is required |
Diffusion limitation | Thickened barrier, short transit time | Responsive | Little effect | Rarely the dominant lesion in adults at rest |
Hypoventilation | Inadequate alveolar ventilation | Responsive | Rises | Reversible without extracorporeal support — exclude it first |
Pitfall — ECMO can worsen native lung shunt
Hypoxic pulmonary vasoconstriction diverts blood away from unventilated lung and is a defence against shunt. Raising mixed venous oxygen content — exactly what VV ECMO does — releases that vasoconstriction, redistributing perfusion back into non-ventilated regions and increasing the native shunt fraction.
The practical consequence is counterintuitive but real: some of the apparent underperformance of the native lung after starting VV ECMO is a direct physiological consequence of the therapy, not deterioration of the underlying disease.
Cardiac Output and the Right Ventricle
The right ventricle is the chamber ECMO most often rescues and most often injures, and it behaves quite unlike the left.
- It is a thin-walled volume pump, tolerant of preload and highly intolerant of afterload. A modest rise in pulmonary vascular resistance can halve stroke volume.
- Pulmonary vascular resistance rises with hypoxaemia, hypercapnia, acidosis, high alveolar pressure and lung overdistension — the exact quartet present in severe ARDS, and each of them correctable by ECMO.
- Ventricular interdependence: RV dilatation shifts the interventricular septum leftward, impairing LV filling. Right heart failure therefore presents as low systemic output with a small, underfilled left ventricle.
- Coronary perfusion of the RV occurs in both systole and diastole normally, but becomes diastole-dependent when RV pressure rises — so systemic hypotension and RV hypertension together create a self-reinforcing spiral of RV ischaemia.
Physiology — how VV ECMO helps the right ventricle without moving any blood
VV ECMO is haemodynamically neutral by construction: the same volume is drained and returned to the venous compartment, so central venous volume, central venous pressure and RV preload are unchanged.
Its benefit to the right ventricle is entirely indirect — by correcting hypoxaemia, hypercapnia and acidosis it lowers pulmonary vascular resistance, and by permitting lower airway pressures and tidal volumes it removes the mechanical component of RV afterload. Improved coronary oxygen content additionally reverses hypoxia-induced myocardial depression. (ECMO: A Practical Guide to Management; ELSO Red Book, 6th ed.)
This is why a patient with severe ARDS and secondary RV failure often improves haemodynamically on VV ECMO despite receiving no circulatory support at all — and why failure to improve should prompt the question of whether the RV failure is primary. See Chapters 3 and 11.
VA ECMO alters preload and afterload directly and in opposite directions; that physiology is the subject of Chapter 13 and is not duplicated here.
The Mixing Equation
This is the single relationship that makes VV ECMO interpretable, and it is a direct application of the parallel-circuit idea from Chapter 1.
With complete native lung failure, arterial saturation is the flow-weighted average of two streams:
Variables. — effective extracorporeal blood flow, that is circuit flow minus recirculated flow, L/min. — total cardiac output, L/min. — post-oxygenator saturation, normally 1.00. — saturation of the blood that bypasses the circuit.
Assumptions. Complete native lung shunt (no oxygenation across the native lung), so this is the worst case. With residual native lung function, arterial saturation is higher.
Evidence — the 0.6 rule
In a 2024 physiological review, Tomarchio and colleagues express pulmonary arterial saturation as the weighted average of circuit and non-circuit blood, and derive a practical anchor: an ECBF/CO ratio of about 0.6, with a venous saturation of 75% and a fully saturating oxygenator, yields an arterial saturation of approximately 90% in a patient with complete native lung dysfunction. They emphasise that this ratio must be revised upward when venous saturation is lower, and that effective rather than total circuit flow must be used when recirculation is present. (Tomarchio E, et al. Perfusion 2024)
Certainty: physiological rationale. This is a modelled relationship, not an outcome-validated target.
Worked example — the desaturation that no circuit setting caused
Complete native lung failure. Effective ECBF 3.6 L/min, cardiac output 6 L/min, so ECBF/CO = 0.6. Post-oxygenator saturation 100%.
Scenario | Calculation | Arterial saturation |
SvO₂ 75% (stable patient) | (0.6 × 1.00) + (0.4 × 0.75) | 90% |
SvO₂ 60% (febrile, agitated, VO₂ risen) | (0.6 × 1.00) + (0.4 × 0.60) | 84% |
Clinical interpretation. A 15-point fall in venous saturation costs 6 points of arterial saturation with no change in pump speed, sweep, FdO₂, cannula position or oxygenator performance. The circuit is behaving perfectly. Treating this desaturation as a circuit problem — escalating flow, exchanging the oxygenator — addresses the wrong term of the equation.
Clinical Pearl — the dilution paradox
Raising cardiac output on VV ECMO lowers the ECBF/CO ratio, and therefore tends to lower arterial saturation, even though total oxygen delivery may improve. This is why inotropes and fluid can be followed by a fall in SaO₂ on VV ECMO, and why that fall is not evidence of harm.
The effect is partly self-limiting: higher cardiac output at constant VO₂ raises SvO₂, which raises the second term of the mixing equation and offsets part of the dilution. Judge the intervention on delivery and lactate, not on the saturation number.
What the Membrane Lung Can and Cannot Do
One piece of device physiology belongs in this chapter because it constrains everything else. Rated flow is the maximum blood flow at which a given oxygenator can raise standardised venous blood (haemoglobin 12 g/dL, inlet saturation 75%) to an outlet saturation of 95%. Above rated flow, blood leaves the membrane incompletely saturated (ECMO: A Practical Guide to Management).
Two consequences follow directly:
- Oxygen transfer is limited by blood flow and the haemoglobin presented to the membrane — so an anaemic patient gets less oxygen transfer at the same pump speed, and the oxygenator is not the thing that failed.
- CO₂ removal rises with blood flow but plateaus according to membrane surface area, and is driven principally by sweep gas (Tomarchio E, et al. Perfusion 2024).
Controversies
Controversy — what saturation is low enough to act on?
Clinical question: What arterial saturation should trigger escalation on VV ECMO?
Evidence supporting permissive hypoxaemia: Delivery, not saturation, determines tissue oxygenation, and the arithmetic in Chapter 1 shows that the last few saturation points contribute little. Escalation is not free: higher flows mean larger cannulae, more haemolysis and more transfusion.
Evidence supporting a higher target: Global delivery can be adequate while regional delivery is not, and the brain is the organ least able to report this. The critical threshold data are population estimates derived in a low-metabolic-rate state and cannot be applied to an individual with confidence.
Current consensus: Saturations in the mid-80s are widely tolerated when delivery is adequate, lactate is falling and there is no evidence of regional ischaemia. There is no validated numerical threshold.
Practical approach: Judge adequacy on the set — lactate trend, venous saturation with a known sampling site, mentation where assessable, urine output, regional perfusion — not on a single number.
Knowledge gap: No randomised trial has compared oxygenation targets on VV ECMO. This is an unresolved question, not a settled practice.
Controversy — is there a haemoglobin target that follows from the physiology?
Clinical question: Given that content is haemoglobin-dominated, should ECMO patients be transfused to a higher threshold than other critically ill patients?
Evidence supporting a higher threshold: Oxygen delivery is linear in haemoglobin; oxygen transfer across the membrane depends on the haemoglobin presented to it; and the ECMO patient often has limited capacity to raise cardiac output in compensation.
Evidence supporting a restrictive threshold: The general critical care transfusion literature consistently favours restrictive strategies, and transfusion on ECMO carries added costs — alloimmunisation in transplant candidates, volume, and inflammatory burden.
Current consensus: Practice varies widely between centres and no consensus threshold exists.
Practical approach: Treat the physiology as a reason to look at haemoglobin when delivery is inadequate, not as a licence for routine liberal transfusion.
Knowledge gap: Directly addressed, with the available evidence, in Chapter 46.
Evidence Summary
Statement | Certainty | Basis |
VO₂ = CO × (CaO₂ − CvO₂) × 10, and O₂ER = VO₂/DO₂ | High — definitional | Fick principle; arithmetic identity |
VO₂ is supply-independent above a critical DO₂ and supply-dependent below it | Moderate | Ronco 1993 (n=18) and the wider oxygen-transport literature; threshold estimates vary by method and population |
Critical DO₂ approximately 3.8–4.5 mL/min/kg, not different between septic and non-septic patients | Low to moderate | Single study, 18 patients, measured during withdrawal of life support — a population estimate, not a bedside target |
CO₂ removal is governed by sweep gas; oxygen transfer by blood flow and haemoglobin | High — physiological rationale, uncontested | Solubility and content-curve differences; ELSO Red Book; Tomarchio 2024 |
Arterial saturation on VV ECMO is the flow-weighted average of circuit and non-circuit blood; ECBF/CO around 0.6 with SvO₂ 75% gives roughly 90% | Physiological rationale | Tomarchio 2024 — a modelled relationship, not an outcome-validated target |
VV ECMO reduces RV afterload indirectly and has no direct haemodynamic effect | High — physiological rationale | Equal drainage and return to the venous compartment; correction of hypoxic pulmonary vasoconstriction and airway pressure |
Raising mixed venous oxygen content releases hypoxic pulmonary vasoconstriction and can increase native shunt fraction | Moderate — physiological rationale with supporting observation | HPV physiology; described in the ECMO literature |
Optimal oxygenation and haemoglobin targets on ECMO | Not established | No randomised comparison; explicit knowledge gap |
Key Takeaways
- Tension, content, delivery and consumption are four different quantities. Only content and flow determine what reaches tissue.
- Every ECMO deterioration reduces to one of four Fick terms: saturation, haemoglobin, flow, or consumption. Ask them in that order.
- VO₂ is treatable. Fever, agitation, shivering and work of breathing are often faster to fix than delivery is to raise.
- There is a critical delivery threshold, but it is a floor to stay above, not a target to maximise — and the published estimates come from a small study in an atypical state.
- Venous saturation is the most useful number on ECMO and the easiest to misread. Know the sampling site relative to the cannulae, or do not interpret it.
- CO₂ and oxygen are not symmetrical. Sweep governs CO₂; blood flow and haemoglobin govern oxygen. The one exception — gas-side water accumulation — is diagnostically valuable.
- Correct hypercapnia slowly. The membrane lung can do it dangerously fast.
- Shunt is why FiO₂ stops working — and VV ECMO can increase native shunt by releasing hypoxic pulmonary vasoconstriction.
- On VV ECMO, arterial saturation is a flow-weighted average. A falling SvO₂ lowers SaO₂ with the circuit behaving perfectly, and raising cardiac output can lower SaO₂ while improving delivery.
Key References
- Ronco JJ, Fenwick JC, Tweeddale MG, et al. Identification of the critical oxygen delivery for anaerobic metabolism in critically ill septic and nonseptic humans. JAMA. 1993;270:1724–1730. DOI: 10.1001/jama.1993.03510140084034. PMID: 8411504
- Tomarchio E, Momigliano F, Giosa L, Collins PD, Barrett NA, Camporota L. The intricate physiology of veno-venous extracorporeal membrane oxygenation: an overview for clinicians. Perfusion. 2024;39(1_suppl):49S–65S. DOI: 10.1177/02676591241238156
- Brodie D, Peek G, MacLaren G, et al. (eds). Extracorporeal Life Support: The ELSO Red Book, 6th edition — chapters on ECLS physiology, oxygen kinetics and CO₂ removal
- Taha AR, Caridi-Scheible M, Leiendecker E, et al. ECMO: A Practical Guide to Management — Chapter 5, Physiology I: Venovenous ECMO (rated flow; membrane lung gas gradients; indirect RV effects)
- Indian Society of Critical Care Medicine. ISCCM Manual of RRT and ECMO in ICU — Chapter 26, Physiology during Extracorporeal Membrane Oxygenation Support
- Cross-references: Chapter 1 (parallel-circuit framework; oxygen content arithmetic), Chapter 6 (VV ECMO physiology in depth), Chapter 13 (VA ECMO haemodynamics), Chapter 27 (blood gas sampling sites), Appendix 1 (equations)
Citation verification status. The Ronco 1993 citation, including DOI, PMID, journal, volume, pages and the quoted critical delivery values, and the Tomarchio 2024 citation with DOI, volume and pages, were verified against source records during the 6 September 2026 search cycle. Book-derived statements are attributed to the specific texts held in the project library and are expert-synthesis sources rather than primary evidence.
An earlier report of critical oxygen delivery in conscious healthy humans could not be retrieved during this cycle (repeated rate limiting) and is therefore not cited; the threshold discussion rests on Ronco alone, with its limitations stated.
Physiological constants (Hüfner constant, plasma solubilities, CO₂ carriage proportions, respiratory quotient) are standard textbook values, given as approximations.
Educational use only. This chapter does not replace institutional ECMO protocols, local policy, specialist consultation, current guidelines or patient-specific clinical judgement.