Modern haemodynamics has produced an impressive collection of numbers. Some are measured, some calculated, some are ratios of other calculated numbers, and some require enough algebra that they acquire an air of authority before anyone has asked whether they are actually useful.
This series is going to look at a few of them. We’ll get to CO₂ gaps, arterial waveform indices, dynamic arterial elastance and ventriculo-arterial coupling later. The common question is whether these increasingly sophisticated numbers actually help us understand what is limiting the circulation and what to do about it.
Or whether they simply help us look clever while treating the number instead of the circulation.
The target is not complexity itself. Complicated measurements can be extremely useful, and simple ones can be useless. Nor is the argument that derived variables contain no physiological information. Most of the numbers in this series are built on perfectly respectable physiology.
The question is narrower: what does the number actually add to the clinical decision?
Does it identify the dominant physiological problem? Does it tell us which treatment to choose? Does a change in the number tell us that the mechanism we intended to alter really changed? And, most importantly, does improving it tell us anything reliable about whether tissue perfusion improved?
Those questions become particularly important when a physiological descriptor is promoted into a treatment target. Once that happens, it is surprisingly easy to stop asking what we are trying to achieve and start trying to make the number look better.
ScvO₂ is a good place to start because I remember when we did exactly that.
There was a period in the early 2000s when central venous oxygen saturation was everywhere. If you trained in intensive care then, the targets are probably still buried somewhere in your memory: CVP 8–12 mmHg, MAP at least 65 mmHg and ScvO₂ at least 70%. Fibreoptic central venous catheters could display the number continuously. A value below 70% looked like unfinished resuscitation.
I barely think about ScvO₂ now.
So what exactly did we think it was telling us?
Why ScvO₂ was so attractive
The underlying idea was entirely reasonable. Shock cannot be understood from blood pressure alone. A patient can have an acceptable MAP but a low cardiac output, profound anaemia or severe hypoxaemia. Metabolic demand can also vary enormously. What we really wanted to know was not simply whether there was pressure in the arterial tree, but whether enough oxygen was being delivered to meet the needs of the tissues.
Oxygen delivery is:
DO₂ = CO × CaO₂
where CO is cardiac output in litres per minute and CaO₂ is arterial oxygen content in mL/dL.
Ignoring the relatively small amount of oxygen dissolved in plasma:
CaO₂ ≈ 1.34 × Hb × SaO₂
The Fick principle then relates oxygen consumption to the difference between arterial and venous oxygen content:
VO₂ = CO × (CaO₂ − CvO₂)
If oxygen delivery falls while metabolic requirements remain unchanged, tissues can initially compensate by extracting a greater proportion of the oxygen delivered to them. Venous oxygen content falls and venous oxygen saturation falls with it.
That is sound physiology. If haemoglobin, arterial oxygen saturation and metabolic demand are stable, a falling venous saturation can certainly tell us that the relationship between delivery and oxygen uptake has changed.
The attraction was obvious. Instead of looking at pressure or cardiac output in isolation, venous oxygen saturation appeared to integrate the whole oxygen transport system into one clinically useful number.
The difficulty starts when we try to work backwards from that number.
The inverse problem
Suppose the ScvO₂ is 55%.
Cardiac output may be low. Haemoglobin may be low. Arterial oxygen saturation may be low. Oxygen consumption may be high because the patient is shivering, febrile or fitting. Several of these may be present simultaneously.
The same ScvO₂ can therefore be produced by completely different physiological states requiring completely different treatments.
There is an additional technical problem. ScvO₂ is not actually mixed venous oxygen saturation. True SvO₂ is measured in the pulmonary artery after venous blood from the superior vena cava, inferior vena cava and coronary circulation has mixed. ScvO₂ is normally sampled from the lower superior vena cava or cavo-atrial region and therefore reflects only part of that mixture. Regional blood-flow redistribution and regional oxygen extraction alter the relationship between the two, and there is no reliable fixed conversion.
That limitation attracted a great deal of attention, but it is not the most important one. Imagine we had a perfect pulmonary artery catheter measuring true SvO₂ continuously with no measurement error at all. We would have removed the sampling problem, but we would still have the same causal problem.
Venous oxygen saturation is the downstream result of cardiac output, arterial oxygen content and realised oxygen consumption. Measuring the result does not uniquely identify which part of the system produced it.
This is the first recurring theme of this series: a variable can describe the state of a system very well while being poor at identifying the mechanism responsible for that state.
Even low flow does not tell you the treatment
Suppose haemoglobin and arterial saturation are normal, the patient is sedated and their metabolic state has not obviously changed. The ScvO₂ falls substantially. Low flow is now a sensible possibility.
But that still leaves the clinically useful question unanswered: why is the flow low?
There are really only two broad possibilities. Delivery may be inadequate, because there is insufficient circulating volume or insufficient venous tone to deliver blood to the heart. Or cardiac acceptance may be limited. Acceptance depends on two reserves: volume reserve (the ability to accommodate more filling) and power reserve (the ability to generate the extra work needed to eject that volume against the prevailing load). I will return to these two constraints, and how to use them at the bedside, in the final article of this series.
ScvO₂ cannot distinguish between them. Even when it correctly alerts us to a possible fall in flow, it does not select the intervention. Depending on the mechanism, the patient might need fluid, a vasopressor, an inotrope, reduction of ventricular afterload, correction of a rhythm problem or relief of an external constraint such as cardiac tamponade.
A low ScvO₂ may tell us that something has changed in the oxygen balance. It does not tell us which part of the circulation to alter.
The deeper problem: global oxygen delivery is not tissue perfusion
The original appeal of venous oxygen saturation rested on a broader model of shock. Oxygen delivery falls, tissues compensate by extracting more oxygen, and eventually delivery becomes so low that oxygen consumption becomes supply-dependent. Oxygen debt accumulates and organ failure follows.
That model works reasonably well at the extremes. In severe haemorrhage or profound low-output cardiac failure, global oxygen delivery can become insufficient to meet metabolic requirements. There really is a point beyond which increased extraction can no longer compensate.
Sepsis is much less tidy.
Once severe hypotension has been treated, many patients with vasodilatory septic shock have a normal or high cardiac output. Their calculated global oxygen delivery can therefore be normal or high as well. Yet they may remain mottled, have a prolonged capillary refill time and continue to develop organ dysfunction. Global oxygen delivery can therefore look perfectly adequate while regional perfusion remains poor, and the normal increase in oxygen extraction may itself be impaired.
The problem is that global oxygen delivery is a bulk quantity. It tells us how much oxygen passes through the systemic circulation each minute. It does not tell us whether that oxygen reaches the right organs, whether it is distributed appropriately within those organs, or whether local blood flow matches local metabolic need.
A large amount of oxygen passing through the aorta is not the same thing as adequate tissue perfusion.
How about O2 extraction?
Macroflow and the microcirculation
Sepsis can produce marked heterogeneity within the microcirculation. Some capillaries remain well perfused while others carry very little flow. Functional capillary density falls, red-cell transit becomes heterogeneous, endothelial behaviour and local vascular regulation are altered, and tissue oedema changes the environment in which exchange takes place.
This can also impair oxygen extraction. Blood passing rapidly through well-perfused channels may return relatively oxygen-rich while neighbouring tissue receives little effective red-cell delivery. Poorly perfused tissue contributes little venous blood to the final mixture, and an unperfused region contributes none at all. The tissues most deprived of flow can therefore have surprisingly little influence on the venous saturation we eventually measure.
There may also be impairment of cellular oxygen utilisation in sepsis, although the contribution of mitochondrial dysfunction is harder to quantify clinically. The important point is simpler: adequate global oxygen delivery does not guarantee that oxygen is distributed, extracted and used appropriately at tissue level.
This helps explain why improving macrocirculatory variables does not always restore microvascular perfusion — the phenomenon usually described as haemodynamic incoherence.
There is also a simple mixing problem that is easy to overlook.
Mixed venous oxygen saturation is weighted by venous flow. A well-perfused region contributes a large volume of blood to the final venous mixture. A poorly perfused region contributes much less. A completely unperfused region contributes none at all.
The tissue receiving the least blood may therefore have remarkably little influence on the final mixed venous saturation. A reassuring global value cannot prove that every regional vascular bed is adequately perfused.
ScvO₂ has the additional limitation that it does not even contain the complete venous mixture.
None of this means that septic organ failure is simply a microcirculatory disease. Inflammation, cellular injury, altered metabolism, congestion and organ-specific mechanisms all contribute. The narrower point is enough: normal global oxygen transport does not establish adequate regional perfusion.
When a high ScvO₂ is not reassuring
A high ScvO₂ can be entirely appropriate. Cardiac output may have increased, arterial oxygen content may be adequate and metabolic demand may have fallen. A previously inadequate oxygen-delivery state may genuinely have been corrected.
But a high value can also occur when oxygen uptake is impaired or flow is distributed poorly. If less oxygen is extracted from the blood, more of it returns in the venous circulation and venous saturation rises. If poorly perfused tissue contributes little venous blood to the sampled mixture, its failure may barely affect the final number.
This produces a fundamental ambiguity. The same increase in ScvO₂ can accompany successful restoration of oxygen delivery or a reduction in oxygen extraction.
A low value and a high value are therefore not mirror images of one another. A sudden fall can be a useful warning that something in the oxygen balance has changed. A normal or high value cannot reliably exclude regional hypoperfusion.
That makes ScvO₂ much easier to defend as a monitoring signal than as a treatment endpoint.
Then we started treating the number
The problem becomes clearer when we consider what happens when oxygen delivery itself becomes the target.
Take fluid.
If a patient is preload responsive, fluid may increase stroke volume and cardiac output. That may increase oxygen delivery. But crystalloid also dilutes haemoglobin and therefore reduces arterial oxygen content.
Suppose haemoglobin falls from 12 to 10 g/dL while cardiac output rises from 5 to 5.5 L/min, with arterial saturation remaining at 98%. Ignoring dissolved oxygen, oxygen delivery falls from roughly 788 to 722 mL/min.
Cardiac output has gone up, but oxygen delivery has gone down.
A 16.7% reduction in oxygen content requires a 20% rise in cardiac output merely to preserve the original oxygen delivery.
This does not necessarily make fluid harmful. If it is correcting a genuine volume deficit, the increase in flow may outweigh the fall in oxygen content. But once additional volume is no longer correcting that deficit, the trade becomes progressively less favourable: haemodilution reduces arterial oxygen content, while rising venous pressure and oedema can impair tissue perfusion. Cardiac output may still increase, but that does not guarantee better oxygen delivery where it is needed.
The intervention has a second cost. Additional volume has to be accommodated somewhere. If the heart cannot accept it without a substantial rise in filling pressure, systemic venous pressure rises. Organ drainage becomes more difficult, capillary hydrostatic pressure rises and oedema increases.
It is therefore entirely possible for an intervention to improve one aspect of global oxygen transport while simultaneously worsening venous congestion or regional tissue conditions.
The relevant question is not whether ScvO₂ rose. It is whether the intervention produced a useful net physiological effect.
More flow is not automatically better microvascular oxygen delivery
Haemodilution also exposes another limitation of thinking purely in terms of bulk flow.
Reducing blood viscosity can lower resistance and increase systemic flow. That is one reason cardiac output may rise as haemoglobin falls. But microvascular oxygen transport depends on more than litres of blood per minute. Red-cell concentration, deformability, microvascular haematocrit, plasma viscosity, endothelial shear and the distribution of red cells through very small vessels all contribute.
Experimental work with extreme haemodilution can therefore produce an apparently paradoxical result: systemic flow rises while effective microvascular red-cell delivery deteriorates.
That does not mean that ordinary crystalloid resuscitation routinely produces microvascular collapse. It simply illustrates the physical limitation of using bulk flow as a complete description of oxygen transport.
More flow is not automatically better capillary oxygen delivery.
A target can improve while the patient does not
Now put the pieces together.
A septic patient receives an intervention intended to improve oxygen transport. Cardiac output rises and ScvO₂ rises. The monitor appears to confirm successful resuscitation.
That may be exactly what happened.
But perhaps haemoglobin fell substantially. Perhaps venous pressure rose. Perhaps oedema increased. Perhaps the additional systemic flow failed to correct the regional perfusion deficit that was causing the clinical problem.
ScvO₂ can accurately report improvement in one component of global oxygen transport while remaining unable to tell us whether the intervention was beneficial overall.
There is an even more difficult possibility. If actual oxygen uptake falls, ScvO₂ may rise further. That fall in uptake could reflect an entirely appropriate reduction in metabolic demand, or it could accompany failure of delivery or utilisation. The number cannot distinguish those possibilities either.
A treatment-responsive variable is not automatically a validated surrogate for successful resuscitation.
Rivers and early goal-directed therapy
This is where ScvO₂ moved from physiological observation to clinical doctrine.
In 2001, Rivers and colleagues published their landmark trial of early goal-directed therapy in 263 patients with severe sepsis or septic shock. The six-hour strategy targeted CVP, MAP and then an ScvO₂ of at least 70%. When ScvO₂ remained low after the earlier targets had been addressed, the protocol included red-cell transfusion to a haematocrit of 30%, followed by dobutamine where necessary.
Hospital mortality was 30.5% in the EGDT group compared with 46.5% with standard treatment.
It was an extraordinary result and it changed practice.
But the trial tested an entire early resuscitation strategy. It did not isolate ScvO₂, and it did not establish that moving an ScvO₂ from, say, 65% to 72% was the component responsible for better outcomes.
Nor was the protocol simply a licence to give fluid until ScvO₂ reached 70%. Fluid was initially linked to CVP targets; persistent low ScvO₂ later in the algorithm could trigger transfusion and inotropic support.
The history is therefore more interesting than a simple story of an incorrect measurement falling out of fashion. A physiological observation became embedded in a successful treatment bundle and gradually acquired the status of an endpoint in its own right.
Seventy per cent began to look like the destination.
What did 70% really mean?
There is nothing magical about 70%.
Different patients have different haemoglobin concentrations, metabolic requirements, extraction reserve, regional blood-flow distribution and microvascular function. There is no reason to think that one venous saturation represents a universal border between adequate and inadequate tissue perfusion.
That does not make thresholds inherently wrong. Medicine uses pragmatic thresholds constantly. A threshold does not need to represent a fundamental physiological boundary to be clinically useful.
But if crossing that threshold becomes a treatment objective, the strategy built around it needs to demonstrate benefit.
Physiological plausibility is not enough.
The later trials
Over the following decade, the unique status of ScvO₂-guided EGDT became increasingly difficult to defend.
In 2010, Jones and colleagues compared an early resuscitation strategy using lactate clearance with one targeting ScvO₂. Lactate clearance met the trial’s criterion for non-inferiority. That did not make lactate a direct measure of tissue oxygenation, but it weakened the claim that ScvO₂ itself was indispensable.
Then came ProCESS, ARISE and ProMISe. The PRISM individual-patient meta-analysis subsequently combined these three major contemporary trials. Ninety-day mortality was 24.9% with EGDT and 25.4% with usual care.
Protocolised EGDT no longer improved outcome over contemporary resuscitation.
That does not prove that ScvO₂ contains no useful information. It does show that mandatory targeting of it was not providing the unique benefit once attributed to the approach.
What is ScvO₂ actually good for?
Suppose a patient has an ScvO₂ of 72% and, without an obvious change in haemoglobin or arterial oxygenation, it falls rapidly to 50%.
I would want to know why.
Perhaps cardiac output has fallen. Perhaps metabolic demand has changed. Perhaps something important has happened that has not yet been recognised.
The number has done something useful: it has created a reason to investigate.
That is different from saying that an ScvO₂ of 50% tells us which treatment the patient needs.
ScvO₂ can be a useful signal. It is a poor diagnosis.
Measure the question you are actually asking
Many of the things we once tried to infer from ScvO₂ can be assessed more directly.
If the question is whether flow is low, measure flow. Echocardiography, LVOT VTI, thermodilution or an appropriate cardiac-output monitor can get much closer to the variable of interest than a venous saturation influenced by haemoglobin, arterial saturation and oxygen consumption.
If the question is whether increasing cardiac filling will increase flow, use a dynamic manoeuvre and measure the change in stroke volume or cardiac output. A vasopressor challenge is one option. Passive leg raising is another. PPV and SVV can be useful under the right conditions. A carefully judged fluid challenge may provide the same information.
If the question is oxygen-carrying capacity, measure haemoglobin and arterial oxygenation.
If the question is tissue perfusion, assess tissue perfusion. Capillary refill time is imperfect, but it is at least looking at a vascular bed rather than trying to infer tissue perfusion from a global oxygen balance. Add mentation, urine output, skin perfusion, lactate trajectory and organ function according to the clinical context.
None of these measurements is perfect, and none should become the next universal number to optimise.
The point is simpler: match the measurement to the question. Before using a surrogate, ask whether you can measure the thing you actually want to know.
Back to the circulation
ScvO₂ gives me an indirect, confounded answer to several important questions and a definitive answer to none of them. That is why I never use it.
The useful bedside question is not “how do I raise the ScvO₂?” It is “what physiological problem am I trying to correct?”
If flow is inadequate, identify the constraint. If you intervene because you think that constraint is limiting flow, measure whether stroke volume responds as predicted. Then ask whether tissue perfusion improved and what the intervention cost.
I will come back to that approach properly in the final article of this series.
ScvO₂ illustrates the central problem rather neatly. Its attraction was that it integrated a great deal of physiology into one number. Cardiac output, oxygen content, metabolic demand and extraction all influence the result.
That integration is also its weakness.
ScvO₂ can tell us that the oxygen balance has changed. It cannot uniquely tell us why. It cannot choose the treatment. It cannot guarantee adequate regional perfusion. And making it rise does not establish that the patient has benefited.
Once the actual clinical question is measured directly, ScvO₂ has very little left to add.
Key references
Rivers E, et al. Early Goal-Directed Therapy in the Treatment of Severe Sepsis and Septic Shock. N Engl J Med. 2001.
https://www.nejm.org/doi/full/10.1056/NEJMoa010307
Jones AE, et al. Lactate Clearance vs Central Venous Oxygen Saturation as Goals of Early Sepsis Therapy. JAMA. 2010.
https://pmc.ncbi.nlm.nih.gov/articles/PMC2918907/
PRISM Investigators. Early, Goal-Directed Therapy for Septic Shock — A Patient-Level Meta-Analysis. N Engl J Med. 2017.
https://www.nejm.org/doi/full/10.1056/NEJMoa1701380
Systematic review of ScvO₂, SvO₂ and cardiac index. 2024.
https://pubmed.ncbi.nlm.nih.gov/39258966/
De Backer D, et al. Microvascular blood flow is altered in patients with sepsis. Am J Respir Crit Care Med. 2002.
https://pubmed.ncbi.nlm.nih.gov/12091178/
Quispe-Cornejo AA, et al. Effect of fluid administration on haemoglobin concentration in critically ill and acutely ill patients. 2022.
https://link.springer.com/article/10.1186/s13054-022-04191-x