Cerebral blood flow autoregulation

This chapter is relevant to Section G4(ii) of the 2017 CICM Primary Syllabus, which expects the exam candidate to "describe the distribution of blood volume and flow in the various regional circulations ... including autoregulation...  To finish butchering that quote, the list of regional circulations has to " ...include, but not limited to, the cerebral...". In short, it appears the syllabus calls for a discussion of the normal mechanisms which maintain stable demand-matched cerebral perfusion in the face of wildly fluctuating systemic conditions. Though we are talking about the brain, it is really the behaviour of its blood vessels we are interested in, which somewhat justifies the author's decision to group this topic with the cardiovascular notes. It has been an extremely popular topic for CICM First Part written questions, and because of its importance, it can be expected to appear again and again. Examples include:

• Question 20 from the first paper of 2021
• Question 14 from the first paper of 2014
• Question 16 from the second paper of 2011
• Question 11 from the second paper of 2009
• Question 5 from the first paper of 2008

There is also some mention of this in the Fellowship exam, but of course at that stage the trainees have already reached the level cap and are preparing for the boss fight - so nobody really cares whether they have any grasp of the fundamentals. Question 1 from the second paper of 2009 briefly touched upon the definition of cerebral perfusion, and then went on to ask more pragmatic details about the utility of using CPP as a therapeutic target. The role of cerebral blood flow autoregulation in the pathogenesis of PRES (posterior reversible leukoencephalopathy syndrome) is also touched upon in Question 14.1 from the first paper of 2016. Neither of these required as much detail as is about to be presented below. For either exam, the following brief summary will easily suffice:

A great review by Busija & Heistadt (2005) would be perfect, but Springer have it under lock and key. Mchedlishvili (1980) is old, but good, and free. Looking at articles which helped to develop this chapter the most, a majority appear to be chapters from Welch's Primer on Cerebrovascular Diseases. From various editions of this book, Yang & Liu (2017),  Chillon & Baumbach (1997), Golanov (1997) and Traysman (2017) were excellent references.

Cerebral blood flow and vascular anatomy

The brain is a hungry organ, to the extent that some authors have described it as having an "avaricious appetite",  implying some sort of corrupt acquisitive materialism.  Its metabolic requirements are discussed elsewhere, and from that discussion the most important take away point is that the most profligate tenants of the skull, the neurons of the grey matter, have a rapacious taste for glucose, and they insist on metabolising it in the most bourgeoise fashion, which is aerobically with oxygen. Clearly this calls for an abundant supply of blood, as it carries both the substrates in question. Thus, blood flow to the brain is quite high: about 50ml per 100g of tissue per minute, or about 700ml/min in total for a standard 1400g brain, which is 14% of a normal cardiac output. In case you are wondering how this blood flow is distributed within the cerebral circulation, an excellent study by Zarrinkoob et al (2015) has the answers. The investigators measured cerebral blood flow in the major arteries of the cerebral circulation using phase-contrast MRI. In the childish diagram below, their data is mapped onto this stock diagram of the cerebral circulation:

This rate of oxygen delivery roughly matches the metabolic demand of the different regions of the brain, which is discussed below.  Of course, these percentages describe blood flow distribution, but not tissue perfusion or oxygen delivery. In case some sadistic SAQ in the future asks to "describe the factors which influence oxygen delivery to the brain", it would probably be important to come out with banal factors such as the oxygen-carrying capacity of the blood (eg. the haemoglobin concentration and its saturation).

Factoring these in, it is possible to calculate the DO₂ for the brain. Quoting some studies by Imre et al which cannot be accessed electronically, Wolff (2008) reports that the DO₂/VO₂ ratio for the brain is normally about 3:1, i.e. the brain is supplied with about three times as much oxygen as is required for its normal resting function. This has implications for transfusion thresholds for traumatic brain injury patients, as one might naturally conclude from this that the oxygen flux to the brain has some sort of built-in buffer and that a relatively large drop in DO₂ could be tolerated. Unfortunately,  that does not appear to be the case. Autoregulatory mechanisms seem to kick in and increase blood flow (East et al, 2018) - which might sound good, but is in fact detrimental, as it increases intracranial pressure. Ergo, though no specific guidelines exist to that effect, upon being questioned about half of the surveyed European intensivists reported using a Hb of 90g/L as their transfusion threshold in TBI (Badenes et al, 2017).

Cerebral perfusion pressure

Working from Ohm's law, pressure is the product of resistance and flow:

Q = (P_a- P_v) / R

where

• P_a- P_v = the pressure difference between the arterial and venous sides of the cerebral circulation, or the cerebral perfusion pressure (CPP)
• Q = blood flow, and
• R = cerebral vascular resistance

This P_a- P_v or CPP is the difference between cerebral arterial and cerebral venous pressure, i.e. the pressure drop across the cerebral circulation. As we have few ways of measuring the pressure in the dural venous sinuses, conventionally we resort to using the intracranial pressure as a surrogate. Thus, cerebral perfusion pressure is the ICP subtracted from the mean arterial pressure (MAP). Or the CVP, for that instance. It is not inconceivable that one's CVP might be higher than one's CSF pressure in the context of some sort of severe right heart problem.

Central venous pressure. Though it was said elsewhere that perfusion pressure (the pressure drop across a vascular bed) should be viewed more as the product of flow and resistance rather than their cause, one should not underestimate the role of pressure in producing flow (flow only exists where there is a pressure difference, so pressure is still pretty important). From this, we can surmise that anything that decreases the pressure difference across the circulation will decrease cerebral blood flow. When CVP is higher than ICP, the CVP becomes the major impedance to blood flow through the brain:

CPP = MAP - CVP

Thus, you can achieve a decreased pressure gradient by either lowering the arterial pressure or by increasing the venous pressure. The effects of increasing the CVP from 0 to 30 mmHg will have the same functional effect as decreasing MAP from 60 to 30 mmHg: the CPP will drop in both cases, and cerebral blood flow will suffer. 

Intracranial pressure is what normally has the effect of impeding blood flow through the brain, because the CVP is usually much lower than the ICP. The normal version of the formula is:

CPP = MAP - ICP

Anyway. Though cerebral perfusion pressure is usually described as the driving gradient for cerebral blood flow, it is probably more logical to discuss pressure as the result of flow, rather than the cause. Pressure is generated when flow is directed into a conduit which resists flow, making resistance the more important factor. And through this clumsy segue, the reader can already see the dim outlines of the Hagen-Poiseuille equation.

Cerebral vascular resistance

As one might recall from virtually every discussion of factors which influence blood pressure, vascular resistance is usually described using the aforementioned equation, which is reproduced below with soul-crushingly tedious predictability:

R = (8 l η) / πr⁴

where 

• l = length of the vessel
• η = viscosity of the fluid
• r = radius of the vessel

Without labouring the issue any further, it is worth considering that the length of cerebral vessels and the viscosity of the blood are not susceptible to rapid adjustment, and so cerebral blood flow autoregulation is really an issue of controlling cerebral vessel diameter. In case anybody ever asks you to quote normal values (they won't), Kety et al (1948) measured a mean value of 1.6 mmHg/ml/100g/min in healthy adults and 3.0 mmHg/ml/100g/min for hypertensive subjects. Those numbers will mean nothing to most normal people, and one should probably clarify that this is a very low value. The brain is an organ which wants flow, and its vessels will put up only a minimal resistance. To use more conventional notation, the usually measured cerebral vascular resistance values are 0.3-1.4 Woods units (Jalan et al, 2001); for comparison, the vascular resistance of the kidney measures  2.8 WU, the myocardium 7.9 WU, and abdominal skin about 200 WU  (Karlsson et al, 2003).

The effect of blood viscosity on cerebral blood flow should probably be discussed at least in passing. It was mentioned briefly in the examiner comments to Question 5 from the first paper of 2008. Blood viscosity is not a completly immutable property, and it can occasionally become relevant - for example, in scenarios where it is dangerous elevated. The higher the viscosity, the lower the flow rate, provided all the other factors remain unchanged. This can be due to either an excess of cells (particularly leucocytes) or due to an excess of protein (eg. in myeloma).  To give you the idea of the scales and magnitudes involved,  Lenz et al (2007), by increasing the viscosity of rat blood threefold using contact lens lubricant, managed to basically halve their cerebral blood flow.

Autoregulation of cerebral blood flow

The brain, like many other tissues, has the ability to control its vascular resistance through vasoconstriction and vasodilation, thus modulating its own blood flow. It is a purely local mechanism, as far as anybody can tell. Cerebral metabolic demand is the main regulator of regional cerebral blood flow, and this regulation occurs automatically, probably in response to the abundance or deficit of various local factors -  mainly metabolic byproducts and metabolic substrates:

• Carbon dioxide concentration in the brain parenchyma
• pH of the blood
• Lactate
• Potassium
• Low oxygen

All of these metabolic factors affect the relationship between systemic blood pressure and cerebral blood flow. In other words, when cerebral metabolic demand is high (substrate levels are low, metabolite levels are high), cerebral blood flow will be higher at any given perfusion pressure because cerebral vascular resistance will decrease. Conversely, where cerebral metabolic demand is stable and perfusion pressure is changing, the same mechanisms ensure that blood flow remains constant and matched to demand. Thus, cerebral blood flow autoregulation can be defined as:

"A homeostatic process that regulates and maintains cerebral blood flow (CBF) constant and matched to cerebral metabolic demand across a range of blood pressures."

- a paraphrase of Armstead, 2016

If this sounds like a difficult concept to explain over a ten-minute written answer, Question 14 from the first paper of 2014 gave specific directions on how the trainees should approach their written synopsis of these relationships."Using a diagram, explain the effect of PaO2, PaCO2 and MAP (Mean Arterial Pressure) on cerebral blood flow (CBF)" was the instruction. That diagram:

As you can see, whoever created the first of these had cleverly used the common units of measurement (mmHg) to combine all the factors using the same x-axis. This might be a case of graph abuse, as one might point out that none of the represented data series have a meaningful relationship to one another (i.e. nothing useful can be derived from the point where the oxygen curve crosses the MAP curve, for example). For example, one might observe that the MAP curve and the PaO₂ curve intersect at 60 mmHg, as this happens to be the value at which both plateau, but if you think about it, this is purely a coincidence.

Anyway, this graph succeeds as a convenient means of displaying all three parameters on one chart, and it is present in many textbooks, which means it is probably an acceptable thing to regurgitate in an exam setting, but for an indepth understanding of the subject, clearly something more detailed is required. 

Effect of MAP on cerebral blood flow

Let us consider a scenario where cerebral metabolic demand is stable, and where systemic blood supply is wildly fluctuating. As systemic arterial blood pressure rises and falls, blood flow to the brain must remain stable in order for normal function to continue. In order to achieve this goal, cerebral vascular resistance varies, to achieve a stable plateau of flow over a fairly wide range of systemic arterial pressure. This relationship is usually depicted in a basic diagram, such as this one:

The shape of these relationships is reasonably stable across textbooks, but that does not mean that it is accurate. Particularly, the lower limit of MAP at which autoregulation fails (usually reported as a CPP of 50 mmHg or a MAP of 60 mmHg) is derived from a diagram by Niels Lassen, who originally developed this triphasic concept in 1959. This ancient carving has subsequently been reproduced in all the major publications with what appears to be uncritical acceptance. However, it is important to point out that Lassen used data from eclampsia patients who were infused with a horrific cocktail of 1950s-era vasodilators (veratrum viride, people, seriously?), all of which probably had a cerebral vasodilatory effect. Modern authors have suggested that this has artificially depressed the lower limit of the autoregulatory plateau, and that in healthy normal adults that plateau may be much higher, perhaps as high as 70 mmHg.

The upper third of the diagram, where autoregulation is apparently lost, also may not exist, or at least not for everybody.  For example, when Harper published a seminal paper on this in 1966 (describing the relationship of cerebral blood flow and blood pressure in normocapnic and hypercapnic animals) the investigators tracked their subjects all the way up to a MAP of 180 mmHg and a systolic of "patent pending", with no apparent loss of plateau:

So, this is clearly a very individual matter, unique as a fingerprint. The interpersonal variability of cerebral blood flow in humans was well demonstrated by Strangdgaard et al (1973), who was able to capture MAP/CBF relationships from a group of severely hypertensive patients. As you can plainly see from these stolen graphs, some had the classical increase in blood flow at the higher range of MAP, whereas others did not.

How does this happen? Multiple mechanisms are probably responsible for this phenomenon, and they probably exert their action simultaneously. Chillon & Baumbach (1997) outline at least four: metabolic, neurogenic, endothelial, and myogenic. Without wasting too much of the reader's time on non-examinable speculation, these theories can be summarised as follows:

• The autonomic neurogenic theory is based on the finding that the cerebral vessels have rich autonomic innervation, so surely the autonomic nervous system must play a role in cerebral vascular control. This clearly is not the only mechanism, as animals seem to be able to autoregulate their cerebral blood flow in spite of total autonomic denervation.  
• Endothelial mechanisms are mainly related to the effects of pressure stretch or shear on cerebrovascular endothelium. The endothelial cells, when harassed in some way, respond by secreting paracrine mediators in a panic, thereby producing vasodilation in response to mechanical stress. Peterson & Wang (2011) have described experiments where removing the endothelium from arteries has the effect of attenuating the vasodilatory response to flow. The paracrine signalling molecules being blamed for these effects are nitric oxide (predictably), as well as endothelium-dependent hyperpolarization factor (EDHF), the eicosanoids, and the endothelins.
• Myogenic autoregulation is the findings that blood vessels have the capacity to respond to pressure stimuli by vasoconstriction, and vice versa. For instance, it appears that even the abovementioned vessels denuded of endothelium can do this, albeit in an attenuated fashion, i.e. it is a property of the smooth muscle fibres themselves. This hypothesis is mainly extrapolated from work done on vessels other than those of the brain, but Busija (1984) quotes several studies which reported the contraction of cerebral vessels in response to stretch or pressure.
• Metabolic autoregulation is a hypothetical mechanism of control which relies on using the metabolic activity of astrocytes and neurons for regulating blood flow to the brain. Logically, the endproducts of metabolic pathways make an attractive molecular explanation for a hypothesis like this, and in the course of their metabolism these cells will have released a whole lot of different molecules, and so there are a large number of potential actors which could be mediating vasodilation. Various authors have proposed CO₂, hydrogen ions (i.e a change in pH), falling oxygen concentration, adenosine, potassium and calcium. Unfortunately, we are yet to see a good study which might put one of these mediators front and centre. Most of the time they seems to be the side dish to some weird unknown main course. For instance, when Wahl & Kuchinsky(1979) tried measuring pH and potassium concentration in the immediate vicinity of pial arteries, they found that these concentrations remained essentially the same while the arteries dilated and contracted impressively in response to pressure.

Effect of PaCO₂ on cerebral blood flow

Carbon dioxide promotes increased cerebral blood flow at any given perfusion pressure. The increase in blood flow can be substantial. Subjects breathing 7% FiCO₂ (corresponding to a PaCO₂ of perhaps 70-80 mmHg) in an experiment by Kety & Schmidt (1947) had basically doubled their cerebral blood flow, without much of an increase in their blood pressure. This increase is almost entirely due to the dilation of cerebral vessels. The pharmacology of how CO₂ produces cerebral vasodilation is explored elsewhere. In summary:

• Change of periarteriolar pH leads to a change in nitric oxide synthase activity;
• Nitric oxide synthase catalyses intracellular cGMP production;
• cGMP acts as a second messenger to affect a change in intracellular ionised calcium availability
• The upshot of all this is a decreased cerebral vascular resistance
• If the resistance is decreased but the pressure difference remains the same, the flow increases.
• The increase in flow is by about 1-2ml/100g/min for every 1mmHg increase in CO₂ (Raichle, Posner & Plum, 1970). Conversely, blood flow decreases as CO₂ decreases. Obviously, this is undesirable if your brain is swollen and/or perfusion-compromised. Hence the neurointensivists' obsession with maintaining a stable (low-normal) CO₂ in patients with various intracranial catastrophes.  

Thus, from what little we know, it appears that (for anaesthetised humans) beyond a CO₂ of 55-60 mmHg, cerebral blood flow autoregulation becomes significantly impaired within a physiologically normal range of blood pressure. To illustrate what happens to the normal autoregulatory relationship between MAP and blood flow in hypercapnia, some sort of a diagram is probably called for. One commonly used image is an original series of dog data from Haeggendal & Johansson (1965); another is a stylised diagram from Meng & Gelb (2015). These are combined and presented here as a cleaned-up version designed to be reproduced in exams, with the numbers stolen from Ekström-Jodal et al (1971). 

Alternatively, if one wanted to plot what happens to cerebral blood flow with a stable perfusion pressure as CO₂ increases, one could produce something like this diagram, modified from Widder & Görtler (2011):

 

This probably needs a finishing statement, for people to skip to. The bottom line is that:

• Hypercapnia:
  • Increases cerebral blood flow at any perfusion pressure
  • Narrows the autoregulation plateau
• Hypocapnia does the opposite
• The lower threshold for autoregulation does not appear to be affected
• The effect has a plateau, i.e. beyond a certain CO₂ level no further increase in cerebral blood flow will occur
• Hypercapnia combined with hypoxia has a magnified effect on cerebral blood flow (i.e under hypoxic conditions, the effect of the same PaCO₂ value will be greater)

Effect of PaO₂ on cerebral blood flow

As already mentioned somewhere above, the total flux of oxygen into the brain is rather high, and probably higher than demand by a factor of three under normal circumstances. From this, logically it should follow that a fairly sizeable drop in arterial oxygen content will be tolerated before cerebral vessels step in and intervene in flow management by vasodilating themselves. This, in fact, is what is observed experimentally.

The clean-looking stylised graph on the left is from Golanov & Reis (1997), and though there is no literature reference attached to it, it closely resembles the same graph you would see in most textbooks. They are all probably plagiarising the same rat study by Jóhannsson and Siesjö (1975), whose graph is represented on the right. The main features which one would need to point out to an examiner would be the relatively stable linear relationship at normoxic or hyperoxic levels. Only at around 50 mmHg PaO₂ does the cerebral vascular resistance finally drop. As hypoxia progresses, cerebral blood flow increases exponentially, and can reach completely insane flow rates; as you can see, some of the experimental animals had a 700% increase in CBF at PaO₂ of around 25 mmHg, corresponding to sats of 50%.

Presumably, this is a desperate tactic to maintain cerebral performance in the face of impending hypoxic unconsciousness and one can see how it might come in handy if one were (for example) drowning. With vascular resistance minimised and cardiac output cranked up by hypoxia, every last remaining molecule of oxygen in being made available to the brain to extend the time of useful consciousness.

Wherever textbook authors are being honest with their readers, they do not claim to understand the physiological processes underlying this reaction to hypoxia. "While it is clear that hypoxemia produces cerebral vasodilation and increased CBF, the precise mechanism by which hypoxemia produces this vasodilation is not", confesses Traystman (2017). Various hypothetical mechanisms include nitric-oxide-mediated vasodilation, or some direct smooth muscle activity, or some mysterious PO₂-sensing chemoreceptors in the vessels. None of these have so far emerged as a clear frontrunner.

Impaired cerebral bloodflow autoregulation

The following is a crudely approximate list of pathophysiological conditions which can impair the autoregulation of cerebral blood flow:

• Hypercapnea
• Ischaemic stroke
• Traumatic brain injury
• Global hypoxic brain injury
• Regionally, surrounding a space-occupying lesion or a hematoma
• Infection, eg. meningitis or encephalitis
• Malignant hypertension
• Diabetic microangiopathy (after many years of uncontrolled diabetes)
• Hepatic encephalopathy
• Septic encephalopathy