The coupling is real and it survives anesthesia. What does not survive inspection is the arithmetic — flow overshoots oxygen use during activation, and the lower limit of autoregulation everyone draws at 50 is a population average that sits below where the average actually is.
Flow tracks metabolism in direction but not in proportion. On activation, oxygen consumption rises about 5% while blood flow rises 30% to 50% — an order of magnitude more delivery than the metabolic demand appears to require, and the reason functional MRI works at all. The same looseness shows up where it costs something: the autoregulation curve drawn as a plateau between 50 and 150 is a statistical description of a population, and one careful review puts the average lower limit in normotensive adults no lower than a MAP of 70.
Charles Roy and Charles Sherrington proposed in 1890 that the brain controls its own blood supply according to local metabolic need. More than 130 years later the proposition still holds and the mechanism still does not have a name. An overwhelming number of candidate mediators have been advanced — hydrogen ion, potassium, adenosine, glycolytic intermediates, phospholipid metabolites, nitric oxide, signals from both neurons and astrocytes — and the honest summary is that no single mechanism appears preeminent. The system has redundancy built into it.
That is not a failure of the physiology; it is a description of it. But it does mean the confident diagram in the lecture is standing on a mechanism nobody has closed, and there is a specific objection worth knowing. If flow rose because a vasoactive metabolite accumulated, the metabolite would have to arrive first. The required timing has never been adequately demonstrated.
Flow-metabolism coupling is quick enough to see. Move a hand and flow rises in the contralateral cortex. Look at something and velocity in the posterior cerebral arteries rises almost immediately. Coupling is preserved through sedation and general anesthesia, and it is visible even during anesthetic-induced burst suppression.
What the coupling does not do is match magnitudes. When flow and oxygen consumption are measured independently by PET, sensory stimulation raises CMRO₂ by around 5% and raises CBF by around 30% to 50%. Glucose metabolism rises with the flow rather than with the oxygen use. This is a large and durable mismatch, and the debate about whether it reflects real anaerobic metabolism has been running since the observation was first made — transient lactate production during photic stimulation on one side, a rapid early rise in tissue deoxyhaemoglobin on the other.
The current reading is that activation triggers immediate anaerobic glucose metabolism to fund glutamate release, and that clearing the glutamate then requires glucose oxidation in excess of oxygen use, producing a net efflux of lactate that is oxidized later.
Two things to take from this. The first is that “flow follows metabolism” is a statement about direction, not about proportion. The second is practical: when CBF and CMRO₂ move together but flow runs far ahead — most starkly during seizure activity — you get hyperemia, and hyperemia in a closed box is a volume problem.
Coupling is generally preserved under anesthesia, but not equally by every agent. Intravenous agents such as propofol appear to preserve flow-metabolism coupling better than the volatiles. The volatiles add a direct vasodilating action on top of whatever metabolic suppression they produce, so flow can be higher than the metabolic rate would predict — visible in the classic finding that for a given CMRO₂, flow is higher under isoflurane than under halothane.
For most cases this is a fine distinction. For a tight brain it is the whole argument, and it is the reason the choice of maintenance agent is a neuroanesthetic decision rather than a preference.
Now the part worth changing your mind about.
The autoregulation curve is drawn as a flat plateau between a cerebral perfusion pressure of about 50 and about 150 mm Hg. Look at how the source text itself qualifies that. The plateau is not actually flat, only relatively flat compared with the slopes at either end. It widens with hypocapnia and with intravenous anesthetics. It is “only a statistical description of how the general population responds,” and a value of 50 mm Hg, even in a normotensive individual, does not guarantee that a particular patient’s circulation is on it. Individual responses vary widely. The relevance of the idealized curve — in particular the lower limit — has been questioned.
Drummond’s 2019 review is that questioning made specific: the average lower limit of cerebral blood flow autoregulation in normotensive adult humans is not less than a MAP of 70 mm Hg, and the average threshold for the onset of central nervous system ischemic symptoms is probably a MAP of 40 to 50 mm Hg.
Put those two figures on a line and the taught number sits between them, closer to the symptom threshold than to the limit it is supposed to represent. That is not a reason to chase pressure, and it is not a new floor to memorize — substituting 70 for 50 makes the same mistake with better arithmetic. It is a reason to know what this patient’s blood pressure normally is before deciding what you are defending, and to treat the round number as inherited rather than established.
Once vasodilation is exhausted, flow falls passively with pressure. Oxygen extraction rises to compensate first; when extraction is maximal, CMRO₂ begins to fall, synaptic transmission fails, and the EEG flattens. Below that comes membrane failure — sodium, calcium and water in, potassium out — and cytotoxic oedema.
The thresholds are worth carrying as a ladder rather than as a single number. Tissue receiving roughly 18 to 23 mL/100 g/min is functionally silent but recovers whenever flow is restored. Below that range, whether it infarcts is a question of duration. And note where the EEG sits on that ladder: it goes quiet above the membrane-failure threshold, which is exactly why any EEG sign of ischemia should be treated as carrying the potential for irreversible damage rather than as a marker that there is still room underneath.
No randomized trial addresses this.
That flow and metabolism are coupled, and that the coupling is generally preserved under sedation and general anesthesia, is established and shown by several independent methods. Nothing on this page disputes it.
Three things beyond that are genuinely unsettled, and it is better to learn them as unsettled than to learn a clean version and unlearn it later.
The mechanism. More than 130 years after Roy and Sherrington proposed local metabolic control, no single mediator has been shown to be preeminent. Hydrogen ion, potassium, adenosine, glycolytic intermediates, phospholipid metabolites, nitric oxide, neuronal and glial signalling have all been implicated; the system has built-in redundancy. A standing objection to the pure metabolic theory is that the required timing — the accumulation of a vasoactive metabolite arriving before the flow increase — has never been adequately demonstrated.
The arithmetic. During activation, CBF and the cerebral metabolic rate for glucose rise far more than CMRO₂ does. Whether that reflects genuine anaerobic metabolism has been argued since the observation was first made, and the evidence still points both ways.
The lower limit. The 50–150 mm Hg plateau is described in the source text itself as "only a statistical description of how the general population responds," with the note that a value of 50 mm Hg does not guarantee any particular patient is on the plateau, and that the relevance of the idealized curve has been questioned. Drummond's review is the sharp version of that questioning: the average lower limit in normotensive adults is not below a MAP of 70.
The failure mode here is subtle because nothing about it looks like an error. You draw the plateau, you know the numbers, you keep the MAP above 50, and the case is uneventful. The graph is not wrong; it is a summary of what a population does, being used as though it described one person.
Two things follow from that, and they run in opposite directions. The first is that the plateau is not flat and the corners are not corners — the "plateau" is only relatively flat compared with the slopes at either end, and it widens with hypocapnia and with intravenous anesthetic agents. The second, and the one worth carrying: the reviewed average lower limit in normotensive adults is a MAP of no less than 70, while the average pressure at which conscious volunteers begin to show central nervous system ischemic symptoms is somewhere around 40 to 50. The number most people are taught sits in the space between those two.
That is not an argument for chasing pressure. It is an argument against the specific comfort of a round number, and for knowing what this patient's pressure normally is before you decide what to defend.
Full text · Preston goes through LibKey to whatever Preston holds — the PDF where there is one — and works on a phone. The doi link always reaches the publisher's record, whether or not we hold a subscription. If you are reading on a computer, LibKey Nomad adds the same one-click PDF button to PubMed and publisher pages everywhere else; it is a desktop browser extension, so it does nothing on a phone or tablet.
Positron emission tomography during sensory stimulation allows cerebral blood flow and oxygen utilization to be measured independently. What does it show?
A textbook describes cerebral autoregulation as maintaining flow across a cerebral perfusion pressure of roughly 50 to 150 mm Hg. A 2019 review argues that the average lower limit of autoregulation in normotensive adults is not below a mean arterial pressure of 70. Which reading is best supported?
A patient with a supratentorial mass has been ventilated to a PaCO₂ of about 25 mm Hg for eight hours in the intensive care unit. The team asks you to return the PaCO₂ to normal. What should you anticipate?
No returning items before day 4. The retrieval window opens once there is something far enough back to be worth retrieving.