The Reflection Coefficient

01

The pearl

In peripheral capillaries sodium crosses freely, so only protein holds water in the vessel and oncotic pressure is what matters. The blood-brain barrier has an estimated pore size of about 7.9 Å and excludes even sodium, so it behaves like the membrane of an osmometer: water follows total osmolality, and each mOsm of difference is worth more than 19 mm Hg. That is the whole mechanism of osmotherapy, the reason albumin has no advantage here, and the reason a litre of lactated Ringer's is not a neutral act.

02

Why it is true

mechanism

Everything you know about fluids was learned on the wrong capillary.

In peripheral tissue the endothelium does not form tight junctions and the pores may be several hundred times larger than those in brain. Sodium, chloride and glucose cross freely; albumin, at roughly 69,000 molecular weight, does not. So the only thing holding water inside a peripheral vessel is the plasma oncotic pressure, produced predominantly by albumin. That is the Starling picture, and in lung, bowel and muscle it is correct.

The cerebral capillary is different in one respect that changes the whole answer. Its estimated pore size is about 7.9 Å, small enough to prevent movement of even very small particles. In effect the blood-brain barrier acts like the semipermeable membrane of an osmometer, and water crosses it according to the relative concentration of all solutes on each side — dominated numerically by sodium, not by protein.

That is the reflection coefficient in a sentence: at this membrane, sodium is an effective osmole. And the leverage is considerable. A pressure of more than 19 mm Hg is generated for each mOsm of difference across a semipermeable membrane, so gradients that look trivial on a laboratory report are not trivial at all.

What follows immediately

Hypertonic crystalloid works. 3% saline at about 1026 mOsm/L and 20% mannitol at about 1098 mOsm/L both establish a large gradient across an intact barrier and pull water out of brain.

Colloid confers no particular advantage. Oncotic pressure was never the operative variable here. In a comparison of hetastarch against saline, no beneficial effect on oedema formation was demonstrated, and the standing recommendation is that colloids should be restricted in patients at risk of intracranial hypertension.

Slightly hypotonic fluids are not neutral. Lactated Ringer’s at about 273 mOsm/L against plasma near 295 moves water the wrong way across a completely intact barrier. One bag is nothing. Six hours of maintenance is a dose.

Where the barrier is broken

The interesting question is what happens after injury. If the barrier is partially disrupted, do cerebral vessels start behaving like peripheral ones?

The honest answer is that the experimental evidence is not conclusive. What is reasonably clear is the clinical consequence: osmotic agents work by dehydrating tissue where the barrier is intact. In effect, normal brain is dehydrated to compensate for the injured region — which is why osmotherapy helps the swollen brain overall without doing much inside the lesion itself.

There is a specific instance worth knowing. Mannitol has been shown to leak into white matter near gliomas, plausibly because of altered permeability around tumours, and that leak can exacerbate peritumoral oedema and produce ICP rebound. The drug that draws water out of normal brain can, in the wrong tissue, become the solute that draws water in.

What the trials say

Three results, and they do not point the same way.

Hypertonic saline versus mannitol. A meta-analysis of 5 randomized trials — 112 patients, 184 episodes of raised pressure — found a relative risk of ICP control of 1.16 (95% CI 1.00–1.33) favouring hypertonic saline, and a difference in mean ICP reduction of 2.0 mm Hg (95% CI −1.6 to 5.7). The authors concluded hypertonic saline is more effective and called for larger trials. Look at those intervals before you repeat the conclusion: the lower bound touches 1.00, and the pressure difference crosses zero.

Continuous hypertonic saline after traumatic brain injury. COBI randomized 370 patients with moderate to severe injury to continuous 20% hypertonic saline plus standard care, or standard care alone. The adjusted common odds ratio for extended Glasgow Outcome Scale at 6 months was 1.02 (95% CI 0.71–1.47), P = 0.92. A strategy of keeping the number low, applied prophylactically, did not change the outcome.

Albumin. In the traumatic brain injury cohort of the SAFE study, 24-month mortality was 33.2% with albumin against 20.4% with saline, relative risk 1.63 (95% CI 1.17–2.26); in severe injury, 41.8% against 22.2%. In moderate injury there was no significant difference. The mechanism is still argued over — 4% albumin is mildly hypo-osmolar, which is the explanation that fits this page, and it has not been proven to be the cause.

Holding the claim at the right size

Osmotherapy is one of the few interventions that reliably does what it is supposed to do. It lowers intracranial pressure, quickly, by a mechanism that is completely understood, and when a brain is swelling in front of you it is the right thing to reach for.

What the trials add is that lowering the pressure and helping the patient are different achievements, and only the first has been demonstrated. Bolus therapy for an acute rise remains well founded; a policy of continuous prophylactic infusion has been tested and did not deliver. Keeping those two apart is most of what a resident can usefully carry from this literature.

And measure the sodium. Every 4 to 6 hours when hypertonic saline is running, because acute rises above 170 mEq/L are likely to produce depressed consciousness or seizures — at which point the treatment has become indistinguishable from the disease.

Small solutes — Na⁺, Cl⁻, glucose, mannitol Large solutes — albumin, starch, dextran

Figure 1. One difference in pore size, two different rules. Because the cerebral capillary retains sodium as well as albumin, the osmotic gradient that moves water is the total one — which is why a 3% saline solution at 1026 mOsm/L is a powerful dehydrating agent and a 5% albumin solution at 290 mOsm/L is not, whatever its oncotic pressure. Particles and pores are not to scale, and the osmolality scale is broken between 320 and 1000.

03

The evidence

2 RCTs (n = 830) · 1 meta-analysis of 5 trials

SAFE, traumatic brain injury cohort

SAFE Study Investigators, Australian and New Zealand Intensive Care Society Clinical Trials Group, Australian Red Cross Blood Service. Saline or albumin for fluid resuscitation in patients with traumatic brain injury. N Engl J Med. 2007;357(9):874-884. Harm

Design: Prespecified analysis of a randomized, blinded resuscitation trial
Population: 460 patients with traumatic brain injury within the SAFE study; 231 received albumin and 229 saline
Intervention: 4% albumin for fluid resuscitation
Control: 0.9% saline
Primary: Death at 24 months
Result: 71/214 (33.2%) with albumin vs 42/206 (20.4%) with saline · relative risk 1.63 (95% CI 1.17–2.26) · P = 0.003
Harm signal: In severe brain injury (GCS 3–8), 61/146 (41.8%) vs 32/144 (22.2%) · relative risk 1.88 (95% CI 1.31–2.70) · P < 0.001. In moderate injury (GCS 9–12) there was no significant difference.

A subgroup, and a convincing one This is a prespecified analysis within a larger resuscitation trial rather than a trial designed in brain injury, so the usual caution about subgroups applies. Two things reduce it: the effect is large, and it concentrates in the severe-injury group where the physiological argument predicts it should. The mechanism remains debated — a 4% albumin solution is mildly hypo-osmolar, which is the explanation that fits this day, but it has not been demonstrated to be the cause.

COBI

Roquilly A, Moyer JD, Huet O, et al. Effect of continuous infusion of hypertonic saline vs standard care on 6-month neurological outcomes in patients with traumatic brain injury: the COBI randomized clinical trial. JAMA. 2021;325(20):2056-2066. No benefit

Design: Multicentre randomized open-label trial
Population: 370 patients with moderate to severe traumatic brain injury
Intervention: Continuous infusion of 20% hypertonic saline plus standard care
Control: Standard care alone
Primary: Extended Glasgow Outcome Scale at 6 months
Result: Adjusted common OR 1.02 (95% CI 0.71–1.47) · P = 0.92

What a null trial of a working drug means Hypertonic saline reliably lowers intracranial pressure; this trial tested whether giving it continuously and prophylactically improves the outcome that matters, and it did not. That is a narrower finding than "hypertonic saline does not work" — it says nothing against bolus therapy for an acute rise in pressure, which is the use most residents will actually see. It does say that a strategy built on keeping the number low, rather than on treating episodes, has been tried.

Hypertonic saline versus mannitol

Kamel H, Navi BB, Nakagawa K, Hemphill JC, Ko NU. Hypertonic saline versus mannitol for the treatment of elevated intracranial pressure: a meta-analysis of randomized clinical trials. Crit Care Med. 2011;39(3):554-559. Benefit

Design: Meta-analysis of randomized clinical trials
Population: 5 trials, 112 patients, 184 episodes of elevated intracranial pressure
Question: Which agent controls raised intracranial pressure more reliably?
Result: Relative risk of ICP control 1.16 (95% CI 1.00–1.33) favouring hypertonic saline · difference in mean ICP reduction 2.0 mm Hg (95% CI −1.6 to 5.7)

How small this actually is 112 patients across five trials, with a confidence interval whose lower bound touches 1.00 and a mean pressure difference whose interval crosses zero. The authors themselves note the constraint and call for larger trials. It is enough to justify hypertonic saline as an alternative to mannitol; it is not enough to justify a policy, and the endpoint is a pressure rather than a patient.

Synthesis. The physiology is settled and the therapeutics are not, and the gap between those two sentences is the day.

What is established: the blood-brain barrier excludes small ions, so water movement into brain follows total osmolality; hypertonic solutions therefore dehydrate brain wherever the barrier is intact; and each mOsm of gradient is worth more than 19 mm Hg of driving pressure. Nothing in the trials disturbs any of that.

What the trials add is a consistent and slightly deflating pattern. Hypertonic saline controls intracranial pressure marginally better than mannitol across five small trials with 112 patients. Giving it continuously to prevent raised pressure after traumatic brain injury did not improve six-month neurological outcome. And resuscitating brain-injured patients with albumin — a colloid, chosen on oncotic reasoning that does not apply to this capillary bed — was associated with substantially higher mortality.

The through-line is that the mechanism reliably predicts what happens to the pressure and unreliably predicts what happens to the patient. That is not a reason to stop using osmotherapy, which remains one of the few genuinely effective things available when a brain is swelling in front of you. It is a reason to keep the claim proportionate: these agents move a number that matters, and only one of the trials here measured whether moving it helped.

04

What you actually do

in the room

Choose fluids by osmolality, not by oncotic pressure, whenever the brain is at risk. The cerebral capillary retains sodium, so the gradient that moves water is the total one. This is why a colloid confers no particular advantage here and why the tonicity of your maintenance crystalloid is a neurological decision.
Treat lactated Ringer's as a hypotonic fluid in this context. At about 273 mOsm/L against plasma near 295, it moves water into brain across an intact barrier. Over a long case that is a dose, not a rounding error.
Reach for hypertonic saline or mannitol for an acute rise, and expect them to work through normal brain. Osmotic agents dehydrate tissue where the barrier is intact — in effect, normal brain is dehydrated to make room. That is also why they do comparatively little inside a region whose barrier has broken down.
Prefer hypertonic saline when the patient is hypovolaemic or haemodynamically unstable. Mannitol produces a pronounced diuresis, which is the last thing an underfilled patient needs. Comparative studies of brain relaxation in craniotomy have found hypertonic saline at least equivalent and better tolerated in this respect.
Measure the sodium every 4 to 6 hours when hypertonic saline is running. That is the recommended interval. Acute rises above 170 mEq/L are likely to produce a depressed level of consciousness or seizures — the treatment becoming the problem.
Expect mannitol to leak where the barrier is broken, and to rebound. Mannitol has been shown to leak into white matter near gliomas, plausibly because of altered permeability around tumours. That leak can exacerbate peritumoral oedema and produce ICP rebound.
Do not resuscitate a brain-injured patient with albumin. In the SAFE traumatic brain injury cohort, mortality at 24 months was 33.2% with albumin against 20.4% with saline, and 41.8% against 22.2% in severe injury. Whatever the mechanism, the direction is not in doubt.
Distinguish treating an episode from running a strategy. Bolus osmotherapy for an acute rise in pressure is well founded. Continuous prophylactic hypertonic saline after traumatic brain injury was tested against standard care and did not improve six-month outcome.

05

The pitfall

Importing the Starling equation from the wrong capillary

Every anaesthetist learns fluid physiology on peripheral capillaries, where the endothelium is leaky to small molecules and the only thing holding water in the vessel is protein. From that comes a set of instincts — colloid stays intravascular, crystalloid leaks into tissue, oncotic pressure is what you manage — and every one of them is misleading at the blood-brain barrier.

The cerebral capillary has an estimated pore size of about 7.9 Å and tight junctions that exclude even sodium and chloride. Peripheral capillary pores may be several hundred times larger. So the cerebral capillary behaves like the membrane of an osmometer, and the gradient that moves water is the total concentration of all solutes — dominated numerically by sodium, not by albumin.

Two practical consequences follow, and they run in opposite directions from the peripheral intuition. A hypertonic crystalloid is a powerful brain dehydrating agent, because sodium is an effective osmole here. And a colloid confers no particular advantage, because the protein it contains was never the operative variable — the SAFE trial found albumin resuscitation associated with substantially higher mortality after brain injury, and 4% albumin is mildly hypo-osmolar.

The version of this error you are most likely to commit is duller than either. It is running maintenance fluid for a long craniotomy without checking its osmolality, and discovering at the end that several litres of something slightly hypotonic have gone in.

06

Sources

primary first

01 SAFE Study Investigators, Australian and New Zealand Intensive Care Society Clinical Trials Group, Australian Red Cross Blood Service. Saline or albumin for fluid resuscitation in patients with traumatic brain injury. N Engl J Med. 2007;357(9):874-884. Full text · Preston doi 10.1056/NEJMoa067514
02 Roquilly A, Moyer JD, Huet O, et al. Effect of continuous infusion of hypertonic saline vs standard care on 6-month neurological outcomes in patients with traumatic brain injury: the COBI randomized clinical trial. JAMA. 2021;325(20):2056-2066. Full text · Preston doi 10.1001/jama.2021.5561
03 Kamel H, Navi BB, Nakagawa K, Hemphill JC, Ko NU. Hypertonic saline versus mannitol for the treatment of elevated intracranial pressure: a meta-analysis of randomized clinical trials. Crit Care Med. 2011;39(3):554-559. Full text · Preston doi 10.1097/CCM.0b013e318206b9be

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.

07

Check yourself

3 items · one attempt each

Item 1 Analysis Neuro 1 · Day 10

Why is total osmolality, rather than oncotic pressure, the variable that governs water movement across the blood-brain barrier?

Item 2 Recall Neuro 1 · Day 10

In the traumatic brain injury cohort of the SAFE study, patients were resuscitated with 4% albumin or 0.9% saline. What was found at 24 months?

Item 3 Application Neuro 1 · Day 10

During a craniotomy for a large tumour the brain is tight. The patient has had a long preoperative fast, is mildly hypotensive and is not obviously volume replete. Which osmotic agent is preferable and why?