Decompressive Craniectomy

01

The pearl

The Brain Trauma Foundation's 2020 update recommends the same operation for one purpose and against it for another, and the distinction is timing. Secondary decompressive craniectomy for late refractory intracranial hypertension is recommended to improve mortality and favourable outcomes; for early refractory hypertension it is not recommended for those endpoints. For either early or late, it is recommended to reduce intracranial pressure and the duration of intensive care. Read that carefully: the guideline separates lowering the pressure from helping the patient, and treats them as different indications.

02

Why it is true

mechanism

This is the last day of Neuro 2, and it is the right one to end on, because it is the day where the curriculum’s recurring theme becomes explicit rather than implied.

Neuro 1 day 01 opened the closed box. This day takes the box apart.

What the operation is, and what it certainly does

Decompressive craniectomy is an advanced option for controlling intracranial pressure in patients with diffuse brain swelling refractory to maximal medical management. It works by the most direct mechanism available anywhere in neurosurgery: it reduces the volume constraint on the cranial contents. A wide bilateral frontotemporal craniectomy, durotomy and duraplasty may be performed.

It lowers the pressure. That is not in dispute in any trial, including the one whose outcome result was negative. It also shortens intensive care stay.

And the source states the tension in a single sentence: although decompressive craniectomy decreases intracranial pressure, it may not improve neurologic outcome and remains controversial.

The two trials, and the variable between them

DECRA randomized 155 adults with severe diffuse traumatic brain injury and intracranial hypertension refractory to first-tier therapies to early bifrontotemporoparietal craniectomy or standard care. Craniectomy decreased intracranial pressure and intensive care stay, and produced worse Extended Glasgow Outcome Scale scores — odds ratio for a worse score 1.84 (95% CI 1.05 to 3.24, P = 0.03).

RESCUEicp randomized 408 patients aged 10 to 65 with intracranial pressure refractory above 25 mm Hg — a last-resort population, against a medical arm that included barbiturates. Mortality at 6 months fell from 48.9% to 26.9%, and the rest of the distribution shifted with it. That distribution is today’s pitfall.

The variable that differs is where in the escalation the operation sits. DECRA tested it comparatively early; RESCUEicp tested it last. And the guideline follows that variable rather than pooling the results.

The recommendation, read carefully

The Brain Trauma Foundation’s 2020 update, issued after RESCUEicp, says three things:

Situation	Endpoint	Recommendation
Late refractory ICP elevation	Mortality and favourable outcomes	Recommended
Early refractory ICP elevation	Mortality and favourable outcomes	Not recommended
Early or late refractory ICP elevation	Reduce ICP and duration of intensive care	Recommended

The third row is the one worth pausing on. A guideline is here willing to recommend an operation for a physiological and logistical endpoint while declining to recommend it for a clinical one at the same timing. That is a formal, published acknowledgement that lowering the number and helping the person are separable — and it is unusual enough that it is worth being able to quote.

The subdural question, answered separately

RESCUE-ASDH asked something different: in a patient already having a traumatic acute subdural haematoma evacuated, should the bone flap be left out? 450 patients randomized to craniotomy or craniectomy with a bone flap of 11 cm or more.

At 12 months the Extended Glasgow Outcome Scale distributions were similar — common odds ratio 0.85 (95% CI 0.60 to 1.18, P = 0.32) — with death in 30.2% after craniotomy and 32.2% after craniectomy. Disability and quality of life were similar.

What differed was the complication profile, and it went both ways:

Additional cranial surgery within 2 weeks: 14.6% after craniotomy versus 6.9% after craniectomy
Wound complications: 3.9% after craniotomy versus 12.2% after craniectomy

So for this decision the evidence is genuinely helpful: it is a surgical trade between a return to theatre and a wound problem, not an outcome question. That is a smaller claim than the other two trials make, and it is more securely held.

Before surgery is on the table

The operation is a last-tier option, which means the tiers above it are yours.

Osmotherapy is unsettled at the top. Two meta-analyses suggest hypertonic saline is more effective than mannitol for raised intracranial pressure. A study comparing them at similar osmotic load found no difference in either the reduction achieved or the duration of action, and it was not included in those meta-analyses. So a larger multicentre trial is still needed to establish the first-line agent and its dose–response. Hypertonic saline may suit patients who need intravascular volume maintained — but long-term elevated serum sodium brings depressed consciousness and seizures of its own.

Perfusion pressure has a ceiling as well as a floor. The current Brain Trauma Foundation level II target is 60 to 70 mm Hg. Studies proposing thresholds above 70 have not shown improved outcome, and higher targets are linked to serious systemic complications — a five-fold increase in the risk of acute respiratory distress syndrome, and possible worsening of cerebral oedema. The evidence does not support a single threshold for every patient, which is the case for multimodal monitoring: brain oxygenation by jugular venous oximetry or brain tissue oxygen tension, autoregulation, microdialysis, continuous EEG.

And the pressure threshold itself is contested, which Neuro 2 day 03 covered: the randomized trial in Bolivia and Ecuador found that care aimed at keeping intracranial pressure below 20 mm Hg was not superior to care based on imaging and clinical examination, with comparable six-month mortality and intensive care stay — while its own authors declined to argue against using monitors, and noted that the finding may not generalise to systems with better prehospital care and rehabilitation.

Seizures, with a clean rule

Post-traumatic seizures occur in 10 to 30% after severe traumatic brain injury and up to 50% in penetrating injury. They matter because they cause metabolic derangement, raise intracranial pressure, and worsen secondary injury.

The classification is by time and it determines the management:

Early — within 7 days. Incidence is reduced by phenytoin. Intravenous levetiracetam is promising with a lower risk profile but is not yet recommended first-line pending further study.
Late — after 7 days. Prophylactic anticonvulsants are not recommended, because they have not been shown to reduce the incidence. A late seizure is treated as a new-onset seizure.

That is one of the cleaner rules in this curriculum: prophylaxis works for one window and not the other, and continuing it past the window is treatment without an indication.

The systemic injury

Severe traumatic brain injury is not confined to the head, and several of its consequences will be presented to you as anaesthetic problems.

Neurogenic pulmonary oedema is the one to recognise on sight. Marked pulmonary vascular congestion, intra-alveolar haemorrhage, protein-rich oedema fluid. Its distinguishing features are rapid onset, a relationship to hypothalamic lesions, and the fact that it can be prevented or attenuated by α-blockers and central nervous system depressants. It is thought to follow a massive sympathetic discharge from the injured brain secondary to intracranial hypertension.

The critical clinical point: traditional therapy for cardiogenic pulmonary oedema is ineffective, and the outcome is often fatal. The treatment is immediate pharmacological or surgical relief of the intracranial hypertension, with supportive respiratory care and careful fluid management. Treating it as a lung problem loses the patient.

Pituitary dysfunction goes both ways and is the day 06 material arriving through trauma. Posterior pituitary dysfunction is commoner. Diabetes insipidus follows craniofacial trauma and basal skull fracture — polyuria, polydipsia, hypernatraemia, high serum osmolality, dilute urine — is usually transient, and is treated with water replacement, with exogenous vasopressin if the patient cannot maintain balance. SIADH presents as euvolaemic or hypervolaemic hyponatraemia. Anterior pituitary insufficiency is rare, but a patient with post-traumatic diabetes insipidus may develop delayed anterior hormone impairment needing replacement.

And the rest of the list, which is worth reading once so that none of it is a surprise: airway obstruction, hypoxaemia, shock, acute respiratory distress syndrome, ECG changes, disseminated intravascular coagulation, non-ketotic hyperosmolar hyperglycaemic coma, stress ulcers and gastrointestinal haemorrhage. Aspiration pneumonia, fluid overload and trauma-related acute respiratory distress syndrome are the common causes of pulmonary dysfunction here.

Ending Neuro 2

Four times in this level the same structure has appeared, and it is worth naming once at the end.

In day 03, managing to an intracranial pressure threshold was not superior to managing to the examination. In day 07, cooling reliably reduced cerebral metabolic rate and did not change outcome. In day 08, every monitor detected ischaemia and none of them improved outcome. And in this day, the operation lowers the pressure in every trial, including the one where the patients did worse.

None of that means measurement is useless. It means that the response of a measurement to a treatment is not evidence that the treatment helped, and that the endpoint has to be chosen and named before the intervention, not inferred from it afterwards.

Which is, in the end, the same discipline as Neuro 1 day 01: know what the number is telling you, and know what it is not.

Death Vegetative state and severe disability Moderate disability and good recovery

Figure 1. The upper panel is the single most important picture in this day and possibly in the curriculum. The mortality bar shrinks by about 22 percentage points and the good-recovery bar does not grow — those survivors appear in vegetative state and severe disability instead. Nothing here says the operation is wrong. It says that "it works" and "it lowers the pressure" are two different claims, and that a family being consented needs the first one described in the terms of the lower half of that bar.

03

The evidence

the mechanism is certain, the trials are the best in this curriculum, and they still do not tell you whether to operate

DECRA

Cooper DJ, Rosenfeld JV, Murray L, et al. Decompressive craniectomy in diffuse traumatic brain injury. N Engl J Med. 2011;364(16):1493-1502. Harm

Design: Multicentre randomized trial, Australia and New Zealand
Population: 155 adults with severe diffuse traumatic brain injury and intracranial hypertension refractory to first-tier therapies
Intervention: Early bifrontotemporoparietal decompressive craniectomy
Control: Standard care
Primary: Extended Glasgow Outcome Scale at 6 months
Result: Craniectomy decreased intracranial pressure and intensive care length of stay, and patients undergoing it had worse Extended Glasgow Outcome Scale scores — odds ratio for a worse score 1.84 (95% CI 1.05 to 3.24, P = 0.03).

A narrow trial that was read too widely The entry criterion was early and comparatively permissive — refractory to first-tier therapy — so the trial tested prophylactic-ish early surgery in diffuse injury, not surgery as a last resort. Baseline pupil reactivity was imbalanced between groups. The result is strong evidence about early craniectomy in diffuse injury and weak evidence about anything else, and the 2020 guideline treats it that way.

RESCUEicp

Hutchinson PJ, Kolias AG, Timofeev IS, et al. Trial of decompressive craniectomy for traumatic intracranial hypertension. N Engl J Med. 2016;375(12):1119-1130. Benefit

Design: International multicentre randomized trial
Population: 408 patients aged 10 to 65 with traumatic brain injury and refractory intracranial pressure above 25 mm Hg
Intervention: Decompressive craniectomy as last-tier therapy
Control: Continued medical management, including barbiturates
Primary: Extended Glasgow Outcome Scale at 6 months
Result: Mortality fell from 48.9% to 26.9%. The rest of the distribution moved with it: vegetative state 8.5% versus 2.1%, lower severe disability 21.9% versus 14.4%, upper severe disability 15.4% versus 8.0%, moderate disability 23.4% versus 19.7%, good recovery 4.0% versus 6.9%.

What "benefit" means here This is a real mortality benefit and the day records it as one. But almost the whole of the survival gain appears in the vegetative and severely disabled categories, and good recovery was lower in the surgical arm at 6 months. Whether that is a benefit is a value judgement the trial cannot make, and it belongs to the patient's family rather than to the theatre.

RESCUE-ASDH

Hutchinson PJ, Adams H, Mohan M, et al. Decompressive craniectomy versus craniotomy for acute subdural hematoma. N Engl J Med. 2023;388(24):2219-2229. No association

Design: International multicentre randomized trial
Population: 450 patients undergoing surgery for traumatic acute subdural haematoma
Intervention: Decompressive craniectomy, bone flap 11 cm or more
Control: Craniotomy with bone flap replacement
Primary: Extended Glasgow Outcome Scale at 12 months
Result: No significant difference — common odds ratio 0.85 (95% CI 0.60 to 1.18, P = 0.32). Death 30.2% after craniotomy versus 32.2% after craniectomy. Additional cranial surgery within 2 weeks was commoner after craniotomy (14.6% versus 6.9%); wound complications were commoner after craniectomy (3.9% versus 12.2%).

A different question with a cleaner answer This trial asks whether to replace the bone flap in a patient who is already having an operation, which is not the DECRA or RESCUEicp question of whether to operate at all. Its usefulness is that it removes a decision from the argument: for acute subdural haematoma, leaving the bone out does not improve outcome, and the trade is between reoperation and wound complications.

Three good trials, and the honest summary is not "it works" or "it does not."

The mechanism was never in question. Removing a large piece of skull relieves the volume constraint on the cranial contents, and the pressure falls. Every trial confirms this, including the one whose outcome result was negative: DECRA found that craniectomy decreased intracranial pressure and shortened intensive care stay while producing worse Extended Glasgow Outcome Scale scores. That combination is the most instructive result in this curriculum, because it separates two things residents are taught to treat as one.

The trials disagree because they asked different questions, and the guideline reflects that rather than averaging them. DECRA randomized patients with diffuse injury refractory to first-tier therapy — early surgery. RESCUEicp randomized patients with pressure above 25 mm Hg refractory to everything short of surgery — last-resort surgery, against a medical arm that included barbiturates. So the 2020 Brain Trauma Foundation update recommends craniectomy for late refractory hypertension and recommends against it for early refractory hypertension, for the same endpoints. Two trials, two answers, one variable between them.

And the third endpoint is separated deliberately. For early or late refractory hypertension, craniectomy is recommended to reduce intracranial pressure and the duration of intensive care. The guideline is willing to recommend an operation for a physiological and logistic endpoint while declining to recommend it for a clinical one, which is unusual and worth noticing — it is a formal acknowledgement that the number and the person can come apart.

RESCUE-ASDH narrows the argument usefully. For a patient already having an acute subdural haematoma evacuated, leaving the bone out did not change disability or quality of life at 12 months. The trade is procedural: fewer return trips to theatre after craniectomy, more wound complications. That is a decision that can be made on surgical grounds without invoking outcome data.

What none of them tells you is what to do at 3 a.m. with this patient. The trials define populations; the guideline defines timing; the decision remains individual, and the honest position in the room is that the evidence supports the operation for a late, refractory, otherwise-untreatable pressure, and does not support reaching for it early.

04

What you actually do

in the room

Separate the two questions the guideline separates. "Will this lower the pressure?" — yes, reliably, early or late. "Will this improve the outcome?" — recommended for late refractory hypertension, not recommended for early. Saying which question is being answered keeps a conversation honest, and the guideline itself models the distinction.
Know what maximal medical management means before surgery is discussed. Decompressive craniectomy is a last-tier option for diffuse swelling refractory to maximal medical management. Two meta-analyses suggest hypertonic saline is more effective than mannitol for raised pressure, though a study comparing them at similar osmotic load found no difference in reduction or duration — so the first-line agent is genuinely unsettled, and long-term hypertonic sodium brings its own depressed consciousness and seizures.
Target a cerebral perfusion pressure of 60 to 70 mm Hg, and resist going higher. That is the current Brain Trauma Foundation level II recommendation. Higher thresholds have not been shown to improve outcome and are linked to serious systemic complications — including a five-fold increase in the risk of acute respiratory distress syndrome and possible worsening of cerebral oedema. No single threshold generalises to every patient, which is the argument for multimodal monitoring.
Give phenytoin for early post-traumatic seizures and stop there. Seizures occur in 10 to 30% after severe traumatic brain injury and up to 50% in penetrating injury. Early seizures — within 7 days — are reduced by phenytoin. Intravenous levetiracetam is promising with a better risk profile but is not first-line yet. Prophylactic anticonvulsants are not recommended for late seizures, because they do not reduce the incidence; a late seizure is treated as new-onset epilepsy.
Recognise neurogenic pulmonary oedema as an intracranial problem. Rapid onset, related to hypothalamic lesions, thought to follow massive sympathetic discharge from intracranial hypertension, with vascular congestion, intra-alveolar haemorrhage and protein-rich oedema fluid. Traditional therapy for cardiogenic pulmonary oedema is ineffective and the outcome is often fatal. Treatment is immediate relief of the intracranial hypertension — pharmacological or surgical — plus supportive respiratory care and careful fluid management.
Expect the pituitary to misbehave in both directions. Posterior pituitary dysfunction is commoner after head trauma. Diabetes insipidus follows craniofacial trauma and basal skull fracture — polyuria, polydipsia, hypernatraemia, high serum osmolality, dilute urine — is usually transient and treated with water replacement, with vasopressin if balance cannot be maintained. SIADH gives euvolaemic or hypervolaemic hyponatraemia. This is Neuro 2 day 06's timeline arriving through a different injury.
For an acute subdural haematoma, treat the bone flap as a surgical decision, not an outcome one. Leaving the bone out did not change disability or quality of life at 12 months. What changed was the complication profile: less additional surgery within 2 weeks after craniectomy (6.9% versus 14.6%), more wound complications (12.2% versus 3.9%).
Be in the consent conversation, or at least know what was said in it. The mortality reduction in RESCUEicp is real and large. So is the fact that good recovery was 4.0% in the surgical arm against 6.9% medically at 6 months. A family told "this operation halves the risk of dying" has been told something true and something incomplete, and the anaesthetist is often the second clinician they speak to.

05

The pitfall

Lower mortality is not more survivors you would recognise

RESCUEicp is a positive trial. Mortality at six months fell from 48.9% to 26.9% — twenty-two percentage points, in a population with refractory intracranial pressure above 25 mm Hg. On any ordinary reading, that is a large benefit, and this curriculum records it as one.

Now look at where those survivors went.

| Outcome at 6 months | Surgical | Medical | | --- | --- | --- | | Death | 26.9% | 48.9% | | Vegetative state | 8.5% | 2.1% | | Lower severe disability | 21.9% | 14.4% | | Upper severe disability | 15.4% | 8.0% | | Moderate disability | 23.4% | 19.7% | | Good recovery | 4.0% | 6.9% |

Read the bottom two rows. Moderate disability moved by less than four percentage points, and good recovery went down. Essentially the entire mortality reduction reappears as vegetative state and severe disability.

That is not an argument against the operation. It is an argument about what sentence you use to describe it, and the wrong sentence gets said in front of families constantly.

"This operation halves the chance of dying" is true.

"This operation improves the chance of a good recovery" is not supported by this trial, and at six months the number ran the other way.

Both sentences describe the same result. Only one of them is what a family hears when a surgeon says the operation might save their relative's life.

Three things follow that are actually actionable.

Do not let the intracranial pressure trace do the arguing. Craniectomy lowers pressure in every trial, including the one where outcomes were worse. DECRA is the proof: it decreased intracranial pressure and shortened intensive care stay while producing worse Extended Glasgow Outcome Scale scores, odds ratio 1.84. If the case for operating is being made on the monitor, the monitor is not the endpoint that was measured.

Timing is the variable that separates the two trials, and the guideline knows it. Recommended for late refractory hypertension. Not recommended for early refractory hypertension. Same operation, same endpoints, opposite recommendations, and the only thing that changed is where in the escalation ladder it sits. Reaching for it early is not a more aggressive version of the right treatment — it is a different intervention with a different trial behind it.

Know what the family was told before you meet them. The anaesthetist is often the second clinician a family speaks to in the hour before a craniectomy, and is sometimes asked the question they did not ask the surgeon. You do not need to run that conversation. You do need to know that "will he survive" and "will he be himself" have different answers here, and to not accidentally answer the second while being asked the first.

One last framing, because it generalises past this operation. This curriculum has now met the same shape four times: the intracranial pressure threshold in Neuro 2 day 03, the cooling in day 07, the monitors in day 08, and the pressure here. Every one of them is a measurement that responds beautifully to treatment, and in none of them does the response guarantee the patient is better. That is not cynicism about monitoring. It is the reason the endpoint has to be named out loud before the intervention is chosen.

06

Sources

primary first

01 Cooper DJ, Rosenfeld JV, Murray L, et al. Decompressive craniectomy in diffuse traumatic brain injury. N Engl J Med. 2011;364(16):1493-1502. Full text · Preston doi 10.1056/NEJMoa1102077
02 Hutchinson PJ, Kolias AG, Timofeev IS, et al. Trial of decompressive craniectomy for traumatic intracranial hypertension. N Engl J Med. 2016;375(12):1119-1130. Full text · Preston doi 10.1056/NEJMoa1605215
03 Hutchinson PJ, Adams H, Mohan M, et al. Decompressive craniectomy versus craniotomy for acute subdural hematoma. N Engl J Med. 2023;388(24):2219-2229. Full text · Preston doi 10.1056/NEJMoa2214172
04 Hawryluk GWJ, Rubiano AM, Totten AM, et al. Guidelines for the management of severe traumatic brain injury: 2020 update of the decompressive craniectomy recommendations. Neurosurgery. 2020;87(3):427-434. Full text · Preston doi 10.1093/neuros/nyaa278

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 2 · Day 13

A family asks whether decompressive craniectomy will help their relative, who has refractory intracranial pressure above 25 mm Hg after traumatic brain injury. Based on RESCUEicp, which statement is most accurate?

Item 2 Recall Neuro 2 · Day 13

Which statement about the 2020 Brain Trauma Foundation decompressive craniectomy recommendations is correct?

Item 3 Application Neuro 2 · Day 13

A patient with severe traumatic brain injury has an intracranial pressure monitor in place. A colleague suggests targeting a cerebral perfusion pressure above 70 mm Hg to maximise oxygen delivery. What is the objection?