Partial removal of skull (decompressive craniectomy) to lower treatment-resistant high pressure in the skull and brain after traumatic brain injury

Review question

This Cochrane Review investigated the effects of a surgical procedure, decompressive craniectomy (DC), on survival and neurological (functional) outcomes for people who have a traumatic brain injury (TBI) that does not penetrate the skull, and high pressure inside the skull that does not respond to medical treatment. In DC part of the skull is removed so the brain has room to expand, and pressure inside the skull can decrease. We compared DC to conventional medical treatments in patients over 12 months old.


The skull is a rigid bone 'box' that protects the brain. Consequently, if an injury causes the brain to swell, this leads to an increase in pressure within the skull. This excess pressure is known as high intracranial pressure (ICP), and is a frequent cause of death and disability in brain-injured people. If ICP cannot be controlled using standard medical measures, then DC may be tried. DC is the surgical removal of a portion of the skull to relieve pressure on the brain. Clinicians are uncertain how effective it is, and do not agree on its role in treatment of TBI.

Search date

This evidence is current to December 2019.

Study characteristics

We identified three relevant trials. One included 27 children (under 18 years); another trial was conducted in multiple sites and recruited 155 adults; and the third was conducted in 24 countries and recruited 408 participants (adolescents and adults). Each compared DC plus standard medical care against standard medical care alone.

In trials, the results are more reliable if patients and medical staff are unaware of which treatment a patient receives. However, the treatment given was evident in these trials, as it is not possible to conceal this type of surgery. The reliability of the results from the trial on children might have been affected because of the methods used to decide which children had DC, and because the trial stopped early. Similarly, the results of another trial might have been affected because of the treatment point at which DC was performed, and because of changes in methods for measuring outcomes. The third trial fit all of our criteria, and was well-conducted.

Study funding sources

Funders of two studies included national research bodies and the third (and smallest) study reported no specific source of funding.

Key results


There is moderate-quality evidence that DC reduces the risk of death slightly at six months compared to standard medical treatment (3 studies), and high-quality evidence that it reduces death at 12 months (1 study).


We analyzed functional outcome in three ways, splitting it into good and bad recovery as follows.

1. Good outcome (including serious disability) versus death or vegetative state: at six months there was no clear benefit for DC (2 studies, low-quality evidence), but at 12 months there was a clear benefit of DC (1 study, high-quality evidence).

2. Good outcome (including moderate disability) versus death, vegetative state or serious disability: at six months there was no clear benefit for DC (2 studies), but at 12 months there was a clear benefit of DC (1 study).

3. Good outcome (including minor disability) versus death, vegetative state, serious or moderate disability: at six months (3 studies) and 12 months (1 study) there was no clear benefit for DC.


DC was superior to standard medical treatment for reducing high ICP within 48 hours (2 studies, moderate-quality evidence).

Adverse events

Adverse events affected more people who have DC, than those who had medical treatment alone (low-quality evidence). These adverse events included problems that occur later, when the skull fragment was surgically replaced.

Quality of the evidence

The quality of our evidence ranges from high to very low. Evidence for outcomes other than mortality was complicated by differences concerning when to perform DC, as the two largest studies used different criteria for this. The small study on children produced low-quality evidence that DC has some benefits; a larger study on children is now in progress, and will improve the quality of our evidence once its results are available.

Authors' conclusions: 

Decompressive craniectomy holds promise of reduced mortality, but the effects of long-term neurological outcome remain controversial, and involve an examination of the priorities of participants and their families. Future research should focus on identifying clinical and neuroimaging characteristics to identify those patients who would survive with an acceptable quality of life; the best timing for DC; the most appropriate surgical techniques; and whether some synergistic treatments used with DC might improve patient outcomes.

Read the full abstract...

High intracranial pressure (ICP) is the most frequent cause of death and disability after severe traumatic brain injury (TBI). It is usually treated with general maneuvers (normothermia, sedation, etc.) and a set of first-line therapeutic measures (moderate hypocapnia, mannitol, etc.). When these measures fail, second-line therapies are initiated, which include: barbiturates, hyperventilation, moderate hypothermia, or removal of a variable amount of skull bone (secondary decompressive craniectomy).


To assess the effects of secondary decompressive craniectomy (DC) on outcomes of patients with severe TBI in whom conventional medical therapeutic measures have failed to control raised ICP.

Search strategy: 

The most recent search was run on 8 December 2019. We searched the Cochrane Injuries Group's Specialised Register, CENTRAL (Cochrane Library), Ovid MEDLINE(R), Ovid MEDLINE(R) In-Process & Other Non-Indexed Citations, Ovid MEDLINE(R) Daily and Ovid OLDMEDLINE(R), Embase Classic + Embase (OvidSP) and ISI Web of Science (SCI-EXPANDED & CPCI-S). We also searched trials registries and contacted experts.

Selection criteria: 

We included randomized studies assessing patients over the age of 12 months with severe TBI who either underwent DC to control ICP refractory to conventional medical treatments or received standard care.

Data collection and analysis: 

We selected potentially relevant studies from the search results, and obtained study reports. Two review authors independently extracted data from included studies and assessed risk of bias. We used a random-effects model for meta-analysis. We rated the quality of the evidence according to the GRADE approach.

Main results: 

We included three trials (590 participants). One single-site trial included 27 children; another multicenter trial (three countries) recruited 155 adults, the third trial was conducted in 24 countries, and recruited 408 adolescents and adults. Each study compared DC combined with standard care (this could include induced barbiturate coma or cooling of the brain, or both). All trials measured outcomes up to six months after injury; one also measured outcomes at 12 and 24 months (the latter data remain unpublished). All trials were at a high risk of bias for the criterion of performance bias, as neither participants nor personnel could be blinded to these interventions. The pediatric trial was at a high risk of selection bias and stopped early; another trial was at risk of bias because of atypical inclusion criteria and a change to the primary outcome after it had started.

Mortality: pooled results for three studies provided moderate quality evidence that risk of death at six months was slightly reduced with DC (RR 0.66, 95% CI 0.43 to 1.01; 3 studies, 571 participants; I2 = 38%; moderate-quality evidence), and one study also showed a clear reduction in risk of death at 12 months (RR 0.59, 95% CI 0.45 to 0.76; 1 study, 373 participants; high-quality evidence).

Neurological outcome: conscious of controversy around the traditional dichotomization of the Glasgow Outcome Scale (GOS) scale, we chose to present results in three ways, in order to contextualize factors relevant to clinical/patient decision-making.

First, we present results of death in combination with vegetative status, versus other outcomes. Two studies reported results at six months for 544 participants. One employed a lower ICP threshold than the other studies, and showed an increase in the risk of death/vegetative state for the DC group. The other study used a more conventional ICP threshold, and results favoured the DC group (15.7% absolute risk reduction (ARR) (95% CI 6% to 25%). The number needed to treat for one beneficial outcome (NNTB) (i.e. to avoid death or vegetative status) was seven. The pooled result for DC compared with standard care showed no clear benefit for either group (RR 0.99, 95% CI 0.46 to 2.13; 2 studies, 544 participants; I2 = 86%; low-quality evidence). One study reported data for this outcome at 12 months, when the risk for death or vegetative state was clearly reduced by DC compared with medical treatment (RR 0.68, 95% CI 0.54 to 0.86; 1 study, 373 participants; high-quality evidence).

Second, we assessed the risk of an 'unfavorable outcome' evaluated on a non-traditional dichotomized GOS-Extended scale (GOS-E), that is, grouping the category 'upper severe disability' into the 'good outcome' grouping. Data were available for two studies (n = 571). Pooling indicated little difference between DC and standard care regarding the risk of an unfavorable outcome at six months following injury (RR 1.06, 95% CI 0.69 to 1.63; 544 participants); heterogeneity was high, with an I2 value of 82%. One trial reported data at 12 months and indicated a clear benefit of DC (RR 0.81, 95% CI 0.69 to 0.95; 373 participants).

Third, we assessed the risk of an 'unfavorable outcome' using the (traditional) dichotomized GOS/GOS-E cutoff into 'favorable' versus 'unfavorable' results. There was little difference between DC and standard care at six months (RR 1.00, 95% CI 0.71 to 1.40; 3 studies, 571 participants; low-quality evidence), and heterogeneity was high (I2 = 78%). At 12 months one trial suggested a similar finding (RR 0.95, 95% CI 0.83 to 1.09; 1 study, 373 participants; high-quality evidence).

With regard to ICP reduction, pooled results for two studies provided moderate quality evidence that DC was superior to standard care for reducing ICP within 48 hours (MD −4.66 mmHg, 95% CI −6.86 to −2.45; 2 studies, 182 participants; I2 = 0%). Data from the third study were consistent with these, but could not be pooled.

Data on adverse events are difficult to interpret, as mortality and complications are high, and it can be difficult to distinguish between treatment-related adverse events and the natural evolution of the condition. In general, there was low-quality evidence that surgical patients experienced a higher risk of adverse events.