Role of Decompressive Craniectomy in Traumatic Brain Injury – A Meta-analysis of Randomized Controlled Trials
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0028-3886.271260
Source of Support: None, Conflict of Interest: None
Keywords: Craniectomy, decompressive, outcome, traumatic brain injuryKey Message: Role of Decompressive craniectomy in TBI has been questioned by many studies. Our meta-analyses of 3 RCTs show that DC improves survival but the functional outcome of survivors is not good.
Severe head injury is a new epidemic. The incidence of head injury is growing due an increasing number of road traffic accidents worldwide. Intracranial pressure (ICP) rises after head injury and high ICP reduces brain perfusion. ICP has been shown to be an independent predictor of mortality following moderate and severe head injury. Hence, all therapeutic measures (medical or surgical) are aimed at reducing ICP following head injury. Nowadays, with the advent of intraparenchymal ICP monitoring systems, patients are managed as per the severity of their ICP., Decompressive craniectomy (DC) is the most commonly used surgical tool to control ICP when medical measures fail. However, the role of DC in patients with head injury has been has been widely debated.[5–8] Recently, trials have shown that DC improves the mortality rate but not the functional outcome of the survivors following traumatic brain injury., We conducted a meta-analysis of all the RCTs published so far on the role of DC in patients with head injury.
This meta-analysis was planned in coherence with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines [Figure 1]. For identification and extraction of the potential studies for inclusion into our analysis, we used PICOS (Population, Intervention, Control and Outcome Study) design [Table 1]. After literature review search, study-related data were abstracted into a standardized PICOS format and assessed for relevance by two independent reviewers. Randomized controlled studies evaluating the use of conventional medical therapy against DC in patients with traumatic brain injury (TBI) were included in our analysis. Overall 6-month mortality was taken as the primary comparative endpoint for the present analysis. Parameters that were consistently compared and documented across various trials were also included as explorative outcomes. The salient features of trials included in the final analysis that met the above criterion are shown in [Table 2].
Literature search strategy
Two independent reviewers searched the online literature available on MEDLINE, Science Citation Index Expanded, Embase, Scopus, Cochrane Central Register of Controlled Trials, clinical trials registry, and meta-register for articles published until Sept. 18, 2016. Bibliographies and references of selected publications, systematic reviews and editorials on treatment regimens for diffuse TBI were also manually screened. The following medical subject heading (MeSH) terms were searched for in the above-said database:
Craniocerebral Trauma, Brain Injury, Head Injury, Brain Trauma, Traumatic Brain Injury, Brain Contusion, Traumatic, Intracranial Hemorrhage, Subdural Hematoma, Diffuse axonal injury, Traumatic Intracranial Hematoma, Traumatic Cerebral Hemorrhages, Traumatic Intracerebral Hemorrhages, Traumatic Brain Hemorrhage, Traumatic Intracranial Edema, Decompressive Craniectomy, Decompressive Hemicraniectomy, Hemicraniectomy, Trauma and Craniectomy.
To obtain a high level/grade of evidence, we included only RCTs in the meta-analysis. Our search extended to peer-reviewed publications that included research articles published either as full articles or meeting abstracts in peer-reviewed journals. The purview of our search extended to include trials published in both English and non-English languages. Once the abstract was analyzed by the searching reviewer and found appropriate, the full text of the article was studied. The decision to include a trial into the final analysis was based on assessments performed by two independent reviewers. Any disagreements between the two were harmonized by consensus and arbitration by a third neutral reviewer. Based on the recommendations by Cochrane Collaboration, another independent researcher assessed the included trials for quality of evidence and possible methodological bias.
Data extraction and principal endpoints
Data were extracted from the full-text article of each included study using a standardized data-extraction form prepared in Microsoft Excel (Microsoft Inc., USA). The following data were extracted from each of the included trial – year and country of publication, study design, patient demographic profile, treatment regimen used for diffuse TBI, patient mortality rate, survival rate, and patient neurological outcome. The scales to assess outcome used in various studies were the Glasgow Outcome Scale (GOS) or Extended Glasgow Outcome Scale (GOS-E). A score of 1–3 is considered a poor outcome in GOS (death, vegetative state, and severe disability), whereas a score of 1–4 is considered a poor outcome in GOS-E (death, vegetative state, upper severe disability, and lower severe disability).,
Eventually, we analyzed the following parameters with consistency:
Statistical analysis of the pooled data was performed using Comprehensive Meta-analysis, Version 2 (Biostat Inc., USA). We first performed meta-analysis using fixed effect modeling and subsequently random-effect modeling if heterogeneity was found to be >40%. Wherever the heterogeneity was found to be greater than 40%, results from “random effect modeling” were reported. The I^2 statistic was used to quantify heterogeneity between the trials. For classification of heterogeneity, values of I2<40% were considered nonsignificant, 40%–60% were considered as moderate heterogeneity, and 60%–90% were reported as high heterogeneity. Pooled results for variables with frequency data were expressed as a Mantel–Haenszel (MH) odds ratio. The number required to treat/harm for each variable with frequency data was also calculated. All MH odds ratios were expressed with their point estimate and 95% confidence interval (CI).A P-value of <0.05 was considered statistically significant.
Risk of bias assessment
All publications found during the search were manually and independently reviewed by two reviewers. Criteria that were used for assessing the risk of bias were based on the recommendations of the Cochrane Collaboration, and included methods of randomization, concealed treatment allocation, blinded data collection and analysis, blinded adjudication of study endpoints, and completeness of data. The graphical synopsis of the above assessment was constructed using the software Review Manager 5 (RevMan) (Cochrane Collaboration).
Studies were also assessed for a possible publication bias initially using a funnel plot and later quantified using Egger's test.
Publication bias was evaluated for reporting of mortality. The possibility of publication bias in the included studies was unlikely [Figure 2]. Egger's regression test showed that X-axis intercept occurred at 0.11 with P value (two-tailed) being 0.93.
Study quality assessment
Quality assessment for bias in the included studies was carried out as per other published meta-analyses and the guidelines laid by the Cochrane Collaboration. These results are shown in [Figure 3]. We used RevMan Version 5 (Cochrane Collaboration) for this evaluation and image generation.
During our preliminary search, we found a total of 741 publications matching our search criterion in the above database. From these, duplicate search results were removed using Endnote (Thompson Reuters, USA). No unpublished/incomplete trial or abstract/meeting proceedings were found suitable during the search. Three trials evaluating adult population were identified from the above resources. One trial by Taylor et al. included a similar intervention and control group in pediatric population. We included this trial in our reporting only for comparative results with adults. No non-English trial was found to be suitable. Two of the trials in adults used GOS scale and two used GOS-E scale for patient evaluation/grading. The values reported in the results below document values from the pooling of only the adult trials. Only one pediatric trial for comparison could be found and the values in the figures are only for the reader's comparison.
Analysis involved 285 and 288 patients in DC group and control groups respectively. The pooled mortality in the decompressive craniectomy group was 24.21% (95% CI, 19.6–29.51) and 38.19% (95% CI, 32.77–43.93). The heterogeneity for this comparison was 34.4% and its MH odds ration demonstrated a clear reduction in mortality with craniectomy [Figure 4]. Patients who underwent DC had a lower mortality association of nearly 50%. The number required to treat for the DC group was found to be 6.49.
Adverse neurological outcomes in survivors
An MH odds ratio of 2.07 (95% CI, 1.77–4.20) was associated with the likelihood of a poor outcome in patients undergoing DC. Thus, patients who underwent DC and showed a 107% increase of a resulting poor neurological outcome. The incidence of poor outcome post survival was 46.32% in the craniectomy group and 37.14% in control group. The heterogeneity associated with this comparison was 48.51%, P = 0.02 [Figure 5]. The number required to harm with craniectomy was found to be 5.20.
Overall adverse outcomes
Adverse outcomes were reported in 201 of 285 patients and 188 of 288 patients in the craniectomy and control groups respectively. The heterogeneity for this comparison was 78.78%. The comparative odds ratio failed to achieve a statistically significant value for this comparison [Figure 6].
Cerebral edema following TBI is a major cause of poor neurological outcomes and ensuing mortality and morbidity. Progressive increases in brain swelling decrease cerebral perfusion and cause mass effect leading to herniation. Therefore, the principle of therapies for TBI has centered upon the control of raised ICP using medical or surgical means. DC, by creation of a cranial defect, allows decompression of the affected cerebral hemispheres and thereby results in lower ICP, facilitating cerebral perfusion and decreasing mass effect of the brain.,
Considerable heterogeneity is present in the published literature of DC in TBI. The variability in the results of such studies can, in part, be explained by the heterogeneity of the population studied, the type of the study design, choice and technique of surgical approach, and monitoring protocols. Despite its widespread use and apparent success, the procedure of DC remains controversial and under scrutiny.,,,,, Indeed, there continue to be uncertainties regarding its appropriate application, which have included the selection criteria, the timing of the craniectomy, the DC approach, the assessment of outcome and potential complications., Early craniectomy has been suggested by many as a means of reducing ICP to prevent secondary insults.,,
The 2009 Cochrane Collaboration literature review did not recommend DC in the adult trauma population for primary treatment because none of the included studies were RCTs. However, the authors suggested that it could be considered a rescue therapy. Since then, three RCTs of DC have been published in adult populations. One RCT in a pediatric population was published in 2001. The aim of this analysis was to fill the void of level I evidence of DC in TBI.
Of the four RCTs published, the RESCUE ICP trial by Hutchinson et al. is the largest and recruited 398 patients. There exists heterogeneity in the selection criteria for enrollment to the studies. Hutchinson et al. recruited patients with TBI with a abnormal computed tomography scan with an ICP value >25 mmHg despite conservative measures. In contrast, the DECRA trial by Cooper et al. recruited only severe TBI with an ICP value for 15 min within a one hour period. Notably, the DECRA trial excluded intracranial hematomas which constituted nearly 20% of the RESCUE ICP trial. The study conducted by Moein et al. recruited only 20 patients with severe TBI. No details about ICP values were available from the article. Taylor included all children over 12 months of age who were admitted to their intensive care unit following TBI and had a functioning intraventricular catheter in situ. Children who had sustained intracranial hypertension (if the ICP was 20–24 mmHg for 30 min or 25–29 mmHg for 10 min or 30 mmHg or more for 1 min, or there was dilatation of one pupil or the presence of bradycardia suggestive of herniation) during the first day after admission.
The neuroimaging finding of diffuse edema without a predominant unilateral focus is not the usual presentation severe TBI. In the RESCUE ICP trial, unilateral frontotemporoparietal DC was done for unilateral hemispheric swelling and a bifrontal DC for diffuse brain swelling. In DECRA trial, patients underwent a bifrontal DC by modified Polin's technique. In contrast, Moein et al. did not describe the technique of DC used in their study. Taylor et al. performed bitemporal craniotomy through a bilateral vertical incision in the mid-temporal region. They removed temporal bones measuring 3–4 cm in diameter and left the dura intact in most of the cases.
The timing of craniectomy also varied among the RCTs. While Hutchinson et al. chose a duration of 1–12 h of ICP >25 mmHg, the DECRA trial had a cut-off ICP >20 mmHg for 15 min in a one hour period within the first 72 hours. Moein et al. performed DC within 24 hours to those randomized to surgical intervention group. Taylor aimed to perform DC within 6 hour of randomization.
Outcome assessment was done at 6 months by Moein et al and Tayloret al., based on GOS at 6 months when compared with GOSE at 6 months, 12 months by Cooper et al., and Hutchinson et al. As mentioned above in the results, the effect of DC in adults with TBI favors craniectomy over standard medical management. However, this is associated with poorer neurological outcomes in surviving patients. In both DECRA and RESCUE ICP, a significant reduction in ICP was noted following DC.
The DECRA trial has received a number of criticisms. First, the patients in the DC interventional arm of the study appeared to have a greater incidence of more severe primary TBI. Second, the ICP treatment threshold >20mmHg for >15 min did not reflect clinical practice, and a high cross-over rate from standard care to DC group. The RESCUE ICP was similarly critized as a signficant proportion of individuals underwent DC after having been assigned to the medical treatment group.
Our review followed the RESCUE ICP and analyzed three RCTS which assessed the role of DC when compared with medical management in the adult population. The pooled Odds Ratio (OR) favored DC as having a favorable effect on mortality but was associated with poorer neurological outcomes in survivors. Both the DECRA and RESCUE ICP tirals demonstrated a significant reduction in ICP, indicating its role as a key neurotoxic factor in TBI. DC should be offered with caution as the number of survivors is increased following DC, but the survivors do not necessarily have good outcomes. Craniectomy commits the patient to a minimum of one additional procedure (cranioplasty) which also can be a cause of complications. In the interim, a helmet or other protective gear must be worn to avoid further brain injury from inadvertent falls or other traumas.
The Brain Trauma Foundation recently published its fourth edition of guidelines for management of traumatic brain surgery. They recommended a large frontotemporoparietal DC over a small frontotemporoparietal craniotomy. They also mentioned that a bifrontal DC is not recommended to improve outcomes as measured by GOS-E score at 6-month post-injury in severe TBI patients with diffuse injury (without mass lesions), and with ICP elevation to values 20 mmHg for more than 15 min within a 1-hour period that are refractory to first-tier therapies. However, they acknowledged that the results of the RESCUE ICP trial were released soon after the completion of these new guidelines, the results of which may affecy these recommendations.
There are few limitations to our study. The number of high-quality studies examining the use of DC after TBI is limited. There was significant heterogeneity in the included studies, especially with respect to the types of DC performed, the medical management administered, and the time points at which ICP was measured and reported. The primary outcome measure in this review was mortality. To influence decision-making for healthcare professionals and policy makers, it is imperative to move toward a more functional assessment such as SF-36 and QOLIBRI. Complications associated with DC were not examined, and it has been shown that occurrence of complications after DC is associated with an increased risk of prolonged hospital or rehabilitation facility stay. The results may have been overtly influenced by RESCUE ICP trial which had 398 patients when compared with 155 in DECRA and 20 in the study by Moein et al.
The results of our meta-analysis show that based on the available RCTs on DC in TBI, there is a mortality benefit by performing a DC over best medical management in patients. However, DC is associated with a higher incidence of poor neurological outcome in the surviving population. In the event of small number of high-quality RCTs, our results must be interpreted with caution.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6]
[Table 1], [Table 2]