Decompressive craniectomy in traumatic brain injury: A single-center, multivariate analysis of 1,236 patients at a tertiary care hospital in India
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0028-3886.156277
Source of Support: None, Conflict of Interest: None
Object: To evaluate the outcome of patients undergoing a decompressive craniectomy (DC) in traumatic brain injury (TBI) and the factors predicting outcome.
Keywords: Decompressive craniectomy; outcome; predictors of outcome; traumatic brain injury
Traumatic brain injury (TBI) is a leading cause of death and disability worldwide. More than 90% of the burden is shared by low- and middle-income countries like India, taking a heavy toll in especially the younger and more productive age groups with grave financial implications.  Outcomes in TBI have improved over the past few decades due to design modifications in the automotive engineering sector and the wider use of preventive strategies like recreational helmet wearing and improved car seat technology. , An elevated intracranial pressure (ICP) is the primary culprit contributing to mortality in more than 50% of patients with TBI.  The first-tier management protocol for reducing ICP includes the use of multiple conservative and minimally-invasive modalities like hyperosmolar agents, analgesia, deep sedation, and a ventriculostomy.  The decompression of the brain started by the ancient Greeks and propagated by modern philosophers in medicine aims at reducing ICP by artificially providing space for the brain to expand. It has been generally accepted that a decompressive craniectomy (DC) does not reverse primary brain injury. It can, however, ameliorate the secondary damage caused by refractory intracranial hypertension (ICH). , Although a DC is helpful in treating uncontrollable ICH, the translation of its effects into an improved outcome is challenged by paucity of robust scientific evidence.  Similar concerns were raised by the multicentric, randomized, controlled DECRA trial.  A review of the published literature clearly highlights the controversies regarding the cost-benefit ratio in performing a DC. The primary drawback in a majority of these studies is their small sample size and heterogeneous study design, thereby raising important ethical and economical questions. ,,,,,,
This single-center, retrospective, multivariate analysis of 1,236 patients operated at a tertiary care hospital in India, was conducted to assess the long-term outcome of patients in terms of the Glasgow outcome score (GOS), and to analyze the factors predicting outcome in patients undergoing DC in TBI.
This is a single-center, retrospective study of TBI patients who underwent DC from January 2008 to December 2013 at a tertiary care hospital in India. The institutional ethical committee clearance was taken before commencement of this study (reference number IESC/T-367/30.08.2013).
Patient population and parameters assessed
The patients presenting to our center with TBI (including mild, moderate, and severe head injury) during the study period of 6 years (from January 2008 to December 2013), and undergoing DC as the surgical procedure, were identified. The records of these patients were extracted retrospectively from the computerized electronic database of the hospital. The baseline demographic profile, mode of injury, clinical status (Glasgow Coma Scale [GCS], pupillary reactivity), radiological findings (intracranial pathology and midline shift as assessed on a head computed tomography (CT)), hemodynamic instability (preoperative blood pressure), intraoperative observations (presence of brain bulge and duraplasty versus slit durotomy), postoperative complications, length of hospital stay and GOS at death/discharge were noted in a pre-prepared proforma.
Among the cohort of 720 alive patients at discharge, the follow-up GOS was assessed in 324 patients (45%) by a telephonic interview using a structured proforma [Figure 1]. The mean follow-up was 25.3 (±12.2 standard deviation; range 3-42) months.
Presurgical intensive care management
The medical management was provided prior to surgical decompression in the neurosurgical intensive care unit (ICU), and was based on the revised Brain Trauma Foundation guidelines (BTF 2007).  The exception to the use of these guidelines was when the patients had frank clinical and radiological signs of brain herniation at the time of admission, in which case, they were immediately taken from the emergency services to the operating room for surgical decompression. The first-line medical management included intubation (depending upon the GCS), fluid resuscitation, ICP monitoring using an intraparenchymal (Codman) catheter, elective ventilation targeting a pCO 2 of 32-35 mmHg, head end elevation (by 30°), heavy sedation (0.05-0.1 mg/kg/h midazolam), analgesia (1 mcg/kg/h fentanyl), hyperosmolar therapy (0.25-1g/kg mannitol every 6-8 h), diuretic therapy (0.5-1 mg/kg furosemide every 8-12 h), and maintenance of normothermia (36.5-37°C), and normoglycemia (80-120 mg/dl). Apart from the baseline radiological investigations at admission (CT head), a repeat CT head, as guided by the patient's clinical status, was obtained using a mobile CT scanner (Schiller, Switzerland) available within the neurosurgical ICU.
DC - Indications and technique
A proper informed consent was taken from the patient's attendants before each surgical procedure. The plan to proceed with a DC was based on the patient's clinical status (deteriorating GCS), radiological status (increasing midline shift), and the presence of a refractory ICH (ICP >20 mmHg lasting for >20 min) despite adequate medical management.
The indications for an urgent DC without prior medical management were diffuse unilateral or bilateral brain edema seen on a CT scan with clinical correlation; effacement of basal cisterns; initial herniation with affected pupils; and, a large intraparenchymal or subdural hematoma (SDH) causing midline shift and a deteriorating GCS. The patients with minor and moderate head injury underwent a DC only when the GCS deteriorated to ≤8, along with radiological evidence of increasing midline shift on CT scan and/or increasing ICP nonresponsive to the medical management for more than 20 min. Delayed surgical decompression in the form of a DC was performed in the cases having a refractory ICP despite adequate medical management owing to progressive diffuse cerebral edema. The absence of brain stem reflexes in a comatose patient was a contraindication for DC. Laterality of DC was decided by the clinical and radiological cues with a bifronto-temporo-parietal craniectomy being done in the selected cases having a symmetrical involvement of both hemispheres and a diffuse brain swelling. The extent of the craniectomy was an approximate anteroposterior length of 15 cm, medially extending until 3 cm lateral to the superior sagittal sinus, and inferiorly reaching till the floor of the middle cranial fossa at the origin of the zygomatic arch [Figure 2] and [Figure 3]. The evacuation of the hematoma was done, if required, followed by a lax duraplasty using an autologous pericranial graft. Postoperative ICP monitoring was done in all recently operated cases. The bone flap, after its removal, was placed either in a subcutaneous pocket in the abdomen or in a bone bank facility (at a temperature <70°C). Representative pre- and postoperative images of patients with TBI who underwent DC at our center are given in [Figure 4].
The statistical analysis was carried out using Stata 11.0 (Stata Corp. LP, Texas, US). A cumulative odds ordinal logistic regression was run to determine the effect of preoperative clinical and radiological findings, intraoperative observations, and post-operative hospital course, on the short-term GOS at the time of discharge. Only the factors with P < 0.25 at univariate analysis were used for multivariate analysis. Results of multivariate and univariate analysis were reported as odds ratio (OR; 95% confidence interval (CI)). The nonparametric data was analyzed by using Kruskal-Wallis and Mann-Whitney tests. The results of the variables including the patient's age, length of hospital stay, GCS at admission, GCS at discharge, and GOS at discharge were reported as median (interquartile range [IQR]); the mortality rate as percentage, and rest of the parametric data as numbers. A P - value of < 0.05 was considered statistically significant.
Out of a total of 1,293 patients who underwent a DC after sustaining a TBI during the period of 6 years, 1,236 patients with complete data were included in the study. Among the cohort of 720 alive patients at discharge, the follow-up GOS was assessed by telephonic interview for 324 patients (45%) using a structured proforma. The higher proportion of patients lost to follow-up in the current study was because many of our patients had financial constraints and lived in remote areas of the country having limited access to any means of communication.
The study cohort comprised of 1,236 patients with 81% males. The median (IQR) age at presentation was 32 years (23-45 years). Road traffic accident (RTA) was the most common mode of injury. The most common radiological finding was an intracerebral contusion (65%), followed closely by a SDH (63%). Preoperatively, the ICP was monitored in 12.2% of patients. This lesser percentage relates to a relatively large proportion of patients taken directly to the operating room from emergency care facility based on their clinical and radiological profile. The median (IQR) GCS at presentation and discharge was 6 (4-8) and 11 (10-15), respectively. The incidence of mild, moderate, and severe TBI at the time of admission was 11.2, 12.4, and 76.4% respectively. However, the decision to perform a DC in patients with mild and moderate TBI was taken only when the patients deteriorated to a GCS <8. There was no statistically significant difference related to the age, mode of injury, preoperative ICP monitoring, and the rate of redo surgery among the three groups, suggesting comparability of results. However, there was a higher proportion of male patients (P = 0.01), along with a higher incidence of pupillary asymmetry, preoperative hypotension, midline shift >10 mm, and associated systemic injuries in those patients who had a severe TBI since the time of admission (P < 0.05). A SDH (P = 0.006) or a subarachnoid hemorrhage (SAH, P = 0.08) were relatively more common radiological observations in the presence of severe TBI as opposed to a higher percentage of contusions and extradural hematoma (EDH) in the minor and moderate TBI groups, although the results were statistically not significant (P > 0.05) [Table 1].
Complications and short-term outcome
The patients in the severe TBI group had a higher incidence of intraoperative brain bulge (P = 0.01) and consequently, a higher proportion of slit durotomy (P = 0.001). On the contrary, a lax duraplasty (P = 0.01) was performed more commonly in patients with minor/moderate TBI. 5.5% of patients were reoperated and a contralateral DC was also performed. There was a significantly higher risk of pulmonary complications (P = 0.04) and prolonged ventilation (P < 0.001) in the severe TBI group. A total of 88 patients (7.5%) developed postoperative hydrocephalus, with 77 of them requiring a cerebrospinal fluid (CSF) diversion procedure either in the form of an endoscopic third ventriculostomy or a ventriculoperitoneal shunt. Posttraumatic seizures occurred in 1.7% of patients until the time of discharge. The median (IQR) duration of hospitalization was 10 (5-18) days with an overall mortality rate of 41.7%. The duration of hospital stay was the shortest in those patients who had initially presented with a minor TBI (P = 0.03). A favorable outcome (GOS 4 and 5) at discharge was observed in 46.5% of the operated patients (39%, 64%, and 78.4% in patients presenting with severe, moderate, and minor TBI, respectively) [Table 1].
Predictors of outcome at discharge
On univariate analysis, the age; severity of head injury; pupillary involvement; presence of preoperative hypotension, midline shift, brain bulge, extradural hematoma, SAH, infarct, or associated injuries (polytrauma); performance of an intraoperative duraplasty or a slit durotomy, carrying out a simultaneous contralateral DC; and, the length of hospitalization were predictors of GOS at discharge (P < 0.05). On a multivariate analysis, the factors predicting a favorable GOS were a younger age (OR 1.03, CI = 1.02-1.04; P < 0.001), no pupillary abnormalities at admission (OR 2.28, CI = 1.72-3.02; P < 0.001), absence of preoperative hypotension (OR 1.91, CI = 1.08-3.38; P = 0.02), an isolated TBI (OR 1.42, CI = 1.08-1.86; P = 0.01), absence of a preoperative infarct (OR 3.68, CI = 1.74-7.81; P = 0.001), presence of a minor head injury (OR 6.33, CI = 4.07-9.86; P < 0.001), performing a duraplasty (OR 1.86, CI = 1.20-2.87; P = 0.005) versus a slit durotomy (OR 3.95, CI = 1.67 - 9.35; P = 0.002), and, avoidance of a contralateral DC (OR 3.58, CI = 1.90-6.73; P < 0.001) [Table 2].
Follow up and long-term outcome
Among the cohort of 720 alive patients at discharge, 324 patients (45%) were assessed for a mean duration of 25.3 (±12.2) months (range, 3-42 months). Nineteen (7.5%) patients were in a persistent vegetative state (PVS), 37 (14.6%) patients had a severe functional disability, and 197 (77.8%) patients had a favorable outcome (GOS 4 and 5) in the severe head injury group at the last known follow-up amongst the surviving patients. A total of 71 patients (21.9%) expired [Figure 1].
Using a retrospective chart review and a prospective telephonic interview, we observed that a total of 410 patients underwent cranioplasty with either an autologous bone flap stored in the abdomen (N = 171), or in a bone bank (N = 209); or, from an artificial bone reconstituted using poly methylmethacrylate bone cement (N = 28) or a titanium mesh (N = 2). The infection rate in the patients undergoing a cranioplasty was 10.5% (43 patients). All these patients required a revision procedure.
Reduced mortality, a long-term favorable outcome, and improved quality of care are all goals to strive for, when opting for a therapeutic approach in patients with TBI. In our series, a favorable outcome (GOS 4 and 5) at discharge was observed in a significant proportion of patients undergoing a DC (46.5% overall; 39%, 64%, and 78.4% in patients presenting with severe, moderate, and minor TBI), which is comparable to the previously reported series where the outcome ranged from 16 to 61%. ,,,, Recent literature has shown the mortality rate in TBI to be ranging from 12 to 55%. ,,,,,, The present cohort also revealed similar results with a 41.7% overall mortality, and a 47.4 % mortality in the severe head injury group at discharge. The risk of post-traumatic seizures was less (1.7%) compared to the previously reported literature, , probably owing to the heterogeneous population represented in the present study. Post-traumatic hydrocephalus was reported in 7.5% patients, a figure similar to that reported by Polin et al.,  and Munch et al.  Prolonged ventilation and the consequent pulmonary complications were the most common causes of morbidity, , particularly in the severe head injury group with a higher incidence of associated chest injuries. The conventional notion that DC shifts the outcome from death to PVS has been challenged by the improved outcome in the current cohort, with only 7.5% of surviving patients being in a PVS and 77.8% having a favorable outcome (GOS 4 and 5) at a long-term follow-up (mean 25.3 months).
The DECRA trial  (Decompressive Craniectomy in Patients with Severe Traumatic Brain Injury) was a multicenter, randomized, controlled trial to test the efficacy of bifrontal DC in adults <60 years with TBI, in whom first-tier intensive care and neurosurgical therapies failed to maintain an ICP <20 mmHg. A total of 155 patients (73 patients in the surgical arm and 82 receiving standard care) with severe diffuse TBI (GCS <9) and CT scan evidence of brain swelling, were recruited from December 2002 to April 2010. The DECRA trial failed to show any benefit for an early bifrontal decompression, with patients in the surgical arm having worse outcomes than those treated medically (70 vs 51%). However, the conclusions of the DECRA trial raised several criticisms. Our study, though a retrospective one, has shown results quite in contrast to the findings reported by the DECRA investigators. These findings may be a significant contribution to the literature and an impetus to the continuation of DC as a treatment modality in the management of refractory ICH in TBI until the time that other robust and well-designed randomized controlled trials further settle the controversy surrounding this highly debatable issue.
The RESCUEicp trial (Randomized Evaluation of Surgery with Craniectomy for Uncontrollable Elevation of Intracranial Pressure) is still ongoing and a total of 400 patients of age range 10-65 years are expected to be recruited until December 2014. This trial is expected to shed some more light on the benefits obtained after a DC in TBI.
The univariate analysis of various independent factors potentially affecting the short-term outcome in patients with TBI revealed 14 variables with statistically significant values (P < 0.05) and 18 variables with P < 0.25. Applying multivariate analysis to these factors (P < 0.25), we finally observed a total of nine truly independent variables significantly affecting the patient's outcome. Severity of head injury at the time of admission had the highest odds (OR 6.33) of predicting the outcome followed by performing slit durotomy (OR 3.95) as compared to duraplasty (OR 1.86). The other factors predicting outcome after a DC were the presence of a preoperative infarct on CT scan (OR 3.68), the need for performing a contralateral DC (OR 3.58), the integrity of the pupillary reflexes (OR 2.28), the presence of preoperative hypotension (OR 1.91), the existence of an isolated TBI (OR 1.42), and the patient's age (OR 1.03). In a developing country like India, the follow-up is usually limited. Hence, the factors predicting outcome in the present study were analyzed for their short-term GOS at discharge rather than analyzing the follow-up GOS that was only available in a limited subgroup. The assumption used here was that there is usually no statistical difference between GOS at discharge and at a follow-up at 6 months as shown in a study by Agrawal et al. 
In a retrospective cohort of 55 patients with severe head injury, Pompucci et al.,  showed that age >65 years and a poor GCS (3-5) at presentation were adversely associated with the functional outcome in TBI patients undergoing a DC. Similarly, Hukkelhoven et al.,  performed a meta-analysis of 5,600 patients with severe TBI, and found that the association between the patient's age and outcome is a continuous one that is inversely related. This observation is supported by hypothesis that the plasticity and repair potential of an adult brain progressively deteriorates as it ages.  Ucar et al.,  found the initial GCS and age as predictors of outcome in a series of 100 DC performed in the severe head injury group. Among the pediatric population, Khan et al.,  published a retrospective study of 25 patients of age range 1-5 years with TBI who underwent a DC. Univariate analysis revealed that the GCS ≤5, a delay in presentation >150 min, a DC performed >4 h post injury, and an intraoperative blood loss >300 ml were significant predictors of a poor outcome. In this study, however, a multivariate analysis was not done.
To the best of author's knowledge, this study is the largest series reported so far with an average long-term follow-up for more than 2 years, evaluating in details, the outcome and prognostic factors in patients with TBI undergoing a DC. Particularly important were the modifiable risk factors such as hemodynamic instability which could be optimized by an adequate resuscitation as early as possible. In a developing country like ours, it is prudent to provide knowledge and resources to the paramedical staff for improving the overall outcome in such patients. Since our center is a tertiary care referral hospital, the time lapse since the sustenance of the injury is also an important confounding factor. This, however, was not accounted for in the current study. Polytrauma has also been shown as a grave prognostic marker in such patients, and it places emphasis on the need for a multispecialty team including an anesthetist, surgeon, orthopedician, physician, and neurosurgeon working in cohesion, following the Advanced Trauma Life Support (ATLS) guidelines.  The dilemma of whether to perform a slit durotomy or a lax duraplasty in the cases having a brain bulge has still not been unresolved. Our study is first to highlight the superior role of duraplasty when compared to a durotomy. In a small subset of patients who develop a contralateral SDH/EDH following a DC, it is not clear whether or not to proceed with a contralateral procedure. The present study highlights the importance of a less radical, conservative management in such patients as opposed to an urgent neurosurgical intervention on the contralateral side. An unique aspect of the present study was the inclusion of patients with minor and moderate head injury (as assessed at the time of admission) who required a DC when they had a subsequent neurological deterioration due to a delayed brain swelling.
A limitation of the present study was related to the fact that it was a single-center, nonrandomized, and retrospective study design. This was an institutional study and many surgeons (with variable experience) have operated on these patients. The data regarding the postoperative ICP monitoring was insufficient. The higher proportion of patients lost to follow-up in the current study was because the study included patients living in remote areas of the country with financial constraints and having limited access to means of communication. Considering the possibility that patients who were lost to follow-up were either dead or in a PVS, we may actually be underestimating the burden of PVS in the present cohort. Since our center is a tertiary care referral hospital, the time lapse since the injury was sustained was also an important influencing factor not accounted for by the current study.
TBI patients undergoing a DV have an overall favorable outcome with an acceptable morbidity and mortality. The severity of head injury, performing a duraplasty rather than a slit durotomy, the avoidance of a contralateral DC, and, the prevention of preoperative hypotension, infarct, or pupillary asymmetry, have the highest odds of influencing the short-term GOS at the time of discharge. Preoperative hemodynamic instability, an important modifiable risk factor, needs to be addressed by an early and adequate resuscitation in the prehospital phase of TBI to improve outcome. The present study also challenges the conventional notion that DC shifts the outcome from mortality to a PVS. The findings in our study may be a significant contribution to the literature and an impetus to the continuation of DC as a last resort in the treatment of refractory ICH in TBI. There is definitely a need for robust and well-designed randomized controlled trials which can further settle this contentious issue of whether or not a DC is useful in improving outcome.
[Figure 1], [Figure 2], [Figure 3], [Figure 4]
[Table 1], [Table 2]