Observed versus Predicted Outcome for DecompressiveCraniectomy: A Population-Based Study

Stephen Honeybul,1 Kwok M. Ho,2 Christopher R.P. Lind,1,3 and Grant R. Gillett4

Abstract

A number of studies have shown that decompressive craniectomy can reduce intracranial pressure and mayimprove outcome for patients with severe head injury. This cohort study assessed the long-term outcome ofneurotrauma patients who had a decompressive craniectomy for severe head injury in Western Australia be-tween 2004 and 2008. The web-based outcome prediction model developed by the CRASH trial collaboratorswas applied to the cohort. Predicted outcome and observed outcome were compared. Characteristics of outcomebetween those who had had a unilateral and those who had had a bilateral decompressive procedure werecompared. All complications were recorded. Among a total of 1,786 adult neurotrauma patients admitted duringthe study period, 147 patients (8.2%) had a decompressive craniectomy. A significant proportion of patients whorequired unilateral (37.3%) and bilateral (46.5%) craniectomy were able to return to work or study at 18 monthsafter the injury. The patients who required bilateral craniectomy more likely to be associated with an unfa-vorable outcome (Glasgow Outcome Scale score �3) than those who had unilateral craniectomy (odds ratio 4.42;95% confidence interval 1.16,16.81; p¼ 0.029), after adjusting for the timing of surgery, mechanism of injury, andthe predicted risk of unfavorable outcome. The functional outcome after either unilateral or bilateral decom-pressive craniectomy was significantly better than that predicted by the CRASH head injury prediction modelwhen the predicted risk was less than 80%. This study has demonstrated that in Western Australia decom-pressive craniectomy is a relatively common surgical procedure for the management of neurotrauma. A sig-nificant proportion of patients had a better-than-predicted long-term functional outcome.

Key words: decompressive craniectomy; outcome; neurotrauma

Introduction

Neurotrauma is a significant public health problem(Andelic et al., 2009; Selassie et al., 2008). It was esti-

mated that among a total of 288,009 hospitalized survivors ofneurotrauma in the United States in 2003, 124,626 patientsdeveloped long-term disability resulting in significant de-mand on rehabilitation services (Selassie et al., 2008). Over thepast two decades there has been a global resurgence of interestin the use of decompressive craniectomy in the managementof severe head injury (Aarabi et al., 2006; Albanese et al., 2003;Coplin et al., 2001; Guerra et al., 1999; Honeybul et al., 2009;Munch et al., 2000; Polin et al., 1997). The procedure can beeither unilateral or bilateral (or bifrontal). A unilateral cra-niectomy is most commonly performed after evacuation of amass lesion, such as a subdural hematoma, when cerebral

swelling is too severe for the bone flap to be safely replaced. Abilateral craniectomy is most commonly performed to controlintractable intracranial hypertension due to diffuse cerebraledema.

A number of studies have demonstrated that these proce-dures are effective in reducing intracranial pressure (Aarabiet al., 2006; Coplin et al., 2001; Guerra et al., 1999; Munchet al., 2000; Polin et al., 1997), and may improve outcome(Aarabi et al., 2006; Albanese et al., 2003; Coplin et al., 2001;Guerra et al., 1999; Honeybul et al., 2009; Munch et al., 2000;Polin et al., 1997). However, many of these reports are limitedby small sample size, inadequate risk adjustment, potentialpublication bias, and limited long-term outcome results.

There are currently two large randomized controlled trialsthat are assessing the effectiveness of decompressive cra-niectomy (Cooper et al., 2008; Hutchinson et al., 2006). Until

1Department of Neurosurgery, Sir Charles Gairdner Hospital and Royal Perth Hospital, Western Australia.2Department of Intensive Care Medicine and School of Population Health, and 3Centre for Neuromuscular and Neurological Disorders,

University of Western Australia, Western Australia.4Dunedin Hospital and Otago Bioethics Centre, University of Otago, Dunedin, New Zealand.

JOURNAL OF NEUROTRAUMA 27:1225–1232 (July 2010)ª Mary Ann Liebert, Inc.DOI: 10.1089/neu.2010.1316

1225

the results of these trials are available, the benefits of de-compressive craniectomy remain unproven (Sahuquillo andArikan, 2006; Servadei et al., 2007).

The main aims of this study were: (1) to assess the popu-lation-based incidence and long-term functional outcome ofpatients who have had a decompressive craniectomy for se-vere neurotrauma in Western Australia; (2) to determinewhether the Corticosteroid Randomisation After SignificantHead Injury (CRASH) collaborators outcome predictionmodel can be used to predict long-term outcome; (3) tocompare characteristics of outcome between those patientswho have had a unilateral with those who have had a bi-frontal decompressive procedure; and (4) to determine thecomplications of the decompressive procedure and subse-quent cranioplasty.

Methods

After obtaining hospital ethics committee approval we re-viewed the case notes and radiology of all patients who hadhad a decompressive craniectomy at the two major traumahospitals in Western Australia between 2004 and 2008. Thesetwo major trauma hospitals together with a pediatric hospitalare the only neurosurgical centers that provide neurotraumaservices in Western Australia. The standard medical man-agement of severe neurotrauma in the intensive care unit wasbased on the Brain Trauma Foundation guidelines (Brattonet al., 2007). This involves protocol-driven, step-wise admin-istration of sedation, ventilation, and cerebrospinal fluiddrainage where possible. Serum sodium concentrations weremaintained between 145 and 150 mmol/L�1. Anticonvulsanttherapy was prescribed for 1–2 weeks. Mannitol, thiopentonecoma, and hyperventilation were not routinely used exceptwhen treating transient rises in intracranial pressure (ICP).Nasogastric feeding was started as soon as possible unlesscontraindicated. All patients managed in the intensive careunit had a parenchymal ICP monitor inserted.

The aim was to maintain the ICP below 20 mm Hg, and thecerebral perfusion pressure above 60 mm Hg. A bifrontaldecompressive craniectomy was considered if the ICP couldnot be maintained below 20 mm Hg despite maximal medicalmanagement. (In the majority of cases the ICP was consis-tently above 30 mm Hg prior to surgery.)

A unilateral decompressive craniectomy was performedfollowing evacuation of a mass lesion when it was not pos-sible to replace the bone flap because the ICP was greater than20 mm Hg. All patients had a parenchymal ICP monitorplaced for postoperative monitoring following evacuation ofthe hematoma and attempted replacement of the bone flap.

The surgical interventions were intended to provide max-imal decompression. In general, the bifrontal craniotomiesextended from the floor of the anterior fossa as far posteriorlyas the coronal suture. The lateral decompression involvedremoval of the squamous temporal bone to decompress themiddle cranial fossa. The unilateral decompressions weredesigned to obtain as large a bone flap as possible within thelimitations of patient positioning. In all cases the dura wasopened (most commonly in a cruciate fashion), and whennecessary a duroplasty was undertaken (usually with a syn-thetic dural substitute).

All patients who were deemed to be suitable for reha-bilitation after their acute care in the two neurotrauma

centers were managed at a centralized rehabilitation hos-pital in Perth. These hospitals serve a population of about2.1 million.

The clinical data analyzed included: demographic factors,mechanism of injury, post-resuscitation Glasgow ComaScale score (GCS), pupillary response, and presence of ex-tracranial injuries. All radiology was reviewed by a con-sultant neurosurgeon (S.H.), and the findings were verifiedby another consultant neurosurgeon blinded to the clinicaldata (C.L.). The specific aims of the study are outlinedbelow.

Outcome

Case notes were used to determine functional outcomes at6, 12, and 18 months after injury. Formal neuropsychologicaltesting was reviewed where appropriate.

The CRASH outcome prediction model

The admission data were entered into the web-based pre-diction model established by the CRASH collaborators.Predicted outcomes were compared with observed outcomes.In this study a Glasgow Outcome Scale score (GOS) �3 wasregarded as an unfavorable outcome. We used the shape,slope, and intercept of a calibration curve, and also Hosmer-Lemeshow C chi-square statistics to assess whether theCRASH head injury prediction model could reliably predictunfavorable outcome over the entire range of predicted risks.The slope of the calibration curve was set to one when theintercept was estimated. A slope >1 indicates that the pre-dicted risks are not different enough across the risk strata, andan intercept <0 indicates that the predicted risks are system-atically too high (Ho, 2007).

Characteristics of outcome of unilateral versusbilateral (bifrontal) decompression

Categorical variables and data that were not normallydistributed were analyzed by the chi-square and Mann-Whitney tests, respectively. Multivariable logistic regressionwas used to assess the difference in outcomes between uni-lateral and bilateral craniectomy. The timing of surgery,mechanism of injury, and the predicted risk of unfavorableoutcome were used as covariates, and no variables were re-moved during the multivariate analysis.

All statistical analyses were two-tailed and performed bySPSS software (version 13; SPSS, Inc., Chicago, IL). A p value<0.05 was regarded as significant. All complications wererecorded.

Results

A total of 2,074 adult and pediatric patients with neuro-trauma were admitted to the two adult neurotrauma centersand the pediatric hospital between 2004 and 2008 in WesternAustralia. The population incidence of neurotrauma was19.47 (95% confidence interval [CI] 16.87,22.26) per 100,000person-years. Among a total of 1,786 neurotrauma adult pa-tients admitted during the study period, 147 patients (8.2%;95% CI 7.0,9.7) required either unilateral (n¼ 73) or bilateral(n¼ 74) decompressive craniectomy (Fig. 1). The incidence ofdecompressive craniectomy was 1.38 (95% CI 1.17,1.62) per100,000 person-years in Western Australia.

1226 HONEYBUL ET AL.

Outcome

At 18-month follow-up (Table 1): 58 patients (39.5%) madea good recovery and could return to work or study; 22 pa-tients (15%) remained moderately disabled; 25 patients (17%)were severely disabled; 5 patients (3.4%) remained in a veg-etative state; and 27 patients (18.4%) had died. Twenty-twopatients died as a direct result of uncontrolled cerebralswelling following the primary traumatic brain injury, 1 pa-tient died following an aortic dissection, and 4 patients diedfollowing complications of the decompressive procedure. Tenpatients did not have 18-month follow-up (6 were repatriatedand 4 were lost to follow-up).

A significant proportion of patients made further neuro-logical recovery between 6 and 18 months after the injury(Table 1). At 6 months, 25 patients had died and 32 patientshad achieved a good outcome. Ten patients did not have an18-month outcome assessment. Of the remaining 80 patients,26 patients (28.8%) improved from moderate to good out-come, 7 patients (8.75%) improved from severe to moderatedisability, 7 patients (8.75%) improved from vegetative stateto severe disability, and 1 patient (1.25%) improved from se-vere disability to good outcome.

The CRASH outcome prediction model

The model significantly overestimated the risk of unfa-vorable neurological outcome at 18 months compared to theoutcomes observed (Hosmer-Lemeshow C statistics¼ 17.5;p¼ 0.025; slope and intercept of the calibration curve¼ 2.01and �0.97, respectively; Fig. 2). The overestimation of riskoccurred predominantly when the predicted risk of an unfa-vorable outcome was less than 80%. The model appeared tobe more accurately calibrated (i.e., the observed risk was veryclose to the predicted risk) when the predicted risk of unfa-vorable outcome was greater than 80% (Fig. 2).

Characteristics of those requiring unilateralversus bilateral (bifrontal) decompression

The patients who required bilateral craniectomy wereyounger, more likely to have extracranial injuries or traumafrom motor vehicle accidents, and had craniectomy later thanthose who required unilateral craniectomy (2.2 versus 1.5days; p¼ 0.001; Table 2).

GOS score was not significantly different between those re-ceiving unilateral and bilateral craniectomy in the univariable

FIG. 1. Flow chart showing the study design (DECRA, Cooper et al., 2008).

DECOMPRESSIVE CRANIECTOMY: A POPULATION STUDY 1227

analyses, but bilateral craniectomy appeared to be associatedwith an unfavorable outcome (GOS� 3) compared to unilat-eral craniectomy (odds ratio 4.42; 95% CI 1.16,16.81; p¼ 0.029)after adjusting for other co-variates (Table 3).

Complications

There was a high incidence of complications. Seventy-fivepatients (51%) developed radiological evidence of a subdur-al/subgaleal effusion greater than 1 cm in thickness, and 5patients had the effusions treated by aspiration. Twenty-onepatients (14%) developed hydrocephalus that required inser-tion of a ventriculoperitoneal shunt. Sixteen patients (11%)developed autologous bone flap site infection that requiredremoval and subsequent titanium cranioplasty. Twenty-onepatients (14%) developed post-traumatic seizures, and 4 pa-tients (2.0%) died directly as a result of a complication of theprocedure. One patient died following a fall onto the unpro-tected cranium (Honeybul, 2009). Three patients sufferedmassive uncontrolled cerebral edema following cranioplasty.

Subdural or subgaleal effusion and hydrocephalus weresignificantly more likely to occur among patients who hadmore severe underlying head injury, as measured by predictedrisk of unfavorable outcome by the CRASH model (Table 4).

Discussion

There now appears little doubt that decompressive cra-niectomy will become a valuable tool in the management ofsevere head injury. However, despite numerous reportsdocumenting successful reductions in ICP (Aarabi et al., 2006;

Table 1. Outcome, Designation, and Complications after Unilateral and Bilateral Decompressive Craniectomy

All patients Unilateral BilateralVariable (n¼ 147) (n¼ 73) (n¼ 74) p Value

Outcome: Glasgow Outcome Scale score, no. (%)At 6 months: 1 32 (21.8) 18 (24.7) 14 (18.9) 0.250a

2 42 (28.6) 15 (20.5) 27 (36.5)3 34 (23.1) 24 (32.9) 10 (13.5)4 14 (9.5) 5 (6.8) 9 (12.1)5 25 (17.0) 11 (15.1) 14 (18.9)

At 12 months: 1 46 (31.3) 22 (31.0) 24 (32.4) 0.405a

2 37 (25.2) 16 (22.5) 21 (28.4)3 28 (19.1) 19 (26.8) 9 (12.2)4 8 (5.4) 3 (4.1) 5 (6.8)5 26 (17.7) 11 (15.5) 15 (20.3)

At 18 months: 1 58 (39.5) 24 (36.4) 34 (47.9) 0.392a

2 22 (15.0) 12 (18.2) 10 (14.1)3 25 (17.0) 17 (25.8) 8 (11.3)4 5 (3.4) 2 (3.0) 3 (4.2)5 27 (18.4) 11 (16.7) 16 (22.5)

Designation at 18 monthsFunctional status (%) 0.179

Died 27 (18.4) 11 (16.4) 16 (22.5)Unemployed or at home 53 (36.1) 31 (46.3) 22 (31.0)Back to work or school 58 (39.5) 25 (37.3) 33 (46.5)

Place of residence (%) 0.103Died 27 (18.4) 11 (15.7) 16 (22.2)Nursing home 25 (17.0) 17 (24.3) 8 (11.1)Home 90 (61.2) 42 (60.0) 48 (66.7)

Wheelchair/bedridden no. (%) 32 (21.8) 19 (27.1) 13 (17.8) 0.233Gastrostomy feeding, no. (%) 20 (13.6) 11 (15.7) 9 (12.3) 1.000

ComplicationsVP shunt, no. (%) 21(14.3) 8 (12.5) 13 (19.7) 0.342Flap site infection, no. (%) 16 (10.9) 7 (11.1) 9 (14.3) 0.790Seizure, no. (%) 21 (14.3) 9 (12.3) 12 (16.7) 0.488Subdural/subgaleal effusion, no. (%) 75 (51.0) 39 (61.9) 36 (57.1) 0.717

aComparing differences in the proportion of patients with favorable (Glasgow Outcome Scale score 1 and 2) and those with unfavorableoutcomes (Glasgow Outcome Scale score 3–5).

VP, ventriculoperitoneal.

FIG. 2. Graph showing the relationship between observedand predicted risk of unfavorable outcomes by the CRASHprediction model at 18 months post-injury.

1228 HONEYBUL ET AL.

Coplin et al., 2001; Guerra et al., 1999; Munch et al., 2000; Polinet al., 1997), and possible improvements in outcome(Aarabi et al., 2006; Albanese et al., 2003; Coplin et al., 2001;Guerra, et al., 1999; Honeybul et al., 2009; Munch et al., 2000;Polin et al., 1997), the benefits of the procedure remain sci-

entifically unproven (Sahuquillo and Arikan, 2006; Servadeiet al., 2007). It is hoped that the results of the two ongoingrandomized controlled trials evaluating the role of decom-pressive craniectomy in severe neurotrauma will provide thenecessary data (Cooper et al., 2008; Hutchinson et al., 2006).

Table 2. Characteristics of the Cohort and the Differences between Unilateral

and Bilateral Decompressive Craniectomy

All patients Unilateral BilateralVariable (n¼ 147) (n¼ 73) (n¼ 74) p Value

Age (SD, median, IQR) 33.1 (15, 28, 20–44) 38.3 (16, 36, 25–51) 28.0 (13, 22, 18–36) 0.001

Male, no. (%) 123 (83.7) 62 (84.9) 61 (82.4) 0.824

Mechanism of injury, no (%): 0.003- Assault 28 (19.0) 22 (30.1) 6 (8.1)- Explosion 2 (1.4) 0 (0) 2 (2.7)- Fell from height 40 (27.2) 20 (27.4) 20 (27.0)- MVA 77 (52.4) 31 (42.5) 46 (62.2)

GCS score (SD, median, IQR) 7.4 (3.3, 7, 4–10) 7.3 (3.4, 7, 4–10) 7.3 (3.1, 7, 5–10) 0.770*

Pupils, no. (%): 0.202- Both reactive 105 (71.4) 50 (68.5) 55 (74.3)- One reactive 19 (12.9) 13 (17.8) 6 (8.1)- Both non-reactive 23 (15.6) 10 (13.7) 13 (17.6)

CT brain findings, no. (%):- SAH 137 (93.2) 68 (93.2) 69 (93.2) 1.000- Midline shift 113 (76.9) 69 (94.5) 44 (59.5) 0.001- Petechial hemorrhage 137 (93.2) 68 (93.2) 69 (93.2) 1.000- Basal cistern effaced 45 (30.6) 28 (38.4) 17 (23.0) 0.043- Intracerebral hematoma (>1 cm) 57 (38.8) 26 (35.6) 31 (41.9) 0.499

Extracranial injuries, no. (%) 61 (41.5) 22 (30.1) 39 (52.7) 0.007

Predicted risk of unfavorable outcomeby the CRASH model, %(SD, median, IQR)

68.8 (76, 62, 46–84) 68.9 (20, 72, 55–87) 68.6 (105, 54, 42–76) 0.001*

Timing of craniectomy fromadmission, days (SD, median, IQR)

1.9 (1.6, 1, 1–2) 1.5 (1.7, 1, 1–1) 2.2 (1.5, 2, 1–3) 0.001*

ICU stay, days (SD, median, IQR) 11.0 (6, 11, 6–15) 9.5 (6, 10, 4–13) 12.5 (6, 12, 9–16) 0.005*

Post-ICU ward stay, days(SD, median, IQR)

52 (65, 30, 12–65) 48 (55, 32,12–63) 56 (74, 28, 14–71) 0.194*

Rehabilitation hospital stay,days (SD, median, IQR)

43 (67, 19, 0–53) 43 (77, 7, 0–46) 42 (55, 28, 0–56) 0.913*

*p Value generated by Mann-Whitney test.ICU, intensive care unit; GCS, Glasglow Coma Scale; SD, standard deviation; IQR, interquartile range; MVA, motor vehicle accident.

Table 3. Multivariate Analysis Comparing the Difference between Unilateral and Bilateral

Decompressive Craniectomy on Risk of Unfavorable Outcome at 18 Months

VariableOdds ratio

(95% confidence interval) p Value

Predicted risk of unfavorable outcomeby the CRASH model (per % increment)

1.13 (1.08,1.17) 0.001

Bilateral decompressive craniectomycompared to unilateral craniectomy

4.42 (1.16,16.81) 0.029

Timing of craniectomy(per-day increment since admission)

0.74 (0.48,1.15) 0.184

Mechanism of neurotraumaa

Motor vehicle accident 1 0.986Assault 0.96 (0.25–3.77) 0.954Fell from height 1.09 (0.31–3.81) 0.896

aTwo patients sustaining neurotrauma by explosion were not included in this multivariate analysis.Hosmer-Lemeshow chi-square and Nagelkerke R2 of the model was 8.59 ( p¼ 0.378) and 0.649, respectively.

DECOMPRESSIVE CRANIECTOMY: A POPULATION STUDY 1229

Nevertheless, the strict inclusion and exclusion criteria ofthese trials may limit the degree to which their findings can beapplied to all categories of head injury.

As such, this population-based study may still provideimportant complementary information on the role of de-compressive craniectomy. This study has shown that de-compressive craniectomy was required in 8% of adult patientswith severe head injury, a result similar to those of otherstudies (Aarabi et al., 2006; Albanese et al., 2003; Guerra et al.,1999). This represents a population-based incidence of de-compressive craniectomy of 1.38 per 100,000 person-years inWestern Australia.

Outcome

There is a wide variation in reported outcomes followingdecompressive craniectomy. This is probably because moststudies have included heterogeneous patient populations andoutcomes have been assessed at different times and withdiffering outcome measures. Nonetheless, the results of thisstudy are comparable with those of some other studies(Aarabi et al., 2006; Albanese et al., 2003; Guerra et al., 1999;Munch et al., 2000). The long-term functional outcome seenafter both unilateral or bilateral craniectomy was good in asignificant proportion of patients; 40% could return to work orstudy by 18 months. A smaller but still significant proportionof patient (20.4%) remained dependent at 18 months. Twenty-five patients (17%) died as a result of the initial injury despitedecompression, and there were four deaths that were as aresult of complications.

While this study has added to the growing literature thatwould seem to support the role of decompressive cra-niectomy, it has clearly demonstrated that neurological re-covery is not fully captured at either 6 or 12 months aftersevere neurotrauma. From 6-month follow-up to 18-monthfollow-up 40 patients improved by one point on the GOSscale and one patient improved by two points. These resultsindicate that interpretation of decompressive craniectomystudies requires caution if they are based on relatively short-term follow-up, and would suggest that a minimum of atleast 12 months of follow-up are required for randomized

controlled trials and for prognostic models of severe neuro-trauma.

The CRASH prediction model

While the prognostic significance of age (Aarabi et al., 2006;Figaji et al., 2003; Guerra et al., 1999; Howard et al., 2008;Steyerberg et al., 2008; Ucar et al., 2005), GCS (Cremer et al.,2006; Guerra et al., 1999; Steyerberg et al., 2008; Ucar et al., 2005),pupillary reaction (Cremer et al., 2006; Steyerberg et al., 2008;Ucar et al., 2005), extracranial injuries (Steyerberg et al.,2008; Ucar et al., 2005), and radiological appearance (Cremeret al., 2006; Munch et al., 2000; Steyerberg et al., 2008), hasbeen demonstrated by a number of studies, the CRASH col-laborators study have developed the first user-friendly web-based outcome prediction model (Perel et al., 2008). Thismodel has combined these prognostic factors, and by usinglogistic regression analysis has provided an index of severityof injury based on the percentage risk of an unfavorableoutcome at 6 months. By applying this model to a smallercohort of patients who had a decompressive craniectomy, wehave demonstrated how this model may be used to highlightthose patients who are very likely to have an unfavorableoutcome at 18 months post-injury (Honeybul, 2009).

Expanding this smaller study has confirmed our earlierfindings, that the model overestimates unfavorable outcome(i.e., outcome in craniectomy patients is better than expectedusing the CRASH prediction model). While it would seemthat the improved outcome is because of the surgical proce-dure, interpretation of these data requires careful consider-ation. First, these results can only be applied to the institutionsin which the patients were managed, and different resultsmay be established in other neurosurgical units. Second,while it would appear that the procedure plays an importantrole in the eventual outcome, there can be little doubt that allaspects of medical and nursing input, from pre-hospital care,emergency resuscitation, and intensive care, through the re-habilitation period provide important contributions. The dif-ference between the predicted and observed outcome couldbe influenced by any of these factors. There was, however, asteep rise in observed unfavorable outcome when the

Table 4. Relation between Predicted Risk of Unfavorable Outcome, Glasgow Coma Scale Score (GCS)and Neurological Complications

Complication Yes No p Value*

Hydrocephalusrequiring shunt

Risk of unfavorableoutcome, % (SD) [median, IQR]

77.1 (16)[84, 62–88]

57.2 (21)[56, 42–73]

0.001

GCS (SD) [median, IQR] 5.0 (2) [4, 3–6] 8.0 (3)[8, 6–11]

0.001

Subdural orsubgalealeffusion

Risk of unfavorable outcome, % (SD)[median, IQR]

63.2 (22)[65, 48–83]

56.1 (18)[54, 42–68]

0.036

GCS (SD) [median, IQR] 7.3 (4) [7, 4–10] 7.9 (3)[8, 6–10]

0.263

Seizure Risk of unfavorable outcome, % (SD)[median, IQR]

63.8 (22)[62, 48–85]

69.6 (82)[62, 46–83]

0.912

GCS (SD) [median, IQR] 7.5 (4)[6, 4–13]

7.3 (3)[7, 5–9]

0.878

*p Values were generated by the Mann-Whitney test.SD, standard deviation; IQR, interquartile range.

1230 HONEYBUL ET AL.

predicted risk was greater than 80%. While it would appearthat poor outcome can be predicted, we would agree with theCRASH collaborators (Perel et al., 2008), that this modelshould only be used to support and not replace clinicaljudgment. We would not advocate using this model to discussstatistical data with relatives of a severely injured patient.Instead the model may be more useful as an index of severityof injury, thereby providing supportive information to facili-tate the discussion of realistic outcome expectations.

Bilateral versus unilateral craniectomy

Our results have shown that the characteristics of patientswho required unilateral decompressive craniectomy werevery different from those who required bilateral decom-pressive craniectomy. However, this comparison has signifi-cant limitations, in that what is really being compared are twodifferent pathologies, namely surgical mass lesion and diffusecerebral edema. The patients who required unilateral de-compressive craniectomy appeared to have a better long-termoutcome than the patients who required bilateral cra-niectomy, after adjusting for timing of surgery and mecha-nism and severity of brain injury. However, while this is aninteresting finding it is not entirely unexpected. At the twoinstitutions examined in this study a unilateral decompres-sion is usually performed following the emergent evacuationof a mass lesion such as an acute subdural hematoma. A bi-frontal decompression procedure is usually performed after aperiod of intracranial pressure monitoring when the patienthas failed to respond to maximal medical management.

There are a number of clinical reports demonstrating thatearly evacuation of a hematoma and decompression reducesmortality (Cremer et al., 2006; Munch et al., 2000). There isalso accumulating experimental evidence that very early de-compression significantly reduces secondary brain injury(Plesnila, 2007; Zweckberger et al., 2006). In view of this itwould seem reasonable to assume that for patients with asimilar index of severity of injury (as determined by theCRASH outcome prediction model), those patients that havean early decompression will have a better outcome than thosethat have a later decompression. Despite this finding theclinical management of patients with severe head injury isunlikely to significantly change. Those patients with an acutesubdural hematoma are likely to have emergent surgery, andthose patients with diffuse edema are likely to have a periodof intracranial pressure monitoring prior to a decompressiveprocedure.

Complications

A number of studies have reported the high incidence ofcomplications associated not only with the decompressiveprocedure, but also with the subsequent cranioplasty (Hon-eybul, 2009; Honeybul (in press), 2009; Stiver, 2009; Yang et al.,2008). Our results showed that there was a high incidence ofpost-traumatic seizures and wound infections, and some ofthese complications such as subdural or subgaleal effusionsand subsequent hydrocephalus were related to the severity ofthe head injury.

Sudden death due to massive uncontrolled cerebral edemafollowing cranioplasty has been reported in one patient in ourprevious study (Honeybul, 2009). It has since occurred in twomore patients. All three patients were making a poor recovery

following very severe head injury. They deteriorated in thefirst few hours following surgery, with fixed dilated pupilsand massive cerebral swelling confirmed on CT scanning.They failed to recover despite removal of the bone flap. Themechanism behind this has not been established, but may berelated to failure of autoregulation (Czosnyka et al., 2001;Sviri et al., 2009). Further studies are required to establishexactly which patients are at risk of this complication.

Summary

In summary, decompressive craniectomy appears to be arelatively common surgical procedure in the management ofneurotrauma in Western Australia. A significant proportionof patients had good long-term functional outcome and couldgo back to work or study by 18 months post-injury. Whilethese results may provide important complementary infor-mation to the ongoing randomized trials, this study has somelimitations. First, observational studies are prone to bias.Second, although our study is one of the largest cohort studiesperformed to date on decompressive craniectomy, the samplesize is still relatively small, and this will affect the precision ofthe results. Finally, while many patients achieved a better-than-predicted outcome, there are many other factors thatmay have had a confounding influence.

Author Disclosure Statement

No conflicting financial interests exist.

References

Aarabi, B., Hesdorffer, D.C., Ahn, E.S., Aresco, C., Scalea, T.M.,and Eisenberg, H.M. (2006). Outcome following decom-pressive craniectomy for malignant swelling due to severehead injury. J. Neurosurg. 104, 469–479.

Albanese, J., Leone, M., Alliez, J.R., Kaya, J.M., Antonini, F.,Alliez, B., and Martin, C. (2003). Decompressive craniectomyfor severe traumatic brain injury: Evaluation of the effects atone year. Crit. Care Med. 31, 2535–2538.

Andelic, N., Hammergren, N., Bautz-Holter, E., Sveen, U.,Brunborg, C., and Røe, C. (2009). Functional outcome andhealth-related quality of life 10 years after moderate-to-severetraumatic brain injury. Acta Neurol. Scand. 120, 16–23.

Bratton, S.L., Chestnut, R.M., Ghajar, J., McConnell Hammond,F.F., Harris, O.A., Hartl, R., Manley, G.T., Nemecek, A.,Newell, D.W., Rosenthal, G., Schouten, J., Shutter, L., Tim-mons, S.D., Ullman, J.S., Videtta, W., Wilberger, J.E., andWright, D.W. (2007). Brain Trauma Foundation; AmericanAssociation of Neurological Surgeons; Congress of Neurolo-gical Surgeons; Joint Section on Neurotrauma and CriticalCare, AANS/CNS. Guidelines for the management of severetraumatic brain injury. J. Neurotrauma 24, Suppl. 1.

Cooper, D.J., Rosenfeld, J.V., Murray, L., Wolfe, R., Ponsford, J.,Davies, A., D’urso, P., Pellegrino, V., Malham, G., and Koss-man, T. (2008). Early decompressive craniectomy for patientswith severe traumatic brain injury and refractory intracranialhypertension—a pilot randomised trial. J. Crit. Care 23, 387–393.

Coplin, W.M., Cullen, N.K., Policherla, P.N., Vinas, F.C.,Wilseck, J.M., Zafonte, R.D., and Rengachary, S.S. (2001).Safety and feasibility of craniectomy with duraplasty as theinitial surgical intervention for severe traumatic brain injury.J. Trauma 50, 1050–1059.

DECOMPRESSIVE CRANIECTOMY: A POPULATION STUDY 1231

Cremer, O.L., Moons, K.G., van Dijk, G.W., van Balen, P., andKalkman, C.J. (2006). Prognosis following severe head injury:Development and validation of a model for prediction ofdeath, disability, and functional recovery. J. Trauma 61, 1484–1491.

Czosnyka, M., Smielewski, P., Piechnik, S., Steiner, L.A., andPickard, J.D. (2001). Cerebral autoregulation following headinjury. J. Neurosurg. 95, 756–763.

Figaji, A.A., Fieggen, A.G., and Peter, J.C. (2003). Early decom-pressive craniectomy in children with severe traumatic braininjury. Childs. Nerv. Syst. 19, 666–673.

Guerra, W.K., Gaab, M.R., Dietz, H., Mueller, J.U., Piek, J., andFritsch, M.J. (1999). Surgical decompression for traumaticbrain swelling: indications and results. J. Neurosurg. 90, 187–196.

Ho, K.M. (2007). Forest and funnel plots illustrated the calibra-tion of a prognostic model: a descriptive study. J. Clin. Epi-demiol. 60, 746–751.

Honeybul, S. (2010). Complications of decompressive cra-niectomy for head injury. J. Clin. Neurosci. 17, 430–435.

Honeybul, S. (2009). Decompressive craniectomy: a new com-plication. J. Clin. Neurosci. 16, 727–729.

Honeybul, S., Ho, K.M., Lind, C.R.P., Corcoran, T., and Gillett,G.R. (2009). The retrospective application of a predictionmodel to patients who have had a decompressive craniectomyfor trauma. J. Neurotrauma 26, 2179–2183.

Howard, J.L., Cipolle, M.D., Anderson, M., Sabella, V., Shol-lenberger, D., Li, P.M., and Pasquale, M.D. (2008). Outcomeafter decompressive craniectomy for the treatment of severetraumatic brain injury. J Trauma 65, 380–385.

Hutchinson, P.J., Corteen, E., Czosnyka, M., Mendelow, A.D.,Menon, D.K., Mitchell, P., Murray, G., Pickard, J.D., Rickels,E., Sahuquillo, J., Servadei, F., Teasdale, G.M., Timofeev, I.,Unterberg, A., and Kirkpatrick, P.J. (2006). Decompressivecraniectomy in traumatic brain injury: the randomized mul-ticenter RESCUEicp study. Acta Neurochir. Suppl. 96, 17–20.

Munch, E., Horn, P., Schure, L., Piepgras, A., Paul, T., andSchmiedek, P. (2000). Management of severe traumatic braininjury by decompressive craniectomy. Neurosurgery 47, 315–323.

Perel, P., Arango, M., Claydon, T., Edwards, P., Komolafe, E.,Poccock, S., Roberts, I., Shakur, H., Steyerberg, E., and Yut-thakasemsunt, S. (2008). Predicting outcome after brain injury:Practical prognostic models based on a large cohort of inter-national patients. BMJ 23, 425–429.

Plesnila, N. (2007). Decompression craniectomy after traumaticbrain injury: recent experimental results. Prog. Brain Res. 161,393–400.

Polin, R.S., Shaffrey, M.E., Bogaey, C.A., Tisdale, N., German-son, T., Bocchicchio, B., and Jane, J.A. (1997). Decompressive

bifrontal craniectomy in the treatment of severe refractoryposttraumatic cerebral edema. Neurosurgery 41, 84–94.

Sahuquillo, J., and Arikan, F. (2006). Decompressive craniectomyfor the treatment of refractory high intracranial pressure intraumatic brain injury. Cochrane Database Syst. Rev. 25,CD003983.

Seelig, J.M., Becker, D.P., Miller, J.D., Greenberg, R.P., Ward,J.D., and Choi, S.C. (1981) Traumatic acute subdural hema-toma: major mortality reduction in comatose patients treatedwithin four hours. N. Engl. J. Med. 18, 1511–1518.

Selassie, A.W., Zaloshnja, E., Langlois, J.A., Miller, T., Jones, P.,and Steiner, C. (2008). Incidence of long-term disability fol-lowing traumatic brain injury hospitalization, United States. J.Head Trauma Rehabil. 23, 123–131.

Servadei, F., Compagnone, C., and Sahuquillo, J. (2007). The roleof surgery in traumatic brain injury. Curr. Opin. Crit. Care. 13,163–168.

Steyerberg, E.W., Mushkudiani, N., Perel, P., Butcher, I., Lu, J.,McHugh, G.S., Murray, G.D., Mararou, A., Roberts, I., Hab-bema, J.D., and Maas, A.I. (2008). Predicting outcome aftertraumatic brain injury: development and international vali-dation of prognostic scores based on admission characteristics.PLoS Med. 5, e165; discussion e165.

Stiver, S.I. (2009). Complications of decompressive craniectomyfor traumatic brain injury. Neurosurg. Focus. 26, E7.

Sviri, G.E., Aaslid, R., Douville, C.M., Moore, A., and Newell,D.W. (2009). Time course for autoregulation recovery follow-ing severe traumatic brain injury. J. Neurosurg. 111, 695–700.

Ucar, T., Akyuz, M., Kazan, S., and Tuncer, R. (2005). Role ofdecompressive surgery in the management of severe headinjuries: prognostic factors and patient selection. J. Neuro-trauma 22, 1311–1318.

Yang, X.F., Wen, L., Shen, F., Li, G., Lou, R., Liu, W.G., andZhan, R.Y. (2008). Surgical complications secondary to de-compressive craniectomy in patients with a head injury: aseries of 108 consecutive cases. Acta Neurochir. (Wien.) 150,1241–1247.

Zweckberger, K., Eros, C., Zimmermann, R., Kim, S.W., Engel, D.,and Plesnila, N. (2006). Effect of early and delayed decom-pressive craniectomy on secondary brain damage after con-trolled cortical impact in mice. J. Neurotrauma 23, 1083–1093.

Address correspondence to:Stephen Honeybul, M.D.

Department of NeurosurgerySir Charles Gairdner Hospital

Hospital Avenue, NedlandsPerth, Western Australia

E-mail: stephen.honeybul@health.wa.gov.au

1232 HONEYBUL ET AL.