British Journal of Neurosurgery
ISSN: 0268-8697 (Print) 1360-046X (Online) Journal homepage: http://www.tandfonline.com/loi/ibjn20
Predicting outcomes of decompressive craniectomy: use of Rotterdam Computed Tomography Classification and Marshall Classification Muhammad Waqas, Muhammad Shahzad Shamim, Syed Faaiz Enam, Mohsin Qadeer, Saqib Kamran Bakhshi, Iqra Patoli & Khabir Ahmad To cite this article: Muhammad Waqas, Muhammad Shahzad Shamim, Syed Faaiz Enam, Mohsin Qadeer, Saqib Kamran Bakhshi, Iqra Patoli & Khabir Ahmad (2016): Predicting outcomes of decompressive craniectomy: use of Rotterdam Computed Tomography Classification and Marshall Classification, British Journal of Neurosurgery, DOI: 10.3109/02688697.2016.1139047 To link to this article: http://dx.doi.org/10.3109/02688697.2016.1139047
Published online: 01 Feb 2016.
Submit your article to this journal
Article views: 5
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=ibjn20 Download by: [Emory University]
Date: 06 February 2016, At: 01:59
BRITISH JOURNAL OF NEUROSURGERY, 2016 http://dx.doi.org/10.3109/02688697.2016.1139047
ORIGINAL ARTICLE
Predicting outcomes of decompressive craniectomy: use of Rotterdam Computed Tomography Classification and Marshall Classification Muhammad Waqas, Muhammad Shahzad Shamim, Syed Faaiz Enam, Mohsin Qadeer, Saqib Kamran Bakhshi, Iqra Patoli and Khabir Ahmad
Downloaded by [Emory University] at 01:59 06 February 2016
Department of Surgery, Section of Neurosurgery, Aga Khan University Hospital, Karachi, Pakistan
ABSTRACT
ARTICLE HISTORY
Background Data on the evaluation of the Rotterdam Computed Tomography Classification (RCTS) as a predictor of outcomes in patients undergoing decompressive craniectomy (DC) for trauma is limited and lacks clarity. Objective To explore the role of RCTS in predicting unfavourable outcomes, including mortality in patients undergoing DC for head trauma. Methods This was an observational cohort study conducted from 1 January 2009 to 31 March 2013. CT scans of adults with head trauma prior to emergency DC were scored according to RCTS. A receiver operating characteristic curve analysis was performed to identify the optimal cut-off RCTS for predicting unfavourable outcomes [Glasgow outcome scale (GOS) ¼ 1–3]. Binary logistic regression analysis was performed to evaluate the relationship between RCTS and unfavourable outcomes including mortality. Results One hundred ninety-seven patients (mean age: 31.4 ± 18.7 years) were included in the study. Mean Glasgow coma score at presentation was 8.1 ± 3.6. RCTS was negatively correlated with GOS (r ¼ 0.370; p50.001). The area under the curve was 0.687 (95% CI: 0.595–0.779; p50.001) and 0.666 (95% CI: 0.589–0.742; p50.001) for mortality and unfavourable outcomes, respectively. RCTS independently predicted both mortality (adjusted odds ratio for RCTS43 compared with RCTS 3: 2.792, 95% CI: 1.235–6.311) and other unfavourable outcomes (adjusted odds ratio for RCTS43 compared with RCTS 3: 2.063, 95% CI: 1.056–4.031). Conclusion RCTS is an independent predictor of unfavourable outcomes and mortality among patients undergoing emergency DC.
Received 26 September 2014 Revised 27 October 2015 Accepted 3 January 2016 Published online 28 January 2016
Introduction Among the radiological scoring systems for traumatic brain injury (TBI), Marshall CT Scan Classification system1 has been widely used for categorisation of severity and prediction of prognosis.2,3 The scoring system has been shown to be an independent predictor of unfavourable outcomes; however, the system is limited by its grouping of majority of the patient requiring DC with a mass lesion into class IV or V.4 Moreover, the Marshall system is a classification system rather than a scoring system, and therefore allows little room to accommodate patients with more complex radiological features.1,3 More recently the Rotterdam Computed Tomography Classification (RCTS),3 which has numerical values of up to 6 has been assessed as a potential tool for predicting outcomes in these patients.4 The published work showed RCTS to have a significant correlation with outcomes although the analysis lacked a specific RCTS cut-off number for prediction of unfavourable outcome or death.4 To explore the role of RCTS in predicting unfavourable outcomes, including mortality in patients with head trauma undergoing DC at a tertiary care setup of a low- and middle-income countries (LMIC).
Materials and methods This was a retrospective observational cohort study conducted at the Section of Neurosurgery, Aga Khan University Hospital Karachi, a CONTACT Muhammad Shahzad Shamim P.O. Box 74800, Karachi, Pakistan ß 2016 Taylor & Francis
[email protected]
KEYWORDS
Decompressive craniectomy; Glasgow outcome scale; Rotterdam Classification; traumatic brain injury; unfavourable outcomes
tertiary care referral hospital with level 1 trauma facilities. The study period was from 1 January 2009 to 31 March 2013. Inclusion criteria All patients aged 16 years, presenting with isolated TBI who required unilateral or bilateral decompressive craniectomy (DC) during the course of hospital stay were included in the study. Exclusion criteria: patients with incomplete medical records, including CT scans were excluded from the study. Data collection procedure Records of all the patients who conformed to the inclusion criteria were retrieved from the hospital database and the operating room (OR) database, that keeps the record of all procedures performed in the OR. Data were collected on a structured pro-forma. Emergency room information on demographics, post-resuscitation Glasgow Coma Score (GCS), revised trauma score (RTS), pupillary response, etc., was gathered from initial trauma assessment forms. Information regarding the surgical intervention and stay in ICU and high dependency unit was obtained from the review of OR and postoperative records available in charts. All individual CT images were reviewed through our picture archiving and communication system (PACS) and each CT scan was classified according to the RCTS by
Department of Surgery, Section of Neurosurgery, Aga Khan University Hospital, Stadium Road,
2
M. WAQAS ET AL.
two of the authors independently. Conflicts in interpretation were resolved by consensus. The authors were blind to the outcomes. Definitions and scales
Downloaded by [Emory University] at 01:59 06 February 2016
RCTS as described originally by Maas et al.3 was used for this purpose (Figure 1). Briefly, it is a standardised CT-based classification system, a number of up to 6 and takes into account status of basal cisterns, midline shift, subarachnoid hemorrhage and extradural hematoma. Patients’ outcomes were determined using the Glasgow outcome scale (GOS) as follows: score of 1: dead; 2: vegetative state; 3: severely disabled (conscious but the patient requires others for daily support due to disability); 4: moderately disabled (the patient is independent but disabled); and 5: good recovery (the patient has resumed most normal activities but may have minor residual problems). Outcomes were dichotomised to indicate favourable or unfavourable as frequently used in literature.5 GOS of 4 or more was considered to be favourable.
Indications and technique of surgery As an institutional protocol, all patients with severe TBI (GCS 8 or less) are managed with immediate intubation and admission in ICU. Intracranial pressure (ICP) monitoring is not routinely employed. Patients with mass lesions such as extradural or subdural hematoma, fulfilling Brain Trauma Foundation guidelines, are taken to the OR for emergency decompressive surgery and in case of diffuse cerebral oedema despite evacuation of mass lesion, the bone flap is not replaced. All the decompressive surgeries are performed by senior residents under direct supervision of one of the six Consultant Neurosurgeons. In a standard DC, the bone flap of at least 12 cm diameter is removed along with expansile duraplasty.
Follow-up Outcomes of the patients were recorded from medical records up till last clinic visit. Information was obtained according to the GOS as described above. The observer of outcomes was masked of the RCTS of individual patients. Analysis Data were analysed using SPSS IBM version 19 (Armonk, NY). Proportions were calculated to describe categorical data. Mean and standard deviation were used for continuous data and scoring variables such as GCS, RTS and GOS were analysed as numerical data after reviewing previous literature and books on statistics.6 A receiver operating characteristic curve analysis was performed to identify the optimal cuto-ff RCT score for predicting such outcomes. Patients were then categorised into two groups based on the basis of ROC cut-off value. Binary logistic regression analysis was performed to evaluate the relationship between RCTS and adverse outcomes including mortality while controlling for the effect of age, RTS and incident to arrival delay in emergency.
Results A total of 235 patients underwent DC in the time period. Thirty-eight patients were excluded from the study based on exclusion criteria. Medical records of 197 patients were thus reviewed. Of these, 174 (88.3%) were male and 23 (11.7%) female. By the end of the study, only 7 patients were lost to follow-up. Demographics, baseline population characteristics and outcomes are shown in Table 1. The mean age of the patients was 31.4 ± 18.7 years. Road traffic accidents were the most common cause of head injury with 129 (65.5%) followed by fall from a height with 36 (18.3%). Mean GCS on arrival of our study subjects was 8.1 ± 3.6. Median duration of follow-up was 6.0 months (range: 3.0–48). The mean RTS was 10.0 ± 1.5. The average RCTS was 3.6 ± 1.2. Figure 2 shows the distribution of patients with different Marshall Classification and RCTS. Binary logistic regression analysis showed a significant independent association between RCTS and mortality (adjusted odds ratio for RCTS 43 compared with RCTS 3: 2.792, 95% CI: 1.235– 6.311) while controlling for the effect of age, RTS and incident to arrival delay (Table 2). A significant independent association was also observed between RCTS and unfavourable outcomes (adjusted odds ratio for RCTS 43 compared with RCTS 3: 2.063, 95% CI: 1.056–4.031) while controlling for the effect of age, RTS and incident to arrival delay (Table 3). The AUC was 0.666 (95% CI: 0.589–0.742; p50.001) for unfavourable outcome and 0.687 (95% CI: 0.595–0.779; p50.001) for mortality. Maximal sensitivity and specificity was obtained with a RCTS of 4 (75% and 55%, respectively) on ROC curve for mortality (Figure 3). ROC analysis for Marshall Classification showed an AUC of 0.48 (95% CI: 0.385–0.591; p ¼ 0.83) for unfavourable outcomes while AUC of 0.452 (95% CI: 0.33–0.57; p ¼ 0.45; Figure 4).
Discussion
Figure 1. Rotterdam Computed Tomography Classification. Adapted from Maas et al.3
TBI will surpass most other diseases as the major cause of death and disability7 by the year 2020. With an estimated 10 million people affected annually, the burden of TBI is especially prominent in LMIC. The cost of managing severe TBI is substantial8 especially when its economic predicament for LMIC is considered. Persistently high ICP after TBI and its correlation with increased mortality has
BRITISH JOURNAL OF NEUROSURGERY
Table 1. Population characteristics and baseline variables.
Downloaded by [Emory University] at 01:59 06 February 2016
Variable Mean age (years) ± SD Sex, no. (%) Male Female Mechanism of head injury, no. (%) Road traffic accident Motorbike Car Pillion Riding Driver Pedestrian Fall Gunshot Others Mean GCS ± SD Mean revised trauma score ± SD Pupil reactivity, no. (%) Equal, reactive Anisocoria Fixed and dilated Pinpoint Not documented Rotterdam CT score, no. (%) 1 2 3 4 5 6 Score 3 Score 4 Mean Rotterdam CT score, mean ± SD Marshall Classification, mean ± SD Median hospital stay, days (IQR) GOS at the end of follow-up, no. (%) (n ¼ 190) Unfavourable outcome Favourable outcome Mortality (n ¼ 190)
Statistic 31.4 ± 18.7 174 (88.3) 23 (11.7) 129 (65.5) 69 (35.0) 15 (7.6) 3 (1.5) 8 (4.1) 34 (17.3) 36 (18.3) 18 (9.1) 13 (6.6) 8.1 ± 3.6 10.0 ± 1.5 105 (53.3) 75 (38.1) 5 (2.5) 1 (0.5) 11 (5.6) 7 (3.6) 28 (14.2) 61 (31.0) 52 (26.4) 37 (18.8) 12 (6.1) 96 (48.8) 101 (51.3) 3.6 ± 1.2 4.6 ± 1.4 14.0 (7.0–23.0) 102 (53.7) 88 (46.3) 44 (23.2)
3
been demonstrated by multiple studies.9–11 DC is emerging as an effective surgical measure for reducing ICP and is now considered a management option for medically refractive ICP.12,13 However, the outcomes of DC tend to vary in different reports and depend greatly on patient selection.12,13 Determining the prognostic indicators in patients with severe TBI undergoing DC can therefore be of great potential value when counseling patients’ families with regards to outcome, clinical decision making and also to help prioritise resources in resource constrained settings such as LMIC. It is generally accepted that the most sensitive prognostic indicator for TBI is the clinical severity of injury.5 However, radiological scoring systems for the prognostication of TBI have also been published, among which the Marshall Classification remains the best known. Rotterdam CT Score (RCTS) was proposed by Maas et al.3 in 2005 and since then, it has been used to predict outcomes after acute TBI and DC.4,14 Although both systems are simple to use and have been shown to have reasonable inter-observer reliability,15 in the context of TBI patients undergoing DC, the RCTS offers theoretical advantage over Marshall Classification because of greater discrimination of qualitative features of CT head, and an ability to quantitatively incorporate more than one radiological feature. The current study has explored the value of RCTS on a cohort of patients undergoing DC for isolated TBI. We attempted to validate RCTS not only as an independent predictor of mortality in TBI patients undergoing DC but also correlated the RCTS with unfavourable outcomes other than death. Moreover, we have also compared the RCTS with other systems used for prognostication such as GCS, RTS, etc., using the same cohort of patients. Our study clearly shows that the estimated marginal means of GOS at a median follow-up of 6 months is inversely related to increasing RCTS (p value50.001). The values are validated for various categories of RCTS even after statistical adjustment for mean age, GCS and incidental to arrival delay (p value50.001).
Figure 2. Graph showing distribution of patients into different Marshall Classification and RCTS.
4
M. WAQAS ET AL.
Table 2. Univariate and multivariate binary logistic regression analysis of factors associated with mortality. Mortality Univariate analysis Variable Age Revised trauma score Incident to arrival delay Rotterdam43
Multivariate analysis
Crude odds ratio
95% CI
p Value
Adjusted odds ratio
95% CI
p Value
1.033 0.682 1.000 3.636
1.014–1.051 0.538–0.865 0.998–1.001 1.707–7.746
0.001 0.002 0.656 0.001
1.043 0.622 0.999 2.792
1.021–1.065 0.473–0.820 0.997–1.002 1.235–6.311
50.001 0.001 0.614 0.014
Table 3. Univariate and multivariate binary logistic regression analysis of factors associated with an unfavourable outcome. Unfavourable outcome Univariate analysis
Downloaded by [Emory University] at 01:59 06 February 2016
Variable Age Revised trauma score Incident to arrival delay Rotterdam 43
Multivariate analysis
Crude odds ratio
95% CI
p Value
Adjusted odds ratio
95% CI
p Value
1.033 0.642 0.999 2.790
1.015–1.051 0.513–0.802 0.995–1.002 1.548–5.029
50.001 50.001 0.438 0.001
1.048 0.571 0.998 2.063
1.026–1.071 0.440–0.743 0.995–1.001 1.056–4.031
50.001 50.001 0.227 0.034
Figure 3. Receiver operator characteristic curve of RCTS for unfavourable outcomes and mortality.
This is not the first paper attempting to validate RCTS. Huang et al. explored this relationship earlier and showed RCTS as an independent predictor of poor outcomes even though the authors did not provide a reference value for RCTS when determining the odds of unfavourable outcome, a point made by Maas himself while commenting on the paper.4 The authors also did not take into account the various confounding factors affecting outcome in a polytrauma patient. It also did not explore Marshall Classification as a predictor of outcomes in DC patients. Current study proves by means of ROC analysis that Marshall Classification is not a good predictor of mortality and outcomes as indicated by AUC of 0.48 (95% CI: 0.385–0.591; p ¼ 0.83) for unfavourable outcomes while AUC of 0.452 (95% CI: 0.33–0.57; p ¼0.45).
We attempted to the best of our ability to take into account the confounders with highest impact such as age, RTS and GCS both in regression analysis and the analysis of estimated means of GOS for individual category of RCTS. Our study is based on the largest series of patients with isolated TBI undergoing DC and to the best of our knowledge, is also the most comprehensive. The confounding effect of systemic injuries was taken care of by inclusion of patients with isolated TBI only, even though other confounders could not be included in regression analysis because of limited sample size. Our study adds to available knowledge and is based on the data from a resource limited country with high burden of TBI. We have divided RCTS into two categories (1–3 and 3–6) for useful interpretation of the classification and
Downloaded by [Emory University] at 01:59 06 February 2016
BRITISH JOURNAL OF NEUROSURGERY
5
Figure 4. Receiver operator characteristic curve of Marshall Classification for unfavourable outcomes and mortality.
found statistically significant odds for mortality and unfavourable outcomes at 3-month follow-up. Other variables such as patients’ age, GCS and RTS were also found to be statistically significant predictors of mortality and unfavourable outcome in both univariate and multivariate regression analysis. These factors have been previously explored with similar results and our study further validates the findings of previous researchers.16 Other than the validity of RCTS for outcomes following TBI, our study also presents some other interesting observations. Our mortality rate (23.2%) and proportion of patients with unfavourable outcome (53.7%) are comparable with most of the published literature on severe TBI and DC.17–19 This is interesting as our indication for DC was not based on formal ICP monitoring, rather was based entirely on clinical and radiological evidence of medically intractable ICP. We would like to clarify here that in the presence of recent evidence and with a background of a resource restricted setting, it is no longer possible for us justify expensive ICP monitoring for our TBI patients.20 Our patient population is also somewhat different from previous publications on TBI as 18.3% of our patients presented with falls, this being the second common mechanism of injury in our series. Moreover, motorbike riders constituted a significant number of our patients (35%) which is not unusual if the data from other LMIC are compared.21 We have previously published on both of these disparities, explaining the cultural peculiarities of our society predisposing to certain mechanisms of injuries.21–23 In a recent survey, it was noted that less than 2% of motorbike riders in our city wear helmets therefore predisposing them to severe TBI.24 The study has several limitations. It has a retrospective design. We do not routinely measure ICP, which limits the generalisability of its results. The study therefore has greater relevance to developing countries where ICP monitoring is not performed routinely. Although there was no major difference in the technique of DC, the procedure was performed by several surgeons which could have a bearing on outcomes.
Conclusion RCTS is an independent predictor of mortality and unfavourable outcomes in patients undergoing DC for isolated TBI. However, its role in clinical setting appears to be limited.
Disclosure statement The authors report no conflict of interest.
References 1. Marshall LF, Marshall SB, Klauber MR, et al. A new classification of head injury based on computerized tomography. J Neurosurg 1991;75:S14–20. 2. Lobato RD, Gomez PA, Alday R, et al. Sequential computerized tomography changes and related final outcome in severe head injury patients. Acta Neurochir (Wien) 1997;139:385–91. 3. Maas AI, Hukkelhoven CW, Marshall LF, Steyerberg EW. Prediction of outcome in traumatic brain injury with computed tomographic characteristics: a comparison between the computed tomographic classification and combinations of computed tomographic predictors. Neurosurgery 2005;57:1173–82. 4. Huang Y-H, Deng Y-H, Lee T-C, Chen W-F. Rotterdam computed tomography score as a prognosticator in head-injured patients undergoing decompressive craniectomy. Neurosurgery 2012;71:80–5. 5. Perel P, Edwards P, Wentz R, Roberts I. Systematic review of prognostic models in traumatic brain injury. BMC Med Inform Decis Mak 2006;6:38. 6. Sabin C, Petrie A. Medical statistics at a glance. Malden, MA: Blackwell Science; 2000. 7. Hyder AA, Wunderlich CA, Puvanachandra P, et al. The impact of traumatic brain injuries: a global perspective. NeuroRehabilitation 2007;22:341–53. 8. McGarry LJ, Thompson D, Millham FH, et al. Outcomes and costs of acute treatment of traumatic brain injury. J Trauma Acute Care Surg 2002;53:1152–9. 9. Balestreri M, Czosnyka M, Hutchinson P, et al. Impact of intracranial pressure and cerebral perfusion pressure on severe disability and mortality after head injury. Neurocrit Care 2006;4:8–13. 10. Bloomfield GL, Ridings PC, Blocher CR, et al. Effects of increased intraabdominal pressure upon intracranial and cerebral perfusion pressure
6
11. 12. 13. 14. 15. 16.
Downloaded by [Emory University] at 01:59 06 February 2016
17.
M. WAQAS ET AL.
before and after volume expansion. J Trauma 1996;40:936–41. discussion 41–3. Saul TG, Ducker TB. Intracranial pressure monitoring in patients with severe head injury. Am Surg 1982;48:477–80. Cooper DJ, Rosenfeld JV, Murray L, et al. Early decompressive craniectomy for patients with severe traumatic brain injury and refractory intracranial hypertension – a pilot randomized trial. J Crit Care 2008;23:387–93. Timofeev I, Santarius T, Kolias AG, Hutchinson PJ. Decompressive craniectomy – operative technique and perioperative care. Adv Tech Stand Neurosurg 2012;38:115–36. Yuh EL, Cooper SR, Ferguson AR, Manley GT. Quantitative CT improves outcome prediction in acute traumatic brain injury. J Neurotrauma 2012;29:735–46. Chun KA, Manley GT, Stiver SI, et al. Interobserver variability in the assessment of CT imaging features of traumatic brain injury. J Neurotrauma 2010;27:325–30. Collaborators MCT, Perel P, Arango M, et al. Predicting outcome after traumatic brain injury: practical prognostic models based on large cohort of international patients. BMJ 2008;336:425–9. Howard JL, Cipolle MD, Anderson M, et al. Outcome after decompressive craniectomy for the treatment of severe traumatic brain injury. J Trauma Acute Care Surg 2008;65:380–6.
18. Aarabi B, Hesdorffer DC, Ahn ES, et al. Outcome following decompressive craniectomy for malignant swelling due to severe head injury. J Neurosurg 2006;104:469–79. 19. Jiang J-Y, Xu W, Li W-P, et al. Efficacy of standard trauma craniectomy for refractory intracranial hypertension with severe traumatic brain injury: a multicenter, prospective, randomized controlled study. J Neurotrauma 2005;22:623–8. 20. Chesnut RM, Temkin N, Carney N, et al. A trial of intracranial-pressure monitoring in traumatic brain injury. N Engl J Med 2012;367:2471–81. 21. Razzak JA, Shamim MS, Mehmood A, et al. A successful model of road traffic injury surveillance in a developing country: process and lessons learnt. BMC Public Health 2012;12:357. 22. Shamim MS, Khan UR, Razzak JA, Rasheed J. Injuries due to fall make summer time power outages: a potential public health issue. J Emerg Trauma Shock 2011;4:147–8. 23. Shamim MS, Ali SF, Enam SA. Non-operative management is superior to surgical stabilization in spine injury patients with complete neurological deficits: a perspective study from a developing world country, Pakistan. Surg Neurol Int 2011;2:166.DOI: 10.4103/2152-7806.90027. 24. Shamim S, Razzak JA, Jooma R, Khan U. Initial results of Pakistan’s first road traffic injury surveillance project. Int J Inj Contr Saf Promot 2011;18:213–17.
ISSN: 0268-8697 (Print) 1360-046X (Online) Journal homepage: http://www.tandfonline.com/loi/ibjn20
Predicting outcomes of decompressive craniectomy: use of Rotterdam Computed Tomography Classification and Marshall Classification Muhammad Waqas, Muhammad Shahzad Shamim, Syed Faaiz Enam, Mohsin Qadeer, Saqib Kamran Bakhshi, Iqra Patoli & Khabir Ahmad To cite this article: Muhammad Waqas, Muhammad Shahzad Shamim, Syed Faaiz Enam, Mohsin Qadeer, Saqib Kamran Bakhshi, Iqra Patoli & Khabir Ahmad (2016): Predicting outcomes of decompressive craniectomy: use of Rotterdam Computed Tomography Classification and Marshall Classification, British Journal of Neurosurgery, DOI: 10.3109/02688697.2016.1139047 To link to this article: http://dx.doi.org/10.3109/02688697.2016.1139047
Published online: 01 Feb 2016.
Submit your article to this journal
Article views: 5
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=ibjn20 Download by: [Emory University]
Date: 06 February 2016, At: 01:59
BRITISH JOURNAL OF NEUROSURGERY, 2016 http://dx.doi.org/10.3109/02688697.2016.1139047
ORIGINAL ARTICLE
Predicting outcomes of decompressive craniectomy: use of Rotterdam Computed Tomography Classification and Marshall Classification Muhammad Waqas, Muhammad Shahzad Shamim, Syed Faaiz Enam, Mohsin Qadeer, Saqib Kamran Bakhshi, Iqra Patoli and Khabir Ahmad
Downloaded by [Emory University] at 01:59 06 February 2016
Department of Surgery, Section of Neurosurgery, Aga Khan University Hospital, Karachi, Pakistan
ABSTRACT
ARTICLE HISTORY
Background Data on the evaluation of the Rotterdam Computed Tomography Classification (RCTS) as a predictor of outcomes in patients undergoing decompressive craniectomy (DC) for trauma is limited and lacks clarity. Objective To explore the role of RCTS in predicting unfavourable outcomes, including mortality in patients undergoing DC for head trauma. Methods This was an observational cohort study conducted from 1 January 2009 to 31 March 2013. CT scans of adults with head trauma prior to emergency DC were scored according to RCTS. A receiver operating characteristic curve analysis was performed to identify the optimal cut-off RCTS for predicting unfavourable outcomes [Glasgow outcome scale (GOS) ¼ 1–3]. Binary logistic regression analysis was performed to evaluate the relationship between RCTS and unfavourable outcomes including mortality. Results One hundred ninety-seven patients (mean age: 31.4 ± 18.7 years) were included in the study. Mean Glasgow coma score at presentation was 8.1 ± 3.6. RCTS was negatively correlated with GOS (r ¼ 0.370; p50.001). The area under the curve was 0.687 (95% CI: 0.595–0.779; p50.001) and 0.666 (95% CI: 0.589–0.742; p50.001) for mortality and unfavourable outcomes, respectively. RCTS independently predicted both mortality (adjusted odds ratio for RCTS43 compared with RCTS 3: 2.792, 95% CI: 1.235–6.311) and other unfavourable outcomes (adjusted odds ratio for RCTS43 compared with RCTS 3: 2.063, 95% CI: 1.056–4.031). Conclusion RCTS is an independent predictor of unfavourable outcomes and mortality among patients undergoing emergency DC.
Received 26 September 2014 Revised 27 October 2015 Accepted 3 January 2016 Published online 28 January 2016
Introduction Among the radiological scoring systems for traumatic brain injury (TBI), Marshall CT Scan Classification system1 has been widely used for categorisation of severity and prediction of prognosis.2,3 The scoring system has been shown to be an independent predictor of unfavourable outcomes; however, the system is limited by its grouping of majority of the patient requiring DC with a mass lesion into class IV or V.4 Moreover, the Marshall system is a classification system rather than a scoring system, and therefore allows little room to accommodate patients with more complex radiological features.1,3 More recently the Rotterdam Computed Tomography Classification (RCTS),3 which has numerical values of up to 6 has been assessed as a potential tool for predicting outcomes in these patients.4 The published work showed RCTS to have a significant correlation with outcomes although the analysis lacked a specific RCTS cut-off number for prediction of unfavourable outcome or death.4 To explore the role of RCTS in predicting unfavourable outcomes, including mortality in patients with head trauma undergoing DC at a tertiary care setup of a low- and middle-income countries (LMIC).
Materials and methods This was a retrospective observational cohort study conducted at the Section of Neurosurgery, Aga Khan University Hospital Karachi, a CONTACT Muhammad Shahzad Shamim P.O. Box 74800, Karachi, Pakistan ß 2016 Taylor & Francis
[email protected]
KEYWORDS
Decompressive craniectomy; Glasgow outcome scale; Rotterdam Classification; traumatic brain injury; unfavourable outcomes
tertiary care referral hospital with level 1 trauma facilities. The study period was from 1 January 2009 to 31 March 2013. Inclusion criteria All patients aged 16 years, presenting with isolated TBI who required unilateral or bilateral decompressive craniectomy (DC) during the course of hospital stay were included in the study. Exclusion criteria: patients with incomplete medical records, including CT scans were excluded from the study. Data collection procedure Records of all the patients who conformed to the inclusion criteria were retrieved from the hospital database and the operating room (OR) database, that keeps the record of all procedures performed in the OR. Data were collected on a structured pro-forma. Emergency room information on demographics, post-resuscitation Glasgow Coma Score (GCS), revised trauma score (RTS), pupillary response, etc., was gathered from initial trauma assessment forms. Information regarding the surgical intervention and stay in ICU and high dependency unit was obtained from the review of OR and postoperative records available in charts. All individual CT images were reviewed through our picture archiving and communication system (PACS) and each CT scan was classified according to the RCTS by
Department of Surgery, Section of Neurosurgery, Aga Khan University Hospital, Stadium Road,
2
M. WAQAS ET AL.
two of the authors independently. Conflicts in interpretation were resolved by consensus. The authors were blind to the outcomes. Definitions and scales
Downloaded by [Emory University] at 01:59 06 February 2016
RCTS as described originally by Maas et al.3 was used for this purpose (Figure 1). Briefly, it is a standardised CT-based classification system, a number of up to 6 and takes into account status of basal cisterns, midline shift, subarachnoid hemorrhage and extradural hematoma. Patients’ outcomes were determined using the Glasgow outcome scale (GOS) as follows: score of 1: dead; 2: vegetative state; 3: severely disabled (conscious but the patient requires others for daily support due to disability); 4: moderately disabled (the patient is independent but disabled); and 5: good recovery (the patient has resumed most normal activities but may have minor residual problems). Outcomes were dichotomised to indicate favourable or unfavourable as frequently used in literature.5 GOS of 4 or more was considered to be favourable.
Indications and technique of surgery As an institutional protocol, all patients with severe TBI (GCS 8 or less) are managed with immediate intubation and admission in ICU. Intracranial pressure (ICP) monitoring is not routinely employed. Patients with mass lesions such as extradural or subdural hematoma, fulfilling Brain Trauma Foundation guidelines, are taken to the OR for emergency decompressive surgery and in case of diffuse cerebral oedema despite evacuation of mass lesion, the bone flap is not replaced. All the decompressive surgeries are performed by senior residents under direct supervision of one of the six Consultant Neurosurgeons. In a standard DC, the bone flap of at least 12 cm diameter is removed along with expansile duraplasty.
Follow-up Outcomes of the patients were recorded from medical records up till last clinic visit. Information was obtained according to the GOS as described above. The observer of outcomes was masked of the RCTS of individual patients. Analysis Data were analysed using SPSS IBM version 19 (Armonk, NY). Proportions were calculated to describe categorical data. Mean and standard deviation were used for continuous data and scoring variables such as GCS, RTS and GOS were analysed as numerical data after reviewing previous literature and books on statistics.6 A receiver operating characteristic curve analysis was performed to identify the optimal cuto-ff RCT score for predicting such outcomes. Patients were then categorised into two groups based on the basis of ROC cut-off value. Binary logistic regression analysis was performed to evaluate the relationship between RCTS and adverse outcomes including mortality while controlling for the effect of age, RTS and incident to arrival delay in emergency.
Results A total of 235 patients underwent DC in the time period. Thirty-eight patients were excluded from the study based on exclusion criteria. Medical records of 197 patients were thus reviewed. Of these, 174 (88.3%) were male and 23 (11.7%) female. By the end of the study, only 7 patients were lost to follow-up. Demographics, baseline population characteristics and outcomes are shown in Table 1. The mean age of the patients was 31.4 ± 18.7 years. Road traffic accidents were the most common cause of head injury with 129 (65.5%) followed by fall from a height with 36 (18.3%). Mean GCS on arrival of our study subjects was 8.1 ± 3.6. Median duration of follow-up was 6.0 months (range: 3.0–48). The mean RTS was 10.0 ± 1.5. The average RCTS was 3.6 ± 1.2. Figure 2 shows the distribution of patients with different Marshall Classification and RCTS. Binary logistic regression analysis showed a significant independent association between RCTS and mortality (adjusted odds ratio for RCTS 43 compared with RCTS 3: 2.792, 95% CI: 1.235– 6.311) while controlling for the effect of age, RTS and incident to arrival delay (Table 2). A significant independent association was also observed between RCTS and unfavourable outcomes (adjusted odds ratio for RCTS 43 compared with RCTS 3: 2.063, 95% CI: 1.056–4.031) while controlling for the effect of age, RTS and incident to arrival delay (Table 3). The AUC was 0.666 (95% CI: 0.589–0.742; p50.001) for unfavourable outcome and 0.687 (95% CI: 0.595–0.779; p50.001) for mortality. Maximal sensitivity and specificity was obtained with a RCTS of 4 (75% and 55%, respectively) on ROC curve for mortality (Figure 3). ROC analysis for Marshall Classification showed an AUC of 0.48 (95% CI: 0.385–0.591; p ¼ 0.83) for unfavourable outcomes while AUC of 0.452 (95% CI: 0.33–0.57; p ¼ 0.45; Figure 4).
Discussion
Figure 1. Rotterdam Computed Tomography Classification. Adapted from Maas et al.3
TBI will surpass most other diseases as the major cause of death and disability7 by the year 2020. With an estimated 10 million people affected annually, the burden of TBI is especially prominent in LMIC. The cost of managing severe TBI is substantial8 especially when its economic predicament for LMIC is considered. Persistently high ICP after TBI and its correlation with increased mortality has
BRITISH JOURNAL OF NEUROSURGERY
Table 1. Population characteristics and baseline variables.
Downloaded by [Emory University] at 01:59 06 February 2016
Variable Mean age (years) ± SD Sex, no. (%) Male Female Mechanism of head injury, no. (%) Road traffic accident Motorbike Car Pillion Riding Driver Pedestrian Fall Gunshot Others Mean GCS ± SD Mean revised trauma score ± SD Pupil reactivity, no. (%) Equal, reactive Anisocoria Fixed and dilated Pinpoint Not documented Rotterdam CT score, no. (%) 1 2 3 4 5 6 Score 3 Score 4 Mean Rotterdam CT score, mean ± SD Marshall Classification, mean ± SD Median hospital stay, days (IQR) GOS at the end of follow-up, no. (%) (n ¼ 190) Unfavourable outcome Favourable outcome Mortality (n ¼ 190)
Statistic 31.4 ± 18.7 174 (88.3) 23 (11.7) 129 (65.5) 69 (35.0) 15 (7.6) 3 (1.5) 8 (4.1) 34 (17.3) 36 (18.3) 18 (9.1) 13 (6.6) 8.1 ± 3.6 10.0 ± 1.5 105 (53.3) 75 (38.1) 5 (2.5) 1 (0.5) 11 (5.6) 7 (3.6) 28 (14.2) 61 (31.0) 52 (26.4) 37 (18.8) 12 (6.1) 96 (48.8) 101 (51.3) 3.6 ± 1.2 4.6 ± 1.4 14.0 (7.0–23.0) 102 (53.7) 88 (46.3) 44 (23.2)
3
been demonstrated by multiple studies.9–11 DC is emerging as an effective surgical measure for reducing ICP and is now considered a management option for medically refractive ICP.12,13 However, the outcomes of DC tend to vary in different reports and depend greatly on patient selection.12,13 Determining the prognostic indicators in patients with severe TBI undergoing DC can therefore be of great potential value when counseling patients’ families with regards to outcome, clinical decision making and also to help prioritise resources in resource constrained settings such as LMIC. It is generally accepted that the most sensitive prognostic indicator for TBI is the clinical severity of injury.5 However, radiological scoring systems for the prognostication of TBI have also been published, among which the Marshall Classification remains the best known. Rotterdam CT Score (RCTS) was proposed by Maas et al.3 in 2005 and since then, it has been used to predict outcomes after acute TBI and DC.4,14 Although both systems are simple to use and have been shown to have reasonable inter-observer reliability,15 in the context of TBI patients undergoing DC, the RCTS offers theoretical advantage over Marshall Classification because of greater discrimination of qualitative features of CT head, and an ability to quantitatively incorporate more than one radiological feature. The current study has explored the value of RCTS on a cohort of patients undergoing DC for isolated TBI. We attempted to validate RCTS not only as an independent predictor of mortality in TBI patients undergoing DC but also correlated the RCTS with unfavourable outcomes other than death. Moreover, we have also compared the RCTS with other systems used for prognostication such as GCS, RTS, etc., using the same cohort of patients. Our study clearly shows that the estimated marginal means of GOS at a median follow-up of 6 months is inversely related to increasing RCTS (p value50.001). The values are validated for various categories of RCTS even after statistical adjustment for mean age, GCS and incidental to arrival delay (p value50.001).
Figure 2. Graph showing distribution of patients into different Marshall Classification and RCTS.
4
M. WAQAS ET AL.
Table 2. Univariate and multivariate binary logistic regression analysis of factors associated with mortality. Mortality Univariate analysis Variable Age Revised trauma score Incident to arrival delay Rotterdam43
Multivariate analysis
Crude odds ratio
95% CI
p Value
Adjusted odds ratio
95% CI
p Value
1.033 0.682 1.000 3.636
1.014–1.051 0.538–0.865 0.998–1.001 1.707–7.746
0.001 0.002 0.656 0.001
1.043 0.622 0.999 2.792
1.021–1.065 0.473–0.820 0.997–1.002 1.235–6.311
50.001 0.001 0.614 0.014
Table 3. Univariate and multivariate binary logistic regression analysis of factors associated with an unfavourable outcome. Unfavourable outcome Univariate analysis
Downloaded by [Emory University] at 01:59 06 February 2016
Variable Age Revised trauma score Incident to arrival delay Rotterdam 43
Multivariate analysis
Crude odds ratio
95% CI
p Value
Adjusted odds ratio
95% CI
p Value
1.033 0.642 0.999 2.790
1.015–1.051 0.513–0.802 0.995–1.002 1.548–5.029
50.001 50.001 0.438 0.001
1.048 0.571 0.998 2.063
1.026–1.071 0.440–0.743 0.995–1.001 1.056–4.031
50.001 50.001 0.227 0.034
Figure 3. Receiver operator characteristic curve of RCTS for unfavourable outcomes and mortality.
This is not the first paper attempting to validate RCTS. Huang et al. explored this relationship earlier and showed RCTS as an independent predictor of poor outcomes even though the authors did not provide a reference value for RCTS when determining the odds of unfavourable outcome, a point made by Maas himself while commenting on the paper.4 The authors also did not take into account the various confounding factors affecting outcome in a polytrauma patient. It also did not explore Marshall Classification as a predictor of outcomes in DC patients. Current study proves by means of ROC analysis that Marshall Classification is not a good predictor of mortality and outcomes as indicated by AUC of 0.48 (95% CI: 0.385–0.591; p ¼ 0.83) for unfavourable outcomes while AUC of 0.452 (95% CI: 0.33–0.57; p ¼0.45).
We attempted to the best of our ability to take into account the confounders with highest impact such as age, RTS and GCS both in regression analysis and the analysis of estimated means of GOS for individual category of RCTS. Our study is based on the largest series of patients with isolated TBI undergoing DC and to the best of our knowledge, is also the most comprehensive. The confounding effect of systemic injuries was taken care of by inclusion of patients with isolated TBI only, even though other confounders could not be included in regression analysis because of limited sample size. Our study adds to available knowledge and is based on the data from a resource limited country with high burden of TBI. We have divided RCTS into two categories (1–3 and 3–6) for useful interpretation of the classification and
Downloaded by [Emory University] at 01:59 06 February 2016
BRITISH JOURNAL OF NEUROSURGERY
5
Figure 4. Receiver operator characteristic curve of Marshall Classification for unfavourable outcomes and mortality.
found statistically significant odds for mortality and unfavourable outcomes at 3-month follow-up. Other variables such as patients’ age, GCS and RTS were also found to be statistically significant predictors of mortality and unfavourable outcome in both univariate and multivariate regression analysis. These factors have been previously explored with similar results and our study further validates the findings of previous researchers.16 Other than the validity of RCTS for outcomes following TBI, our study also presents some other interesting observations. Our mortality rate (23.2%) and proportion of patients with unfavourable outcome (53.7%) are comparable with most of the published literature on severe TBI and DC.17–19 This is interesting as our indication for DC was not based on formal ICP monitoring, rather was based entirely on clinical and radiological evidence of medically intractable ICP. We would like to clarify here that in the presence of recent evidence and with a background of a resource restricted setting, it is no longer possible for us justify expensive ICP monitoring for our TBI patients.20 Our patient population is also somewhat different from previous publications on TBI as 18.3% of our patients presented with falls, this being the second common mechanism of injury in our series. Moreover, motorbike riders constituted a significant number of our patients (35%) which is not unusual if the data from other LMIC are compared.21 We have previously published on both of these disparities, explaining the cultural peculiarities of our society predisposing to certain mechanisms of injuries.21–23 In a recent survey, it was noted that less than 2% of motorbike riders in our city wear helmets therefore predisposing them to severe TBI.24 The study has several limitations. It has a retrospective design. We do not routinely measure ICP, which limits the generalisability of its results. The study therefore has greater relevance to developing countries where ICP monitoring is not performed routinely. Although there was no major difference in the technique of DC, the procedure was performed by several surgeons which could have a bearing on outcomes.
Conclusion RCTS is an independent predictor of mortality and unfavourable outcomes in patients undergoing DC for isolated TBI. However, its role in clinical setting appears to be limited.
Disclosure statement The authors report no conflict of interest.
References 1. Marshall LF, Marshall SB, Klauber MR, et al. A new classification of head injury based on computerized tomography. J Neurosurg 1991;75:S14–20. 2. Lobato RD, Gomez PA, Alday R, et al. Sequential computerized tomography changes and related final outcome in severe head injury patients. Acta Neurochir (Wien) 1997;139:385–91. 3. Maas AI, Hukkelhoven CW, Marshall LF, Steyerberg EW. Prediction of outcome in traumatic brain injury with computed tomographic characteristics: a comparison between the computed tomographic classification and combinations of computed tomographic predictors. Neurosurgery 2005;57:1173–82. 4. Huang Y-H, Deng Y-H, Lee T-C, Chen W-F. Rotterdam computed tomography score as a prognosticator in head-injured patients undergoing decompressive craniectomy. Neurosurgery 2012;71:80–5. 5. Perel P, Edwards P, Wentz R, Roberts I. Systematic review of prognostic models in traumatic brain injury. BMC Med Inform Decis Mak 2006;6:38. 6. Sabin C, Petrie A. Medical statistics at a glance. Malden, MA: Blackwell Science; 2000. 7. Hyder AA, Wunderlich CA, Puvanachandra P, et al. The impact of traumatic brain injuries: a global perspective. NeuroRehabilitation 2007;22:341–53. 8. McGarry LJ, Thompson D, Millham FH, et al. Outcomes and costs of acute treatment of traumatic brain injury. J Trauma Acute Care Surg 2002;53:1152–9. 9. Balestreri M, Czosnyka M, Hutchinson P, et al. Impact of intracranial pressure and cerebral perfusion pressure on severe disability and mortality after head injury. Neurocrit Care 2006;4:8–13. 10. Bloomfield GL, Ridings PC, Blocher CR, et al. Effects of increased intraabdominal pressure upon intracranial and cerebral perfusion pressure
6
11. 12. 13. 14. 15. 16.
Downloaded by [Emory University] at 01:59 06 February 2016
17.
M. WAQAS ET AL.
before and after volume expansion. J Trauma 1996;40:936–41. discussion 41–3. Saul TG, Ducker TB. Intracranial pressure monitoring in patients with severe head injury. Am Surg 1982;48:477–80. Cooper DJ, Rosenfeld JV, Murray L, et al. Early decompressive craniectomy for patients with severe traumatic brain injury and refractory intracranial hypertension – a pilot randomized trial. J Crit Care 2008;23:387–93. Timofeev I, Santarius T, Kolias AG, Hutchinson PJ. Decompressive craniectomy – operative technique and perioperative care. Adv Tech Stand Neurosurg 2012;38:115–36. Yuh EL, Cooper SR, Ferguson AR, Manley GT. Quantitative CT improves outcome prediction in acute traumatic brain injury. J Neurotrauma 2012;29:735–46. Chun KA, Manley GT, Stiver SI, et al. Interobserver variability in the assessment of CT imaging features of traumatic brain injury. J Neurotrauma 2010;27:325–30. Collaborators MCT, Perel P, Arango M, et al. Predicting outcome after traumatic brain injury: practical prognostic models based on large cohort of international patients. BMJ 2008;336:425–9. Howard JL, Cipolle MD, Anderson M, et al. Outcome after decompressive craniectomy for the treatment of severe traumatic brain injury. J Trauma Acute Care Surg 2008;65:380–6.
18. Aarabi B, Hesdorffer DC, Ahn ES, et al. Outcome following decompressive craniectomy for malignant swelling due to severe head injury. J Neurosurg 2006;104:469–79. 19. Jiang J-Y, Xu W, Li W-P, et al. Efficacy of standard trauma craniectomy for refractory intracranial hypertension with severe traumatic brain injury: a multicenter, prospective, randomized controlled study. J Neurotrauma 2005;22:623–8. 20. Chesnut RM, Temkin N, Carney N, et al. A trial of intracranial-pressure monitoring in traumatic brain injury. N Engl J Med 2012;367:2471–81. 21. Razzak JA, Shamim MS, Mehmood A, et al. A successful model of road traffic injury surveillance in a developing country: process and lessons learnt. BMC Public Health 2012;12:357. 22. Shamim MS, Khan UR, Razzak JA, Rasheed J. Injuries due to fall make summer time power outages: a potential public health issue. J Emerg Trauma Shock 2011;4:147–8. 23. Shamim MS, Ali SF, Enam SA. Non-operative management is superior to surgical stabilization in spine injury patients with complete neurological deficits: a perspective study from a developing world country, Pakistan. Surg Neurol Int 2011;2:166.DOI: 10.4103/2152-7806.90027. 24. Shamim S, Razzak JA, Jooma R, Khan U. Initial results of Pakistan’s first road traffic injury surveillance project. Int J Inj Contr Saf Promot 2011;18:213–17.
