- Email: [email protected]
Predictors of Failure of Nonoperative Management Following Subaxial Spine Trauma and Creation of Modified Subaxial Injury Classification System
Original Article
Predictors of Failure of Nonoperative Management Following Subaxial Spine Trauma and Creation of Modified Subaxial Injury Classification System Frederick L. Hitti1, Brendan J. McShane1, Andrew I. Yang1, Cole Rinehart1, Ahmed Albayar1, Marc Branche1, Neil R. Malhotra1, M. Burhan Janjua1,2, Zarina S. Ali1, James M. Schuster1, Ali K. Ozturk1
BACKGROUND: Subaxial cervical spine injuries may be treated with either nonoperative stabilization or surgical fixation. The subaxial injury classification (SLIC) provides 1 method for suggesting the degree of necessity for surgery. In the current study, we examined if the SLIC score, or other preoperative metrics, can predict failure of nonoperative management.
-
METHODS: We performed a retrospective chart review to identify patients who presented with acute, nonpenetrating, subaxial cervical spine injury within our health system between 2007 and 2016. Patient demographics, medical comorbidities, injuries, and treatments were collected. Logistic regression analysis was used to determine potential predictors of failure of nonoperative management.
-
RESULTS: During the study period, 40 patients met the inclusion criteria. A small subset of patients failed nonoperative management (n [ 5, 12.5%). The mean SLIC score was 3.9 1.9; however, 14 (35%) patients had scores >4. Neither total SLIC score (P [ 0.68) nor SLIC subscores (morphology [P [ 0.96], discoligamentous complex [P [ 0.83], neurologic status [P [ 0.60]) predicted failure of nonoperative treatment. Time to evaluation/treatment did predict failure of nonoperative management. Evaluation within 8 hours of injury was a negative predictor of failure (odds ratio [ 0.03, P [ 0.001) and evaluation 24 hours or more after injury was a positive predictor of failure (odds ratio [ 66.00, P < 0.001). We created a modified SLIC score
-
Key words mSLIC - Nonoperative failure - Predictors - Subaxial cervical spine - Trauma SLIC -
Abbreviations and Acronyms CI: Confidence interval DLC: Diskoligamentous complex mSLIC: Modified subaxial injury classification OR: Odds ratio
on the basis of these findings, which significantly predicted failure of nonoperative management (P [ 0.044). CONCLUSIONS: Management of subaxial spine injuries is complex. In our cohort, SLIC scoring did not adequately predict odds of failure of nonoperative management. Time to evaluation, however, did. We created a modified SLIC score that significantly predicted failure of nonoperative management.
-
INTRODUCTION
A
lmost half of all spinal cord injuries (SCIs) in North America are a result of cervical spine trauma, and the incidence of these injuries is increasing.1,2 Importantly, these spinal cord injuries can have profound neurologic ramifications.3 Subaxial cervical spine trauma includes dislocations, ligamentous disruption, and/or fractures from C3C7. Traumatic injuries of the cervical spine can lead to neurologic impairment and/or instability.4,5 It is critical to quickly diagnose and treat these injuries to prevent or ameliorate neurologic decline. Subaxial spine trauma may be treated with observation, nonoperative stabilization, or operative intervention with internal fixation and/or decompression. External devices such as cervical collars or a cervical halo may be used to achieve stabilization in patients managed nonoperatively. Patients are typically managed
ROC: Receiver-operating characteristic SLIC: Subaxial injury classification From the Departments of 1Neurosurgery and 2Orthopaedic Surgery, Pennsylvania Hospital, University of Pennsylvania, Philadelphia, Pennsylvania, USA To whom correspondence should be addressed: Frederick L. Hitti, M.D., Ph.D. [E-mail: [email protected]] Citation: World Neurosurg. (2018). https://doi.org/10.1016/j.wneu.2018.11.048 Journal homepage: www.WORLDNEUROSURGERY.org Available online: www.sciencedirect.com 1878-8750/$ - see front matter ª 2018 Elsevier Inc. All rights reserved.
WORLD NEUROSURGERY -: e1-e6, - 2018
www.WORLDNEUROSURGERY.org
e1
ORIGINAL ARTICLE FREDERICK L. HITTI ET AL.
PREDICTORS OF FAILURE OF NONOPERATIVE MANAGEMENT
nonoperatively if the cervical spine is not grossly unstable or if they are not surgical candidates. Scoring systems have been developed to aid in diagnosis and treatment of cervical spine injury. The AOSpine subaxial injury classification system4 may be used to aid in morphologic classification of the injury, and the decision to pursue operative versus nonoperative treatment may be aided by the use of the subaxial injury classification (SLIC) system.1,6 The SLIC score uses morphology, integrity of the diskoligamentous complex (DLC), and neurologic status to guide treatment decisions.5 Some have developed treatment guidelines on the basis of a combination of SLIC scores and expert consensus.7 One study followed patients prospectively and found that none of those in their cohort exhibited neurologic deterioration if SLIC criteria were used to guide management.8 The authors argued in favor of SLIC-based management. Although the aforementioned scoring systems and treatment guidelines are available, there are limited studies on the utility of these systems to predict failure of nonoperative management. Here, we examined a cohort of patients who presented with subaxial cervical spine trauma and were treated nonoperatively to assess factors that could predict failure of nonoperative management. METHODS Study Population In this Institutional Review Boardapproved study, we enrolled consecutive patients with acute nonpenetrating subaxial cervical spine injuries referred for neurosurgical consultation from June 1, 2007eJune 1, 2016. This patient population included patients with fractures or dislocations of C3-C7. Ten surgeons at our institution participated in this study. Their clinical decision making was not bound by SLIC scores.
patients, 40 (37%) were initially managed nonoperatively with cervical spine collars. The demographics of this cohort of patients are shown in Table 1. The mean age was 50.5 21.5, and the majority of patients were male (n ¼ 27, 67.5%). With regards to race, the highest percentage of patients were Caucasian (n ¼ 19, 47.5%). A large percentage of the patients were African American (n ¼ 17, 42.5%). We tracked patient comorbidities and found osteoporosis, diabetes, coronary artery disease, peripheral arterial disease, and cancer in 10.0%, 7.5%, 7.5%, 5.0%, and 2.5% of patients, respectively. Regarding morphology of injury, 25 (62.5%) patients presented with compression fractures, 12 (30.0%) presented with distraction injuries, and 3 (7.5%) presented with rotation/translation injuries (see Table 1). Five patients (12.5%) failed nonoperative management and required surgical stabilization. Subluxation/anterolisthesis on repeat cervical spine radiograph (4 patients) or dynamic instability on flexion/extension cervical spine radiograph (1 patient) were deemed indications for surgical fixation due to failure of nonoperative management. The effect of age, sex, and race on failure of nonoperative management was investigated using logistic regression. None of these demographic factors was significant in predicting treatment failure (age: P ¼ 0.92; sex: P ¼ 0.71; race: P ¼ 0.07). The effect of patient comorbidities on failure of nonoperative stabilization could not be reliably analyzed due to cohort sample size. Subaxial Injury Classification Scores as Predictor of Failure of Nonoperative Management In our cohort of patients initially managed nonoperatively, the mean SLIC score was 3.9 1.9 (n ¼ 40, see Table 1). The mean SLIC score of patients initially managed operatively was 6.67 1.65 (n ¼ 69). The treating physicians in this study were not
CLINICAL DATA COLLECTION Once the target patient population was identified, we collected relevant data from medical records—both paper and electronic (Epic, Epic Systems Corporation, Madison, Wisconsin, USA). Demographic information was collected, and medical comorbidities at the time of initial injury were noted. We also recorded the type of injury, mechanism of injury, time to neurosurgical evaluation from initial injury, and length of hospitalization.
Table 1. Demographics Age
50.5 21.5 years old
Sex
Males, 27 (67.5%) Females, 13 (32.5%)
Race
Caucasians, 19 (47.5%) African-Americans, 17 (42.5%) Asians, 3 (7.5%) Other, 1 (2.5%)
Comorbidities
Osteoporosis, 4 (10.0%) Diabetes, 3 (7.5%) Coronary artery disease, 3 (7.5%) Peripheral arterial disease, 2 (5.0%) Cancer, 1 (2.5%)
Morphology of injury
Compression fracture, 25 (62.5%) Distraction, 12 (30.0%) Rotation/Translation 3 (7.5%)
STATISTICAL ANALYSIS Statistical analysis was performed using Microsoft Excel and R. Logistic regression analysis was performed using the generalized linear model in R. Type II analysis of variance was then used to compute P values. Only univariate logistic regression was performed due to the sample size of the cohort. Results were considered significant if P < 0.05. Receiver-operating characteristic (ROC) analysis and curves were generated using Prism (GraphPad software). Averages are presented as mean standard deviation unless otherwise indicated. RESULTS Study Population We first identified 109 patients who presented to our institution following acute nonpenetrating subaxial spine trauma. Of these
e2
www.SCIENCEDIRECT.com
SLIC score
3.9 1.9
SLIC > 4
14 (35%)
Patients failing nonoperative management
5 (12.5%)
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2018.11.048
ORIGINAL ARTICLE FREDERICK L. HITTI ET AL.
PREDICTORS OF FAILURE OF NONOPERATIVE MANAGEMENT
bound by SLIC treatment guidelines. Fourteen patients in the nonoperative group (35%) had a SLIC score >4 (i.e., SLIC scores high enough to recommend operative intervention). We examined whether or not the SLIC score was predictive of failure of nonoperative management by logistic regression (Table 2). Total SLIC score was not predictive of treatment failure (odds ratio [OR] 1.11, 95% confidence interval [CI] 0.65e 1.79, P ¼ 0.68). Furthermore, univariate logistic regression of the SLIC subscores demonstrated that none were predictive of failure of nonoperative management: morphology (OR 0.98, 95% CI 0.36e2.24, P ¼ 0.96), DLC disruption (OR 1.19, 95% CI, 0.23e6.49, P ¼ 0.83), and neurologic status (OR 1.20, 95% CI 0.58e2.32, P ¼ 0.60). EFFECT OF TIME TO STABILIZATION ON ODDS OF FAILING NONOPERATIVE MANAGEMENT We hypothesized that time to cervical spine stabilization would have an impact on failure of nonoperative management. To examine this, we noted the time to cervical spine stabilization from initial injury. We also analyzed length of hospital stay because medically complex patients tend to have longer hospitalizations. The average length of the acute hospitalization was 7.8 7.1 days in this cohort of patients. We performed univariate logistic regression followed by type II analysis of variance to examine the effect of rapid time to nonoperative stabilization and length of hospitalization on odds of requiring surgical stabilization after initial nonoperative management. We found that patients who were stabilized <8 hours after initial injury had a decreased incidence of future surgery with an OR of 0.03 (0.001e0.28, 95% CI), P ¼ 0.001 (Table 3). The length of hospitalization was not associated with failure of nonoperative management. We also hypothesized that a delay in nonoperative stabilization might increase the odds of failure of nonoperative management. Using logistic regression, we found that stabilization 24 hours after the initial injury resulted in an increased odds of failure of nonoperative management with an OR of 66.00 (6.48e1729.44, 95% CI), P < 0.001 (see Table 3). Reasons for delayed stabilization included failure to diagnose a cervical spine injury at an outside institution and delayed presentation to the hospital. DEVELOPMENT OF THE MODIFIED SUBAXIAL INJURY CLASSIFICATION SYSTEM Given that the SLIC score did not predict likelihood of failure of nonoperative management, we aimed to modify this scoring system using the prognostic indicators we found to determine if we
Table 2. Subaxial Injury Classification Score (SLIC) as Predictor of Failure of Nonoperative Management Odds Ratio 95% Confidence Interval P Value Total SLIC score
1.11
0.65e1.79
0.68
Table 3. Effect Time to Evaluation on Odds of Failing Nonoperative Management Odds Ratio 95% Confidence Interval P Value <8 Hours from injury
0.03
0.001e0.28
0.001
>24 Hours from injury
66.00
6.48e1729.44
<0.001
Length of hospitalization
1.00
0.84e1.12
0.95
could improve the prediction accuracy of the SLIC score. We first added time to evaluation to create a modified SLIC system (Tables 4 and 5). Time to stabilization of <8 hours was given a value of 1 point because it was a negative predictor of failure and time to stabilization >24 hours was given a value of 2 points because it was a strong predictor of failure. We used logistic regression to test how well the SLIC score modified to include time to stabilization, mSLIC (time only), could predict failure of nonoperative management and found a strong trend toward significance (OR 1.39, CI 0.998e2.07, P ¼ 0.052, see Table 4). Given the biological significance of osteoporosis for bone healing, we considered adding the presence of this comorbidity to the SLIC score. The mSLIC score including osteoporosis as the sole additional modifier performed modestly better than the original SLIC score; however, it did not reach significance (OR 1.14, CI 0.68e1.83, P ¼ 0.60, see Table 4). We then added both time to evaluation and osteoporosis as modifiers to the original SLIC scoring system (see Table 5) and found that the mSLIC score with both factors included (mSLIC total) was able to significantly predict failure of nonoperative management (OR 1.41, CI 1.01e2.08, P ¼ 0.044, see Table 4). To assess the performance of this scoring system as a clinical tool, we used ROC analysis and found that area under the curve was 0.82 0.09 (mean standard error of the mean) (95% CI 0.65e0.98, P ¼ 0.02). ROC analysis of the original SLIC scoring system revealed an area under the curve of 0.59 0.12 (mean standard error of the mean) (95% CI 0.37e0.82, P ¼ 0.50). The performance of the 2 scoring systems on predicting failure of nonoperative management is shown in Figure 1. Regarding optimal cutoff score for operative versus nonoperative management, the ROC analysis showed that a cutoff score of 4 for operative management would yield a sensitivity of 80% (95% CI 28.36%99.49%) and specificity of 65.71% (95% CI 47.79%80.87%) (Table 6).
Table 4. Modified Subaxial Injury Classification (mSLIC) Score as Predictor of Failure of Nonoperative Management Odds Ratio
95% CI
P Value
Morphology
0.98
0.36e2.24
0.96
mSLIC score (time only)
1.39
0.998e2.07
0.052
Diskoligamentous complex
1.19
0.23e6.49
0.83
mSLIC score (osteoporosis only)
1.14
0.68e1.83
0.60
Neurologic status
1.20
0.58e2.32
0.60
mSLIC score (total)
1.41
1.01e2.08
0.044
WORLD NEUROSURGERY -: e1-e6, - 2018
www.WORLDNEUROSURGERY.org
e3
ORIGINAL ARTICLE FREDERICK L. HITTI ET AL.
PREDICTORS OF FAILURE OF NONOPERATIVE MANAGEMENT
Table 5. Modified Subaxial Injury Classification (mSLIC) System Characteristic
Points
Table 6. Receiver Operating Curve Data for Modified Subaxial Injury Classification (mSLIC) System mSLIC Score
Sensitivity (95% CI)
Specificity (95% CI)
0
2
100% (47.82%e100%)
37.14% (21.47%e55.08%)
Compression
1
3
100% (47.82%e100%)
48.57% (31.38%e66.01%)
Burst
2
4
80% (28.36%e99.49%)
65.71% (47.79%e80.87%)
Distraction
3
5
60% (14.66%e94.73%)
74.29% (56.74%e87.51%)
4
6
60% (14.66%e94.73%)
85.71% (69.74%e95.19%)
7
60% (14.66%e94.73%)
94.29% (80.84%e99.30%)
20% (0.51%e71.64%)
94.29% (80.84%e99.30%)
Morphology No abnormality
Rotation/translation Diskoligamentous complex Intact
0
8
Intermediate
1
CI, confidence interval.
Disrupted
2
Neurologic status Intact
0
Root injury
1
Complete cord injury
2
Incomplete cord injury Ongoing cord compression
3 þ1
Time to stabilization <8 hours
1
>24 hours
2
Comorbidities Osteoporosis
1
DISCUSSION A large number of cases of SCI are the result of cervical spine trauma. These injuries can lead to devastating neurologic dysfunction.1,2 Trauma can cause fractures, dislocations, or ligamentous disruption in any region of the cervical spine; however, the diagnosis and management of C1 and C2 injuries are
Figure 1. Receiver operating characteristic curves of the modified subaxial injury classification (mSLIC) and SLIC systems. The area under the curve is greater with the mSLIC system compared with the SLIC system.
e4
www.SCIENCEDIRECT.com
considered separately from the diagnosis and management of subaxial spine (C3C7) injuries given the anatomic and functional variation between these regions. Subaxial spine injuries may be treated with expectant management, nonoperative external stabilization, or operative internal stabilization/fixation. Although our understanding is improving, the ideal treatment for various fractures can remain elusive at times and nonoperative treatment may fail in some patients resulting in the necessity for delayed operative intervention. Furthermore, such failure can lead to neurologic consequences in a delayed fashion. The various factors contributing to this failure are not well understood. Here, we examined predictors of failure of nonoperative management of subaxial spine trauma. The decision to treat subaxial cervical spine injury nonoperatively versus operatively is subject to many considerations including pattern/extent of injury, spinal stability, neurologic injury, and necessity of spinal canal decompression. Although most clinicians are guided by their experience, several classification systems have been developed to aid in diagnosis and treatment of these injuries. The AOSpine classification system aids in diagnosis by classifying injuries into diagnostic categories, and the SLIC system allows for the generation of a score that guides management.1,4,5 The SLIC score is computed by scoring the morphology of the injury, status of the DLC, and neurologic function of the patient.5 We found that SLIC score did not predict the likelihood of failure of nonoperative management. Indeed, a substantial portion of the patients in our cohort (35%) had SLIC scores >4 (i.e., SLIC scores in which operative treatment is recommended), but only 12.5% of patients in our cohort failed nonoperative management. Interestingly, SLIC subscores (e.g., morphology, DLC, neurologic status) also did not predict likelihood of failure. These findings question the utility of SLIC scoring in guiding treatment decision making in the clinical population of 1 large medical center in the northeast United States. Although we observed only 65% concordance between SLIC scores and actual recommended treatment in our cohort, other groups have found higher concordance. One group retrospectively calculated SLIC scores and found that prior treatment of cervical spine trauma at
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2018.11.048
ORIGINAL ARTICLE FREDERICK L. HITTI ET AL.
PREDICTORS OF FAILURE OF NONOPERATIVE MANAGEMENT
their center (nonoperative vs. operative) agreed with SLIC scoring guidelines in >90% of cases.9 These differences may simply be due to differences in clinician experience and management; however, other factors such as differences in patient populations may also play a role. Scoring systems such as the SLIC and AOSpine systems are also subject to interobserver and intraobserver variation. One group found poor reliability of the AOSpine subaxial cervical injury classification system in terms of both interobserver and intraobserver variation.4 Some have shown poor interrater agreement on the morphologic component of the SLIC score.10 However, other studies have demonstrated that interobserver agreement is high with some variability in rating of DLC injury.11,12 Proponents of the SLIC scoring system, on the other hand, have reported that the system is at least as reliable as prior scoring systems using intraclass coefficient analysis.13 The relative ease of calculation bolsters the popularity of this scoring rubric, and mobile phone applications are even available to aid in calculation of the SLIC score.14 Although SLIC scores did not predict failure of nonoperative management, we found that time to evaluation and treatment was an important predictor of treatment failure. Namely, those patients evaluated and treated <8 hours from injury were unlikely to fail nonoperative management, whereas those with delayed stabilization (>24 hours after injury) were likely to fail nonoperative management. These findings prompted us to develop the mSLIC. This modified scoring system was a significant predictor of failure of nonoperative management. Furthermore, the area under the ROC curve was significant. This was not true of the original SLIC score. Using ROC analysis, we determined that a score of 4 for operative management would yield fair sensitivity and specificity. This modified scoring system could be applied for real-time clinical decision making by using the time to evaluation for scoring purposes.
2. Passias PG, Poorman GW, Segreto FA, et al. Traumatic fractures of the cervical spine: analysis of changes in incidence, cause, concurrent injuries, and complications among 488,262 patients from 2005 to 2013. World Neurosurg. 2018;110: e427-e437. 3. Grauer JN, Vaccaro AR, Lee JY, et al. The timing and influence of MRI on the management of patients with cervical facet dislocations remains highly variable: a survey of members of the Spine Trauma Study Group. J Spinal Disord Tech. 2009;22: 96-99. 4. Silva OT da, Sabba MF, Lira HIG, et al. Evaluation of the reliability and validity of the newer AOSpine
CONCLUSIONS Diagnosis and treatment of subaxial cervical spine injuries is difficult given the variety of injury patterns and neurologic consequences. Scoring systems have been developed to aid in clinical decision making, but we found that the SLIC score did not predict failure of nonoperative management. On the other hand, delayed time to stabilization increased the likelihood of treatment failure. We used these findings to generate the mSLIC scoring system. This modified system may serve to better guide clinical decision making regarding operative versus nonoperative treatment. ACKNOWLEDGMENTS We thank the members of the Neurosurgery Clinical Research Division for their assistance with Institutional Review Board approval and data collection. We thank senior statistician Michael J. Kallan for his careful review of the data analysis and helpful suggestions.
subaxial cervical injury classification (C-3 to C-7). J Neurosurg Spine. 2016;25:303-308.
REFERENCES 1. Patel AA, Hurlbert RJ, Bono CM, Bessey JT, Yang N, Vaccaro AR. Classification and surgical decision making in acute subaxial cervical spine trauma. Spine. 2010;35:S228.
Interestingly, one group found that younger patients were more likely to fail nonoperative management.15 We did not observe this effect in our cohort. The authors of the previous study interpreted their findings as selection bias, so this may account for the difference between our studies. We must acknowledge some limitations of our study. Other than the inherent limitations of a retrospective review, our study included a relatively small number of patients from a single center. These factors may result in sampling bias and may affect generalizability of our findings. Despite these limitations, we believe our results merit further study. Specifically, the effect of time to evaluation and stabilization should be examined in other cohorts to replicate and confirm the findings of this study. Furthermore, these findings would be greatly strengthened by replication in a prospective study.
5. Patel AA, Dailey A, Brodke DS, et al. Subaxial cervical spine trauma classification: the Subaxial Injury Classification system and case examples. Neurosurg Focus. 2008;25:E8. 6. Joaquim AF, Patel AA, Vaccaro AR. Cervical injuries scored according to the Subaxial Injury Classification system: an analysis of the literature. J Craniovertebral Junction Spine. 2014;5:65-70. 7. Dvorak MF, Fisher CG, Fehlings MG, et al. The surgical approach to subaxial cervical spine injuries: an evidence-based algorithm based on the SLIC classification system. Spine. 2007;32: 2620-2629. 8. Joaquim AF, Ghizoni E, Tedeschi H, da Cruz HYF, Patel AA. Clinical results of patients with subaxial cervical spine trauma treated according to the SLIC score. J Spinal Cord Med. 2014;37:420-424.
WORLD NEUROSURGERY -: e1-e6, - 2018
9. Samuel S, Lin J-L, Smith MM, et al. Subaxial injury classification scoring system treatment recommendations: external agreement study based on retrospective review of 185 patients. Spine. 2015;40:137-142. 10. van Middendorp JJ, Audigé L, Bartels RH, et al. The Subaxial Cervical Spine Injury Classification System: an external agreement validation study. Spine J Off J North Am Spine Soc. 2013;13:1055-1063. 11. Stone AT, Bransford RJ, Lee MJ, et al. Reliability of classification systems for subaxial cervical injuries. Evid Based Spine Care J. 2010;1:19-26. 12. Lee WJ, Yoon SH, Kim YJ, Kim JY, Park HC, Park CO. Interobserver and intraobserver reliability of sub-axial injury classification and severity scale between radiologist, resident and spine surgeon. J Kor Neurosurg Soc. 2012;52:200-203. 13. Whang PG, Patel AA, Vaccaro AR. The development and evaluation of the subaxial injury classification scoring system for cervical spine trauma. Clin Orthop. 2011;469:723-731.
www.WORLDNEUROSURGERY.org
e5
ORIGINAL ARTICLE FREDERICK L. HITTI ET AL.
14. Kubben PL. SLIC 2: improved decision support for subaxial cervical spine injury. Surg Neurol Int. 2012; 3. 15. Aarabi B, Mirvis S, Shanmuganathan K, et al. Comparative effectiveness of surgical versus nonoperative management of unilateral, nondisplaced, subaxial cervical spine facet fractures
e6
www.SCIENCEDIRECT.com
PREDICTORS OF FAILURE OF NONOPERATIVE MANAGEMENT
without evidence of spinal cord injury: clinical article. J Neurosurg Spine. 2014;20:270-277.
Citation: World Neurosurg. (2018). https://doi.org/10.1016/j.wneu.2018.11.048 Journal homepage: www.WORLDNEUROSURGERY.org
Conflict of interest statement: The authors report no conflict of interest concerning the materials or methods used in this study or the findings described in this paper. Received 16 August 2018; accepted 7 November 2018
Available online: www.sciencedirect.com 1878-8750/$ - see front matter ª 2018 Elsevier Inc. All rights reserved.
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2018.11.048
Predictors of Failure of Nonoperative Management Following Subaxial Spine Trauma and Creation of Modified Subaxial Injury Classification System Frederick L. Hitti1, Brendan J. McShane1, Andrew I. Yang1, Cole Rinehart1, Ahmed Albayar1, Marc Branche1, Neil R. Malhotra1, M. Burhan Janjua1,2, Zarina S. Ali1, James M. Schuster1, Ali K. Ozturk1
BACKGROUND: Subaxial cervical spine injuries may be treated with either nonoperative stabilization or surgical fixation. The subaxial injury classification (SLIC) provides 1 method for suggesting the degree of necessity for surgery. In the current study, we examined if the SLIC score, or other preoperative metrics, can predict failure of nonoperative management.
-
METHODS: We performed a retrospective chart review to identify patients who presented with acute, nonpenetrating, subaxial cervical spine injury within our health system between 2007 and 2016. Patient demographics, medical comorbidities, injuries, and treatments were collected. Logistic regression analysis was used to determine potential predictors of failure of nonoperative management.
-
RESULTS: During the study period, 40 patients met the inclusion criteria. A small subset of patients failed nonoperative management (n [ 5, 12.5%). The mean SLIC score was 3.9 1.9; however, 14 (35%) patients had scores >4. Neither total SLIC score (P [ 0.68) nor SLIC subscores (morphology [P [ 0.96], discoligamentous complex [P [ 0.83], neurologic status [P [ 0.60]) predicted failure of nonoperative treatment. Time to evaluation/treatment did predict failure of nonoperative management. Evaluation within 8 hours of injury was a negative predictor of failure (odds ratio [ 0.03, P [ 0.001) and evaluation 24 hours or more after injury was a positive predictor of failure (odds ratio [ 66.00, P < 0.001). We created a modified SLIC score
-
Key words mSLIC - Nonoperative failure - Predictors - Subaxial cervical spine - Trauma SLIC -
Abbreviations and Acronyms CI: Confidence interval DLC: Diskoligamentous complex mSLIC: Modified subaxial injury classification OR: Odds ratio
on the basis of these findings, which significantly predicted failure of nonoperative management (P [ 0.044). CONCLUSIONS: Management of subaxial spine injuries is complex. In our cohort, SLIC scoring did not adequately predict odds of failure of nonoperative management. Time to evaluation, however, did. We created a modified SLIC score that significantly predicted failure of nonoperative management.
-
INTRODUCTION
A
lmost half of all spinal cord injuries (SCIs) in North America are a result of cervical spine trauma, and the incidence of these injuries is increasing.1,2 Importantly, these spinal cord injuries can have profound neurologic ramifications.3 Subaxial cervical spine trauma includes dislocations, ligamentous disruption, and/or fractures from C3C7. Traumatic injuries of the cervical spine can lead to neurologic impairment and/or instability.4,5 It is critical to quickly diagnose and treat these injuries to prevent or ameliorate neurologic decline. Subaxial spine trauma may be treated with observation, nonoperative stabilization, or operative intervention with internal fixation and/or decompression. External devices such as cervical collars or a cervical halo may be used to achieve stabilization in patients managed nonoperatively. Patients are typically managed
ROC: Receiver-operating characteristic SLIC: Subaxial injury classification From the Departments of 1Neurosurgery and 2Orthopaedic Surgery, Pennsylvania Hospital, University of Pennsylvania, Philadelphia, Pennsylvania, USA To whom correspondence should be addressed: Frederick L. Hitti, M.D., Ph.D. [E-mail: [email protected]] Citation: World Neurosurg. (2018). https://doi.org/10.1016/j.wneu.2018.11.048 Journal homepage: www.WORLDNEUROSURGERY.org Available online: www.sciencedirect.com 1878-8750/$ - see front matter ª 2018 Elsevier Inc. All rights reserved.
WORLD NEUROSURGERY -: e1-e6, - 2018
www.WORLDNEUROSURGERY.org
e1
ORIGINAL ARTICLE FREDERICK L. HITTI ET AL.
PREDICTORS OF FAILURE OF NONOPERATIVE MANAGEMENT
nonoperatively if the cervical spine is not grossly unstable or if they are not surgical candidates. Scoring systems have been developed to aid in diagnosis and treatment of cervical spine injury. The AOSpine subaxial injury classification system4 may be used to aid in morphologic classification of the injury, and the decision to pursue operative versus nonoperative treatment may be aided by the use of the subaxial injury classification (SLIC) system.1,6 The SLIC score uses morphology, integrity of the diskoligamentous complex (DLC), and neurologic status to guide treatment decisions.5 Some have developed treatment guidelines on the basis of a combination of SLIC scores and expert consensus.7 One study followed patients prospectively and found that none of those in their cohort exhibited neurologic deterioration if SLIC criteria were used to guide management.8 The authors argued in favor of SLIC-based management. Although the aforementioned scoring systems and treatment guidelines are available, there are limited studies on the utility of these systems to predict failure of nonoperative management. Here, we examined a cohort of patients who presented with subaxial cervical spine trauma and were treated nonoperatively to assess factors that could predict failure of nonoperative management. METHODS Study Population In this Institutional Review Boardapproved study, we enrolled consecutive patients with acute nonpenetrating subaxial cervical spine injuries referred for neurosurgical consultation from June 1, 2007eJune 1, 2016. This patient population included patients with fractures or dislocations of C3-C7. Ten surgeons at our institution participated in this study. Their clinical decision making was not bound by SLIC scores.
patients, 40 (37%) were initially managed nonoperatively with cervical spine collars. The demographics of this cohort of patients are shown in Table 1. The mean age was 50.5 21.5, and the majority of patients were male (n ¼ 27, 67.5%). With regards to race, the highest percentage of patients were Caucasian (n ¼ 19, 47.5%). A large percentage of the patients were African American (n ¼ 17, 42.5%). We tracked patient comorbidities and found osteoporosis, diabetes, coronary artery disease, peripheral arterial disease, and cancer in 10.0%, 7.5%, 7.5%, 5.0%, and 2.5% of patients, respectively. Regarding morphology of injury, 25 (62.5%) patients presented with compression fractures, 12 (30.0%) presented with distraction injuries, and 3 (7.5%) presented with rotation/translation injuries (see Table 1). Five patients (12.5%) failed nonoperative management and required surgical stabilization. Subluxation/anterolisthesis on repeat cervical spine radiograph (4 patients) or dynamic instability on flexion/extension cervical spine radiograph (1 patient) were deemed indications for surgical fixation due to failure of nonoperative management. The effect of age, sex, and race on failure of nonoperative management was investigated using logistic regression. None of these demographic factors was significant in predicting treatment failure (age: P ¼ 0.92; sex: P ¼ 0.71; race: P ¼ 0.07). The effect of patient comorbidities on failure of nonoperative stabilization could not be reliably analyzed due to cohort sample size. Subaxial Injury Classification Scores as Predictor of Failure of Nonoperative Management In our cohort of patients initially managed nonoperatively, the mean SLIC score was 3.9 1.9 (n ¼ 40, see Table 1). The mean SLIC score of patients initially managed operatively was 6.67 1.65 (n ¼ 69). The treating physicians in this study were not
CLINICAL DATA COLLECTION Once the target patient population was identified, we collected relevant data from medical records—both paper and electronic (Epic, Epic Systems Corporation, Madison, Wisconsin, USA). Demographic information was collected, and medical comorbidities at the time of initial injury were noted. We also recorded the type of injury, mechanism of injury, time to neurosurgical evaluation from initial injury, and length of hospitalization.
Table 1. Demographics Age
50.5 21.5 years old
Sex
Males, 27 (67.5%) Females, 13 (32.5%)
Race
Caucasians, 19 (47.5%) African-Americans, 17 (42.5%) Asians, 3 (7.5%) Other, 1 (2.5%)
Comorbidities
Osteoporosis, 4 (10.0%) Diabetes, 3 (7.5%) Coronary artery disease, 3 (7.5%) Peripheral arterial disease, 2 (5.0%) Cancer, 1 (2.5%)
Morphology of injury
Compression fracture, 25 (62.5%) Distraction, 12 (30.0%) Rotation/Translation 3 (7.5%)
STATISTICAL ANALYSIS Statistical analysis was performed using Microsoft Excel and R. Logistic regression analysis was performed using the generalized linear model in R. Type II analysis of variance was then used to compute P values. Only univariate logistic regression was performed due to the sample size of the cohort. Results were considered significant if P < 0.05. Receiver-operating characteristic (ROC) analysis and curves were generated using Prism (GraphPad software). Averages are presented as mean standard deviation unless otherwise indicated. RESULTS Study Population We first identified 109 patients who presented to our institution following acute nonpenetrating subaxial spine trauma. Of these
e2
www.SCIENCEDIRECT.com
SLIC score
3.9 1.9
SLIC > 4
14 (35%)
Patients failing nonoperative management
5 (12.5%)
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2018.11.048
ORIGINAL ARTICLE FREDERICK L. HITTI ET AL.
PREDICTORS OF FAILURE OF NONOPERATIVE MANAGEMENT
bound by SLIC treatment guidelines. Fourteen patients in the nonoperative group (35%) had a SLIC score >4 (i.e., SLIC scores high enough to recommend operative intervention). We examined whether or not the SLIC score was predictive of failure of nonoperative management by logistic regression (Table 2). Total SLIC score was not predictive of treatment failure (odds ratio [OR] 1.11, 95% confidence interval [CI] 0.65e 1.79, P ¼ 0.68). Furthermore, univariate logistic regression of the SLIC subscores demonstrated that none were predictive of failure of nonoperative management: morphology (OR 0.98, 95% CI 0.36e2.24, P ¼ 0.96), DLC disruption (OR 1.19, 95% CI, 0.23e6.49, P ¼ 0.83), and neurologic status (OR 1.20, 95% CI 0.58e2.32, P ¼ 0.60). EFFECT OF TIME TO STABILIZATION ON ODDS OF FAILING NONOPERATIVE MANAGEMENT We hypothesized that time to cervical spine stabilization would have an impact on failure of nonoperative management. To examine this, we noted the time to cervical spine stabilization from initial injury. We also analyzed length of hospital stay because medically complex patients tend to have longer hospitalizations. The average length of the acute hospitalization was 7.8 7.1 days in this cohort of patients. We performed univariate logistic regression followed by type II analysis of variance to examine the effect of rapid time to nonoperative stabilization and length of hospitalization on odds of requiring surgical stabilization after initial nonoperative management. We found that patients who were stabilized <8 hours after initial injury had a decreased incidence of future surgery with an OR of 0.03 (0.001e0.28, 95% CI), P ¼ 0.001 (Table 3). The length of hospitalization was not associated with failure of nonoperative management. We also hypothesized that a delay in nonoperative stabilization might increase the odds of failure of nonoperative management. Using logistic regression, we found that stabilization 24 hours after the initial injury resulted in an increased odds of failure of nonoperative management with an OR of 66.00 (6.48e1729.44, 95% CI), P < 0.001 (see Table 3). Reasons for delayed stabilization included failure to diagnose a cervical spine injury at an outside institution and delayed presentation to the hospital. DEVELOPMENT OF THE MODIFIED SUBAXIAL INJURY CLASSIFICATION SYSTEM Given that the SLIC score did not predict likelihood of failure of nonoperative management, we aimed to modify this scoring system using the prognostic indicators we found to determine if we
Table 2. Subaxial Injury Classification Score (SLIC) as Predictor of Failure of Nonoperative Management Odds Ratio 95% Confidence Interval P Value Total SLIC score
1.11
0.65e1.79
0.68
Table 3. Effect Time to Evaluation on Odds of Failing Nonoperative Management Odds Ratio 95% Confidence Interval P Value <8 Hours from injury
0.03
0.001e0.28
0.001
>24 Hours from injury
66.00
6.48e1729.44
<0.001
Length of hospitalization
1.00
0.84e1.12
0.95
could improve the prediction accuracy of the SLIC score. We first added time to evaluation to create a modified SLIC system (Tables 4 and 5). Time to stabilization of <8 hours was given a value of 1 point because it was a negative predictor of failure and time to stabilization >24 hours was given a value of 2 points because it was a strong predictor of failure. We used logistic regression to test how well the SLIC score modified to include time to stabilization, mSLIC (time only), could predict failure of nonoperative management and found a strong trend toward significance (OR 1.39, CI 0.998e2.07, P ¼ 0.052, see Table 4). Given the biological significance of osteoporosis for bone healing, we considered adding the presence of this comorbidity to the SLIC score. The mSLIC score including osteoporosis as the sole additional modifier performed modestly better than the original SLIC score; however, it did not reach significance (OR 1.14, CI 0.68e1.83, P ¼ 0.60, see Table 4). We then added both time to evaluation and osteoporosis as modifiers to the original SLIC scoring system (see Table 5) and found that the mSLIC score with both factors included (mSLIC total) was able to significantly predict failure of nonoperative management (OR 1.41, CI 1.01e2.08, P ¼ 0.044, see Table 4). To assess the performance of this scoring system as a clinical tool, we used ROC analysis and found that area under the curve was 0.82 0.09 (mean standard error of the mean) (95% CI 0.65e0.98, P ¼ 0.02). ROC analysis of the original SLIC scoring system revealed an area under the curve of 0.59 0.12 (mean standard error of the mean) (95% CI 0.37e0.82, P ¼ 0.50). The performance of the 2 scoring systems on predicting failure of nonoperative management is shown in Figure 1. Regarding optimal cutoff score for operative versus nonoperative management, the ROC analysis showed that a cutoff score of 4 for operative management would yield a sensitivity of 80% (95% CI 28.36%99.49%) and specificity of 65.71% (95% CI 47.79%80.87%) (Table 6).
Table 4. Modified Subaxial Injury Classification (mSLIC) Score as Predictor of Failure of Nonoperative Management Odds Ratio
95% CI
P Value
Morphology
0.98
0.36e2.24
0.96
mSLIC score (time only)
1.39
0.998e2.07
0.052
Diskoligamentous complex
1.19
0.23e6.49
0.83
mSLIC score (osteoporosis only)
1.14
0.68e1.83
0.60
Neurologic status
1.20
0.58e2.32
0.60
mSLIC score (total)
1.41
1.01e2.08
0.044
WORLD NEUROSURGERY -: e1-e6, - 2018
www.WORLDNEUROSURGERY.org
e3
ORIGINAL ARTICLE FREDERICK L. HITTI ET AL.
PREDICTORS OF FAILURE OF NONOPERATIVE MANAGEMENT
Table 5. Modified Subaxial Injury Classification (mSLIC) System Characteristic
Points
Table 6. Receiver Operating Curve Data for Modified Subaxial Injury Classification (mSLIC) System mSLIC Score
Sensitivity (95% CI)
Specificity (95% CI)
0
2
100% (47.82%e100%)
37.14% (21.47%e55.08%)
Compression
1
3
100% (47.82%e100%)
48.57% (31.38%e66.01%)
Burst
2
4
80% (28.36%e99.49%)
65.71% (47.79%e80.87%)
Distraction
3
5
60% (14.66%e94.73%)
74.29% (56.74%e87.51%)
4
6
60% (14.66%e94.73%)
85.71% (69.74%e95.19%)
7
60% (14.66%e94.73%)
94.29% (80.84%e99.30%)
20% (0.51%e71.64%)
94.29% (80.84%e99.30%)
Morphology No abnormality
Rotation/translation Diskoligamentous complex Intact
0
8
Intermediate
1
CI, confidence interval.
Disrupted
2
Neurologic status Intact
0
Root injury
1
Complete cord injury
2
Incomplete cord injury Ongoing cord compression
3 þ1
Time to stabilization <8 hours
1
>24 hours
2
Comorbidities Osteoporosis
1
DISCUSSION A large number of cases of SCI are the result of cervical spine trauma. These injuries can lead to devastating neurologic dysfunction.1,2 Trauma can cause fractures, dislocations, or ligamentous disruption in any region of the cervical spine; however, the diagnosis and management of C1 and C2 injuries are
Figure 1. Receiver operating characteristic curves of the modified subaxial injury classification (mSLIC) and SLIC systems. The area under the curve is greater with the mSLIC system compared with the SLIC system.
e4
www.SCIENCEDIRECT.com
considered separately from the diagnosis and management of subaxial spine (C3C7) injuries given the anatomic and functional variation between these regions. Subaxial spine injuries may be treated with expectant management, nonoperative external stabilization, or operative internal stabilization/fixation. Although our understanding is improving, the ideal treatment for various fractures can remain elusive at times and nonoperative treatment may fail in some patients resulting in the necessity for delayed operative intervention. Furthermore, such failure can lead to neurologic consequences in a delayed fashion. The various factors contributing to this failure are not well understood. Here, we examined predictors of failure of nonoperative management of subaxial spine trauma. The decision to treat subaxial cervical spine injury nonoperatively versus operatively is subject to many considerations including pattern/extent of injury, spinal stability, neurologic injury, and necessity of spinal canal decompression. Although most clinicians are guided by their experience, several classification systems have been developed to aid in diagnosis and treatment of these injuries. The AOSpine classification system aids in diagnosis by classifying injuries into diagnostic categories, and the SLIC system allows for the generation of a score that guides management.1,4,5 The SLIC score is computed by scoring the morphology of the injury, status of the DLC, and neurologic function of the patient.5 We found that SLIC score did not predict the likelihood of failure of nonoperative management. Indeed, a substantial portion of the patients in our cohort (35%) had SLIC scores >4 (i.e., SLIC scores in which operative treatment is recommended), but only 12.5% of patients in our cohort failed nonoperative management. Interestingly, SLIC subscores (e.g., morphology, DLC, neurologic status) also did not predict likelihood of failure. These findings question the utility of SLIC scoring in guiding treatment decision making in the clinical population of 1 large medical center in the northeast United States. Although we observed only 65% concordance between SLIC scores and actual recommended treatment in our cohort, other groups have found higher concordance. One group retrospectively calculated SLIC scores and found that prior treatment of cervical spine trauma at
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2018.11.048
ORIGINAL ARTICLE FREDERICK L. HITTI ET AL.
PREDICTORS OF FAILURE OF NONOPERATIVE MANAGEMENT
their center (nonoperative vs. operative) agreed with SLIC scoring guidelines in >90% of cases.9 These differences may simply be due to differences in clinician experience and management; however, other factors such as differences in patient populations may also play a role. Scoring systems such as the SLIC and AOSpine systems are also subject to interobserver and intraobserver variation. One group found poor reliability of the AOSpine subaxial cervical injury classification system in terms of both interobserver and intraobserver variation.4 Some have shown poor interrater agreement on the morphologic component of the SLIC score.10 However, other studies have demonstrated that interobserver agreement is high with some variability in rating of DLC injury.11,12 Proponents of the SLIC scoring system, on the other hand, have reported that the system is at least as reliable as prior scoring systems using intraclass coefficient analysis.13 The relative ease of calculation bolsters the popularity of this scoring rubric, and mobile phone applications are even available to aid in calculation of the SLIC score.14 Although SLIC scores did not predict failure of nonoperative management, we found that time to evaluation and treatment was an important predictor of treatment failure. Namely, those patients evaluated and treated <8 hours from injury were unlikely to fail nonoperative management, whereas those with delayed stabilization (>24 hours after injury) were likely to fail nonoperative management. These findings prompted us to develop the mSLIC. This modified scoring system was a significant predictor of failure of nonoperative management. Furthermore, the area under the ROC curve was significant. This was not true of the original SLIC score. Using ROC analysis, we determined that a score of 4 for operative management would yield fair sensitivity and specificity. This modified scoring system could be applied for real-time clinical decision making by using the time to evaluation for scoring purposes.
2. Passias PG, Poorman GW, Segreto FA, et al. Traumatic fractures of the cervical spine: analysis of changes in incidence, cause, concurrent injuries, and complications among 488,262 patients from 2005 to 2013. World Neurosurg. 2018;110: e427-e437. 3. Grauer JN, Vaccaro AR, Lee JY, et al. The timing and influence of MRI on the management of patients with cervical facet dislocations remains highly variable: a survey of members of the Spine Trauma Study Group. J Spinal Disord Tech. 2009;22: 96-99. 4. Silva OT da, Sabba MF, Lira HIG, et al. Evaluation of the reliability and validity of the newer AOSpine
CONCLUSIONS Diagnosis and treatment of subaxial cervical spine injuries is difficult given the variety of injury patterns and neurologic consequences. Scoring systems have been developed to aid in clinical decision making, but we found that the SLIC score did not predict failure of nonoperative management. On the other hand, delayed time to stabilization increased the likelihood of treatment failure. We used these findings to generate the mSLIC scoring system. This modified system may serve to better guide clinical decision making regarding operative versus nonoperative treatment. ACKNOWLEDGMENTS We thank the members of the Neurosurgery Clinical Research Division for their assistance with Institutional Review Board approval and data collection. We thank senior statistician Michael J. Kallan for his careful review of the data analysis and helpful suggestions.
subaxial cervical injury classification (C-3 to C-7). J Neurosurg Spine. 2016;25:303-308.
REFERENCES 1. Patel AA, Hurlbert RJ, Bono CM, Bessey JT, Yang N, Vaccaro AR. Classification and surgical decision making in acute subaxial cervical spine trauma. Spine. 2010;35:S228.
Interestingly, one group found that younger patients were more likely to fail nonoperative management.15 We did not observe this effect in our cohort. The authors of the previous study interpreted their findings as selection bias, so this may account for the difference between our studies. We must acknowledge some limitations of our study. Other than the inherent limitations of a retrospective review, our study included a relatively small number of patients from a single center. These factors may result in sampling bias and may affect generalizability of our findings. Despite these limitations, we believe our results merit further study. Specifically, the effect of time to evaluation and stabilization should be examined in other cohorts to replicate and confirm the findings of this study. Furthermore, these findings would be greatly strengthened by replication in a prospective study.
5. Patel AA, Dailey A, Brodke DS, et al. Subaxial cervical spine trauma classification: the Subaxial Injury Classification system and case examples. Neurosurg Focus. 2008;25:E8. 6. Joaquim AF, Patel AA, Vaccaro AR. Cervical injuries scored according to the Subaxial Injury Classification system: an analysis of the literature. J Craniovertebral Junction Spine. 2014;5:65-70. 7. Dvorak MF, Fisher CG, Fehlings MG, et al. The surgical approach to subaxial cervical spine injuries: an evidence-based algorithm based on the SLIC classification system. Spine. 2007;32: 2620-2629. 8. Joaquim AF, Ghizoni E, Tedeschi H, da Cruz HYF, Patel AA. Clinical results of patients with subaxial cervical spine trauma treated according to the SLIC score. J Spinal Cord Med. 2014;37:420-424.
WORLD NEUROSURGERY -: e1-e6, - 2018
9. Samuel S, Lin J-L, Smith MM, et al. Subaxial injury classification scoring system treatment recommendations: external agreement study based on retrospective review of 185 patients. Spine. 2015;40:137-142. 10. van Middendorp JJ, Audigé L, Bartels RH, et al. The Subaxial Cervical Spine Injury Classification System: an external agreement validation study. Spine J Off J North Am Spine Soc. 2013;13:1055-1063. 11. Stone AT, Bransford RJ, Lee MJ, et al. Reliability of classification systems for subaxial cervical injuries. Evid Based Spine Care J. 2010;1:19-26. 12. Lee WJ, Yoon SH, Kim YJ, Kim JY, Park HC, Park CO. Interobserver and intraobserver reliability of sub-axial injury classification and severity scale between radiologist, resident and spine surgeon. J Kor Neurosurg Soc. 2012;52:200-203. 13. Whang PG, Patel AA, Vaccaro AR. The development and evaluation of the subaxial injury classification scoring system for cervical spine trauma. Clin Orthop. 2011;469:723-731.
www.WORLDNEUROSURGERY.org
e5
ORIGINAL ARTICLE FREDERICK L. HITTI ET AL.
14. Kubben PL. SLIC 2: improved decision support for subaxial cervical spine injury. Surg Neurol Int. 2012; 3. 15. Aarabi B, Mirvis S, Shanmuganathan K, et al. Comparative effectiveness of surgical versus nonoperative management of unilateral, nondisplaced, subaxial cervical spine facet fractures
e6
www.SCIENCEDIRECT.com
PREDICTORS OF FAILURE OF NONOPERATIVE MANAGEMENT
without evidence of spinal cord injury: clinical article. J Neurosurg Spine. 2014;20:270-277.
Citation: World Neurosurg. (2018). https://doi.org/10.1016/j.wneu.2018.11.048 Journal homepage: www.WORLDNEUROSURGERY.org
Conflict of interest statement: The authors report no conflict of interest concerning the materials or methods used in this study or the findings described in this paper. Received 16 August 2018; accepted 7 November 2018
Available online: www.sciencedirect.com 1878-8750/$ - see front matter ª 2018 Elsevier Inc. All rights reserved.
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2018.11.048