Anda di halaman 1dari 7

Journal of Clinical Neuroscience 42 (2017) 122128

Contents lists available at ScienceDirect

Journal of Clinical Neuroscience


journal homepage: www.elsevier.com/locate/jocn

Case study

Prognostic model for patients with traumatic brain injuries and


abnormal computed tomography scans
Jefferson Rosi Junior a, Leonardo C. Welling b,, Marcelo Schafranski b, Lin Tchia Yeng a,
Rogrio Ruscito do Prado a, Edwin Koterba a, Almir Ferreira de Andrade a, Manoel Jacobsen Teixeira a,
Eberval Gadelha Figueiredo a
a
Division of Neurological Surgery, University of Sao Paulo, Brazil
b
Division of Neurological Surgery, State University of Ponta Grossa, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: Traumatic brain injury (TBI) is an important cause of death and disability worldwide. The prognosis eval-
Received 24 October 2016 uation is a challenge when many variables are involved. The authors aimed to develop prognostic model
Accepted 5 March 2017 for assessment of survival chances after TBI based on admission characteristics, including extracranial
injuries, which would allow application of the model before in-hospital therapeutic interventions. A
cohort study evaluated 1275 patients with TBI and abnormal CT scans upon admission to the emergency
Keywords: unit of Hospital das Clinicas of University of Sao Paulo and analyzed the final outcome on mortality. A
Brain injury
logistic regression analysis was undertaken to determine the adjusted weigh of each independent vari-
Prognostic
Survival
able in the outcome. Four variables were found to be significant in the model: age (years), Glasgow
Coma Scale (315), Marshall Scale (MS, stratified into 2,3 or 4,5,6; according to the best group positive
predictive value) and anysochoria (yes/no). The following formula is in a logistic model (USP index to head
injury) estimates the probability of death of patients according to characteristics that influence on mor-
tality. We consider that our mathematical probability model (USP Index) may be applied to clinical prog-
nosis in patients with abnormal CT scans after severe traumatic brain injury.
2017 Elsevier Ltd. All rights reserved.

1. Introduction Specific patient characteristics may affect the outcome in opposite


directions [8,9].
Traumatic brain injury (TBI) usually presents with a broad spec- The importance of baseline clinical and computed tomographic
trum of symptoms and disabilities [1,2]. Worldwide, every year (CT) characteristics for survival probability and prediction of long-
millions require treatment after TBI. The disability rates is related term outcome has been investigated by many studies [1014].
to the severity and mechanisms of the TBI and unfavorable out- Such prediction of outcome may support clinical decision-making
comes (Glasgow Outcome Scale 4,5 or 6) following TBI are quite and resources allocation. When patients are classified according
prevalent [15]. to risks, the data may be useful to compare outcomes, surgical
Creation of a prognostic model, with many variables, to stratify results or even, stratify patients for future clinical trials [7,8].
the risk for an individual patient is a challenge [6]. The accurate The majority of models was developed using small sample sizes
prediction of long-term outcome, after emergency neurological and includes few data obtained after admission [15]. The signifi-
assessment combining individual biological predictors, tomo- cance of extracranial injuries has no consensus. While many stud-
graphic findings and extracranial injuries is very complex [68]. ies demonstrated that outcome is related to the severity of the
primary cerebral damage others emphasize the importance of
extracranial injuries [6,7,9]. Besides all variables included in differ-
ent models, the external validation of these are critical to clinical
Corresponding author at: Division of Neurological Surgery, State University of
applicability [10].
Ponta Grossa, Rua Tiradentes, 976. Centro, 84010-190 Ponta Grossa, Brazil.
E-mail addresses: jefrosijr@gmail.com (J.R. Junior), leonardowelling@yahoo.com.
In this paper the authors aimed to develop a prognostic model
br (L.C. Welling), marceloschafranski@yahoo.com.br (M. Schafranski), linyeng@ for assessment of survival chances after TBI based on admission
gmail.com (L.T. Yeng), ruscittousp@yahoo.com.br (R.R. do Prado), koterba@uol.com. characteristics, including extracranial injuries and, in addition,
br (E. Koterba), lcwelling9@gmail.com (A.F. de Andrade), manoeljacobsen@gmail. demonstrate that many variables that are included in well-
com (M.J. Teixeira), ebgadelha@yahoo.com (E.G. Figueiredo).

http://dx.doi.org/10.1016/j.jocn.2017.03.012
0967-5868/ 2017 Elsevier Ltd. All rights reserved.
J.R. Junior et al. / Journal of Clinical Neuroscience 42 (2017) 122128 123

known clinical models are not significant. The purpose of this 3.3. Computed tomography findings
model is to be applied before in-hospital therapeutic interventions.
Marshall degree is statistically associated with mortality
(p < 0.001). Midline deviation is a significant factor in increasing
2. Methods mortality (p < 0.001). Despite exclusion of analysis Marshall I
patients, any category above Marshall II increases the risk of death.
Our cohort study evaluated 1275 patients with TBI and abnor- This was especially observed in Marshall VI when compared with
mal CT scans upon admission to the emergency unit of Hospital Marshall V patients (Table 2). When grouped the overall mortality
das Clinicas of University of Sao Paulo between September 1, of Marshall II and III was 12% and Marshall IV, V and VI was 26%
2003 and December 31, 2009. The primary objective was to ana- (p < 0.001).
lyze the final outcome on mortality. Clinical and radiographic data
were prospectively collected and entered into our database. The
3.4. Glasgow Coma Scale and anisocoria
demographic data, mechanism of cranial trauma, Glasgow Coma
Scale, anisocoria, presence of thoracic lesions, abdominal injuries,
The average Glasgow Coma Scale of patients who die is statisti-
upper limbs fractures, lower limbs fractures, hip fractures and neu-
cally lower than the patients who survived (p < 0.001) (table 3).
rosurgical intervention were compiled. The Marshall scale and
The 1-point increase in Glasgow scale causes a 19% reduction in
midline deviation were documented. Sedation before neurological
the chance of mortality of the patient. Pupillary abnormalities were
assessment, hemodynamic instability or CT scan without abnor-
associated with higher mortality (p < 0.001). The mortality of
malities (Marshall I patients) were excluded from our analysis.
anisocoric patients was 38% and 16% for isocoric patients
Multivariate logistic regression analysis, chi-square test or the
(p < 0.001). Patients with anisocoria have 67% higher chance of
likelihood ratio tests were used to evaluate the effects of patients
mortality (Table 2).
demographics, type of TBI, and extracranial injuries in mortality.
Age, ethnics, gender, etiology of trauma, spinal fractures, thoracic
trauma, abdominal lesions, upper and lower extremities fractures, 3.5. Neurosurgical intervention
GCS score, fixed pupillary response on hospital admission, Marshall
classification, midline shift deviation were included in our analysis. Patients (n = 601) were operated on following current literature
The influence of age and Glasgow Coma Scale on mortality was recommendations for the neurosurgical management of patients
estimated with the use of summary measures (mean, standard with traumatic brain injury. For lesions with more than 50 cm3,
deviation, median, minimum and maximum) and Students t-test. the decision were easier, and usually in favor of surgery. For lesions
Logistic regression analysis was used to analyze the adjusted with volume less than 25 cm3 the conservative management were
weigh of each independent variable in the outcome. According to prefered. When the lesion volume was between 25 and 50 cm3
our results in the multivariate analysis and adjusted weight, four associated factors, as midline shift, cisterns aspects, and Glasgow
variables were found to be significant in the model: age (years), Coma Scale were considered for surgical indication.
Glasgow Coma Scale (315), Marshall Scale (MS, stratified into
2,3 or 4,5,6; according to the best group positive predictive value) 3.6. Multivariate analysis
and anysochoria (yes/no).
After interpretation of the adjusted weight of significant vari- A logistic regression analysis was undertaken to define the
ables a mortality probability calculator was developed using as adjusted weigh of each independent variable in the outcome. Four
input the logit (p) obtained from the analysis applying those four variables were found to be significant in the model: age (years),
independent variables. The tests were performed at a significance Glasgow Coma Scale (315), Marshall Scale (MS, stratified into
level of 5%. 2,3 or 4,5,6; according to the best group positive predictive value)
and anysochoria (yes/no). Data from the regression analysis, odds
ratio with IC 95%, beta-coefficients and p-values from the four
3. Results independent variables included in the model are presented in
Table 4.
3.1. General characteristics The following formula is in a logistic model (we call it USP index
to head injury) estimates the probability of death of patients
Baseline data is presented in Table 1. Most patients were men according to characteristics that influence on mortality:
(81%). More than half (75%) of participants were caucasians. Road Logistic regression equation: logit (p) = 1.36 + (age * 0.03)
traffic crashes represented the most common cause of injury + (GCS * 0.2) + (MS 4,6 or 6 * 0.72) + (anysochoria * 0.58). Proba-
(54%). Young people (less than 30 years) were involved in 39% of bility = 1/1 + elogit (p).
hospital admission. The general mortality was 19.1%. The mortality Constant: 2.04. Hosmer and Lemeshow test: p = 0.24, area
increases with increasing age of the patient (p < 0.001). Ethnics under the curve (AUC): 0.77 ( 0.740.79).
were statistically associated with mortality, 26% of Caucasian peo- Based on this logistic equation, a mortality probability calcula-
ple died versus 16% of others (p = 0.03). tor (Supplementary material) was developed using as input the
logit (p) obtained from the analysis applying those four indepen-
3.2. Extracranial injuries and mortality dent variables. As shown in the example (Fig. 1), the death proba-
bility of a patient aged 65, with a GCS of 7, a MS of 4, with no
The presence of spinal fracture was associated with lower mor- anysochoria is estimated to be around 47.7%. The significance level
tality in the univariate analysis. Of 169 spinal fractures in this ser- of 5% was considered.
ies only 8 patients (5%) died, while 21% of patients without spinal
fractures died (p < 0.001). Spinal fractures were detected in 20% of 4. Discussion
patients classified as Marshall 4, 5 or 6 and 6% of patients classified
as Marshall 2 or 3 (p < 0.001). Thoracic, abdominal, pelvic lesions Neurotrauma is an expressive public health problem. It is a
and upper or lower limbs fractures were not associated with higher heterogeneous disease and outcomes for individual patients are
mortality in these patients (Table 2). very difficult to predict. There are many variables as mechanism
124 J.R. Junior et al. / Journal of Clinical Neuroscience 42 (2017) 122128

Table 1
Demographic data.

Death Total p
No Yes
n % n %
Sex 0.650
Female 199 81.9 44 18.1 243
Male 832 80.6 200 19.4 1032
Age (years) <0.001
<20 158 89.3 19 10.7 177
2029 280 85.9 46 14.1 326
3039 195 81.3 45 18.8 240
4049 153 84.1 29 15.9 182
5059 82 69.2 44 30.8 143
6069 73 77.7 21 22.3 94
7079 46 73.0 17 27.0 63
>80 27 54.0 23 46.0 50
Ethnia 0.034
Caucasian 763 79.3 199 20.7 962
Mixed 203 86.4 32 13.6 235
Black 50 80.6 12 19.4 62
Oriental 15 93.8 1 6.3 16
Causes <0.001
Running over 233 74.2 81 25.8 314
Car accident 116 89.2 14 10.8 130
Motorcycle accident 213 89.5 25 10.5 238
Bike accident 31 86.1 5 13.9 36
Fall from a height 220 79.7 56 20.3 276
Fall from own height 93 75.0 31 25.0 124
Others 125 79.6 32 20.4 157
Total 1031 80.9 244 19.1 1275

Chi-square test.

Table 2
Relationship between extracranial injuries, Marshall classification, midline deviation anisocoria and mortality.

Variable Death Total p


No Yes
Spinal fracture <0.001
No 870 78.7 236 21.3 11.6
Yes 161 95.3 8 4.7 169
Thoracic trauma 0.997
No 883 80.9 209 19.1 1092
Yes 148 80.9 35 19.1 183
Abdominal trauma 0.570
No 957 80.7 229 19.3 1186
Yes 74 83.1 15 16.9 89
Pelvic trauma 0.101
No 970 80.4 236 19.6 1206
Yes 61 88.4 8 11.6 69
Upper extremities fracture 0.218
No 943 80.5 229 19.5 1172
Yes 88 85.4 15 14.6 103
Lower extremities fracture 0.170
No 930 80.4 227 19.6 11.57
Yes 101 85.6 17 14.4 118
Marshall classification <0.001#
Diffuse injury II 520 90.6 54 9.4 574
Diffuse injury III (swelling) 37 63.8 21 36.2 58
Diffuse injury IV (shift) 5 29.4 12 70.6 17
Evacuated mass lesion 463 77.0 138 23.0 601
Midline desviation <0.001#
00.5 896 82.6 189 17.4 1085
0.6 1 50 86.2 8 13.8 58
1.11.5 46 70.8 19 29.2 65
1.62.5 34 63.0 20 37.0 54
>2.5 5 38.5 8 61.5 13
Anisocoria <0.001
No 946 83.3 190 16.7 1136
Yes 85 61.2 54 38.8 139
Total 1031 80.9 244 19.1 1275t

Chi-square test.
# Likelihood ratio test.
J.R. Junior et al. / Journal of Clinical Neuroscience 42 (2017) 122128 125

Table 3
Relationship between Glasgow Coma Scale and Age on mortality.

Variable Death Average SD Median Minimum Maximum N p


GCS No 10.18 3.81 10 3 15 1031 <0.001
Yes 7.46 3.61 6.5 3 15 244
Total 9.66 3.92 9 3 15 1275
Age No 36.78 18.78 32 1 92 1031 <0.001
Yes 45.38 21.45 43 6 98 244
Total 38.43 19.60 34 1 98 1275

t-Student test.
Bold values represent p < 0.05.

Table 4
Variables included in the model: adjusted weigh of each independent variable in the outcome.

OR IC 95% Coefficient p
Age (years) 1.03 1.021.03 0.03 0.0001
Glasgow Coma Scale (315) 0.81 0.770.84 0.2 0.0001
Marshall Scale 4, 5, 6 (0: no; 1: yes) 2.07 1.502.85 0.72 0.0001
Anysochoria (0: no; 1: yes) 1.79 1.172.72 0.58 0.006

higher income countries, which compromises its worldwide


generalization.
In the CRASH prediction model, the authors included patients
with milder injuries (ECG > 12). In this study, the variables associ-
ated with poor prognosis were: lower Glasgow Coma Score, older
age, major extracranial injury and absent pupil reactivity. Some
CT findings were associated with poorer outcomes also. These
comprise petechial hemorrhages, effacement of the third ventricle
or basal cisterns, traumatic subarachnoid hemorrhage, midline
shift, or non-evacuated hematoma [19]. In the CRASH cohort a high
proportion of patients from low- and middle-income countries
were included.
The main criticism to these models is that they include different
populations, in different regions that had been submitted to differ-
ent treatment protocols. They also included patients with milder
traumatic brain injuries. In our series we included only cases with
tomographic abnormalities and excluded Marshall I patients. The
Fig. 1. An example of mortality probability calculator. As shown, the death main reason is that many Marshall I patients were victims of
probability of a patient aged 65, with a GCS of 7, a MS of 4, with no anysochoria
anoxic encephalopathy, status epilepticus, concussion and poison-
is estimated to be around 47.7%.
ing among others factors.

of injury, patients comorbidities, pre-hospitalar assistance and 4.1. Glasgow Coma Scale
several others factors that should be analyzed jointly in a prognos-
tic model [68,11,13,1618]. Since the pioneering study of Jennett Glasgow Coma Scale (GCS) is applied to assess the clinical
et al. [12], for the prediction of functional outcome and mortality severity of intracranial injury [20]. The verification of GCS should
after severe TBI, manifold prognostic models have been developed. be performed in the first hour, usually after respiratory and hemo-
The main objective of these models is to afford an unbiased mea- dynamic stabilization, although, it could be confounded by medical
sure of the precocious individual outcome in the hospital admis- sedation, paralysis or intoxication [21]. In our study, GCS assess-
sion. Despite great prediction models (actually there are more ment was carried out after initial stabilization, in patients without
than one hundred prediction models for neurotrauma patients sedation. When sedation was necessary for transportation, the pre-
published in the literature) only few have been developed in large hospitalar team GCS scores were used for analysis. The average
groups of TBI [9]. The flaws of small samples studies are related to Glasgow Coma Scale of patients who die was statistically lower
poor statistical power what prevents accurate determination of than the patients who survived (p < 0.001). The 1-point decrease
outcomes. in Glasgow scale lead to a 19% increase in the chance of mortality
Recently, based on large and generalizable populations, the of the patient. Confirming our data there are many studies demon-
International Mission for Prognosis and Analysis of Clinical Trials strating a relation between low score on GCS and poorer outcomes
in TBI (IMPACT) [16] and the Medical Research Council Corticos- [21,22]. According to Narayan et al. the GCS score (the sum score
teroid Randomization After Significant Head Injury (CRASH) [19] and the motor component score) has been shown to have a signif-
study groups were included to formulate the prediction models, icant correlation with outcome following severe TBI. The predictive
which are currently used. value for poor outcome in patients with initial GCS 68 was 26%
In the IMPACT model, which lay emphasis on in severe TBI and for patients with GCS 53 was 77% [23]. Another study by Gale
patients, the data were collected from eight clinical trials and three et al. found that the mortality rate was 88% for those with a GCS
observational series. The Glasgow Coma Scale motor score, age and score of 35 [24].
pupillary reactivity were the best predictors of mortality and 6- One of the difficulties of GCS is the poor interobserver reliability
month outcome [16]. However, the patients monitored were from and imprecise assignment of score in prehospital and emergency
126 J.R. Junior et al. / Journal of Clinical Neuroscience 42 (2017) 122128

departments [7,25,26]. A simple measure of neurological examina- which is recognized as prognostic factors were inadequate
tion have been considered by some authors According to Healey markers of extracranial injuries. In this context, the Injury Severity
et al. the motor component (from Glasgow Coma Scale) preserves Score (ISS) is more appropriate for extracranial injury assessment
the predictive values of the complete GCS [26]. The simplified [35]. In our paper the presence of extracranial injury, not assessed
motor score (SMS) which is a three-point measure based upon by ISS, was associated with poorer outcome in patients with low
the GCS motor response have been studied also [27]. According Glasgow Coma Scale in the univariate analysis and the tomo-
to Caterino et al. SMS has the same predictive value for neurolog- graphic aspects were not significant in the first moment. This
ical outcomes as the entire GCS [25]. However, in a recent meta- adventitious association was observed only in patients with low
analysis that compared the complete GCS and simplified motor GCS, it was not proper to include in our prognostic model, since
score, it was observed that in the prediction of outcome both scales we have included patients with higher GCS also.
were similar. However, GCS better predicted the mortality [28]. In
our study, despite of criticisms related to interrater reliability we
do not consider entire GCS as a complex scale since it is used 4.4. Age
worldwide.
According to Chesnut et al., age is a stronger predictor of func-
4.2. Anisocoria tional outcome and mortality in TBI [30]. Many studies demon-
strated that older age is associated with poorer outcome and the
Usually pupillary diameter and reactivity is not influenced from threshold values varying between 30 and 60 years. The chance of
sedation or paralysis. It is well recognized that pupillary reactivity an adverse outcome increases with patient age in a stepwise man-
suggest brainstem lesion or compression and are related with ner, especially after the 60 years [7,17,18]. According to Ono et al.
adverse outcome [29]. In our series, the presence of anisocoria [14], the threshold of 40 years was used as distinction between
was related to 38% of mortality. According to Balestreri et al. poor and good outcomes when the intracranial injuries were sep-
patients who present with anisocoria have 67% higher chance of arated in epidural, subdural and parenchymal haematomas. One
mortality. The presence of anisocoria may be reliable to predict of the criticisms of this paper, made by the own authors, was the
mortality and the presence of intracranial lesion [22]. As demon- small sample size (272 patients). In the same way, in pediatric
strated by Chesnut et al. 43% of patients with pupillary asymme- patients (<15 years) with GCS = 3 and bilaterally fixed and dilated
try > 3 mm had intracranial mass lesions [30]. Recently Sobuwa pupils, the mortality is lower and functional outcome is better than
et al. developed a predictive model for severe TBI and pupil reactiv- adults with same neurological status [14]. Luerssen et al. [38] pub-
ity was independently significant predictor of outcome. In their lished a relevant study that compared the impact of age on patients
study, the odds ratio for good outcome in cases of bilateral pupil- with severe TBI and pediatric patients had lower mortality when
lary reactivity was 4.40 [31]. In a retrospective cohort of 24.115 compared to adults (28% versus 47%).
patients the best accuracy for outcome prediction was found for The observation that poorer survival was associated with
GCS motor component and pupil reactivity. When both (pupil reac- increasing age beyond a certain threshold has been published for
tivity and GCS motor component) were combined the predictive age cut off points of either 40 or 55 years, and requires further
accuracy of GCS alone was outmatched [32]. Majdan et al. observed interpretation [7]. Our results were quite different since our higher
the same results, in which 445 patients were analyzed and the GCS mortality occurred between 5059 years and after 80 years old.
motor score and pupillary reactivity obtained the best prognostic There were no positive linear correlation between age and mortal-
value to predict 6-month mortality in patients with severe TBI. In ity. However young people (<20 years), who suffered TBI, the mor-
pediatric patients, non-reactive pupils are also related to higher tality reached approximately 10%.
mortality but in lower percentage than adults [33]. In our study
the patients under 20 were not analyzed as a separate group.

4.3. Extracranial injury 4.5. Gender

The significance of extracranial injury in TBIs patients is contro- The association of gender and outcome is described in many
versial. There were studies which demonstrated that the presence studies, however there are discrepancies between them [7,19,39].
of major extracranial injuries was associated with poorer outcome, It should be observed that men are more likely to have severe
however in others, the outcome depends on the severity of the pri- TBI in road traffic accidents and assaults. Interestingly there is evi-
mary cerebral damage and is not worsened by the presence of dence that women who survived after severe TBI have poorer qual-
extracranial injuries [34,35]. According to Van Leeuwen et al. the ity of life and worse functional outcomes when compared with
prognostic effect of extracranial injury relates to brain injury sever- men [7]. Despite this, gender has not been well established as a
ity, and the larger effects are observed milder TBI patients [36]. prognostic factor [7,39]. In our series we did not find any associa-
Confirming this data in the POCON [37] study (507 patients), tion between sex and outcome (p = 0.65).
Lingsma et al. [38] demonstrated that extracranial injury did not
added prognostic value in the general patient population. Never-
theless, in patients with less severe intracranial injury (Marshall 4.6. Ethnics
CT Classification I or II or higher GCS), the extracranial injury adds
some prognostic value. So, in severe TBI patients the brain injury Ethnics were statistically associated with mortality, 26% of Cau-
itself was responsible for poor prognosis and in less severe TBI casian people died versus 16% of other ethnias (p = 0.03). It was not
the presence of extracranial injury was the fundamental cause of included in our prognostic model since is difficult to define what is
poor outcome [38]. This has been observed when patients classi- a particular ethnia in our series because of the mixed races
fied as Marshall 4, 5 or 6 were diagnosed with more spinal frac- observed. According to Sorani et al. [40] the ethnic origin has an
tures than patients classified as Marshall 2 or 3. In this situation association with outcome after TBI. They demonstrated that black
probably the mortality was related to the brain injury and the people have a worse outcome than Caucasian or Asian people. The
spinal fracture was not a protective factor. In a study that reasons for this relation can be speculative and may include access
enrolled 683 patients, hypotension and hypoxemia on admission to emergency and rehabilitation care.
J.R. Junior et al. / Journal of Clinical Neuroscience 42 (2017) 122128 127

4.7. Marshall classification individually. It can therefore be included in decision algorithm to


the best treatment for each patient.
The classification of head injury based on CT findings was intro-
duced in 1991. According to Marshall et al. [10], the presence or Author contribution
absence of a mass lesion, compression of basal cisterns and midline
shift were the three variables analyzed. Since then, several papers Jefferson Rosi Junior: literature search, study design, data col-
have analyzed their data according to this classification. According lection, data analysis, critical revision.
to Brain Trauma Foundation, the effacement of the cisterns and the Leonardo C. Welling: literature search, data analysis, data inter-
presence of subarachnoid hemorrhage were the robust CT predic- pretation, writing.
tors of outcome [2]. It should be emphasized that many studies Marcelo Schafranski: data analysis, data interpretation.
have focused on CT abnormalities and included broad categoriza- Lin Tchia Yeng: study design, data collection, and critical
tions. For example, in traumatic subarachnoid hemorrhage, most revision.
studies describe the presence or absence of this abnormality with- Rogrio Ruscito do Prado: study design, critical revision.
out differentiating the location or extension. In our paper any cat- Edwin Koterba: study design, critical revision.
egory of Marshall, except those classified as Marshall I, increases Almir Ferreira de Andrade: study design, critical revision.
the risk of death compared to patients who only shows foci of hem- Manoel Jacobsen Teixeira: study design, critical revision.
orrhage, with an increased chance of mortality in patients who Eberval Gadelha Figueiredo: study design, data analysis, data
were not operated on. Maybe, the combination of individual CT interpretation, writing, critical revision.
characteristics in a model, such as in Rotterdam CT score, would
provide better discrimination in outcomes.
Conflict of interest

4.8. Limitations of the study None.

One criticism of our series, which can be extended to IMPACT Disclosures of funding received
[16] and CRASH [19] models, is that secondary injuries (mainly rep-
resented by hypoxia, hyperglycemia, fever and coagulopathy) is None.
linked to worse outcome and has not been included in the analyi-
sis. On the other hand, treatments that are promptly implemented Acknowledgements
(represented as ICP monitoring, multimodal monitoring, surgical
intervention which includes also decompressive craniectomy) We would like to thank to all staff neurosurgeons from Hospital
have been associated with better outcome and lower mortality das Clnicas of University of Sao Paulo involved in the care of
[7]. It should be highlighted that the CRASH model overestimates patients.
mortality and poor outcome in patients who are submitted to neu-
romonitoring and management of elevated intracranial pressure
Appendix A. Supplementary data
after TBI. However, the CRASH prognostic model overrates the
probability of adverse outcome in patients who are submitted to
Supplementary data associated with this article can be found, in
decompressive craniectomy.
the online version, at http://dx.doi.org/10.1016/j.jocn.2017.03.012.
The application of models to earlier predicts individual outcome
and the use of prognostic models to define patient populations and
to compare different patient series should be distinguished. An References
additional study that compares our model with other more com-
[1] Langlois A, Rutland-Brown W, Wald MM. The epidemiology and impact of
plexes ones are underway. traumatic brain injury: a brief overview. J Head Trauma Rehabil 2006;21
The implantation of prognostic models in neurotraumatology is (5):3758.
a challenge. The clinicians should apply these models in general [2] Brain Trauma Foundation. American association of Neurological Surgeons,
Congress of Neurological Surgeons. Guidelines for the management of severe
clinical application with observation of their own limitations. They traumatic brain injury. J Neurotrauma 2007;24(Suppl. 1):S1106.
should consider that many studies of traumatic brain injury, espe- [3] Stein SC, Georgoff P, Meghan S, et al. 150 years of treating severe traumatic
cially prognostic models, which has so many variables because of brain injury: a systematic review of progress in mortality. J Neurotrauma
2010;27(7):134353.
the heterogeneity of traumatic brain injury per se and presence [4] Thornhill S, Teasdale GM, Murray GD, et al. Disability in young people and
of type II errors (i.e. failing to detect an effect that is present) is adults one year after head injury: prospective cohort study. BMJ 2000;320
very common. Besides research applications and methodological (7250):16315.
[5] Faul M, Coronado V. Epidemiology of traumatic brain injury. Hand Clin Neurol
limitations, our main objective was to improve the prognostic esti- 2015;127:313.
mate and improve quality of care. [6] Steyerberg EW, Mushkudiani N, Perel P, et al. Predicting outcome after
traumatic brain injury: development and international validation of prognostic
scores based on admission characteristics. PLoS Med 2008;5(8):125161.
[7] Lingsma HF, Roozenbeek B, Steyerberg EW, et al. Early prognosis in traumatic
5. Conclusion brain injury: from prophecies to predictions. Lancet Neurol 2010;9(5):54354.
[8] Perel P, Edwards P, Wentz R, et al. Systematic review of prognostic models in
traumatic brain injury. BMC Med Inform Decis Mak 2006;6:38.
Estimation of prognosis is the keystone of neurological surgery. [9] Lefering R, Paffrath T, Linker R, et al. German Society for Trauma Surgery. Head
In our study, age, presence of anisocoria, Marshall classification injury and outcome what influence do concomitant injuries have? J Trauma
and GCS were significantly related to poor outcome. On the other 2008;65(5):103643.
[10] Marshall LF, Marshall SB, Klauber MR, et al. The diagnosis of head injury
hand, extracranial injuries demonstrated no relevant influence on
requires a classification based on computed axial tomography. J Neurotrauma
mortality in a multivariate model. Our mathematical probability 1992;9(Suppl.):S28792.
model (USP Index) is a simple and effective predictor of the mortal- [11] Maas AI, Hukkelhoven CW, Marshall LF, et al. Prediction of outcome in
ity in patients with severe traumatic brain injury and can be easily traumatic brain injury with computed tomographic characteristics: a
comparison between the computed tomographic classification and
applied to the clinical decision-making process. The prediction rule combinations of computed tomographic predictors. Neurosurgery 2005;57
may never replace and only complement the clinical judgment (6):117382.
128 J.R. Junior et al. / Journal of Clinical Neuroscience 42 (2017) 122128

[12] Jennett B, Teasdale G, Braakman R, et al. Predicting outcome in individual [27] Healey C, Osler TM, Rogers FB, et al. Improving the Glasgow coma scale score:
patients after severe head injury. Lancet 1976;1(7968):10314. motor score alone is a better predictor. J Trauma 2003;54(4):6718.
[13] Andrews PJ, Sleeman DH, Statham PF, et al. Predicting recovery in patients [28] Gill M, Windemuth R, Steele R, et al. A comparison of the Glasgow coma scale
suffering from traumatic brain injury by using admission variables and score to simplified alternative scores for the prediction of traumatic brain
physiological data: a comparison between decision tree analysis and logistic injury outcomes. Ann Emerg Med 2005;45(1):3742.
regression. J Neurosurg 2002;97(2):32636. [29] Mauritz W, Leitgeb J, Wilbacher I, et al. Outcome of brain trauma patients who
[14] Ono J, Yamaura A, Kubota M, et al. Outcome prediction in severe head injury: have a Glasgow Coma Scale score of 3 and bilateral fixed and dilated pupils in
analyses of clinical prognostic factors. J Clin Neurosci 2001;8(2):1203. the field. Eur J Emerg Med. 2009;16(3):1538.
[15] Cremer OL, Moons KG, van Dijk GW, et al. Prognosis following severe head [30] Farahvar A, Gerber LM, Chiu YL, et al. Response to intracranial hypertension
injury: development and validation of a model for prediction of death, treatment as a predictor of death in patients with severe traumatic brain
disability, and functional recovery. J Trauma 2006;61(6):148491. injury. J Neurosurg. 2011;114(5):14718.
[16] Maas AI, Marmarou A, Murray GD, et al. Prognosis and clinical trial design in [31] Sobuwa S, Hartzenberg HB, Geduld H, et al. Predicting outcome in severe
traumatic brain injury: the IMPACT study. J Neurotrauma 2007;24(2):2328. traumatic brain injury using a simple prognostic model. S Afr Med J 2014;104
[17] Murray GD, Butcher I, McHugh GS, et al. Multivariable prognostic analysis in (7):4924.
traumatic brain injury: results from the IMPACT study. J Neurotrauma 2007;24 [32] Hoffmann M, Lefering R, Rueger JM, et al. Pupil evaluation in addition to
(2):32937. Glasgow Coma Scale components in prediction of traumatic brain injury and
[18] Bahloul M, Chelly H, Ben Hmida M, et al. Prognosis of traumatic head injury in mortality. Br J Surg 2012 Jan;99(Suppl. 1):12230.
South Tunisia: a multivariate analysis of 437 cases. J Trauma 2004;57 [33] Majdan M, Steyerberg EW, Nieboer D, et al. Glasgow Coma Scale motor score
(2):25561. and pupillary reaction to predict six-month mortality in patients with
[19] MRC Crash Trial Collaborators, Perel P, Arango M, et al. Predicting outcome traumatic brain injury: comparison of field and admission assessment. J
after traumatic brain injury: practical prognostic models based on large cohort Neurotrauma 2015 Jan;32(2):1018.
of international patients. BMJ 2008;336(7641):4259. [34] Holland MC, Mackersie RC, Morabito D, et al. The development of acute lung
[20] Teasdale G, Jennett B. Assessment of coma and impaired consciousness. A injury is associated with worse neurologic outcome in patients with severe
practical scale. Lancet 1974;2(7872):814. traumatic brain injury. J Trauma 2003;55(1):10611.
[21] Marmarou A, Lu J, Butcher I, et al. Prognostic value of the Glasgow Coma Scale [35] Ho KM, Burrel M, Rao S. Extracranial injuries are important in determining
and pupil reactivity in traumatic brain injury assessed pre-hospital and on mortality of neurotrauma. Crit Care Med 2010;38(7):15628.
enrollment: an IMPACT analysis. J Neurotrauma 2007;24(2):27080. [36] van Leeuwen N, Lingsma HF, Perel P, et al. Prognostic value of major
[22] Balestreri M, Czosnyka M, Chatfield DA, et al. Predictive value of Glasgow extracranial injury in traumatic brain injury: an individual patient data
Coma Scale after brain trauma: change in trend over the past ten years. J meta-analysis in 39,274 patients. Neurosurgery 2012;70(4):8118.
Neurol Neurosurg Psychiatry 2004 Jan;75(1):1612. [37] Andriessen TM, Horn J, Franschman G, et al. Epidemiology, severity
[23] Narayan RK, Greenberg RP, Miller JD, et al. Improved confidence of outcome classification, and outcome of moderate and severe traumatic brain injury: a
prediction in severe head injury. A comparative analysis of the clinical prospective multicenter study. J Neurotrauma 2011;28(10):201931.
examination, multimodality evoked potentials, CT scanning, and intracranial [38] Lingsma H, Andriessen TM, Haitsma I, et al. Prognosis in moderate and severe
pressure. J Neurosurg 1981;54(6):75162. traumatic brain injury: external validation of the IMPACT models and the role
[24] Gale JL, Dikmen S, Wyler A, et al. Head Injury in the Pacific Northwest. of extracranial injuries. J Trauma Acute Care Surg 2013;74(2):63946.
Neurosurgery 1983;12(5):48791. [39] Mushkudiani NA, Engel DC, Steyerberg EW, et al. Prognostic value of
[25] Emami P, Czorlich P, Frtizsche FS, et al. Impact of Glasgow Coma Scale score demographic characteristics in traumatic brain injury: results from the
and pupil parameters on mortality rate and outcome in pediatric and adult IMPACT study. J Neurotrauma 2007;24(2):25969.
severe traumatic brain injury: a retrospective multicenter cohort study. J [40] Sorani MD, Lee M, Kim H, et al. Race/ethnicity and outcome after traumatic
Neurosurg 2016;1:18 [Epub ahead of print]. brain injury at a single, diverse center. J Trauma 2009;67(1):7580.
[26] Caterino JM, Raubenolt A. The prehospital simplified motor score is accurate as
the prehospital Glasgow coma scale: analysis of a statewide trauma registry.
Emerg Med J 2012;29(6):4926.

Anda mungkin juga menyukai