Content uploaded by Humberto Kluge Schroeder
Author content
All content in this area was uploaded by Humberto Kluge Schroeder on May 08, 2020
Content may be subject to copyright.
ORIGINAL ARTICLE
Mortality in Severe Traumatic Brain Injury: A Multivariated
Analysis of 748 Brazilian Patients From Floriano´polis City
Evandro Tostes Martins, MD, Marcelo Neves Linhares, MD, PhD, Daniel Santos Sousa, MD,
Humberto Kruger Schroeder, MD, Jardel Meinerz, MD, Luís Antoˆnio Rigo, MD, Melina More´ Bertotti,
Jackson Gullo, Alexandre Hohl, MD, Felipe Dal-Pizzol, MD, PhD, and Roger Walz, MD, PhD
Background: Traumatic brain injury (TBI) is a major cause of incapacity and
mortality worldwide, with most of the burden occurring in low-income and
middle-income countries. A number of clinical, demographic, and neurosur-
gical variables of patients with TBI were associated with their outcome.
Methods: We investigated the mortality of Brazilian patients with severe
TBI at the time of discharge, using a multiple logistic regression analysis.
Clinical, demographic, radiologic, and neurosurgical variables, and mortality
at time of discharge of all consecutive patients (n ⫽748) with severe TBI
(admission Glasgow scale ⱕ8) treated in our intensive care unit were
analyzed. The variables were collected in a prospective manner between January
1994 and December 2003.
Results: Eighty-four percent (n ⫽631) of the patients were men. The mean age
was 34.8 (⫾16.3) years and the mortality was 33.3%. After the multiple logistic
regression, the adjusted odds ratio (OR) for death was higher in older (⬎60
years) than younger (up to 30 years) patients (OR ⫽2.51, 95% confidence interval
[CI] 1.31– 4.79, p⫽0.006). The mortality was also associated with sub-
arachnoid hemorrhage (OR ⫽1.86, 95% CI ⫽1.23–2.81, p⫽0.003) on
computed tomography (CT) scan; admission Glasgow Scale of 3 or 4 in
comparison to 7 or 8 (OR ⫽3.97, 95% CI ⫽2.49 – 6.31, p⬍0.001);
bilateral midryasis (OR ⫽11.52, 95% CI ⫽5.56 –23.87, p⬍0.0001), or
anisocoria (OR ⫽2.65, 95% CI ⫽1.69 – 4.17, p⬍0.0001) in comparison to
isocoric pupils. There was a trend for higher mortality in patients with type
III injury on the Marshall classification of CT (OR ⫽3.63, 95% CI ⫽
0.84 –15.76, p⫽0.08) than in patients with normal CT. Patients without
thoracic trauma disclose higher mortality than patients with associated
thoracic trauma do (OR ⫽2.02, 95% CI ⫽1.19 –3.41, p⫽0.009). The final
model presented disclosed 76.9% of overall correct prediction with the
survival and death predicted at 87.6% and 55.6%, respectively.
Conclusion: Age, CT findings, Glasgow coma scale, pupil examination, and
the presence of thoracic trauma at admission were independently associated
with mortality at the time of discharge in Brazilian patients with severe TBI.
Key Words: Severe traumatic brain injury, Prognostic models, Mortality,
Logistic regression.
(J Trauma. 2009;67: 85–90)
Traumatic brain injury (TBI) is a critical public health
problem that deserves the attention of the world’s health
community. TBI is the leading worldwide cause of morbidity
and mortality of young people.
1
TBI is highly frequent in
low-income and middle-income countries, including Brazil.
2
There are few studies concerning the statistics of TBI in
Brazil. Koizumi et al.
3
estimated that in 1997 the mortality
related to TBI in Sao Paulo City was between 26.2 and 39.3
per 100,000 inhabitants.
Diagnostic and therapeutic decisions are based on the
patient’s prognosis. Prognostic models are statistical models that
combine two or more variables of patient’s data to predict
clinical outcome. Recently, the International Mission on Prog-
nosis and Analysis of Clinical Trials in TBI reported a series
of articles derived from a cohort patients drawn from eight
randomized controlled trials and three observational stud-
ies.
4–8
The International Mission on Prognosis and Analysis
of Clinical Trials in TBI study showed that age, Glasgow
coma scale (GCS), pupil response, and computed tomography
(CT) characteristics are the most powerful independent prognos-
tic variables for the outcome measured 6 months after injury.
7
Perel et al.
2
analyzed 31 articles published since 1990
that gave an overall prognostic estimation for patients with
TBI combining the predictive information using logistic re-
gression. Similar analyses were performed by Mushkudiani et
al.
9
They suggested that studies of prognostic models in TBI
need better description of the measurement and validity of
variables included in the model, large sample sizes, adequate
handling of continuous variables and missing data, assessment
of interaction in the multivariable analysis, clear description of
the calculation of the prognostic score, external validation, and
adequate report of model performance measures.
2,9
The authors
also point out the necessity of studies including populations from
low-income and middle-income countries, where most of the
burden of TBI occurs.
2
There is no prospective study in Brazil concerning
prognostic estimation of patients with TBI. Here, we inves-
Submitted for publication March 31, 2008.
Accepted for publication July 22, 2008.
Copyright © 2009 by Lippincott Williams & Wilkins
From the Unidade de Terapia Intensiva (E.T.M., J.M., L.A.R.), Hospital Governador
Celso Ramos; Departamento de Clínica Me´dica (E.T.M., M.N.L., H.K.S.,
M.M.B., J.G., A.H., R.W.), Nu´cleo de Pesquisas em Neurologia Experimental e
Clínica (NUPNEC-UFSC), Universidade Federal de Santa Catarina (UFSC);
Servic¸o de Neurocirurgia (M.N.L., D.S.S.), Hospital Governador Celso Ramos;
Departamento de Cirurgia (M.N.L.), UFSC, Floriano´polis; and Laborato´rio de
Laborato´rio de Fisiopatologia Experimental (F.D.P.), Universidade do Extremo
Sul Catarinense (UNESC), Criciu´ma, SC, Brasil.
Supported by CNPq (Brazilian Council for Scientific and Technologic Develop-
ment, Brazil), FAPESC (Foundation for Scientific Research and Technology
of Santa Catarina State). The FEESC (Foundation for Education and Engi-
neering of Santa Catarina) supported the technician work of Shumaia Deguer
from The Cyclops Project, UFSC.
Address for reprints: Roger Walz, MD, PhD, NUPNEC, Departamento de Clínica
Me´dica, Hospital Universita´rio, 3 andar, Universidade Federal de Santa
Catarina—UFSC, Campus Universita´rio, Trindade, Floriano´polis, 88.040-
970, SC, Brasil; email: rogerwalz@hotmail.com.
DOI: 10.1097/TA.0b013e318187acee
The Journal of TRAUMA
®
Injury, Infection, and Critical Care • Volume 67, Number 1, July 2009 85
tigated the mortality of patients with severe TBI at the time of
discharge using multiple logistic regression analysis.
PATIENTS AND METHODS
We included 748 consecutive patients with severe TBI
admitted to the intensive care unit of the Hospital Governador
Celso Ramos between January 1, 1994 and December 31,
2003. This is a public reference hospital for TBI covering a
population of approximately 1 million, in the metropolitan
area of Florianopolis city. The neurosurgical team and most
staffs of the intensive care unit were essentially the same
during all the time of study.
Inclusion criteria were GCS score 8 or lower after acute
neurosurgical resuscitation, or deterioration to that level
within 48 hours of impact and admission to the intensive care
unit. Victims of gunshot injury and patients who evolved to
brain death before 24 hours of admission were excluded.
Death at the time of discharge was the primary end-
point. Secondary endpoint was not evaluated. The indepen-
dent variables analyzed were age, sex, cause of TBI, admis-
sion GCS, CT findings, presence of associated trauma (face,
spinal, thorax, abdomen or limbs), admission glucose levels,
and pupil examination. Face trauma included isolated or
combined skin lacerations, visible hematoma, and eye le-
sions. Patients with limb fractures or major articulation le-
sions of limbs were classified as having limb trauma. Verte-
bral fractures with or without spinal cord-associated injury
were classified as spinal trauma. Isolated or combined lung
contusion, pneumothorax, or hemothorax were considered
thoracic trauma. Isolated or combined pneumoperitoneum,
hemoperitoneum, or visceral lesions were considered as hav-
ing abdominal trauma. Brain CT findings were classified in
the six categories according to Marshall et al.
10,11
Patients
with isolated brain stem lesion on CT were classified as
another category. Patients showing brain stem lesions asso-
ciated with encephalic mass effect lesion (Marshall CT clas-
sification other than type I or II) were classified according to
the Marshall classification only. The presence of traumatic
sub-arachnoid hemorrhage was another independent variable.
CT analysis was performed by one of researchers (ETM) and
confirmed by another one, when necessary. Most of these
isolated or combined variables have been proved to be asso-
ciated with TBI outcome in the previous earlier studies and
were included.
1,2,12–15
The year of admission was included as an independent
variable to control possible effects of the learning curve, CT
resolution, and changes in the protocol treatment, occurring
during the study period that could modify the outcome. The
clinical, demographic, radiologic, and neurosurgical variables
were prospectively collected using the research protocol by
one investigator (ETM). In 63 cases, the admission serum
glucose levels were obtained retrospectively by two investi-
gators (LAR and JM). Age was not confirmed in six patients.
Admission levels of glucose were not measured in six pa-
tients. CT Marshall Scale was not evaluated in one patient.
Six patients had ocular trauma and pupils were not adequately
evaluate.
Statistical Analysis
Univariate analysis was performed to investigate the
association between the independent variables and the mor-
tality at time of discharge. Continuous variables were ana-
lyzed by Student’s ttest. Categorical variables were analyzed
by binary logistic regression. The continuous variables were
also categorized to be analyzed by binary logistic regression.
The magnitude of association between death and the inde-
pendent variables was measured by the odds ratio (OR) and
respective 95% confidence interval (CI) estimated by uncon-
ditional logistic regression.
To identify variables that were independently associated
with death we performed a multiple logistic regression using the
forward conditional method. The probability of stepwise entry
was 0.05 and removal was 0.10. The classification cutoff was 0.5
with maximum iterations of 20. “p” levels lower than 0.01 were
considered significant. This more stringent criterion for the “p”
level of significance was based on the Bonferroni adjustment for
multiple tests.
16
Statistical analysis was performed using the
SPSS program 10.0 (Chicago, IL).
RESULTS
Eighty-four percent (n ⫽631) of patients were men.
The mean age of patients was 34.8 (⫾16.3) years and the
overall mortality was 33.3%. The causes of TBI were road
accident (30.1%), automobile accident (23%), fall (12.8%),
motorcycle accident (24.3%), aggression (3.7%), bicycle ac-
cident (3.2%), others (2.8%).
Results of the univariate prognostic analysis are shown
in Table 1. There were no associations between death at time
of discharge and associated spine injury, abdominal trauma,
or gender. There was a trend for lower mortality of patients
with TBI related to automobile accidents. The mortality was
associated with old age, higher glucose levels, Marshall CT
classification of type III injury, nonevacuated lesions or
presence of traumatic sub-arachnoid hemorrhage on CT,
absence of face or thoracic trauma, lower scores in the
admission GCS, and anisocoric or midriatic pupils. The
mortality was higher in the years between 1994 and 1997 than
between 2001 and 2003.
Results of multiple logistic regressions are shown in
Table 2. The adjusted OR for death was higher in older
patients (⬎60 years) in comparison to younger patients (up to
30 years) (OR ⫽2.51, 95% CI ⫽1.31– 4.79, p⫽0.006). The
mortality was also associated with sub-arachnoid hemorrhage
(OR ⫽1.86, 95% CI ⫽1.23–2.81, p⫽0.003) on CT scan;
Glasgow Scale of 3 or 4 in comparison to 7 or 8 (OR ⫽3.97,
95% CI ⫽2.49 – 6.31, p⬍0.001); pupil examination with
bilateral midryasis (OR ⫽11.52, 95% CI ⫽5.56 –23.87, p⬍
0.0001) or anisocoric (OR ⫽2.65, 95% CI ⫽1.69 – 4.17, p⬍
0.0001) in comparison to isocoric pupils. The mortality was
also associated with year of attendance between 1994 and
1995 (OR ⫽3.17, 95% CI ⫽1.71–5.88, p⬍0.0001) or 1996
and 1997 (OR ⫽2.17, 95% CI ⫽1.16 – 4.04, p⫽0.01) in
comparison to 2002 to 2003. There was a trend for higher
mortality in patients with type III injury on the Marshall CT
classification (OR ⫽3.63, 95% CI ⫽0.84 –15.76, p⫽0.08)
than in patients with normal CT. Patients without thoracic
Martins et al. The Journal of TRAUMA
®
Injury, Infection, and Critical Care • Volume 67, Number 1, July 2009
© 2009 Lippincott Williams & Wilkins86
TABLE 1. Death of Patients With Severe Head Injury at Time of Discharge According to Clinical, Demographic, Radiologic,
and Neurosurgical Variables
Variables
All Patients,
Nⴝ748 (%)
Outcome
Crude OR (95% CI) pSurvivors, n ⴝ499 (%) Death, n ⴝ249 (%)
Gender
Male 631 (84.4) 428 (67.8) 203 (32.2) 1.0
Female 117 (15.6) 71 (60.7) 46 (39.3) 1.36 (0.91–2.05) 0.13
Age (yr)*
Mean (⫾SD) 34.8 (⫾16.3) 33.04 (⫾15.5) 36.16 (17.6) NA 0.01
12–30 391 (52.7) 270 (69.0) 121 (31.0) 1.0
31–45 187 (25.2) 124 (66.3) 63 (33.7) 1.13 (0.78–1.64) 0.51
46–60 94 (12.7) 62 (65.9) 32 (34.1) 1.15 (0.71–1.86) 0.56
Higher than 60 70 (9.4) 39 (55.7) 31 (44.3) 1.77 (1.06–2.98) 0.03
Glucose
†
Mean (⫾SD) 163.9 (65.5) 156.8 (55.9) 178.2 (79.5) ⬍0.0001
61–110 93 (12.4) 65 (69.9) 28 (30.1) 1.0
111–220 516 (69) 357 (69.2) 159 (30.8) 1.03 (0.64–1.67) 0.89
221–300 96 (12.8) 55 (57.3) 41 (42.7) 1.73 (0.95–3.15) 0.07
⬎300 27 (3.6) 11 (40.7) 16 (59.3) 3.38 (1.39–8.19) 0.007
⬍60 10 (1.3) 6 (60) 4 (40) 1.55 (0.40–5.91) 0.51
Year of the attendance
2003–2002 132 (17.7) 101 (76.6) 31 (23.3) 1.0
2001–2000 142 (19.0) 100 (70.4) 42 (29.6) 1.38 (0.80–2.37) 0.24
1999–1998 133 (17.8) 94 (70.7) 39 (29.3) 1.36 (0.78–2.36) 0.27
1997–1996 162 (21.7) 105 (64.8) 57 (35.2) 1.78 (1.07–2.99) 0.03
1995–1994 178 (23.8) 98 (55.1) 80 (44.9) 2.68 (1.63–4.42) 0.001
Cause of TBI
Road accident 225 (30.1) 143 (63.6) 82 (36.4) 1.0
Automobile accident 172 (23.0) 123 (71.5) 49 (28.5) 0.69 (0.45–1.07) 0.09
Falls 96 (12.8) 52 (54.2) 44 (45.8) 1.47 (0.91–2.40) 0.12
Motorcycle accident 182 (24.3) 128 (70.3) 54 (29.7) 0.74 (0.48–1.12) 0.15
Agression 28 (3.7) 21 (75.0) 7 (25.0) 0.58 (0.24–1.43) 0.24
Bicycle accident 24 (3.2) 16 (66.7) 8 (33.3) 0.87 (0.36–2.13) 0.76
Others 21 (2.8) 16 (76.2) 5 (23.8) 0.55 (0.19–1.54) 0.25
Marshall CT classification‡
Type I injury 22 (2.9) 19 (86.4) 3 (13.6) 1.0
Type II injury 175 (23.4) 145 (82.9) 30 (17.1) 1.31 (0.36–4.71) 0.68
Type III injury 172 (23.0) 107 (62.2) 65 (37.8) 3.84 (1.10–13.51) 0.03
Type IV injury 58 (7.8) 19 (32.8) 39 (67.2) 13.0 (3.42–49.42) ⬍0.001
Evacuated mass lesion 240 (32.1) 154 (64.2) 86 (35.8) 3.54 (1.02–12.29) 0.05
Nonevacuated lesion 30 (4.0) 14 (46.7) 16 (53.3) 7.24 (1.76–29.75) 0.006
Brainstem lesion 50 (6.7) 41 (82.0) 9 (18.0) 1.39 (0.34–5.73) 0.65
Sub-arachnoids hemorrhage
No 481 (64.3) 340 (70.7) 141 (29.3) 1.0
Yes 267 (35.7) 159 (59.6) 108 (40.4) 1.64 (1.20–2.24) 0.002
Associated trauma
Yes 323 (43.2) 234 (72.4) 89 (27.6) 1.0
No 425 (56.8) 265 (62.4) 160 (37.6) 1.59 (1.16–2.17) 0.004
Associated trauma type
Face
Yes 108 (14.4) 82 (75.9) 26 (24.1) 1.0
No 640 (85.6) 417 (65.2) 223 (34.8) 1.69 (1.05–2.70) 0.03
Cervical spine
Yes 27 (3.6) 20 (74.1) 7 (25.9) 1.0
No 721 (96.4) 479 (66.4) 242 (33.6) 1.44 (0.60–3.46) 0.41
The Journal of TRAUMA
®
Injury, Infection, and Critical Care • Volume 67, Number 1, July 2009 Mortality in Severe Traumatic Brain Injury
© 2009 Lippincott Williams & Wilkins 87
trauma disclosed higher mortality than patients with associ-
ated thoracic trauma did (OR ⫽2.02, 95% CI ⫽1.19 –3.41,
p⫽0.009). The final model presented in Table 2 disclosed
76.9% of overall correct prediction. Survival and death were
correctly predicted in 87.6% and 55.6%, respectively.
DISCUSSION
The present work demonstrates that old age, CT find-
ings, GCS, pupil examination, and presence of thoracic
trauma at admission of patients with severe TBI are indepen-
dently associated with mortality at time of discharge. There
are few previous prospective studies with similar large sam-
ple of cases in the medical literature, and this is the first
prospective study of patients with TBI in Brazil. The sim-
plicity and objectiveness of the applied research protocol, the
prospective data collection by one single researcher, and the
maintenance of almost the same medical staff involved with
the patients’ care support the internal validity of the study.
Replication of this model in other populations is desirable to
determine its external validity. Although some differences in
race distribution can be observed among different Brazilian
regions, we believe, considering the socio-economic status,
gender, age, and the mechanisms of TBI, that our findings
probably disclose an external validity for other populations in
Brazil. In our sample, there are no victims of gunshot injury,
which are important causes of TBI in the favelas of big
Brazilian cities. We believe this is an important variable to be
investigated in further studies with other specific populations.
The final model presented here is user friendly; how-
ever, its overall prediction capacity needs to be improved.
The inclusion of other clinical or laboratorial variables, such
as the presence and treatment of hemodynamic instability,
hypoxia, anemia, fever, seizures, infections, elevated intra-
cranial pressure, renal, hepatic, or respiratory failure, would
certainly improve the accuracy of our model. These abnor-
malities can occur at admission or even during patient care. It
has been recently identified that much of the brain damage
after TBI develops over time with the primary injury initiat-
ing a secondary injury cascade made up of deleterious patho-
physiological and biochemical reactions.
1,17
Some neuro-
chemical signals related to TBI may activate both
neuroprotective and neurotoxic pathways,
18,19
depending on
the final effects on the complex biochemical network modu-
lation at the cellular level.
20,21
Identification of these bio-
chemical markers and their association with clinical, labora-
torial, radiologic, and neurosurgical variables are an
important scientific challenge in the identification of prog-
nostic markers and possible therapeutic targets in TBI.
Several patients who survived severe or even moderate
TBI evolve with significant impairment in quality of life. The
TABLE 1. (continued)
Variables
All Patients,
Nⴝ748 (%)
Outcome
Crude OR (95% CI) pSurvivors, n ⴝ499 (%) Death, n ⴝ249 (%)
Dorsal-lombar spine
Yes 7 (0.9) 4 (57.1) 3 (42.9) 1.0
No 741 (99.1) 495 (66.8) 246 (33.2) 0.66 (0.47–2.98) 0.59
Thorax
Yes 141 (18.9) 109 (77.3) 32 (22.7) 1.0
No 607 (81.1) 390 (64.3) 217 (35.7) 1.89 (1.24–2.91) 0.003
Abdominal
No 678 (90.6) 456 (67.3) 222 (32.7) 1.0
Yes 70 (9.4) 43 (61.4) 27 (38.6) 1.29 (0.78–2.14) 0.32
Limbs
Yes 204 (27.3) 146 (71.6) 58 (28.4) 1.0
No 544 (72.7) 353 (64.9) 191 (35.1) 1.36 (0.96–1.94) 0.08
Others
No 740 (98.9) 495 (66.9) 245 (33.1) 1.0
Yes 8 (1.1) 4 (50.0) 4 (50.0) 2.02 (0.50–8.15) 0.32
Glasgow Coma Scale
7 or 8 311 (41.6) 255 (82.0) 56 (18.0) 1.0
5 or 6 192 (25.7) 135 (70.3) 57 (29.7) 1.93 (1.26–2.94) ⬍0.0001
3 or 4 243 (32.5) 108 (44.4) 135 (55.6) 5.69 (3.87–8.36) ⬍0.0001
Pupils§
Isocorics 283 (37.8) 239 (84.4) 44 (15.6) 1.0
Miotics 30 (4.0) 23 (76.7) 7 (23.3) 2.00 (0.79–5.04) 0.27
Anisocorics 347 (46.4) 216 (62.2) 131 (37.8) 3.29 (2.23–4.86) ⬍0.00001
Midriatics 83 (11.1) 17 (20.5) 66 (79.5) 21.09 (11.32–39.30) ⬍0.00001
* Age was not confirmed in six cases.
†
Glucose were not measured in the admission in six patients.
‡
Tomographic Marshall scale was not evaluated in one patient.
§
Six patients had ocular trauma and pupils were not adequately evaluated.
Martins et al. The Journal of TRAUMA
®
Injury, Infection, and Critical Care • Volume 67, Number 1, July 2009
© 2009 Lippincott Williams & Wilkins88
development of prognostic models to predict the level of quality
of life of patients with different levels of TBI, including cases of
severe, moderate, or even slight injury, are clinically relevant.
To our knowledge, there is no study in literature about prognos-
tic models for predicting quality of life, neuropsychological, or
psychiatric consequences after TBI.
In the present study, the initial association between
glucose levels at admission and mortality was not confirmed
after the multiple logistic regression analysis, indicating that
this was a confound bias because of imbalances of other
variables analyzed. We cannot exclude that modifications in
glucose levels during patient care might be associated with
the prognosis. The association between thoracic trauma and
lower mortality was also an intriguing finding and it remains
to be confirmed in other populations. Patients with associated
thoracic trauma could have a dissipated impact all over the
body and, consequently, a lower energy impact on the head.
However, if lower brain impact was the cause of lower
mortality in patients with thoracic trauma, this was not
associated with bad GCS and pupil examination as supported
by the multiple logistic regression analysis. We believe this
may be a confound bias related to early airway support
applied to patients who present lung contusion, pneumotho-
rax, or hemothorax. This variable was not adequately con-
trolled in our analysis. If this hypothesis is correct, one can
conclude that ventilation support needs to be indicated early
TABLE 2. Final Model of Multiple Logistic Regression That Better Explain the Mortality of Patients With Severe Head Injury at
Time of Discharge
Variables
All Patients,
Nⴝ748 (%)
Outcome
Adjusted OR* (95% CI) pSurvivors, n ⴝ499 (%) Death, n ⴝ249 (%)
Age (yr)
†
Mean (⫾SD) 34.8 (⫾16.3) 33.04 (⫾15.5) 36.16 (17.6) NA 0.01
12–30 391 (52.7) 270 (69.0) 121 (31.0) 1.0
31–45 187 (25.2) 124 (66.3) 63 (33.7) 1.05 (0.66–1.67) 0.84
46–60 94 (12.7) 62 (65.9) 32 (34.1) 1.61 (0.90–2.88) 0.11
Higher than 60 70 (9.4) 39 (55.7) 31 (44.3) 2.51 (1.31–4.84) 0.006
Year of the attendance
2002–2003 132 (17.7) 101 (76.6) 31 (23.3) 1.0
2001–2000 142 (19.0) 100 (70.4) 42 (29.6) 1.00 (0.51–1.98) 0.99
1999–1998 133 (17.8) 94 (70.7) 39 (29.3) 1.15 (0.58–2.25) 0.68
1997–1996 162 (21.7) 105 (64.8) 57 (35.2) 2.17 (1.16–4.04) 0.01
1995–1994 178 (23.8) 98 (55.1) 80 (44.9) 3.17 (1.71–5.88) ⬍0.0001
Marshall scale (CT)
‡
Type I injury 22 (2.9) 19 (86.4) 3 (13.6) 1.0
Type II injury 175 (23.4) 145 (82.9) 30 (17.1) 0.55 (0.13–2.23) 0.40
Type III injury 172 (23.0) 107 (62.2) 65 (37.8) 1.05 (0.26–4.21) 0.94
Type IV injury 58 (7.8) 19 (32.8) 39 (67.2) 3.63 (0.84–15.76) 0.08
Evacuated mass lesion 240 (32.1) 154 (64.2) 86 (35.8) 0.81 (0.21–3.16) 0.76
Nonevacuated mass lesion 30 (4.0) 14 (46,7) 16 (53.3) 3.18 (0.66–15.26) 0.15
Brainstem lesion 50 (6.7) 41 (82.0) 9 (18.0) 0.34 (0.07–1.60) 0.17
Sub-arachnoid hemorrhage
No 481 (64.3) 340 (70.7) 141 (29.3) 1.0
Yes 267 (35.7) 159 (59.6) 108 (40.4) 1.86 (1.23–2.81) 0.003
Thoracic trauma
Yes 141 (18.9) 109 (77.3) 32 (22.7) 1.0
No 607 (81.1) 390 (64.3) 217 (35.7) 2.02 (1.19–3.41) 0.009
Glasgow Coma Scale
7 or 8 311 (41.6) 255 (82) 56 (18) 1.0
5 or 6 192 (25.7) 135 (70.3) 57 (29.7) 1.68 (1.03–2.75) 0.04
3 or 4 243 (32.5) 108 (44.4) 135 (55.6) 3.97 (2.49–6.31) ⬍0.0001
Pupils
§
Isocorics 283 (37.8) 239 (84.4) 44 (15.6) 1.0
Miotics 30 (4.0) 23 (76.7) 7 (23.3) 1.47 (0.53–4.07) 0.40
Anisocorics 347 (46.4) 216 (62.2) 131 (37.8) 2.65 (1.69–4.17) ⬍0.0001
Midriatics 83 (11.1) 17 (20.5) 66 (79.5) 11.24 (5.42–23.30) ⬍0.0001
* Indicate the magnitude of association among each categorical variables and death independently of imbalances in the distribution of the other clinical, demographic, radiologic,
and neurosurgical variables studied.
†
Age was not confirmed in six cases.
‡
Tomographic Marshall scale was not evaluated in one patient.
§
Six patients had ocular trauma and pupils were not adequately evaluated.
The Journal of TRAUMA
®
Injury, Infection, and Critical Care • Volume 67, Number 1, July 2009 Mortality in Severe Traumatic Brain Injury
© 2009 Lippincott Williams & Wilkins 89
in selected cases. In these patients, complications related to
mechanical ventilation will be supplanted by the supposed
neuroprotection provided by early ventilation support. This is
currently being evaluated by our research group.
The mortality of patients with brain stem lesion was
similar to that observed in patients with normal CT (Table 2),
an unexpected findings that needs to be interpreted with
caution. We believe that lower sensitivity of CT to detect
brain stem lesions was a probable confounding bias for this
finding, not controlled in the present study.
22
In one decade (between 1994 and 2003), the mortality
of patients admitted with severe TBI was three times lower
(Table 2) probably due to better quality of diagnosis and
treatment by our health care team, including the prehospital
assistance, which also improved along. Interestingly, the
population and the number of vehicles increased in the
metropolitan region covered by our hospital, but the number
of patients with severe TBI admitted to our hospital decreased
around 25% (Table 2). The most significant drop was ob-
served in the years 1997 and 1998. These findings can be
explained by the implementation of more severe traffic laws
in Brazil as well as the obligatory use of seatbelts for all car
passengers, and helmets for motorcycle users, in 1997. The
number of accidents, and consequently of patients admitted
with severe TBI, may have also dropped due to the modern-
ization finished in 1997 of one of the busiest Brazilian
highways that run through the metropolitan area of Flori-
anopolis.
Here, we demonstrated that age, CT findings, GCS,
pupil examination, and the absence of thoracic trauma at
admission are independently associated with mortality at time
of discharge in a sample of Brazilian patients. Our findings
support the obvious inference that modernization of highways
and the obligatory use of seatbelts and helmets contribute to
decrease the number and severity of TBI.
Improvement in this prediction model and its replica-
tion in other populations are necessary to determine its
external validity. The development of prognostic models to
predict the level of cognitive impairment, development of
psychiatric illness and their effect on quality of life after
severe, moderate, or even slight TBI will be clinically rele-
vant.
ACKNOWLEDGMENTS
This work is in memoriam to our co-worker Jardel
Meinerz who died after the manuscript was submitted.
REFERENCES
1. Marion DW, Shartz MC, Tyler-Kabara EC. Trauma of the nervous
system. Craniocerebral trauma. In: Bradley WG, Daroff RB, Fenichel
GM, Jankovic J, eds. Neurology in Clinical Practice. Philadelphia, PA:
Butterworth Heinemann; 2004:1127–1147.
2. 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.
3. Koizumi MS, Lebrao ML, Mello-Jorge MH, Primerano V. [Morbidity
and mortality due to traumatic brain injury in Sao Paulo City, Brazil,
1997.] Arq Neuropsiquiatr. 2000;58:81– 89.
4. Maas AI, Steyerberg EW, Butcher I, et al. Prognostic value of comput-
erized tomography scan characteristics in traumatic brain injury: results
from the IMPACT study. J Neurotrauma. 2007;24:303–314.
5. Marmarou A, Lu J, Butcher I, et al. Prognostic value of the Glasgow Coma
Scale and pupil reactivity in traumatic brain injury assessed pre-hospital and
on enrollment: an IMPACT analysis. J Neurotrauma. 2007;24:270 –280.
6. McHugh GS, Engel DC, Butcher I, et al. Prognostic value of secondary
insults in traumatic brain injury: results from the IMPACT study.
J Neurotrauma. 2007;24:287–293.
7. Murray GD, Butcher I, McHugh GS, et al. Multivariable prognostic
analysis in traumatic brain injury: results from the IMPACT study.
J Neurotrauma. 2007;24:329 –337.
8. Van Beek JG, Mushkudiani NA, Steyerberg EW, et al. Prognostic value
of admission laboratory parameters in traumatic brain injury: results
from the IMPACT study. J Neurotrauma. 2007;24:315–328.
9. Mushkudiani NA, Hukkelhoven CW, Hernandez AV, et al. A systematic
review finds methodological improvements necessary for prognostic
models in determining traumatic brain injury outcomes. J Clin Epide-
miol. 2008;61:331–343.
10. Marshall LF, Marshall SB, Klauber MR, et al. A new classification of
head injury based on computerized tomography. J Neurosurg. 1991;75:
S14 –S20.
11. Marshall LF, Marshall SB, Klauber MR, et al. The diagnosis of head
injury requires a classification based on computed axial tomography.
J Neurotrauma. 1992;9(suppl 1):S287–S292.
12. Eisenberg HM, Gary HE Jr, Aldrich EF, et al. Initial CT findings in 753
patients with severe head injury. A report from the NIH Traumatic Coma
Data Bank. J Neurosurg. 1990;73:688 – 698.
13. Foulkes MA, Eisenberg HM, Jane JA, Marmarou A, Marshall LF; and
the Traumatic Coma Data Bank Research Group. The Traumatic Coma
Data Bank: design, methods, and baseline characteristics. J Neurosurg.
1991;75:S8 –S13.
14. Marshall LF, Becker DP, Bowers SA, et al. The National Traumatic
Coma Data Bank. Part 1: Design, purpose, goals, and results. J Neuro-
surg. 1983;59:276 –284.
15. Marshall LF, Toole BM, Bowers SA. The National Traumatic Coma
Data Bank. Part 2: Patients who talk and deteriorate: implications for
treatment. J Neurosurg. 1983;59:285–288.
16. Perneger TV. What’s wrong with Bonferroni adjustments. BMJ. 1998;
316:1236 –1238.
17. Raghupathi R. Cell death mechanisms following traumatic brain injury.
Brain Pathol. 2004;14:215–222.
18. Ikonomidou C, Turski L. Why did NMDA receptor antagonists fail
clinical trials for stroke and traumatic brain injury? Lancet Neurol.
2002;1:383–386.
19. Papadia S, Soriano FX, Leveille F, et al. Synaptic NMDA receptor
activity boosts intrinsic antioxidant defenses. Nat Neurosci. 2008;11:
476 – 487.
20. Bhalla US, Iyengar R. Emergent properties of networks of biological
signaling pathways. Science. 1999;283:381–387.
21. Weng G, Bhalla US, Iyengar R. Complexity in biological signaling
systems. Science. 1999;284:92–96.
22. Mannion RJ, Cross J, Bradley P, et al. Mechanism-based MRI classifi-
cation of traumatic brainstem injury and its relationship to outcome.
J Neurotrauma. 2007;24:128 –135.
Martins et al. The Journal of TRAUMA
®
Injury, Infection, and Critical Care • Volume 67, Number 1, July 2009
© 2009 Lippincott Williams & Wilkins90
