J Rehabil Med 2010; 42: 425436

Review Article

PROGNOSIS OF SIX-MONTH FUNCTIONING AFTER MODERATE TO SEVERE

TRAUMATIC BRAIN INJURY: A SYSTEMATIC REVIEW OF PROSPECTIVE
COHORT STUDIES

Els C. Husson, MD1, Gerard M. Ribbers, MD, PhD1,2, Agnes H. P. Willemse-van Son, PhD1,2,
Arianne P. Verhagen, PhD3 and Henk J. Stam, MD, PhD1
From the Rotterdam Neurorehabilitation Research (RoNeRes),1Department of Rehabilitation Medicine, Erasmus MC,
University Medical Centre, 2Rijndam Rehabilitation Centre and 3Department of General Practice, Erasmus MC, Univer-
sity Medical Centre, Rotterdam, The Netherlands

Objective: To systematically review which determinants, as- discharged home, often without any follow-up treatment, and
sessed within the first month after a moderate to severe trau- many experience long-term consequences without receiving
matic brain injury, predict 6-month functional outcome. adequate help (13).
Methods: Databases were searched for relevant publica- In the hospital phase, a global prognosis on outcome is made
tions between 1995 and August 2008. Selection criteria were: and follow-up treatment is initiated. Early prognostication of
prospective cohort studies; determinants associated with outcome is relevant for directing rehabilitation treatment and
functional outcome 6 months after moderate to severe trau- for informing patients and relatives about the prognosis. A
matic brain injury in adult patients; determinants assessed broad variety of possible determinants of prognosis after TBI
within the first month post-injury. Two reviewers independ- has been investigated. However, the studies are of varying
ently performed the selection and quality assessment. A best-
quality and often generate conflicting results. In a systematic
evidence synthesis was performed for prognostic factors as-
review on prognostic models, the authors established that: (i)
sessed in 2 or more studies.
as many as 89 determinants were included in prognostic mod-
Results: Twenty-eight studies were included, 27 of which
els; (ii) the methodological quality of many of these models
were high quality. Most studies used the Glasgow Outcome
Score at 6 months post-injury as outcome measure, some- was poor; and (iii) that they were rarely validated on external
times in combination with other outcome measures. Strong populations (4). Therefore, well-accepted algorithms to deter-
evidence for predicting outcome at 6 months was found for mine which patients will recover well and which will be at risk
the Glasgow Coma Scale (GCS), GCS admission, motor for long-term disabilities are currently not available. Little is
score, midline shift on computed tomography scan, sub- known about predicting outcome at the level of resuming daily
dural haematoma and pulsatility index. Strong evidence of activities and social participation.
no association was found for gender and intraventricular In a systematic review of long-term prognosis after TBI,
haemorrhage. For other determinants, inconclusive or no Willemse-van Son et al. (5) established that the level of disabil-
evidence was found. ity at discharge from rehabilitation predicts ongoing disability
Conclusion: GCS, GCS on admission, motor score, midline 1 year or more after TBI. In a second study Willemse-van Son
shift, subdural haematoma and pulsatility index predicted et al. (6) established that the Barthel Index at hospital discharge
outcome 6 months after traumatic brain injury. Gender predicts community integration 3 years post-onset. Optimiz-
and intraventricular haemorrhage did not have predictive ing a patients level of activities may contribute to an optimal
value. level of participation in the longer run. From this perspective
Key words: traumatic brain injury; outcome; systematic review; it may be pivotal to identify those patients who are in need of
rehabilitation; prognosis. subacute rehabilitation.
J Rehabil Med 2010; 42: 425436 The objective of the present systematic review was therefore
to identify which determinants, assessed within the first month
Correspondence address: Gerard M. Ribbers, MD, PhD, post-injury, predict functioning 6 months post-injury.
Rijndam Rehabilitation Centre, Westersingel 300, 3015 LJ
Rotterdam, The Netherlands. E-mail: g.ribbers@rijndam.nl
Submitted June 26, 2009; accepted March 10, 2010
METHODS

Search strategy
INTRODUCTION
In 1995 the Brain Trauma Foundation in the United States developed
Traumatic brain injury (TBI) is a major cause of death and the first TBI Guidelines with the assistance of a group of international
experts (7). This set the benchmark for evidence-based guidelines in
disability. The incidence rate in Europe, reporting hospital- neurosurgery and other surgical specialties regarding TBI. After an
ized patients and patients who die before reaching hospital, is update in 2000 and 2007, the guidelines were endorsed by the American
approximately 243 per 100,000 per year (1). The majority is Association of Neurological Surgeons and the World Health Organi-

2010 The Authors. doi: 10.2340/16501977-0566 J Rehabil Med 42

Journal Compilation 2010 Foundation of Rehabilitation Information. ISSN 1650-1977
426 E. C. Husson et al.

zations Committee on Neurotrauma. We performed a computerized (9)); (vii) the majority (at least 80%) of the patients in the studies was
literature search of PubMed and PsychINFO from 1995 to August 2008. in the age range 1865 years; (viii) the article was written in English,
Additionally, reference lists of identified publications were checked. French, German or Dutch; (ix) the article was a full-text article; (x)
The search strategy was developed and tested for PubMed and the study design was a prospective cohort study.
adapted for PsychINFO. To describe the population the MeSH term A study was excluded if: (i) the study population had additional seri-
craniocerebral trauma was used. To describe the design the follow- ous neurological, oncological or systemic impairments; (ii) outcome
ing key terms were used: the MeSH term predictive value of tests, was presented only as a dichotomous distinction on the GOS between
prognos* and predict*. To select the adult population the MeSH terms death/vegetative state and severe disability/moderate disability/
adult and middle aged were used. Two reviewers independently good recovery.
screened titles and abstracts to identify relevant articles. Relevant Two reviewers assessed all criteria independently in the full-text
articles were retrieved in full text. Full papers were also retrieved articles. Reviewers were not blinded as to the names of author(s),
when abstracts were absent or if they provided insufficient informa- institution(s) or journal. In case of disagreement, consensus was
tion to enable selection. sought. If disagreements were not resolved a third reviewer made
the final decision.
Selection criteria
A study was included if all of the following criteria were met: (i) the Quality assessment
study investigated determinants of functional outcome after TBI; (ii) The methodological quality of the articles was assessed with a modi-
TBI was defined as an alteration in brain function as a result of an fied version of an established criteria list for prospective cohort stud-
acute external violent force to the head; (iii) outcome was described ies (10). The criteria list was modified for the topic of the review in
as functional outcome, measured with at least the Glasgow Out- concordance with the framework for assessing validity in prognostic
come Score (GOS) (8) or a comparable measure describing activity studies (11).
limitations or participation restrictions; (iv) the study investigated the The criteria list consisted of 16 items (footnote Table I). The item
association between determinants, measured in the first month post- was scored positive (yes), if it fulfilled the criterion. If a criterion was
injury, and outcome measures as defined; (v) outcome was assessed not fulfilled, the item was scored negative (no). If there was insufficient
at 6 months post-injury; (vi) the study population consisted of mode information, the item was scored unclear (?). The total sum of positive
rate and/or severe TBI patients or a separately analysed subgroup of items was calculated as the quality score (maximum 16 points). A study
moderate to severe TBI patients (Glasgow Coma Scale (GCS) 312 that scored 8 or more points was considered high quality.

Table I. Results of methodological assessment

Article a b c d e f g h i j k l m n o p Quality score
Wang et al. (27) 1 1 1 1 1 0 1 0 1 1 1 0 1 1 0 0 11
Signoretti et al. (12) 1 0 0 1 0 0 1 0 1 1 1 0 1 0 0 0 7
Fabbri et al. (36) 1 1 1 1 1 0 1 0 1 1 1 0 1 1 0 1 12
Hebb et al. (37) 1 0 1 1 1 0 1 1 1 1 1 0 1 0 0 0 10
IMPACT study (1323) 1 1 0 1 1 0 1 0 1 1 1 0 1 0 1 1 11
Pang et al. (50) 1 1 0 1 1 1 1 1 1 1 1 0 1 0 0 0 11
Tanriverdi et al. (63) 1 1 0 1 1 1 1 1 1 1 0 0 1 1 0 1 12
Oh et al. (29) 1 0 0 1 0 0 1 0 1 1 1 0 1 0 0 1 8
Hiler et al. (24) 1 1 0 1 1 0 1 1 0 1 0 0 1 1 0 1 10
Shutter et al. (30, 33) 1 0 1 1 1 0 1 1 1 1 1 1 1 1 0 0 12
Vos et al. (44) 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 15
Glenn et al. (34) 1 1 1 1 1 1 1 1 1 1 1 0 1 1 0 0 13
Lew et al. (28) 0 0 1 1 1 1 1 0 1 1 1 0 1 1 0 0 10
Mattioli et al. (40) 1 1 1 1 1 0 1 0 0 1 0 0 1 0 0 1 9
Tan et al. (32) 1 1 0 1 1 1 1 0 1 1 1 0 1 1 0 1 12
Sarrafzadeh et al. (51) 1 1 1 1 1 0 1 1 1 1 1 0 1 1 0 0 12
Rovlias & Kotsou (39) 1 1 1 1 0 0 1 0 1 1 1 0 1 1 0 1 11
Moreno et al. (38) 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 15
Struchen et al. (25) 1 0 1 1 0 0 1 0 1 1 1 1 1 1 0 1 11
Theilen et al. (31) 1 0 0 1 0 0 1 1 0 1 0 1 1 1 0 0 8
Lannoo et al. (26) 1 0 1 1 1 0 1 0 1 1 1 1 1 1 0 1 12
Van den Brink et al. (52) 1 1 0 1 1 1 1 1 1 1 1 0 1 1 0 1 13
Lubillo et al. (41) 1 1 0 1 1 1 1 1 1 1 1 0 1 1 1 1 14
Della Corte et al. (48) 1 1 1 1 1 1 1 1 1 1 1 0 1 1 0 0 13
Le Roux et al. (43) 1 0 1 1 1 1 1 1 1 1 1 0 1 1 0 0 12
Kelly et al. (42) 1 1 0 1 1 0 1 1 1 1 1 0 1 1 0 0 11
Ellenberg et al. (35) 1 1 1 1 1 1 1 0 1 1 0 0 1 0 0 0 10
Rae-Grant et al. (49) 1 1 1 1 1 0 1 0 1 1 1 0 1 1 0 1 12
Criteria; yes = 1, no/? = 0: (a) inception cohort; (b) description of source population; (c) description of relevant inclusion and exclusion criteria; (d) follow-
up at least 6 months; (e) drop-outs/loss to follow-up < 20%; (f) information completers vs loss to follow-up/drop-outs; (g) prospective data collection;
(h) treatment in cohort is fully described/standardized; (i) clinically relevant potential prognostic factors; (j) standardized or valid measurements; (k)
data presentation of most important prognostic factors; (l) clinically relevant outcome measures; (m) standardized or valid measurements; (n) data
presentation of most important outcome measures; (o) appropriate univariate crude estimates; (p) appropriate multivariate analysis techniques.

J Rehabil Med 42
 Prognosis of 6-month outcome after TBI 427

Two reviewers independently scored the quality. In case of disagree- trials (RCT) and observational cohorts. We included studies
ment, consensus was sought. If consensus could not be reached, a third with patients with moderate to severe TBI (GCS 312). Some
reviewer made the final decision. Overall inter-observer agreement
studies used other classifications for severity: the necessity
was tested with kappa statistics; the agreement on the 448 individual
items was compared with the expected agreement. of full intensive care unit (ICU) treatment (24), a motor GCS
< 6 (25), or a motor GCS < 6 for more than 24 h (26). These
Data extraction populations were considered to meet the inclusion criteria
Data on study cohort, inclusion and exclusion criteria, number of and were therefore included in this review. More than 40
participants, time post-injury, loss to follow-up, outcome measure- prognostic determinants were studied. All studies used the
ments, prognostic factors, and results on associations were extracted, GOS at 6 months after injury as an outcome measure, except
using a standardized form.
for 2 studies that used the extended GOS (GOS(E)) (27, 28),
Analysis and 1 study that used the Disability Rating Scale (DRS) (29).
A best-evidence synthesis was performed, in which 4 levels of evi- Several studies used additional outcome measures besides the
dence (10) (Table II) were defined to determine the strength of the GOS(E): the DRS (25, 2931), the employability component
association of prognostic factors with functioning after 6 months. of the DRS (EMP) (30) the Sickness Impact Profile (SIP) (26),
Significant relative risk ratios (RRs), odds ratios (ORs), or significant a neuropsychological test battery (26) the Cognitive Ability
associations (p<0.05) that were provided by the studies were used to
Scale (29) and mortality. For this review only the GOS(E)
determine the levels of evidence. We used the results of multivariate
analyses preferentially to establish levels of evidence. When multivari- and DRS were considered. Many of the articles did not report
ate data on a determinant were not available, we used the univariate descriptive statistics regarding the determinants, or effect sizes
results. We also checked all studies for univariate results to rule out such as an odds ratio. This complicated the estimation of the
the chance that we missed relevant information. For our conclusion impact of the determinants.
and analysis we only present results on determinants investigated in
2 or more studies.
Levels of evidence
We only established the levels of evidence for prognostic fac-
RESULTS tors analysed in 2 or more studies. Table IV presents the results
Selection of studies of the best-evidence synthesis.
In total, 770 non-duplicate citations were identified and 106
full-text articles were retrieved. Agreement was initially Patient characteristics
reached in 85% (90 of 106) of articles, and consensus was There was strong evidence of no relationship between gender
sought and reached for 16 articles. Finally, 39 articles were and outcome. Gender was analysed in 9 studies (21, 22, 27,
selected, investigating 28 cohorts. 3238), 5 of which used gender as a determinant in the mul-
tivariate analysis (21, 22, 32, 3537). None of these 9 studies
Methodological quality found a relationship between gender and outcome.
The overall inter-observer agreement of the methodological The evidence for the prognostic value of age was incon-
quality assessment was K=0.61, representing substantial clusive. Nine studies found that an older age was related to
agreement. Disagreement occurred mainly because of read- a worse outcome (21, 22, 2426, 29, 35, 37, 39, 40), and 4
ing errors and was easily resolved. For 9 individual items studies found no relationship between age and outcome (31, 32,
disagreement persisted and a third reviewer made the final 36, 41). In contrast, one study found an inverse relationship;
decision. Table I shows the final results of the methodological an older age was associated with a better outcome (33). After
assessment. One study was of lower quality and scored 7 points taking the univariate results into account, a larger proportion
(12); all other studies were of high quality. of studies found no association between age and outcome (27,
3134, 38, 4244).
Study characteristics
The study characteristics are presented in Table III. All studies Cerebral oxygenation
were prospective observational cohort studies except for one The evidence for the predictive value of pulsatility index (PI,
(1323), which used the data of several randomized controlled difference between systolic and diastolic blood flow velocities
divided by mean blood flow velocity) on transcranial Doppler
(TCD) was strong. Two high-quality studies found an associa-
Table II. Levels of evidence tion between higher PI and worse outcome. ORs were high:
21.42 (95% confidence interval (CI) 3.81183.08) (38) and
Strong Consistent findings ( 80%) in at least 2 high-
quality cohorts
25.69 (95% CI 4.95113.26) (32).
Moderate One high-quality cohort and consistent findings The evidence for the prognostic value of cerebral blood flow
( 80%) in one or more low-quality cohorts (CBF), blood flow velocity, cerebral perfusion pressure (CPP)
Limited Findings of one cohort or consistent findings and intracranial pressure (ICP) was inconclusive.
( 80%) in one or more low-quality cohorts CPP and ICP were analysed in 7 studies (24, 25, 32, 34, 38,
Inconclusive Inconsistent findings irrespective of study quality
42, 43), 4 of which presented a multivariate analysis (24, 25,

J Rehabil Med 42
 Table III. Data extraction
428

Article Patient characteristics Outcome Univariate results Multivariate results

Wang et al. US, 12 patients, 11 completing. Mean age 26 years. GOSE Significant (p<0.005): mean FA CC, mean FA CC Significant (p<0.005): mean FA CC, mean FA CC
2008 (27) Inclusion: severe closed head TBI, injury mechanism splenium, fibre count CC, mean length CC anterior body, splenium, fibre count CC anterior body, mean length

J Rehabil Med 42
consistent with DAI, informed consent (legal guardian), FDI CC anterior body. Age (p=0.046) CC anterior body, FDI CC anterior body, fibre
8th grade education level, at least 16 years old. Non-significant: DAI index in FLAIR imaging volume PVF.
Exclusion: previous TBI or neurological disorder. (p=0.010), sex, initial GCS, TCDB, all other fibre Non-significant: all other fibre measurements.
measurements.
Signoretti et US, 25 patients, ? completing. Inclusion: severe TBI. GOS 13 Significant: single voxel and multi voxel MRS ratios (15
E. C. Husson et al.

al. 2008 (12) vs 45 patients) NAA/Cho, NAA/Cr.

Fabbri et al. Italy, 309 patients, 289 completing. Median age 50 GOS 13 Significant: GCS, age, high-risk mechanism, ISS, CT Significant: Basal skull fracture (OR 8.89), SAH (OR
2008 (36) years. Inclusion: Moderate TBI, time since injury < 24 vs 45 Marshall category. 4.50), coagulopathy (OR 4.48), subdural haematoma
h, age > 10 years, local inhabitants. Exclusion: sedation/ Non-significant: Gender, co morbidity, door-to-CT (OR 3.04), CT Marshall category (OR 1.82), GCS
intubation before ER, hypotension through extra cranial time, injury-to-CT time, road accidents, falls. (OR 0.58).
injuries, cardiopulmonary resuscitation, penetrating head Non-significant: Gender, age, co morbidity, ISS,
injury. drugs/alcohol, high risk mechanism, hypotension,
hypoxia, depressed skull fracture, intracranial
haematoma, epidural haematoma, intraventricular
haematoma.
Hebb et al. US, 53 patients, no dropouts. Mean age 41.5 years. GOS Significant: 3 day PAV I and II vs III and IV and Significant: age, pupillary reaction, presenting GCS,
2007 (37) Inclusion: GCS 313, age 1685 years, abnormal CT (PAV 3 days, correlation, diffuse oedema, multiple subcortical lesions, diffuse oedema, multiple lesion sites, white matter, 3
scan, direct admission to study hospital with diagnosis 7 days, brain brain stem injuries. day PAV.
TBI. Exclusion: pentobarbital/propofol induced burst injury Non-significant: basal ganglia, thalamus, deep white Non-significant: gender, basal ganglia, thalamus,
suppression, pre-existing neurological disorder, hepatic/ profiles) matter. brain stem injuries.
metabolic encephalopathy, brain death, GCS increase to
14/15 within 8 h.
IMPACT Different countries, 8989 patients. Number of patients GOS Significant: (proportional ORs) age (2.14), Black vs Results Model A: adjusted for age, motor score,
study 2007 analysed were different for the various determinants. Caucasian (1.30), education >12 years vs < 8 years (0.70), pupils.
(1323) Median age (8,719 patients) 30 years. Inclusion: patients CT Marshall score (reference class II) I (0.45), III (2.50), Significant: p<0.01: age, GCS motor, pupillary
from 8 RCTs and 3 observational studies, moderate- IV (3.03), V/VI (2.18), compressed/absent cisterns (2.45), reactions, hypoxia (1.65), hypotension (2.06),
severe TBI. Exclusion: penetrating head injury, not midline shift 15 mm (1.36), midline shift > 5 mm (2.20), Marshall class I (0.47), III/IV (2.23), V/VI (1.48)),
surviving transport to hospital. tSAH (2.64), EDH (0.64), SDH (2.14, contusion (1.34), compressed/absent cisterns (1.83), midline shift 15
SBP < 120 mmHg (1.53), SBP >150 mmHg (1.42), mm (1.31), tSAH (2.01), EDH (0.63), SDH (1.33),
MBP < 85 mmHg (1.30), MBP > 110 mmHg (1.45), contusion (1.40), GCS eye (none) (1.54), GCS verbal
hypoxia (2.08), hypotension (2.67), hypothermia (2.21), (none) (1.51), SBP < 120 mmHg (1.28), SBP >150
GCS motor (vs localizes/obeys) enrolment (1.747.48), mmHg (1.30), pH (0.84), Hb (0.76), glucose (1.45),
pre-hospital (1.713.82), 1st in hospital (1.304.14), GCS platelets (0.79), prothrombin time (1.63)
eye enrolment (0.36), pre-hospital (0.37), 1st in hospital p<0.05: black race (1.44), >12 years education
(0.53), GCS verbal enrolment (0.38), pre-hospital (0.39), (0.74), hypothermia (1.63), midline shift > 5 mm,
1st in hospital (0.48), pupil reactivity (vs both reacting) (1.38), MBP < 85 mmHg (1.14), MBP >110 mmHg
1 reacting enrolment (2.71), pre-hospital (2.70), 1st in (1.27).
hospital (2.49), neither reacting enrolment (7.31), pre- Non-significant: Asian race, gender, education 912
hospital (4.77), 1st in hospital (5.50), cause (vs falls) road years, cause of injury (falls, road, assault, work,
traffic (0.66), assault (0.66), sports (0.45), glucose (1.68), sports, other), high/low sodium.
sodium low (1.40), pH. (0.80) platelets (0.70), Hb (0.69)
 Table III. contd.

Article Patient characteristics Outcome Univariate results Multivariate results

Non-significant: gender, Asian vs Caucasian, education
912 years vs < 8 years, cause work-related, other causes,
sodium high, prothrombin time.
Pang et al. Singapore, 513 patients, no dropouts. Mean age 44.4 GOS No analysis of individual determinants. No analysis of individual determinants.
2007 (50) years. Inclusion: severe TBI, GCS 38.
Tanriverdi et Turkey, 71 patients, no dropouts. Inclusion: acute head GOS 13 vs Non-significant: IL1A gene. Non-significant: IL1A gene.
al. 2006 (63) injury. 45
Oh et al. Korea, 82 patients, no dropouts? Mean age 50 years. DRS No analysis. Subgroup of severe TBI
2006 (29) Inclusion: brain injury, admission to ICU. (Cognitive Significant: 6 month DRS: systolic blood pressure,
ability) age, intracranial haematoma, GCS motor score, heart
rate.
Non-significant: DRS: pupils.
Hiler et al. United Kingdom, 126 patients, 107 completing. Age GOS Significant: correlations: CT Marshall class, midline shift, Significant: age, midline shift, PRx (1st 24 h). GOS
2006 (64) 1474 years. Inclusion: head injury, necessity of full volume of lesion, 1st 24 h: CPP, PRx, total period: ICP, = 3.990.0086 age 0.947 midline shift 1.54
ICU. CPP, PRx. PRx.
Non-significant: 1st 24 h: ICP, total period: GCS. Non-significant: admission GCS, CT scan type, ICP
and CPP (1st 24 h)
Shutter et al. US, 45 patients, 42 completing. Mean age 30 years. GOS 13 vs Significant for GOS: motor GCS, SSEP present, MRS Significant for GOS: admission GCS (2 12.4), Motor
2004 (65) Inclusion >14 years, presenting in ER, severe TBI 45 variables GCS (2 13.3), age (2 18.7), best early MRS levels.
Shutter et al. (GCS 38). Exclusion: previous TBI, other neurological DRS Non-significant: age, gender, field GCS, admission GCS, Non-significant: field GCS, best early MRS ratios
2006 (30) disease, ineligibility for MRI scanning. EMP ER pupil, diffuse axonal injury on MRI, contusion on Significant DRS: PWM Cho, PWM Cho/Cr.
MRI. MRS variables
Significant for DRS: motor GCS, SSEP present, MRS
variables
Significant for EMP: motor GCS, SSEP present, MRS
variables.

Vos et al. The Netherlands, 85 patients, no dropouts. Median age GOS 13 vs Significant: GCS (OR 0.65), Injury severity score (OR Significant: GFAP, GCS after resuscitation (OR
2004 (44) 32 years. Inclusion within 36 h, severe TBI (GCS < 8). 45 1.05), CT (OR 1.79), hypotension (OR 2.85), hypoxia 0.51), CT (OR 2.15), S100b. NSE, S100b, ISS equals
Exclusion: penetrating injury, no possibility for follow- (OR 1.74), GFAP, S100b, NSE CT (in model, no OR).
up, no blood sample, no informed consent Non-significant: age, pupillary reactions. Non-significant: hypotension, hypoxia.

Glenn et al. US, 49 patients, no dropouts. Mean age 36 years. GOS 13 vs Significant: linear trend: artO2sat, CMRO2, art glucose, Significant: linear trend: CMRO2, art glucose, art
2003 (66) Inclusion > 14 years, within 24 h, moderate-severe TBI 45 metabolic ratio, art lactate, AVDlac, CMRlac, CBF, lactate, CSFlac. Good/bad: artO2sat, art glucose, art
(GCS 13 or deterioration to GCS 13 within 24 h) CSFlac, pupils %any abnormal, GCS, CT score. Good/ lactate, CSFlac
requiring mechanical ventilation and ICP monitoring. GOS linear bad: artO2sat, CMRO2, art glucose, art lactate, CBF, Non-significant: Linear trend: JugO2sat, artO2sat,
Exclusion: terminal illness, severe neurological illness, trend CSFlac, pupils %any abnormal, GCS, CT score BBB artpO2, AVDO2, AVDglc, CMRglc, metabolic ratio,
acute complete SCI. damage. AVDlac, CMRlac, CBF, CSFglc. Good/bad: CMRO2,
Non-significant: linear trend: hypotension-hypoxia JugO2sat, artpO2, AVDO2, AVDglc, CMRglc,
episode, CT score BBB damage, ISS, age, gender, mean metabolic ratio, AVDlac, CMRlac, CBF, CSFglc.
Prognosis of 6-month outcome after TBI

CPP, %CPP < 60 mmHg, mean ICP, %ICP>20 mmHg, (controlled for age, injury, ICP, CPP).
JugO2sat, artpO2, AVDO2, AVDglc, CMRglc, CSFglc.
Good/bad: hypotension-hypoxia, episode, ISS, age,

J Rehabil Med 42
429

gender, mean CPP, %CPP < 60 mmHg, mean ICP,

 Table III. contd.
430

Article Patient characteristics Outcome Univariate results Multivariate results

%ICP>20 mmHg,
JugO2sat, artpO2, AVDO2, AVDglc, CMRglc, CSFglc,
metabolic ratio, AVDlac, CMRlac, CT score.

J Rehabil Med 42
Lew et al. US, 22 patients, no dropouts. Mean age 36 years. GOSE Significant: SEP bilaterally absent vs normal/present but No analysis.
2003 (28) Inclusion: acute severe TBI (GCS 8), 1770 years. (mean) abnormal, ERP absent vs normal/present but abnormal
Exclusion: pre-existing neurological disorder, median Non-significant: SEP present but abnormal vs bilaterally
neuropathy, hearing loss, dementia, psychiatric disorder, normal, ERP present but abnormal vs normal.
E. C. Husson et al.

SCI, barbiturate use.

Mattioli et al. Italy, 169 patients (out of 282 eligible), subgroup of 117 GOS: 1- step- Subgroup with GCS < 9: no univariate results. Significant: age (OR 1.051), admission GCS (OR
2003 (40) with severe TBI (GCS < 9). 5 dropouts. Mean age 42 worsening 1.723), compressed/absent cisterns (OR 2.914),
years. Inclusion: head injured patients in ICU. Exclusion: major lesions on CT (OR 1.790).
gunshot, SCI, CT scan not available. Non-significant: tSAH, Morris-Marshall-grade
(distribution of haemorrhage).
Tan et al. China, 96 patients, no dropouts. Median age 35 years. GOS 13 Significant: non-reactive pupils at admission, GCS < 5 at Significant: non-reactive pupils at admission (OR
2001 (32) Inclusion: Severe TBI (GCS 38). vs 45 admission, shock at admission, CT class 5/6 (not Marshall 1.32), GCS < 5 at admission (OR 4.56), CT class
classification), multiple sites of tSAH, initial ICP, initial 5/6 (OR 5.24), initial ICP (OR 11.78), initial CPP
CPP, mean blood flow velocity, PI, high-density lesions. (OR 3.54), mean blood flow velocity (OR 15.43), PI
Non-significant: age, gender, convulsions, (25.69).
hyperglycaemia, pO2, pCO2, Hb. Non-significant: age, gender, convulsions,
hyperglycaemia, pO2, pCO2, Hb, tSAH, high-density
lesions. Shock?
Sarrafzadeh Germany, 119 patients, of which 39 non head injury GOS 12 GOS: no difference between head injured patients with No analysis for GOS, only for mortality.
et al. 2001 (control), 110 completing. Median age 32 years. vs 3 vs 45 or without extra cranial injuries.
(51) Inclusion: severe TBI (GCS < 8), age 675 years,
Exclusion: haemodynamically unstable, death within
24 h, penetrating head injury, fixed, dilated pupils on
admission.
Rovlias et al. Greece, 125 patients, all completing. Mean age 42 years. GOS 13 Significant: mean WBC count in favourable vs Significant: GCS after resuscitation (OR 1.7818),
2001 (39) Inclusion: non-penetrating head injury, age 1670 years. vs 45 unfavourable outcome. pupillary reaction (OR 0.2833), age (OR 0.9622),
Exclusion: previous TBI, drug/alcohol abuse, brain intracranial diagnosis, WBC count.
death, factors that might alter WBC count, associated
trauma.
Moreno et al. Spain, 125 patients, no dropouts. Mean age 34 years. GOS 13 Significant: pupillary reaction (OR 16.74), shock (OR Significant: GCS (OR 0.24), pupillary reaction (OR
2000 (38) Inclusion: severe head injury, GCS < 9, cranial lesions vs 45 0.20), CT class III (cisterns) (OR 17.23), SAH absent (OR 3.86), shock (OR 0.13), PI (OR 21.42)
on CT, stay in hospital > 24 h. Exclusion: inability to 0.45), APACHE II (OR 1.35), GCS (OR 0.25), ICP (OR Non-significant: CT class III, SAH, Hb, APACHE II,
perform TCD, irregular left ventricle ejection volume, 1.08), CPP (OR 0.94), blood flow velocity (OR 0.96), PI ICP, CPP, blood flow velocity.
heart rate > 140 bpm. (OR 8.50), Hb (OR 0.76).
Non-significant: gender, associated lesions, focal injury,
convulsions, otorrhagia, surgery, age, pH, PaCO2, PaO2,
glucose.
Struchen et USA, 184 patients, 99 completing for DRS, 127 for GOS Significant for GOS: duration of: intracranial Significant for DRS (Sr2, incremental explained
al. 2001 (25) GOS. Mean age 34 years. Inclusion: GCS motor DRS hypertension (2 14.2), systemic hypotension (2 8.8), variance) Age (0.0840.101), ER-GCS (0.084
< 6, non-penetrating TBI. Exclusion: unstable cerebral perfusion hypotension (2 16.2), jugular venous 0.090), log-ICP (0.244), log-MAP (0,044), log-CPP
cardiopulmonary status, brain death, participation in oxygen desaturation. (0.197), log-VO2 (0.063).
experimental condition of clinical trial.
 Table III. contd.

Article Patient characteristics Outcome Univariate results Multivariate results

Theilen et al. Germany, 45 patients, 32 completing. Mean age 35,3 GOS Significant: ESR (r2 GOS 0.67, DRS 0.66, PPV 0.83) Significant: ESR (accuracy 90.6%), SSEP (accuracy
2000 (31) years (range 1875). Inclusion: Severe TBI (GCS 38) DRS Non-significant: age. 84.4%)
after resuscitation or deteriorating within 6 h. Exclusion: GCS no analysis Non-significant: age, GCS after resuscitation.
penetrating head injury, EEG inactivity or burst- SSEP (PPV 0.66), BAEP (PPV 0.47) no p-value.
suppression pattern.
Lannoo et al. Belgium, 158 patients, no dropouts for GOS, 80 patients GOS No analysis. Significant: age (r2 0.159, F 7.92)
2000 (26) died, completing impairment index 64, SIP 68. Inclusion: (SIP, Non-significant: GCS (lowest before sedation),
closed head injury, coma duration (GCS motor < 6) at Neuro- pupils, base excess, platelet count, temperature,
least 24 h, age > 15 years, no medical history of CNS psychological ventricular haemorrhage, PaO2.
disease or mental retardation. test battery
(impairment
index))

Van den The Netherlands, 101 patients, no dropouts. Mean age 34 GOS 13 Significant: low PbrO2 >30 min (unfavourable outcome, Significant: low PbrO2> 30 min, remains independent
Brink et al. years (range 1182). Inclusion: GCS 8, non-penetrating vs 45 OR 2.8). prognostic factor in logistic regression models,
2000 (52) head injury. only status of perimesencephalic cisterns reduced
prognostic value.

Lubillo et al. Spain, 82 patients, no dropouts. Mean age 36 years. GOS 13 Significant: motor GCS 3 (OR 25.6), pupils (OR 51.1; Significant: Motor GCS (OR 10.8), bilaterally non-
1999 (41) Inclusion: GCS 8, non-penetrating head injury, vs 45 0.57), systolic blood pressure (OR 11.81), CT score post- reactive pupils (OR 31.8), post-operative CT III/IV
removal of intracranial haematoma. operative I/II (OR 0.04), III/IV (OR 21.95), associated (OR 8.9).
intracranial lesions (OR 7.15). Non-significant: hypotension, intracranial lesions,
Non-significant: unilaterally reactive pupils, basal cisterns age.
status.
Della Corte Italy, 22 patients, no dropouts. Mean age 29 years. GOS 13 Significant: GH response to GHRH on day 7, GH No analysis.
et al. 1998 Inclusion: severe head injury. Exclusion: females, age vs 45 response to TRH, TSH response to TRH, PRL response to
(67) < 16 years or > 60 years, GCS 3, steroid use, endocrine GHRH on day 7 and 15
abnormalities, severe cardiac/pulmonary disorders, Non-significant: basal values of hormones, GH response
requirement of dopamine or > 2 units of blood. to GHRH on day 2 and 15, PRL response to TRH, PRL
response to GHRH on day 2.
Le Roux et US, 32 patients, no dropouts. Median age 35 years. GOS 13 Significant: delayed cerebral infarction, elevated AVDO2 No analysis.
al. 1997 (43) Inclusion: GCS 8, admission within 8 h, isolated vs 45 post- treatment, limited improvement in AVDO2 post-
non-penetrating head injury, normal blood pressure, treatment, admission GCS.
normothermia, haematocrit > 30%, survival > 24 h. Non-significant: age, ICP (post-treatment), CPP (post-
treatment), AVDO2 pre-treatment.
Kelly et al. US, 54 patients, no dropouts. Mean age 35. Inclusion: GOS 13 Significant: CBF (3 groups), age, h of high ICP, CPP. No analysis.
1997 (42) GCS 13, serial CBF measurement and ICP monitoring. vs 45 Non-significant: GCS, no of CT diagnoses, mean ICP,
CBF (mean).
Ellenberg et US, 314 patients, 259 completing. Mean age 29 years. GOS 13 No analysis. Significant: age (OR 0.96), time in coma (OR 0.93),
al. 1996 (68) Inclusion: initial GCS 8, age > 16 years, survival of vs 45 time in PTA (OR 0.98), dexamethason use (OR 0.29).
Prognosis of 6-month outcome after TBI

initial evaluation, out of coma before hospital discharge. Non-significant: first GCS score, race, sex, pupillary
Exclusion: penetrating injury. response, CT abnormalities, use of phenytoin, use of
morphine sulphate.

J Rehabil Med 42
431
 432 E. C. Husson et al.

32, 38). Two studies found a lower CPP and a higher ICP to

Lac: lactate; MAP: mean arterial pressure; MBP: mean arterial blood pressure; MRS: magnetic resonance spectroscopy; NSE: neuron specific enolase; OPG: ocular plethysmography; PaO2: arterial oxygen
pressure; PAV: percent alpha variability; PbrO2: brain oxygen pressure; PRL: prolactin; PRx: pressure reactivity; PVF peduncular projections to ventral frontal cortex; S100b: calcium-binding protein; Sat:
saturation; SBP: systolic blood pressure; SCI: spinal cord injury; TCDB: trauma coma data bank; TRH: thyrotropin releasing hormone; tSAH: traumatic SAH; TSH: thyroid stimulating hormone; WBC:
Art: arterial; AVD: arteriovenous difference; BAEP: brainstem auditory evoked potentials; BBB: blood brain barrier; CC: corpus callosum; CMR: cerebral metabolic ratio; CSF: cerebral spinal fluid; DAI:
diffuse axonal injury; EEG: electro-encephalogram; ER: emergency room; ERP: cognitive event related potentials; ESR: electroencephalogram silence ratio; FA: fractional anisotropy; FDI: fibre intensity
index; FLAIR: fluid attenuation and inversion recovery; GFAP: glial fibrillary acidic protein; GH: growth hormone; GHRH: growth hormone releasing hormone; Glc: glucose; Hb: haemoglobin; Jug: jugular;
be associated with worse outcome (25, 32). Univariate results

Data on SSEP: the study reported a significant relationship in the univariate analysis, but the presented figures of 1 month and 6 month outcomes differed in the total number of abnormal SSEPs. We
*ORs are mentioned for determinants that are investigated in more than one study. Combinations of factors or models are not mentioned, except when details about the contribution of the individual
were consistent with multivariate results.
Significant: EEG, TCD (R = 0.54). GCS day 7.

The evidence for the prognostic value of hypotension and

hypoxia was also inconclusive. In the multivariate analysis of
4 studies, hypotension was associated with a worse outcome
Non-significant: BAEP, SSEP, OPG.

(21, 25, 29, 38). However, in 4 other studies no significant

relationship was found (32, 36, 41, 44). Hypoxia was tested in
a multivariate model in 6 studies (21, 26, 32, 34, 36, 44). One
study found a correlation between hypoxia and unfavourable
Multivariate results

outcome (45), and another study reported a significant result

for oxygen saturation, but not for oxygen pressure (34). The
others found no significant results.

Imaging studies
Computed tomography (CT) scanning was assessed in most
studies. Different classification methods were used, but the
Marshall classification (46) was used most commonly. There-
Non-significant: CT, age, (hypertension (1 patient)).

fore, we made subgroups corresponding to this classification.

For the separate category midline shift (Marshall IV) we found
strong evidence. Midline shift was analysed in 2 studies and
both found an association with worse outcome (16, 21, 24).
Significant: GCS day 7 (LR 34.04)

The evidence for the prognostic value of the total Marshall

classification and for the separate categories mass lesion (cor-
responding to Marshall V/VI) and compressed/absent cisterns
(Marshall III) was inconclusive, in the analysis of multivariate
results. In the univariate analysis, however, compressed/absent
Univariate results

cisterns were associated with worse outcome in all 3 studies

Studies also report results on subgroups, not specified in this table. Results regard all patients together.

(16, 21, 38, 40).

GOS: 1=death, 2=vegetative state, 3=severe disability, 4=moderate disability, 5=good recovery.

The evidence for the prognostic value of subdural hae-

matoma (SDH) was strong. Two studies found an association
have assumed that this was due to a typing error, and that the p-value of < 0.001 was correct.

with worse outcome (16, 21, 47). There was strong evidence
determinants are reported. Only outcome measures relevant to this review are mentioned.

of no relationship between intraventricular haemorrhage (IVH)

and outcome. Two studies found no association (26, 47). The
Outcome

evidence for subarachnoid haemorrhage (SAH), epidural

GOS

haematoma (EDH) and intracranial haemorrhage (ICH) was

inconclusive (Tables III and IV).
al. 1996 (49) Inclusion: GCS 8, age 15 years, coma > 48 h, closed

Magnetic resonance spectroscopy (MRS) was investigated in

2 studies (12, 30, 33). Regarding the items investigated in both
Rae-Grant et US, 69 patients, 3 completing. Mean age 36 years.

studies, only univariate data were available. These items were

N-acetylaspartate/creatine (NAA/Cho) and N-acetylaspartate/
head injury, consent form, no brain death.

choline (NAA/Cr) ratios. NAA/Cho had an association with

outcome in both studies. Because one study was of low quality
(12), this evidence was moderate. The evidence for NAA/Cr
Specified in article, too detailed for this table.

was inconclusive.

Physical examination
Patient characteristics

We found strong evidence for the prognostic value of the GCS

overall (all GCS measurements together), the GCS on admis-
sion to the hospital and the GCS motor score. The GCS was
analysed in 21 different studies (17, 21, 2427, 29, 3134,
3744, 4749), 5 of which presented only univariate results
white blood cell.
Table III. contd.

(27, 34, 42, 43, 48). Of the remaining 16 studies, 13 found that
a lower GCS was related to worse outcome. All 4 studies on
Article

motor GCS (17, 21, 29, 33, 41) found that lower motor score
was related to worse outcome. Five studies (32, 33, 37, 40,
J Rehabil Med 42
 Prognosis of 6-month outcome after TBI 433

47) found an association between lower admission GCS and in 1 of 2 studies (31, 49). Two other studies presented only a
worse outcome, 1 found no association (24). The evidence for univariate analysis (28, 33), one found a relationship between
the prognostic value of GCS after resuscitation and field GCS presence of abnormal SSEP and worse outcome, and one found
was inconclusive (Tables III and IV). There was inconclusive a relationship between absent vs (ab)normal SSEP and lower
evidence for the prognostic value of pupillary reactions. Six mortality, but not for abnormal vs normal SSEP.
studies found a relation between absence of or abnormal pupil-
lary reactions and worse outcome in the multivariate analysis, Laboratory parameters
(17, 21, 32, 3739, 41), 3 studies found no association (26).
However, univariate analysis showed an association in 6 out The evidence for haemoglobin, platelet count and hypergly-
of 7 studies. caemia was inconclusive (Tables III and IV).
The evidence for the prognostic value of body temperature
and injury severity score (ISS) was inconclusive (Tables III Subgroup analysis of larger studies
and IV). To estimate the influence of sample size on our results, a
subgroup analysis of studies including more than 100 patients
Electrophysiological data was performed. In these 12 studies (21, 2426, 35, 3840,
The evidence for the prognostic value of somatosensory evoked 47, 5052) we found strong evidence that midline shift and
potentials (SSEP) was inconclusive. In a multivariate analysis, subdural haematoma predict outcome. Strong evidence of no
a relation between normal SSEP and better outcome was found association was found for gender and intraventricular haemor-

Table IV. Results of qualitative analysis

Association with Association with
Determinant Number of studies better outcome Non-significant worse outcome Inconsistent Level of evidence
Older age 14 1 4 9 0 Inconclusive
Body temperature 2 0 1 1 0 Inconclusive
Lower blood flow velocity 3 0 1 2 0 Inconclusive
Lower CBF* 2 0 0 1 1 Inconclusive
Lower CPP 4 0 2 2 0 Inconclusive
CT 21
Marshall ordinal 4 0 2 2 0 Inconclusive
Compressed/absent cisterns 3 0 1 2 0 Inconclusive
Midline shift 2 0 0 2 0 Strong
Mass lesion 4 0 3 1 0 Inconclusive
Traumatic SAH 6 0 4 2 0 Inconclusive
EDH 2 1 1 0 0 Inconclusive
SDH 2 0 0 2 0 Strong
ICH 2 0 1 1 0 Inconclusive
IVH 2 0 2 0 0 Strong no
Lower GCS 16 0 3 13 0 Strong
Admission 6 0 1 5 0 Strong
After resuscitation 3 0 1 2 0 Inconclusive
Motor 4 0 0 4 0 Strong
Field 2 0 1 1 0 Inconclusive
Male gender 5 0 5 0 0 Strong no
MRS NAA/Cr* 2 0 1 1 0 Inconclusive
MRS NAA/Cho* 2 0 0 2 0 Moderate
Haemoglobin low 3 0 2 1 0 Inconclusive
Hyperglycaemia 3 0 1 2 0 Inconclusive
Hypotension 8 0 4 4 0 Inconclusive
Hypoxia 6 0 4 1 1 Inconclusive
Higher ICP 4 0 2 2 0 Inconclusive
Lower ISS 2 1 1 0 0 Inconclusive
Higher PI 2 0 0 2 0 Strong
Platelet count low 2 0 1 1 0 Inconclusive
Abnormal/absent pupillary 9 0 3 6 0 Inconclusive
reactions
Abnormal SSEP 2 0 1 1 0 Inconclusive
*Univariate analysis.
CBF: cerebral blood flow; CPP: cerebral perfusion pressure; CT: computed tomography; EDH: epidural haematoma; SDH: subdural
haematoma; ICH: intracranial haemorrhage; IVH: intraventricular haemorrhage; GCS: Glasgow Coma Scale; MRS: Magnetic resonance
spectroscopy; NAA/Cr: N-acetylaspartate/choline; NAA/Cho: N-acetylaspartate/creatine ; ICP: intracranial pressure; ISS: injury severity
score; PI: pulsatility index ; SSEP: somatosensory evoked potentials.
J Rehabil Med 42
434 E. C. Husson et al.

rhage. Inconclusive evidence was found for GCS, age, hypoten- however, suggest that age might not be a strong predictor of
sion, hypoxia, epidural haematoma, SAH, pupillary reactions, outcome. PI has not been included in previous prognostic
state of basal cisterns, haemoglobin, platelet count, body models as far as we know. Given the high odds ratios found
temperature, ICP and CPP. In a subgroup analysis of studies in 2 studies in this review, the PI may well be a promising
with more than 300 patients (3 studies (21, 35, 47)) we found determinant of outcome.
strong evidence for the predictive value of subarachnoid haem- In a systematic review on SSEP, the authors concluded that
orrhage and subdural haematoma. We found strong evidence SSEP are the best single overall predictor of outcome after TBI
of no association between gender and outcome. Inconclusive (56). We could not find any evidence to support this conclu-
evidence was found for GCS, age, hypotension, hypoxia, sion. It is possible that SSEP are more useful for predicting
epidural haematoma, and pupillary reactions. mortality than for predicting functional outcome.

Limitations of the review

DISCUSSION
We searched studies published between the presentation of
The objective of this systematic review was to identify which the TBI guidelines in 1995 and August 2008. It is possible
determinants, assessed within the first month after TBI, predict that relevant publications before or after that time were not
functioning 6 months post-onset. Providing new information included in this review, such as the Traumatic Coma Data Bank
or explaining the findings in terms of pathophysiology is not studies (57). However, we think that it would be inappropri-
the primary goal of a systematic review. Therefore, although ate to include older studies in a review on prognostic factors,
relevant for clinicians, we were cautious not to speculate about because there have been substantial changes in treatment over
underlying mechanisms or explanations for which no evidence the last decades (7, 58). This most likely has had an effect on
was given in the included studies. the general prognosis after TBI, thus hampering comparability
Our study established strong evidence that low GCS overall, between older and more recent studies. Furthermore, publica-
low GCS on admission, low motor score, presence of midline tion bias might have occurred. Studies with significant results
shift on CT scan, presence of subdural haematoma and high might be easier to publish and therefore easier to find. Also,
PI on transcranial Doppler were predictors of poor outcome we did not include studies published in languages other than
6 months post-TBI. We found strong evidence that gender English, French, German or Dutch. We studied only moderate
and intraventricular haemorrhage do not predict outcome. to severe TBI patients. Therefore it is not clear whether the
We found moderate evidence that a lower ratio NAA/Cho findings can also be generalized to mild TBI patients.
on MRS predicts poor outcome. However, this was based on Besides the GOS, only a few other measures of functional
univariate data. For all other determinants the evidence was outcome were used. The GOS is a rather crude measure, with
inconclusive. limited sensitivity to change (59, 60). There is some observer
variation in outcome assessment with the GOS(E), which might
Comparison with other literature have influenced the results of the studies (61, 62). The effect of
We compared our findings with several other reviews and this on our results is not clear. The prognostic value of a deter-
relevant publications on prognostic factors and prognostic minant might be underestimated by using such a crude measure.
models after TBI. In a systematic review on the long-term However, underestimation might also happen when more catego-
prognosis (1 year or more) after TBI, Willemse et al. (5) es- ries are used, as a result of a higher misclassification rate (61).
tablished that older age, pre-injury unemployment, substance In planning individualized long-term rehabilitation programmes
abuse, and more severe disability at rehabilitation discharge more detailed information on outcome is pivotal.
were strong predictors for ongoing disability. Inconclusive Some of the included studies excluded penetrating head
evidence was found for female gender, and lower GCS. Our injury, other studies did not. Little information was given in
results were different with regard to age and GCS. Possibly the studies about how many patients had closed or open head
long-term functioning depends less on initial severity scores injuries, or about the prognosis of these subgroups. It is not
such as the GCS and more on psychosocial variables such as clear whether this has influenced our results.
age, coping style, social support and, for example, financial Some variables might have a U-shape correlation with out-
resources. In the studies included in the current review, few come. For example: in the IMPACT study both very high and
other sociodemographic factors were investigated besides very low blood pressure were associated with a worse outcome
age. In contrast, the studies included in the review on the (13). In many other studies, blood pressure was dichotomized
longer term by Willemse et al., investigated primarily socio- on a certain threshold. Therefore, the adverse effect of a very
demographic factors, and few basic medical and neurological high blood pressure might obscure the prognostic value of
variables. hypotension in the studies analysed in this review. It was not
As mentioned in the introduction, several prognostic models possible to calculate a correction for this effect.
have been developed. Most models predict mortality or GOS In the analysis, the results of large studies were given equal
after 6 or 12 months and include age, GCS or motor score, weight to the results of small studies. Therefore negative
pupillary reactivity and some other factors, such as CT pa- findings in smaller studies might influence the results dispro-
rameters, hypotension or hypoxia (45, 5355). Our findings, portionately. However, based on the separate analysis of the

J Rehabil Med 42
 Prognosis of 6-month outcome after TBI 435

results of studies with more than 100 and 300 patients, this Bullock RM. Assessment of mitochondrial impairment in traumatic
has not resulted in false negative conclusions. The evidence brain injury using high-resolution proton magnetic resonance
spectroscopy. J Neurosurg 2008; 108: 4252.
for determinants such as age, pupillary reactions, hypotension
13. Butcher I, Maas AI, Lu J, Marmarou A, Murray GD, Mushkudiani
and hypoxia remained inconclusive in both subgroup analyses, NA, et al. Prognostic value of admission blood pressure in trau-
indicating that our results are robust. matic brain injury: results from the IMPACT study. J Neurotrauma
2007; 24: 294302.
14. Butcher I, McHugh GS, Lu J, Steyerberg EW, Hernandez AV,
Recommendations
Mushkudiani N, et al. Prognostic value of cause of injury in trau-
In predicting outcome after brain injury there is a reasonable matic brain injury: results from the IMPACT study. J Neurotrauma
amount of uncertainty. A multitude of determinants may con- 2007; 24: 281286.
15. Maas AI, Marmarou A, Murray GD, Teasdale SG, Steyerberg EW.
tribute to the prognosis. In planning rehabilitation treatment,
Prognosis and clinical trial design in traumatic brain injury: the
clinicians should pay attention to the contextual factors as well IMPACT study. J Neurotrauma 2007; 24: 232238.
as to the early medical information. Because TBI is a lifelong 16. Maas AI, Steyerberg EW, Butcher I, Dammers R, Lu J, Marmarou
disorder in which changing contextual demands may generate A, et al. Prognostic value of computerized tomography scan
new needs for professional support, and because we are unable characteristics in traumatic brain injury: results from the IMPACT
study. J Neurotrauma 2007; 24: 303314.
to accurately predict who is at risk of incurring restrictions of
17. Marmarou A, Lu J, Butcher I, McHugh GS, Murray GD, Steyerberg
activities or participation, TBI patients should be involved in EW, et al. Prognostic value of the Glasgow Coma Scale and
a life-long, well-coordinated programme. pupil reactivity in traumatic brain injury assessed pre-hospital
Furthermore, outcome measurement should be more specific and on enrollment: an IMPACT analysis. J Neurotrauma 2007;
than just the GOS(E) and should encompass more detailed 24: 270280.
18. Marmarou A, Lu J, Butcher I, McHugh GS, Mushkudiani NA,
functional outcome measures, for instance on participation in Murray GD, et al. IMPACT database of traumatic brain injury:
leisure and professional activities, caregiver strain and the risk design and description. J Neurotrauma 2007; 24: 239250.
of developing mood disorders. 19. McHugh GS, Butcher I, Steyerberg EW, Lu J, Mushkudiani N,
Much of the literature available is written from an isolated Marmarou A, et al. Statistical approaches to the univariate prog-
neurological, neurosurgical or rehabilitation perspective. To nostic analysis of the IMPACT database on traumatic brain injury.
J Neurotrauma 2007; 24: 251258.
improve outcome prediction, multidisciplinary research should 20. McHugh GS, Engel DC, Butcher I, Steyerberg EW, Lu J,
take place, and knowledge should be integrated. Mushkudiani N, et al. Prognostic value of secondary insults in
traumatic brain injury: results from the IMPACT study. J Neuro-
trauma 2007; 24: 287293.
REFERENCES 21. Murray GD, Butcher I, McHugh GS, Lu J, Mushkudiani NA,
Maas AI, et al. Multivariable prognostic analysis in traumatic
1. Tagliaferri F, Compagnone C, Korsic M, Servadei F, Kraus J. A brain injury: results from the IMPACT study. J Neurotrauma
systematic review of brain injury epidemiology in Europe. Acta 2007; 24: 329337.
Neurochir (Wien) 2006; 148: 255268; discussion 268. 22. Mushkudiani NA, Engel DC, Steyerberg EW, Butcher I, Lu J,
2. Ribbers GM. Traumatic brain injury rehabilitation in the Nether Marmarou A, et al. Prognostic value of demographic characteris-
lands: dilemmas and challenges. J Head Trauma Rehabil 2007; tics in traumatic brain injury: results from the IMPACT study. J
22: 234238. Neurotrauma 2007; 24: 259269.
3. Fleminger S, Ponsford J. Long term outcome after traumatic brain 23. Van Beek JG, Mushkudiani NA, Steyerberg EW, Butcher I,
injury. BMJ 2005; 331: 14191420. McHugh GS, Lu J, et al. Prognostic value of admission laboratory
4. Perel P, Edwards P, Wentz R, Roberts I. Systematic review of parameters in traumatic brain injury: results from the IMPACT
prognostic models in traumatic brain injury. BMC Med Inform study. J Neurotrauma 2007; 24: 315328.
Decis Mak 2006; 6: 38. 24. Hiler M, Czosnyka M, Hutchinson P, Balestreri M, Smielewski P,
5. Willemse-van Son AH, Ribbers GM, Verhagen AP, Stam HJ. Matta B, et al. Predictive value of initial computerized tomography
Prognostic factors of long-term functioning and productivity after scan, intracranial pressure, and state of autoregulation in patients
traumatic brain injury: a systematic review of prospective cohort with traumatic brain injury. J Neurosurg 2006; 104: 731737.
studies. Clin Rehabil 2007; 21: 10241037. 25. Struchen MA, Hannay HJ, Contant CF, Robertson CS. The relation
6. Willemse-van Son AH, Ribbers GM, Hop WC, Stam HJ. Com- between acute physiological variables and outcome on the Glas-
munity integration following moderate to severe traumatic brain gow Outcome Scale and Disability Rating Scale following severe
injury: a longitudinal investigation. J Rehabil Med 2009; 41: traumatic brain injury. J Neurotrauma 2001; 18: 115125.
521527. 26. Lannoo E, Van Rietvelde F, Colardyn F, Lemmerling M, Vandeker-
7. Bullock R, Chesnut RM, Clifton G, Ghajar J, Marion DW, Narayan ckhove T, Jannes C, et al. Early predictors of mortality and
RK, et al. Guidelines for the management of severe head injury. morbidity after severe closed head injury. J Neurotrauma 2000;
Brain Trauma Foundation. Eur J Emerg Med 1996; 3: 109127. 17: 403414.
8. Jennett B, Bond M. Assessment of outcome after severe brain 27. Wang JY, Bakhadirov K, Devous MD Sr, Abdi H, McColl R, Moore
damage. Lancet 1975; 1: 480484. C, et al. Diffusion tensor tractography of traumatic diffuse axonal
9. Teasdale G, Jennett B. Assessment of coma and impaired con- injury. Arch Neurol 2008; 65: 619626.
sciousness. A practical scale. Lancet 1974; 2: 8184. 28. Lew HL, Dikmen S, Slimp J, Temkin N, Lee EH, Newell D, et
10. Scholten-Peeters GG, Verhagen AP, Bekkering GE, van der Windt al. Use of somatosensory-evoked potentials and cognitive event-
DA, Barnsley L, Oostendorp RA, et al. Prognostic factors of related potentials in predicting outcomes of patients with severe
whiplash-associated disorders: a systematic review of prospective traumatic brain injury. Am J Phys Med Rehabil 2003; 82: 5361;
cohort studies. Pain 2003; 104: 303322. quiz 6254, 80.
11. Altman DG. Systematic reviews of evaluations of prognostic vari- 29. Oh HS, Seo WS, Lee S, Song H. Comparisons of the prognostic
ables. BMJ 2001; 323: 224228. predictors of traumatic brain injury according to admission Glas-
12. Signoretti S, Marmarou A, Aygok GA, Fatouros PP, Portella G, gow Coma Scale scores-based on 1- and 6-month assessments.

J Rehabil Med 42
436 E. C. Husson et al.

Taehan Kanho Hakhoe Chi 2006; 36: 621629. WohlbergCJ, et al. Outcome of severe brain injury: a multimodality
30. Shutter L, Tong KA, Lee A, Holshouser BA. Prognostic role of neurophysiologic study. J Trauma 1996; 40: 40l407.
proton magnetic resonance spectroscopy in acute traumatic brain 50. Pang BC, Kuralmani V, Joshi R, Hongli Y, Lee KK, Ang BT, et
injury. J Head Trauma Rehabil 2006; 21: 334349. al. Hybrid outcome prediction model for severe traumatic brain
31. Theilen HJ, Ragaller M, Tscho U, May SA, Schackert G, Albrecht injury. J Neurotrauma 2007; 24: 136146.
MD. Electroencephalogram silence ratio for early outcome progno- 51. Sarrafzadeh AS, Peltonen EE, Kaisers U, Kuchler I, Lanksch
sis in severe head trauma. Cri Care Med 2000; 28: 35223529. WR, Unterberg AW. Secondary insults in severe head injury
32. Tan H, Feng H, Gao L, Huang G, Liao X. Outcome prediction in do multiply injured patients do worse? Cri Care Med 2001; 29:
severe traumatic brain injury with transcranial Doppler ultrasonog- 11161123.
raphy. Chin J Traumatol 2001; 4: 156160. 52. van den Brink WA, van Santbrink H, Steyerberg EW, Avezaat CJ,
33. Shutter L, Tong KA, Holshouser BA. Proton MRS in acute trau- Suazo JA, Hogesteeger C, et al. Brain oxygen tension in severe head
matic brain injury: role for glutamate/glutamine and choline for injury. Neurosurgery 2000; 46: 868876; discussion 876878.
outcome prediction. J Neurotrauma. 2004; 21: 16931705. 53. Cremer OL, Moons KG, van Dijk GW, van Balen P, Kalkman CJ.
34. Glenn TC, Kelly DF, Boscardin WJ, McArthur DL, Vespa P, Oertel Prognosis following severe head injury: Development and valida-
M, et al. Energy dysfunction as a predictor of outcome after moder- tion of a model for prediction of death, disability, and functional
ate or severe head injury: indices of oxygen, glucose, and lactate recovery. J Trauma 2006; 61: 14841491.
metabolism. J Cereb Blood Flow Metab 2003; 23: 12391250. 54. Choi SC, Narayan RK, Anderson RL, Ward JD. Enhanced spe-
35. Ellenberg JH, Levin HS, Saydjari C. Posttraumatic amnesia as a cificity of prognosis in severe head injury. J Neurosurg 1988;
predictor of outcome after severe closed head injury. Prospective 69: 381385.
assessment. Arch Neurol 1996; 53: 782791. 55. Andrews PJ, Sleeman DH, Statham PF, McQuatt A, Corruble V,
36. Fabbri A, Servadei F, Marchesini G, Stein SC, Vandelli A. Early Jones PA, et al. Predicting recovery in patients suffering from
predictors of unfavourable outcome in subjects with moderate head traumatic brain injury by using admission variables and physi-
injury in the emergency department. J Neurol Neurosurg Psychiatr ological data: a comparison between decision tree analysis and
2008; 79: 567573. logistic regression. J Neurosurg 2002; 97: 326336.
37. Hebb MO, McArthur DL, Alger J, Etchepare M, Glenn TC, 56. Carter BG, Butt W. Are somatosensory evoked potentials the best
Bergsneider M, et al. Impaired percent alpha variability on con- predictor of outcome after severe brain injury? A systematic review.
tinuous electroencephalography is associated with thalamic injury Intensive Care Med 2005; 31: 765775.
and predicts poor long-term outcome after human traumatic brain 57. Marshall LF, Becker DP, Bowers SA, Cayard C, Eisenberg H, Gross
injury. J Neurotrauma 2007; 24: 579590. CR, et al. The National Traumatic Coma Data Bank. Part 1: design,
38. Moreno JA, Mesalles E, Gener J, Tomasa A, Ley A, Roca J, et al. purpose, goals, and results. J Neurosurg 1983; 59: 276284.
Evaluating the outcome of severe head injury with transcranial 58. Carney NA, Ghajar J. Guidelines for the management of severe
Doppler ultrasonography. Neurosurg Focus 2000; 8: e8. traumatic brain injury. Introduction. J Neurotrauma 2007; 24
39. Rovlias A, Kotsou S. The blood leukocyte count and its prognostic Suppl 1: S12.
significance in severe head injury. Surg Neurol 2001; 55: 190196. 59. Levin HS, Boake C, Song J, McCauley S, Contant C, Diaz-Marchan
40. Mattioli C, Beretta L, Gerevini S, Veglia F, Citerio G, Cormio M, P, et al. Validity and sensitivity to change of the extended Glas-
et al. Traumatic subarachnoid hemorrhage on the computerized gow Outcome Scale in mild to moderate traumatic brain injury. J
tomography scan obtained at admission: a multicenter assessment Neurotrauma 2001; 18: 575584.
of the accuracy of diagnosis and the potential impact on patient 60. van Baalen B, Odding E, van Woensel MP, Roebroeck ME. Reliability
outcome. J Neurosurg 2003; 98: 3742. and sensitivity to change of measurement instruments used in a trau-
41. Lubillo S, Bolanos J, Carreira L, Cardenosa J, Arroyo J, Manzano matic brain injury population. Clin Rehabil 2006; 20: 686700.
J. Prognostic value of early computerized tomography scanning 61. Choi SC, Clifton GL, Marmarou A, Miller ER. Misclassification
following craniotomy for traumatic hematoma. J Neurosurg 1999; and treatment effect on primary outcome measures in clinical trials
91: 581587. of severe neurotrauma. J Neurotrauma 2002; 19: 1722.
42. Kelly DF, Martin NA, Kordestani R, Counelis G, Hovda DA, 62. Wilson JT, Slieker FJ, Legrand V, Murray G, Stocchetti N, Maas AI.
Bergsneider M, et al. Cerebral blood flow as a predictor of outcome Observer variation in the assessment of outcome in traumatic brain
following traumatic brain injury. J Neurosurg 1997; 86: 633641. injury: experience from a multicenter, international randomized clini-
43. Le Roux PD, Newell DW, Lam AM, Grady MS, Winn HR. Cerebral cal trial. Neurosurgery 2007; 61: 123128; discussion 128129.
arteriovenous oxygen difference: a predictor of cerebral infarc- 63. Tanriverdi T, Uzan M, Sanus GZ, Baykara O, Is M, Ozkara C,
tion and outcome in patients with severe head injury. J Neurosurg et al. Lack of association between the IL1A gene (889) poly-
1997; 87: 18. morphism and outcome after head injury. Surg Neurol 2006; 65:
44. Vos PE, Lamers KJ, Hendriks JC, van Haaren M, Beems T, 710; discussion 10.
ZimmermanC, et al. Glial and neuronal proteins in serum predict 64. Hiler M, Czosnyka M, Hutchinson P, Balestreri M, Smielewski P,
outcome after severe traumatic brain injury. Neurology 2004; 62: Matta B, et al. Predictive value of initial computerized tomography
13031310. scan, intracranial pressure, and state of autoregulation in patients
45. Hukkelhoven CW, Steyerberg EW, Habbema JD, Farace E, Marmarou with traumatic brain injury. J Neurosurg 2006; 104: 731737.
A, Murray GD, et al. Predicting outcome after traumatic brain 65. Shutter L, Tong KA, Holshouser BA. Proton MRS in acute trau-
injury: development and validation of a prognostic score based on matic brain injury: role for glutamate/glutamine and choline for
admission characteristics. J Neurotrauma 2005; 22: 10251039. outcome prediction. J Neurotrauma 2004; 21: 16931705.
46. Marshall LF. A new classification of head injury based on compu- 66. Glenn TC, Kelly DF, Boscardin WJ, McArthur DL, Vespa P, Oertel
terized tomography. J Neurosurg 1991; 75: S14S20. M, et al. Energy dysfunction as a predictor of outcome after moder-
47. Fabbri A, Servadei F, Marchesini G, Stein SC, Vandelli A. Early ate or severe head injury: indices of oxygen, glucose, and lactate
predictors of unfavourable outcome in subjects with moderate head metabolism. J Cereb Blood Flow Metab 2003; 23: 12391250.
injury in the emergency department. J Neurol Neurosurg Psychiatr 67. Della Corte F, Mancini A, Valle D, Gallizzi F, Carducci P, Mignani
2008; 79: 567573. V, et al. Provocative hypothalamopituitary axis tests in severe head
48. Della Corte F, Mancini A, Valle D, Gallizzi F, Carducci P, Mignani injury: correlations with severity and prognosis. Crit Care Med
V, et al. Provocative hypothalamopituitary axis tests in severe head 1998; 26: 14191426.
injury: correlations with severity and prognosis. Cri Care Med 68. Ellenberg JH, Levin HS, Saydjari C. Posttraumatic Amnesia as a
1998; 26: 14191426. predictor of outcome after severe closed head injury. Prospective
49. Rae-Grant AD, Eckert N, Barbour PJ, Castaldo JE, Gee W, assessment. Arch Neurol 1996; 53: 782791.

J Rehabil Med 42