Systematic review of multivariable prognostic models for mild traumatic brain injury.

Journal of Neurotrauma

rR ee

Systematic review of multivariable prognostic models for mild traumatic brain injury

ie ev Journal:

Manuscript ID:

Manuscript Type:

Journal of Neurotrauma NEU-2014-3600.R1 Reviews

w

Date Submitted by the Author:

Complete List of Authors:

26-Aug-2014 Silverberg, Noah; UBC, Medicine Gardner, Andrew; University of Newcastle, School of Medicine and Public Health Brubacher, Jeff; UNiversity of British Columbia, Emergency Medicine Panenka, William; UNiversity of British Columbia, Li, Jun Jian; Simon Fraser University, Iverson, Grant; Harvard Medical School, Department of Physical Medicine and Rehabilitation; Harvard Medical School, Red Sox Foundation and Massachusetts General Hospital Home Base Program

Keywords:

/N

ly

On

HEAD TRAUMA, TRAUMATIC BRAIN INJURY, RECOVERY

ot fo bu

tri

is

rD Mary Ann Liebert, Inc, 140 Huguenot Street, New Rochelle, NY 10801

n

tio

Journal of Neurotrauma Systematic review of multivariable prognostic models for mild traumatic brain injury (doi: 10.1089/neu.2014.3600) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

rP Fo Journal of Neurotrauma: http://mc.manuscriptcentral.com/neurotrauma

Page 1 of 61


Systematic review of multivariable prognostic models for mild traumatic brain injury

rR ee

Noah D. Silverberg, PhD* University of British Columbia & GF Strong Rehab Centre Address: 4255 Laurel St., Vancouver, British Columbia, V5Z 2G9, Canada. Phone: 1-604-734-1313. Email: [email protected]

ie ev

Andrew J. Gardner, PsyD

w

Centre for Translational Neuroscience and Mental Health, School of Medicine and Public Health, University of Newcastle

On

Address: Level 5, McAuley Building, Calvary Mater Hospital, Waratah, NSW 2298, Australia Phone: +61 2 4921 5000; Fax: +61 2 4985 4200; Email: [email protected].

Jeffrey R. Brubacher, MD

ot

/N

ly

Department of Emergency Medicine, University of British Columbia, Address: Room 3300 910 West 10th Avenue Vancouver, British Columbia, Canada, V5Z 1M9

fo rD

Phone: 1-604-875-5242; Fax: 1-604-875-4872; Email: [email protected]

William J. Panenka, MD

Vancouver, British Columbia, Canada V6T 2A1

bu

Address: Detwiller Pavilion, 2255 Wesbrook Mall

tri

Department of Psychiatry, University of British Columbia,

is

Mary Ann Liebert, Inc, 140 Huguenot Street, New Rochelle, NY 10801

n

tio


rP Fo 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Page 2 of 61


rR ee Jun Jian Li, BSc

Faculty of Health Sciences, Simon Fraser University Address: Blusson Hall, Room 11300, 8888 University Drive, Simon Fraser University, Burnaby, British Columbia, Canada, V5A 1S6

ie ev


Grant L. Iverson, PhD

w

Department of Physical Medicine and Rehabilitation, Harvard Medical School; Spaulding

On

Rehabilitation Hospital; Massachusetts General Hospital Sport Concussion Clinic; & Red Sox Foundation and Massachusetts General Hospital Home Base Program

ly

Address: Center for Health and Rehabilitation; 79/96 Thirteenth Street, Charlestown Navy Yard, Charlestown, MA, USA 02129

ot

/N

Phone: 517-952-6194; Email: [email protected]

*Corresponding author

Keywords: traumatic brain injury, concussion, prognosis, systematic review

bu

tri

is

rD

Running title: Mild traumatic brain injury prognostic models

fo


n

tio


rP Fo 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 3 of 61


Abstract

rR ee

Prognostic models can guide clinical management and increase statistical power in clinical trials. The availability and adequacy of prognostic models for mild traumatic brain injury (MTBI) is uncertain. The present study aimed to (i) identify and evaluate multivariable prognostic models for MTBI, and (ii) determine which pre-, peri-, and early post-injury variables have independent

ie ev

prognostic value in the context of multivariable models. An electronic search of MEDLINE, PsycINFO, PubMed, EMBASE, and CINAHL databases for English-language MTBI cohort

w

studies from 1970-2013 was supplemented by Web of Science citation and hand searching. This search strategy identified 7,789 articles after removing duplicates. Of 182 full-text articles

On

reviewed, 26 met eligibility criteria including (i) prospective inception cohort design, (ii) prognostic information collected within one month post-injury, and (iii) 2+ variables combined

ly

to predict clinical outcome (e.g., postconcussion syndrome) at least one month later. Independent

/N

reviewers extracted sample characteristics, study design features, clinical outcome variables,

ot

predictor selection methods, and prognostic model discrimination, calibration, and crossvalidation. These data elements were synthesized qualitatively. The present review found no

fo

multivariable prognostic model that adequately predicts individual patient outcomes from MTBI.

rD

Suboptimal methodology limits their reproducibility and clinical usefulness. The most robust prognostic factors in the context of multivariable models were pre-injury mental health and early

is

post-injury neuropsychological functioning. Women and adults with early post-injury anxiety

tri

also have worse prognoses. Relative to these factors, the severity of MTBI had little long-term prognostic value. Future prognostic studies should consider a broad range of biopsychosocial predictors in large inception cohorts.

bu


n

tio


rP Fo 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Page 4 of 61

Introduction

rR ee

Mild traumatic brain injuries (MTBIs) are very common. They account for at least 80% of all TBIs and have an annual incidence of more than 500 per 100,000.1–3 There is a large and

compelling literature on the clinical course of MTBI in civilian trauma patients, athletes, active duty military service members, and veterans, summarized in several recent systematic reviews.4– 11

ie ev

It is now well-established that patients with MTBI typically recover within days to weeks but a

minority (of debated size) experience protracted or incomplete recovery. Persistent symptoms are

w

associated with disability and high health service use.12–15 Identifying individuals on a trajectory of poor outcome soon after MTBI is therefore essential.

On

As with other health conditions, flagging patients who are likely to have a good or poor outcome

ly

enables identification of those who require monitoring or who can benefit from early therapeutic

/N

interventions. Prognostic models can also enhance statistical power in randomized controlled

ot

trials through risk stratification and covariate adjustment.16,17 A single predictive variable is rarely sufficient, but multiple predictors can be combined into a multivariable prognostic model

fo

to accurately gauge risk for a clinical outcome of interest.18 By evaluating multiple predictor

rD

variables in the same sample, prognostic models can also clarify the relative importance of each and suggest causal links to good or poor outcome.19

tri

is A systematic review of multivariable prognostic models for moderate-to-severe traumatic brain

bu

injury identified the Glasgow Coma Scale, pupil reactivity, and computed tomography (CT)

findings as the best predictors of outcome.20 These crude indicators lose their prognostic power


n

tio


rP Fo 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 5 of 61


when applied to patients with mild TBI21,22. MTBI is by definition associated with a ceiling or near ceiling score on the Glasgow Coma Scale.23,24 Pupil and CT abnormalities are typically

rR ee

absent. Clearly, more sensitive biomarkers will be necessary to improve prognostics for MTBI, and several (e.g., advanced neuroimaging techniques) appear promising in exploratory studies.25

Psychosocial factors such as early post-injury anxiety also appear to influence outcome from MTBI.9,26 However, the relative importance of these biopsychosocial factors and their joint

ie ev

ability to predict outcome from MTBI is not yet clear.

w

The primary aim of the present study is to identify and evaluate existing multivariable prognostic models appropriate for clinical and research applications. To the authors’ knowledge, this is the

On

first systematic review of multivariable prognostic models for MTBI. A further aim is to clarify which individual factors independently predict MTBI outcome. Other recent systematic reviews

ly

extracted information about correlates of MTBI outcome,19,27 most notably those performed by

/N

the International Collaboration on MTBI Prognosis and reported in a special issue of the

ot

Archives of Physical Medicine and Rehabilitation.19 but bBy focusing on multivariable models derived from inception cohorts, the present study adds knowledge about the unique and relative prognostic value of each factor.

fo

bu

tri

is

rD Mary Ann Liebert, Inc, 140 Huguenot Street, New Rochelle, NY 10801

n

tio


rP Fo 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Page 6 of 61

Materials and Methods

rR ee

The protocol for the present study was registered with PROSPERO and can be accessed here: http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42013003631.

Eligibility criteria

ie ev

Studies with participants who sustained an MTBI according to the most widely used diagnostic criteria23,24 or a compatible definition of MTBI were included. Studies could have included

w

patients with MTBI exclusively or reported separate analyses for an MTBI subgroup. Studies that analyzed a combined sample of MTBI and non-head injury controls were also eligible, if

On

diagnosis (MTBI vs. non-head injury) was included as a covariate. Eligible studies must have reported on a global clinical outcome (e.g., symptomatic status). Narrow “outcomes” such as a

ly

medical complication (e.g., seizures) or specific symptom (e.g., fatigue) were excluded.

ot

/N

To maximize internal validity, only studies that had a prospective inception cohort design, recruiting patients from an acute care setting (e.g., Emergency Department), and whose cohort

fo

had at least 30 participants were included. Studies could have focused on school-aged children,

rD

adolescents, and/or adults for inclusion, enabling a comparison of similarities and differences in prognostic factors across the age span. Eligible studies must have combined at least two

is

variables collected in the acute to subacute phase ( 0.80) by standard interpretive criteria. The model predicting PCS in Stujelmeir et al42 fell to 0.70 after correcting for optimism with bootstrapping (an internal validation

fo

technique). The prognostic model reported by Topolovec-Vranic et al44 may have been

rD

artificially inflated because the physician classifying cases as normal or abnormal at follow-up appears to have been blind to only one of the predictors in the model.

tri

is An additional four models with dichotomous outcomes in four studies reported total

classification accuracy but not the area under the Receiver Operating Characteristic curve.33,45–47

bu

Total classification accuracy ranged from 66% to 85%. Additionally, the logistic regression


n

tio


rP Fo 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 13 of 61


model predicting one month PCS outcome in Bazarian et al35 achieved 90% classification

accuracy but only in the 42% of the sample that could be classified according to a subsequent

rR ee

publication.36 A prognostic model is considered effective when its classification error rate is less than the base rate of the outcome of interest. There was only one study47 (with high attrition and low event:predictor ratio) where the prognostic model was more than 5% better than the default assumption that all participants will achieve a good outcome.

ie ev

Six studies analyzed a continuous outcome and all reported total R2 values for their final models,

w

which ranged widely from 0.06 to 0.89. The only two studies achieving a total R2 > 0.4041,48 shared several features: large number of predictors, small sample sizes (i.e., inadequate

On

participant:variable ratio), stepwise predictor selection, and postconcussion symptom severity score as the outcome.

/N

ly

Creating clinician-friendly prediction rules from regression coefficients can facilitate translation

ot

of prognostic models into patient care.19 This was done in only one eligible study.42

Independent prognostic value of predictors

fo rD

Table 2 summarizes the pre-, peri-, and early post-injury prognostic variables considered across the eligible studies and their unique contributions to multivariable prediction. Many pre- (e.g.,

is

age, gender, prior MTBIs), peri- (e.g., LOC, CT findings, mechanism of injury), and post-injury

tri

variables (e.g., anxiety, neuropsychological testing, serum biomarkers) were considered. The

most consistent independent predictors of poor outcome were female gender, pre-injury mental health, post-injury anxiety and traumatic stress, and post-injury cognitive functioning.

bu


n

tio


rP Fo 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Page 14 of 61

Discussion

rR ee

The present study systematically reviewed the literature for multivariable prognostic models for MTBI. The primary aim was to identify models that can be used in clinical practice (e.g., triaging

patients at risk for poor outcome) and research (e.g., risk stratification in clinical trials). The review focused on inception cohort designs to provide the strongest level evidence and because

ie ev

using information collected soon after the injury to predict long-term outcome is most clinically relevant. Although we identified 26 eligible studies involving 6,939 participants aged 5 to 80+

w

from three continents, the methodology used to develop them was largely suboptimal and their predictive accuracy was low. In line with our secondary aim, we identified several pre-, peri-,

On

and early post-injury prognostic factors that independently predicted MTBI outcome in multivariable models.

ot

/N

Multivariable prognostic models

ly

Eligible studies almost universally used bivariate screening and stepwise regression approaches to select variables into a prognostic model from a larger set of candidate predictors. These

fo

methods result in unstable predictor selection and bias in the covariate coefficients, especially at

rD

lower subject per predictor ratios.49,50 Subject per predictor ratios were low (10)?

Outcome

Anxiety, HADSDepression, Behavioral Response to Illness Questionnaire All-ornothing scale, BRIQ limiting, litigation, Brief Social Support Questionnaire availability and satisfaction scales.

stepwise backward logistic regression

Perception Questionnaire (negative illness perceptions composite) [OR=1.05, 1.011.10]

RPQ

iew

n:k > 10

On

Assessed within 2 weeks.

1999, 47% at 6 months

ly/

Age, gender, GCS, abnormal pupillary response, hypotensive episode, hypoxic episode, Abbreviated Injury Scale of the Head, Injury Severity Score, Injury Severity ScoreExtracranial, PTA (y/n), LOC (y/n), alcohol intoxication, use of anticoagulants, 17 CT parameters. Assessed in ED.

No

Stepwise forward logistic regression

CT only model: Number of hemorrhagic contusions [OR=1.9, 1.2-3.1], presence of facial fractures [OR=1.8, 1.23.0]. Combined model: Age [OR 1.02, 1.01-1.03], Injury Severity Score-Extracranial [OR=1.07, 1.04-1.1], ethanol intoxication [OR=0.5,0.3-0.8], number of hemorrhagic contusions [OR=1.9, 1.2-3.1], facial fractures [OR=1.5, 0.9-2.5].

Unfavorable outcome = GOS-E < 7 vs 7-8

tf

or

Di

n:k > 10 for both models

Note: 99% confidence intervals reported.

GCS 15: 75% +LOC: 43% +CT:20% Mechanism of injury was Page 4 of 16

Overall model accuracy


CT only model: AUC=0.57, 95% CI= 0.52-0.61; Combined model: AUC=0.69 95% CI = 0.65-0.73

st

rib

ut

io

n

Journal of Neurotrauma Systematic review of multivariable prognostic models for mild traumatic brain injury (doi: 10.1089/neu.2014.3600) as been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ fro

Fo 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48


rP

First author, year, country, age group King, 199610 England Adult

ee

Sample characteristics

rR

MVA:58% MTBI definition: Head injury with PTA 10 for both models McLean, 200913 USA Adult

MTBI definition: Traumatic injury with GCS 13-15 and any of: LOC for ≤ 30 min due to trauma, PTA, or ≥ 2 postconcussive symptoms. Excluded: -Prior MTBI? N -Psychiatric history? N -Extracranial injury? N

Page 6 of 16

507, 29% lost at 12 months (includes non-head injury controls)

NR. Interactions between predictors considered. Assessed within 24 hours post-injury.

Univariate screening, then enter all.

Pre-injury mental health (SF-36 Mental) [B=-0.058, p 10 for all models

Di

st

Assessed within 3 weeks (80% within 1-2 weeks) MTBI definition: Head trauma with GCS=13-15 at ED presentation, no


n:k NR

Excluded: -Prior MTBI? N -Psychiatric history? Y -Extracranial injury? Y

Meares, 201115

Outcome

Severity Score [ns], head injury x RPQ interaction [ns]

ev

McNally, 201314

Page 52 of 61

NR, 62 retained at 3 months

Diagnosis (MTBI vs nonhead injury), time, age, gender, IQ, Consistent Long-Term Retrieval,

Univariate screening

Acute Stress Disorder Scale Dissociative Total [OR=1.05, 1.00-1.10], Pre-injury mood or anxiety disorder on Mini-


rib

PCS (y/n), defined as 3+ symptoms rated as at least of

ut NR

io

n


Fo

Page 53 of 61

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

rP


First author, year, country, age group Adult

ee


rR

intracranial lesion on CT, and any of: confusion or disorientation, LOC (no more than 30 minutes), PTA (no greater than 24 hours), or transient neurological abnormalities.

N, attrition, timing of followup

ev

Excluded: -Prior MTBI? N -Psychiatric history? N -Extracranial injury? N

Norway Adult

MTBI definition: Head trauma with GCS=13-15 at ED admission, no focal neurological deficit, and any of: LOC or retrograde amnesia. Excluded: -Prior MTBI? Y -Psychiatric history? Y -Extracranial injury? Y

Predictor selection method

Symbol Digit Modalities Test-oral, Sequential Reaction Time 1 of the California Computerized Assessment Package, acute pain, Acute Stress Disorder Scale Dissociative cluster and total score, MiniInternational Neuropsychiatric Inventory pre-injury depression/anxiety disorder and substance use disorder, interactions between predictors.

iew

On

ly/

GCS 15: 86% +LOC: NR +CT: 0% Mechanism of injury was MVA: 82% Muller, 200916

Candidate predictors, timing of assessment

Assessed within 2 weeks (mean = 4.9 days) 59, 7% lost at 6 months

Neuropsychological impairment index at baseline, GCS (15 vs 14 vs 13), CT (normal vs abnormal), MRI (normal vs abnormal), S-100B (3 different levels), APOE-4 allele (present vs absent) Assessed within 48 hours post-injury.

Predictors in final model, adequate ratio of participants to predictors (n:k >10)?

Outcome

International Neuropsychiatric Inventory [OR=2.99, 1.38-6.45], females in non-head injury group [OR = 10.05, 2.71-37.30], MTBI in females [OR=0.17, 0.04-0.68], acute pain [ns], Acute Stress Disorder Scale Dissociative cluster [ns], Symbol Digit Modalities Testoral [ns], time by Symbol Digit Modalities Test-oral [ns]

"often."

Low n:k

No

Enter all

tf

Neuropsychological impairment index [coefficient of change = 0.16, -0.18 to -0.14]; GCS [ns], CT [ns], MRI [ns], S100B [ns], APOE-4 [coefficient of change = 0.05, 0.001-0.11]

or

Change in neuropsychologi cal impairment index from baseline to 6 months post

Di

Note: For all predictors, model was adjusted for age and all other predictors except APOE-4

NR

st

rib

ut

GCS 15: 73% +LOC: NR +CT: 18.6% Mechanism of injury was MVA:NR Page 8 of 16



io

n


Fo 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48


rP

First author, year, country, age group

ee


rR

Olsson, 201317

MTBI definition: ACRM

Australia


Pediatric





Outcome


150, 5% lost at 6 months

Pre-injury PCS symptoms, pre-injury physical healthrelated quality of life, preinjury psychosocial healthrelated quality of life, Wechsler Abbreviated Scale of Intelligence IQ, pre-injury anxiety, preinjury depression, preinjury somatic, pre-injury social, pre-injury Impact of Event Scale–Revised avoidance, hyperaoursal, and intrusion scales.

Stepwise forward logistic regression in each subgroup

In subgroup with no imaging: Pre-injury psychosocial healthrelated quality of life [B=-0.55, p 10

No

Stepwise discriminate function analysis

Assessed within 1 week.

Australia Page 9 of 16

MTBI definition: ACRM with no focal neurological signs or intracranial abnormality

tf

Prior TBI [25% variance], preinjury stressors [10% variance]. n:k > 10

or

Di

st

GCS 15: NR +LOC: NR +CT: 0% Mechanism of injury was MVA:10% Ponsford, 201219

Page 54 of 61


Gender, age, pre-injury psychiatric disorder, preinjury physical healthrelated quality of life (SF-

Univariate screening, then enter all in ordinal

Model predicting 3 month PCS from pre-injury and acute: Preinjury psychiatric disorder [OR=2.56, p=.006], pre-injury


rib

PCS severity (ordinal), where continuous PCS symptom scores

ut NR

io

n


Fo

Page 55 of 61

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

rP



ee


on CT.

Adult

rR


ev


GCS 15: NR +LOC: NR +CT: 0% Mechanism of injury was MVA:10%




Outcome

36), PTA duration, ImPACT Verbal and Visual Memory acute, diagnosis (MTBI vs nonhead injury), ImPACT Verbal and Visual Memory at 1 week, Pain visual analog scale at 1 week, SF36 Physical at 1 week, HADS-Anxiety at 1 week, narcotics/ analgesics use

logistic regression

physical health-related quality of life [OR=1.09, p=.004], diagnosis [ns], ImPACT Verbal Memory [ns], ImPACT Visual Memory [ns], gender [ns], age [ns], PTA [ns]

were divided into 3 categories with equal frequency in each

iew

On

ly/

Assessed within 48 hours and again at 1 week postinjury.

Savola, 200320 Finland Adult

MTBI definition: Definite trauma to head or altered consciousness (LOC no more than 30 min), with GCS=13-15 on admission and no focal neurological deficits. Excluded: -Prior MTBI? N -Psychiatric history? N -Extracranial injury? N

199, 14% lost at 1 month

Skull fracture, S100B >0.49 ug/l, dizziness in ED, headache in ED, age, LOC, PTA, extracranial injury, prior head injury, employment status, insurance, psychotropic drugs, heavy alcohol use, smoking prior use of illicit drugs. Assessed in ED.

Model predicting PCS at 3 months from injury and 1-week variables: HADS-Anxiety [OR=1.18, p.49 [OR=5.5, 1.6-18.6], dizziness in ED [OR=3.1, 1.2-8.0], headache in ED [OR=2.6, 1.0-6.5], age [OR=1.05, 1.01-1.10]

tf

Low n:k

or

Di

Note: Model was adjusted for adjusted for gender, educational status, and “other variables”

GCS 15: NR +LOC: 40% +CT: 5% Mechanism of injury was MVA:30% Page 10 of 16



PCS (y/n), defined as 1+ symptom that started the day of the injury, lasted at least one month, and was more severe vs pre-injury.

NR

st

rib

Note: Interviewer blinded to S100B value only.

ut

io

n


Fo 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48


rP

First author, year, country, age group Stalnacke, 200521 Sweden Adult

ee


rR

MTBI definition: Blunt head trauma with GCS 13–15 on arrival to ED and diagnosis of MTBI leading to hospitalization for observation.





Outcome


88, 22% at 15 months

Serum S100 B, serum neuron-specific enolase, dizziness in ED, headache in ED, nausea in ED.

Univariate screening followed by stepwise forward logistic regression

PCS model: None selected.

PCS (y/n), defined as 1+ symptom

NR

ev

Exclusion criteria NR.

iew

Assessed in ED.

GCS 15: 86% +LOC: 61% +CT: NR Mechanism of injury was MVA: 2%

Stulemeijer, 200822 Netherlands Adult

MTBI definition: Head trauma with or without LOC (up to 30 min) or PTA, hospital admission GCS of 13–15. Excluded: -Prior MTBI? N -Psychiatric history? N -Extracranial injury? N GCS 15: 75% +LOC: 64% +CT: 20% Mechanism of injury was MVA: 55%

Page 11 of 16

Page 56 of 61

Disability model: S100B [OR = 10.0, 1.46-68.91], dizziness in ED [OR=4.2, 1.01-17.32], nausea [ns] Life satisfaction model: nausea in ED [OR=0.2, 0.06-0.53]

Disability (y/n), defined as any vs no disability on the Rivermead Head Injury Follow-Up Questionnaire

On

n:k > 10 for both models.

ly/


Age, gender, education, pre-injury emotional problems, pre-injury physical comorbidity, prior head injury, admission GCS (15 or 26 or below), selfefficacy ( 10 for PCS model, low n:k for return to work model


st

Return to work, defined as full vs less than full

rib

PCS model: AUC=0.82, correction with bootstrap AUC=0.73

Note: Favorable outcomes coded as 1.

Return to work model: AUC=0.79, correction with bootstrap AUC=0.70

ut

io

n


Fo

Page 57 of 61

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

rP



ee


rR



Outcome


Stepwise logistic regression

Male gender [OR=0.20, 0.050.81], +LOC [OR = 8.11, 1.9533.74], RPQ at 3 days postinjury [OR = 1.13, 1.05-1.21], serum neuron-specific enolase level >14.5 ug/L [OR=5.32, 1.42-19.91]

Physician clinical assessment (abnormal vs normal)

AUC = 0.90

Assessed within 1 month post-injury (mean = 9 days)

iew

TopolovecVranic, 201123

MTBI definition: Non-penetrating head trauma with GCS 13-15

141, 33% lost at 6 weeks

Canada

Excluded: -Prior MTBI? N -Psychiatric history? Y -Extracranial injury? Y

Note: 95 cases at followup includes those discharge d for being normal

GCS 15: 93% +LOC: 35% +CT: NR Mechanism of injury was MVA: 37%


Questionnaire (>20 or below)

ev

Adult


Gender, age, marital status, years of education, employment status, time to blood draw, serum S100B level, serum neuronspecific enolase level, mechanism of injury, LOC (y/n), prior head trauma, Galveston Orientation and Amnesia Test score, dizziness in ED, headache in ED, vomiting in ED, RPQ score at day 3.

On

ly/

No

tf

Assessed in ED and at 3 days post-injury. van Veldhoven, 201124 USA Adult

MTBI definition: ACRM Excluded: -Prior MTBI? Y -Psychiatric history? Y -Extracranial injury? N GCS 15: NR +LOC: NR +CT: 49% Mechanism of injury was MVA: NR


Age, CT (abnormal finding vs not), preinjury depression on Center for Epidemiological StudiesDepression Scale, preinjury exposure to traumatic events rated on the Stressful Life Events Questionnaire, pre-injury posttraumatic stress symptoms on the PTSD Checklist-Civilian version. Assessed within 2 weeks

Page 12 of 16

Low n:k

Note: Physicians were blinded to the biomarker values but likely not other predictors.

Hierarchical entry (age, then CT, then preinjury depression, then preinjury stressful events and posttraumatic stress symptoms).

or

Di

SF-36 Physical model: Age [ns], CT findings [ns], pre-injury depression [R2 change = .038], pre-injury stressful life events [R2 change = .067], pre-injury post-traumatic stress [R2 change = 0.021] SF-36 Mental model: Age [ns], CT findings [ns], pre-injury depression [ns], pre-injury stressful life events [R2 change = .028], pre-injury posttraumatic stress [ns]


SF-36 Physical (continuous), SF-36 Mental (continuous)

st

rib

SF-36 Physical model: R2=0.15; SF36 Mental model: R2=0.06

ut

io

n


Fo 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48


rP


Whittaker, 200725

ee


rR

England

MTBI definition: Head injury with GCS 13-15, LOC < 20 min, PTA < 24 hours.

Adult

Exclusion criteria NR.


USA Adult




PCS (y/n), defined as 1+ cognitive symptom, 3+ somatic/ emotional symptoms, and functional impairment on the Sydney Psychosocial Reintegration Scale.

85% total classification accuracy, Nagelkerke R2 = 50.1%

GOS-E (ordinal) for models, but GOS-E (dichotomous, 8 vs. 10 for both models

ev

MTBI definition: Head trauma with GCS 13-15 upon admission to ED and triage to CT; LOC no more than 30 min and PTA no greater than 24 hours. Excluded: -Prior MTBI? N -Psychiatric history? N -Extracranial injury? N GCS 15: 79% +LOC: NR +CT: 27% Mechanism of injury was MVA:NR


GCS, LOC, PTA, HADSdepression, posttraumatic stress (Impact of Events Scale), RPQ, and 5 scales from a modified version of the Illness Perception Questionnaire-Revised.

iew

On

Assessed 1-3 weeks postinjury.

135 at 3 months (attrition NR)

Screening with bivariate and hierarchical regression, then Purposeful Selection (Hosmer & Lemeshow)

ly/

Age, gender, GCS at ED arrival, presence of LOC or PTA, prior TBI, education (10)?

outcome clinical/demographic and clinical/demographic+CT models. Low n:k for others. Parental anxiety in ED [ns], history of Attention Deficit Hyperactivity Disorder [ns] n:k > 10

Outcome


PCS (y/n), defined as 3+ symptoms in comparison to pre-injury

NR

On

Assessed in ED.

Excluded: -Prior MTBI? N -Psychiatric history? N -Extracranial injury? Y GCS 15: NR +LOC: NR +CT: 0% Mechanism of injury was MVA:3%

ly/

No

tf

or

ACRM = Mild Traumatic Brain Injury Committee of the American Congress of Rehabilitation Medicine (1993) definition of MTBI as head trauma with GCS=13-15 and any of: LOC (30 min or less), amnesia before or after (PTA no greater than 24 hours), alteration in mental status, focal neurological deficit(s). AUC = Area under the curve (aka. C-statistic) ED = Emergency Department GCS = Glasgow Coma Scale score GOS-E = Glasgow Outcome Scale – Extended HADS = Hospital Anxiety and Depression Scale LOC = Loss of consciousness. NA = Not applicable. NR = Not reported or sufficiently described in article to enable data element extraction. n:k = ratio of participants to prognostic variables. For continuous outcomes, n = N = total sample size. For binary and ordinal outcomes, n = number of cases with least frequent outcome. ns = non-significant (e.g., 95% confidence interval for odds ratio includes 1.00) OR = Odds ratio with 95% confidence interval, where available, otherwise with p value. PCS = Postconcussion Syndrome PTA = Post-traumatic amnesia RPQ = Rivermead Postconcussion Symptom Questionnaire

Di

st

rib

ut

Page 14 of 16


io

n


Fo 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48


Page 60 of 61

rP

References

ee

1. Babcock, L., Byczkowski, T., Wade, S.L., Ho, M., Mookerjee, S., and Bazarian, J.J. (2013). Predicting postconcussion syndrome after mild traumatic brain injury in children and adolescents who present to the emergency department. JAMA Pediatr 167, 156-161. 2. Babikian, T., McArthur, D., and Asarnow, R.F. (2013). Predictors of 1-month and 1-year neurocognitive functioning from the UCLA longitudinal mild, uncomplicated, pediatric traumatic brain injury study. J Int Neuropsychol Soc 19, 145-154. 3. Bazarian, J.J., Wong, T., Harris, M., Leahey, N., Mookerjee, S., and Dombovy, M. (1999). Epidemiology and predictors of post-concussive syndrome after minor head injury in an emergency population. Brain Inj 13, 173-189. 4. Bazarian, J.J., and Atabaki, S. (2001). Predicting postconcussion syndrome after minor traumatic brain injury. Acad Emerg Med 8, 788-795. 5. Dischinger, P.C., Ryb, G.E., Kufera, J.A., and Auman, K.M. (2009). Early predictors of postconcussive syndrome in a population of trauma patients with mild traumatic brain injury. J Trauma 66, 289-296. 6. Faux, S., Sheedy, J., Delaney, R., and Riopelle, R. (2011). Emergency department prediction of post-concussive syndrome following mild traumatic brain injury--an international cross-validation study. Brain Inj 25, 14-22. 7. Heitger, M.H., Jones, R.D., Dalrymple-Alford, J.C., Frampton, C.M., Ardagh, M.W., and Anderson, T.J. (2007). Mild head injury--a close relationship between motor function at 1 week post-injury and overall recovery at 3 and 6 months. J Neurol Sci 253, 34-47. 8. Hou, R., Moss-Morris, R., Peveler, R., Mogg, K., Bradley, B.P., and Belli, A. (2012). When a minor head injury results in enduring symptoms: a prospective investigation of risk factors for postconcussional syndrome after mild traumatic brain injury. J Neurol Neurosurg Psychiatry 83, 217-223. 9. Jacobs, B., Beems, T., Stulemeijer, M., van Vugt, A.B., van der Vliet, T.M., Borm, G.F., and Vos, P.E. (2010). Outcome prediction in mild traumatic brain injury: age and clinical variables are stronger predictors than CT abnormalities. J Neurotrauma 27, 655-668. 10. King, N.S. (1996). Emotional, neuropsychological, and organic factors: their use in the prediction of persisting postconcussion symptoms after moderate and mild head injuries. J Neurol Neurosurg Psychiatry 61, 75-81. 11. King, N.S., Crawford, S., Wenden, F.J., Caldwell, F.E., and Wade, D.T. (1999). Early prediction of persisting post-concussion symptoms following mild and moderate head injuries. Br J Clin Psychol 38, 15-25. 12. Lannsjö, M., Backheden, M., Johansson, U., Af Geijerstam, J.L., and Borg, J. (2013). Does head CT scan pathology predict outcome after mild traumatic brain injury? Eur J Neurol 20, 124-129. 13. McLean, S.A., Kirsch, N.L., Tan-schriner, C.U., Sen, A., Frederiksen, S., Harris, R.E., Maixner, W., and Maio, R.F. (2009). Health status, not head injury, predicts concussion symptoms after minor injury. Am J Emerg Med 27, 182-190. 14. McNally, K.A., Bangert, B., Dietrich, A., Nuss, K., Rusin, J., Wright, M., Taylor, H.G., and Yeates, K.O. (2013). Injury versus noninjury factors as predictors of postconcussive symptoms following mild traumatic brain injury in children. Neuropsychology 27, 1-12. 15. Meares, S., Shores, E.A., Taylor, A.J., Batchelor, J., Bryant, R.A., Baguley, I.J., Chapman, J., Gurka, J., and Marosszeky, J.E. (2011). The prospective course of postconcussion syndrome: the role of mild traumatic brain injury. Neuropsychology 25, 454-465.

rR

ev

iew

On

ly/

No

tf

or

Di

st

rib

ut

Page 15 of 16


io

n


Fo

Page 61 of 61

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48


rP

16. Müller, K., Ingebrigtsen, T., Wilsgaard, T., Wikran, G., Fagerheim, T., Romner, B., and Waterloo, K. (2009). Prediction of time trends in recovery of cognitive function after mild head injury. Neurosurgery 64, 698-704. 17. Olsson, K., Lloyd, O.T., Lebrocque, R.M., McKinlay, L., Anderson, V., and Kenardy, J. (2013). Predictors of child post-concussion symptoms at 6 and 18 months following mild traumatic brain injury. Brain Inj 27, 145-157. 18. Ponsford, J., Willmott, C., Rothwell, A., Cameron, P., Ayton, G., Nelms, R., Curran, C., and Ng, K.T. (1999). Cognitive and behavioral outcome following mild traumatic head injury in children. J Head Trauma Rehabil 14, 360-372. 19. Ponsford, J., Cameron, P., Fitzgerald, M., Grant, M., Mikocka-Walus, A., and Schönberger, M. (2012). Predictors of postconcussive symptoms 3 months after mild traumatic brain injury. Neuropsychology 26, 304-313. 20. Savola, O., and Hillbom, M. (2003). Early predictors of post-concussion symptoms in patients with mild head injury. Eur J Neurol 10, 175181. 21. Stålnacke, B., Björnstig, U., Karlsson, K., and Sojka, P. (2005). One-year follow-up of mild traumatic brain injury: Post-concussion symptoms, disabilities and life satisfaction in relation to serum levels of S-100B and neurone-specific enolase in acute phase. J Rehabil Med 37, 300-305. 22. Stulemeijer, M., van der Werf, S., Borm, G.F., and Vos, P.E. (2008). Early prediction of favourable recovery 6 months after mild traumatic brain injury. J Neurol Neurosurg Psychiatry 79, 936-942. 23. Topolovec-Vranic, J., Pollmann-Mudryj, M.A., Ouchterlony, D., Klein, D., Spence, J., Romaschin, A., Rhind, S., Tien, H.C., and Baker, A.J. (2011). The value of serum biomarkers in prediction models of outcome after mild traumatic brain injury. J Trauma 71, S478-86. 24. van Veldhoven, L.M., Sander, A.M., Struchen, M.A., Sherer, M., Clark, A.N., Hudnall, G.E., and Hannay, H.J. (2011). Predictive ability of preinjury stressful life events and post-traumatic stress symptoms for outcomes following mild traumatic brain injury: analysis in a prospective emergency room sample. J Neurol Neurosurg Psychiatry 82, 782-787. 25. Whittaker, R., Kemp, S., and House, A. (2007). Illness perceptions and outcome in mild head injury: a longitudinal study. J Neurol Neurosurg Psychiatry 78, 644-646. 26. Yuh, E.L., Mukherjee, P., Lingsma, H.F., Yue, J.K., Ferguson, A.R., Gordon, W.A., Valadka, A.B., Schnyer, D.M., Okonkwo, D.O., Maas, A.I., Manley, G.T., and TRACK-TBI Investigators. (2013). Magnetic resonance imaging improves 3-month outcome prediction in mild traumatic brain injury. Ann Neurol 73, 224-235. 27. Zemek, R., Clarkin, C., Farion, K.J., Vassilyadi, M., Anderson, P., Irish, B., Goulet, K., Barrowman, N., and Osmond, M.H. (2013). Parental anxiety at initial acute presentation is not associated with prolonged symptoms following pediatric concussion. Acad Emerg Med 20, 10411049.

ee

rR

ev

iew

On

ly/

No

tf

or

Di

st

rib

ut

Page 16 of 16


io

n

Development of prognostic models for patients with traumatic brain injury: a systematic review.

Systematic review of the clinical course, natural history, and prognosis for pediatric mild traumatic brain injury: results of the International Collaboration on Mild Traumatic Brain Injury Prognosis.

Mild traumatic brain injury.

Mild traumatic brain injury.

Systematic review of the risk of Parkinson's disease after mild traumatic brain injury: results of the International Collaboration on Mild Traumatic Brain Injury Prognosis.

Systematic review of the risk of dementia and chronic cognitive impairment after mild traumatic brain injury: results of the International Collaboration on Mild Traumatic Brain Injury Prognosis.

Systematic review of the prognosis after mild traumatic brain injury in adults: cognitive, psychiatric, and mortality outcomes: results of the International Collaboration on Mild Traumatic Brain Injury Prognosis.

Systematic review of self-reported prognosis in adults after mild traumatic brain injury: results of the International Collaboration on Mild Traumatic Brain Injury Prognosis.

Systematic review of return to work after mild traumatic brain injury: results of the International Collaboration on Mild Traumatic Brain Injury Prognosis.

Psychosocial consequences of mild traumatic brain injury in children: results of a systematic review by the International Collaboration on Mild Traumatic Brain Injury Prognosis.

Systematic review of prognosis after mild traumatic brain injury in the military: results of the International Collaboration on Mild Traumatic Brain Injury Prognosis.

Nonsurgical interventions after mild traumatic brain injury: a systematic review. Results of the International Collaboration on Mild Traumatic Brain Injury Prognosis.

Neurobehavioral Outcomes of Mild Traumatic Brain Injury: A Mini Review.

Systematic search and review procedures: results of the International Collaboration on Mild Traumatic Brain Injury Prognosis.

Systematic review of depression in mild traumatic brain injury: study protocol.

Concussion: A Systematic Review.

Mild traumatic brain injury and ADHD: a systematic review of the literature and meta-analysis.

Decompressive Craniectomy for Severe Traumatic Brain Injury: A Systematic Review.

Cervical Spine Involvement in Mild Traumatic Brain Injury: A Review.

Neuroimaging after mild traumatic brain injury: review and meta-analysis.

Factors Associated with Mild Traumatic Brain Injury in Veterans and Military Personnel: A Systematic Review.

Models of mild traumatic brain injury: translation of physiological and anatomic injury.

Advanced neuroimaging of mild traumatic brain injury.

Epidemiology of work-related traumatic brain injury: a systematic review.