Summary of Evidence Table– Definition of DBI
1. What is the definition of DBI?
Reference Number / Source / Study Design / Study Participants / Treatment Interventions / Outcome Measures / Relevant Finding / Limitations/Assessment of Study Quality1 / Dixon 2009 / Review / None / Analysis & Description of treatment algorithm / None / Previously employed definition of devastating brain injury / No study patients ; expert opinion on definition
2. How accurate is prognostication in DBI?
Reference Number / Source / Study Design / Study Participants / Treatment Interventions / Outcome Measures / Relevant Finding / Limitations/Assessment of Study Quality5 / Kotwica 1995 / Retrospective case series / 111 patients with Glasgow Coma Score (GCS) < 3, and CT scans, without significant extracranial injuries. / Relation of CT lesions to outcome; no ICP monitoring. / CT lesions; 2 month GOS / 89% mortality; poor outcomes associated with CT lesions / Case series – no randomization.
Chance of treatment bias.
6 / Towfighi 2011 / Secondary analysis of National Inpatient Database / 2, 537 097 patients with diagnosis of stroke, aged 35-64 / Analysis of demographic patterns / Demographic and clinical characteristics, mortality / Decline in mortality from stroke over 1997 to 2006 .
Better survival among women compared with men. / Large national database. Non-randomized study. Possibility of treatment bias and between center variation.
7 / Ovbiagele 2010 / Secondary analysis of National Inpatient Database / Patients with stroke from 1997-2006 / Analysis of demographic patterns / Demographic and clinical characteristics, mortality / Stroke hospitalizations decreased from 1997 till 2006. Mortality decreased from 11.5% in 1997-98 to 10.3% in 2005-06. / Large national database. Non-randomized study. Possibility of treatment bias and between center variation.
8 / Arabi, 2010 / Single center, comparison of mortality pre- and post- implementation of treatment protocol / 434 comatose patients with TBI, older than 12 years / Implementation of Brain Trauma Foundation guidelines based Protocol / Hospital mortality, ICU mortality, tracheotomy rate, duration of mechanical ventilation, length of ICU and hospital stays. / Use of protocol associated with reduction in hospital and ICU mortality, without increase in survival with severe disability / Standardized methods of therapy, without specifying degree of adherence - potential for treatment bias.
Comparison between historical cohorts – time effect.
Prospectively collected data.
9 / Sturgeon 2007 / Secondary analysis of regional hospital association database / Stroke cases in Minneapolis-St Paul , older than 29 years, in 1980-2002 / Analysis of demographics and outcome / Hospitalization, demographics, mortality / Total stroke mortality and hospital case fatality declined by up to 50% over the period / Retrospectively collected data. Large regional database, with potential for treatment and ICD-9 diagnosis variability
10 / Andaluz 2008 / Secondary analysis of National Inpatient Database / Patients with SAH from 1993 to 2003 / Analysis of demographic patterns / In hospital mortality, routine discharge (good outcome), non-routine discharge (poor outcome), Length of stay / Mortality declined by 20% for SAH and by 50% for unruptured aneurysms. Clipping procedures did not increase, but endovascular procedures doubled. / Large national database. Non-randomized study. Possibility of treatment bias and between center variation. Discharge disposition used as surrogate for outcome
11 / Parry-Jones 2013 / Retrospective review of prospective database / 1364 patients with ICH from Jan 2008 till Oct 2010 / Testing accuracy of ICH grading scales against outcome / Clinical and demographic factors, CT scan – ICH (& modified) score, ICH grading scale. 30 day mortality / Mortality reported at around 50%. Age was poor predictor. GCS most predictive part of scale. / Single center referral database with possibility of bias. Blinded to ICH scores during treatment.
12 / Lu 2005 / Secondary analysis of combined multiple databases / 1839 patients, GCS 3-8, aged 16-65 yrs over period of 1984-1996. / Assessment of changes in outcome over time / Clinical and demographic characteristics, mortality rate. / Mortality in the year 1984 equaled 39% and gradually decreased
to a level of 27% in 1996. Significant change despite controlling for age, gender, race, cause of injury, admission
pupillary response and motor score / Amalagam of different databases, including clinical trial. Differing protocols of care, and possibility of between center bias.
3. What is the impact of early prognostication (before 72 hours)?
Reference Number / Source / Study Design / Study Participants / Treatment Interventions / Outcome Measures / Relevant Finding / Limitations/Assessment of Study Quality14 / Massagli 1996 / Retrospective cohort / 75 patients less than 17 years, with GCS < 8, over Jan 1985 to Dec 1986 / Association of indices of traumatic brain injury in children with outcome at hospital discharge, and 5 to 7 years later. / Abstracted GOS at 5- 7 yrs after injury. / Late outcome significantly associated with head AIS score, ISS score, pupils in the field, GCS scores at 24 and 72 hours, length of coma, site
of discharge, and early GOS scores. / Small cohort in study. Retrospective indirect determination of GOS. Single center.
15 / Gatson 2013 / Prospective observational cohort study / 18 patients with TBI and GCS<9, with ventriculostomy / Correlation between CSF beta-amyloid oligomers and GOS at 6 months / 6 month Extended GOS, Disability rating scale (DRS) / CSF oligomer levels within 72 hours correlated with GOS-E, and DRS / Small study. Single center. No breakdown of CSF sampling times.
16 / Chabok 2012 / Prospective observational study / 28 patients GCS < 9, with diffuse axonal injury, over March 2008 to May 2009 / Serum S100B and Neuron specific enolase (NSE) levels in relation to GOS / CSF sampling at 6,24, 48 and 72 hours. Mortality by time of discharge, GOS at 3 months and 2 years / Increased levels of NSE and S100B at 72 hours are associated with poor outcome / Small study – nonrandomized, selected study population, with possibility of bias.
17 / Settervall 2011 / Prospective cohort study / 277 patients > 14 years old with TBI admitted to trauma center within 12 hours of injury, over December 2006 and October 2007 / Comparison of predictive accuracy of initial, best, and worst GCS over first 72 hours. / Clinical demographics, hospital mortality / Hospital mortality related to all three scores – strongest association with best score, but no significant differences between scores. / Single center study, exclusion of extracranial injuries in model
18 / Eriksson 2012 / Retrospective case series / 32 patients with severe TBI and brain tissue oximetry monitoring (pBrO2), over January to July 2008 / Comparison of pBrO2, ICP and CPP in survivors and non-survivors / Clinical demographics , hospital mortality / pBrO2 less than 299 mmm Hg over first 72 hours was predictive of mortality / Small study –heterogenous injuries, dispersed ISS, variability in time of insertion
Single center study
Retrospective data
19 / Kahraman 2011 / Retrospective case series review of prospectively collected data / 30 patients >14 years with severe TBI, GCS < 9, admitted within 6 hours of injury, with ICP monitoring / Assessment of effect of accumulated ICP and CPP insults over time (pressure x time dose [PTD]– measured automatically and manually) , with outcome / Clinical demographics , in hospital mortality, discharge GCS, length of ICU and hospital stays, 3 month extended GOS / PTD for CPP > 100 after first 24 hours was most significantly predictive of mortality / Standardized protocol of care
Retrospective data
Model did not include therapeutic intensity
20 / Van Santbrink 2002 / Prospective cohort study / 57 TBI patients with GCS<9 / Serial assessment of transcranial Doppler estimated cerebral blood flow velocity(CBFV), and relationship with outcome, concentrating on first 72 hours / Clinical demographics
6 month GOS. / CBFV reduction associated with poor outcome. Reduced flow is most evident ipsilateral to focal injury / Standardized treatment protocol.
Nonrandomized.
Possible selection bias. Small study.
21 / Wolach 2001 / Prospective cohort study / 14 males aged 16-65 years with TBI, and GCS < 8 / Evaluation of immunological defects within 72 hours of TBI, compared healthy controls and with previous study on persistent vegetative state / Humoral and cellular immunological function, complicating infections, mortality and consciousness (up to 6 months) / Severe brain injury induces significant immune deficiency within 72 hours. Cellular function most compromised. Effect does not persist beyond months. / Small numbers, non-randomized with possibility of selection bias .
Single center.
22 / Togha 2004 / Retrospective case series / 122 patients with primary intracerebral hemorrhage(ICH), admitted between May 1999 and April 2002 / Exploration of mortality and associated features in patients with ICH. / Clinical demographics, in-hospital mortality / One third of deaths occurred within first two days after brain injury. In hospital mortality associated with admission GCS, hematoma volume, diabetes, and intraventricular hematoma / Single center. No withdrawal of support.
23 / Becker 2001 / Retrospective case series. / 87 patients with supratentorial ICH over 1994-97. / Exploration of mortality and associated features in patients with ICH. / Demographics, radiologic and clinical variables, discharge status, in hospital mortality, surgical intervention, withdrawal of support / Clinical communication of prognosis was at odds with prognostic models and most significantly affected withdrawal of support decisions and subsequent outcome / CT scan reviewers blinded to clinical data.
Single center study, with no controls to accurately assess variability between providers. Surgery very likely subject to selection bias
24 / O’Callahan 1995 / Prospective cohort study / 47 patients aged 18 to 83 years, with severe head injury over 12 month study period / To determine frequency of withdrawal of support in patients with severe head injury, with relationship to prognostic accuracy and communication with families / Clinical variable, predicted prognosis and contributing factors, outcome / Life support withdrawn in over 50% of patients. Prognosis based on judgment (as opposed to scoring systems) of neurologic function. Families generally agreed with physicians assessment. / Single center study. Interviewed residents rather than attendings. Non-randomized study.
25 / Tien 2006 / Retrospective case series review / 245 patients , aged 15-90 years admitted with blunt injury and an admission GCS of 3 over January 2001 to December 2003 / Comparison of mortality with reactive pupils / Clinical demographics; in-hospital death / Patients with GCS 3 and bilateral fixed dilated pupils had 100% mortality. Patients with reactive pupils had 42% mortality, but were more likely to receive surgical treatment / Excluded use of paralytics and legal intoxicants only. Single center. No standardized protocol. Probable selection and treatment bias.
26 / Brody 2010 / Retrospective case series review / 68 patients with fatal gunshot wounds(GSW), over September 2003 to September 2008 / Factors affecting frequency of organ donation / Clinical demographics, hospital disposition, organ donation / 48% not admitted to ICU and expired in ED. Only 18% of those received vasopressors. Identified missed donor opportunities. / Small sample size. Concerns of inconsistent record keeping.
4. What factors identify patients at high risk for death due to brain injury?
Reference Number / Source / Study Design / Study Participants / Treatment Interventions / Outcome Measures / Relevant Finding / Limitations/Assessment of Study Quality27 / Combes 1996 / Prospective cohort / 198 patients less than 17 years, with GCS < 8, over Jan 1985 to Dec 1986 / Association of indices of traumatic brain injury in children with outcome at hospital discharge, and 5 to 7 years later. / Abstracted GOS at 5- 7 yrs after injury. / Late outcome significantly associated with head AIS score, ISS score, pupils in the field, GCS scores at 24 and 72 hours, length of coma, site
of discharge, and early GOS scores. / Small cohort in study. Retrospective indirect determination of GOS. Single center.
28 / Murano 2005 / Retrospective case series / 298 patients with GSW to head, presenting to a level 1 trauma center between January 1992 to December 2003 / Analysis of clinical data to determine predictors of death / Clinical demographics, trajectory / In-hospital mortality was
51 per cent. Admission GCS <5 and ISS >25 were associated
with mortality. 7% of survivors had admission GCS of 3. Respiratory arrest or hypotension on admission, along with transhemispheric
and transventricular trajectories were all
predictors of death, but not absolute. / Single trauma center – referral pattern may confer selection bias. Extracranial injuries contribute to mortality and may confound. No exact data on mode of death (withdrawal vs progression to death)
29 / Bershad 2008 / Retrospective case series / 248 patients with subdural hemorrhage (SDH) admitted to Neurosciences ICU over January 1997 to December 2001 / Analysis of effect of coagulopathy on outcome / Clinical demographics, in hospital mortality / Acute Physiology And Chronic Health Evaluation (APACHE) III score and coagulopathy independently predicted in-hospital death, while surgical evacuation was associated with reduced in-hospital
deaths / Retrospective observations. No standardized protocol for reversal of coagulopathy
Single center. Aggressive INR target. No CT scan data.
30 / Park 2009 / Retrospective case series / 672 patients with SAH or TBI admitted to Neurosurgical ICU over July 2003 to June 2005 / Accuracy of Simplified Acute Physiology Score (SAPS)II and APACHE II scoring systems in predicting mortality in 1st 24 hours of admission / Clinical demographics; In-hospital mortality / Both scales overpredicted mortality. SAPS II was independent predictor in SAH only, while both scales were predictive in TBI. TBI patients are
more affected by systemic insults than SAH patients, / Retrospective data. Single tertiary care center. No data on state prior to admission including transfer from other centers.
31 / Labib 2011 / Retrospective Registry review / 276 patients over 65 with ISS > 15 admitted to Canadian Level 1 trauma center between January 2004 to December 2006 / Analysis of geriatric mortality after severe injury / Clinical demographics; In-hospital mortality / Falls were principal cause of injury, while intubation, blood transfusions, and trauma to head, C-spine, or chest were all associated with risk of death.
Subsequent respiratory, gastrointestinal, or infectious
complications were associated with in-hospital mortality (26.8%), which was comparable with
US National Trauma Data Bank. Mortality was not different either side of 80 years of age. / Retrospective data. Single tertiary care center. No data on state prior to admission including transfer from other centers.
32 / Cho 1997 / Retrospective case series / 200 patients . 13 years with acute head injury, admitted to neurosurgical ICU over September 1992 to December 1994. / Analysis of predictive power of GCS, and APACHE II and III, in relation to hospital mortality and functional outcome. / Clinical demographics; In-hospital mortality / No differences between scoring systems on hospital mortality, but APACHE III and II predicted death and functional outcome more accurately after 14 days. Physiologic derangement and comorbidities seem to be most relevant in later period after injury. / Retrospective data. Single center. Possible selection bias for surgical treatment but no influence of surgery on outcome model.
33 / Alvarez 1998 / Secondary analysis of European – North American ICU study / Secondary analysis of 401 patients with head trauma, admitted to 12 European/ North American ICU’s. / Comparison of APACHE II, SAPS II, and Mortality Probability Models (MPM) II, as well as GCS, in predicting mortality / Clinical demographics; In-hospital mortality / MPM II provide best prediction. APACHE II and SAPS II did not calibrate well, and underpredicted mortality / Same database used to create MPM II. No uniform protocol of care. Possible selection and treatment bias.
34 / Hyam 2006 / Secondary analysis of Intensive Care National Audit and Research Centre (ICNARC) database / 11,021 patients with TBI within ICNARC database, composed of 374,594 admissions to 169 general ICUs and 5,743
admissions to two neurosurgical units, over
December 1995 to May 2005 / Comparison of SAPS II, MPM II, APACHE II and III and the ICNARC model plusraw GCS / Clinical demographics, outcome on discharge from ICU, and discharge from hospital / The ICNARC model, SAPS II and MPM II have superior
calibration and discrimination compared to APACHE II
and III in TBI although none had perfect calibration
Lowest GCS in the first 24 hours in ICU discriminated well when available, but could not be objectively
assessed for 57% of admissions / Explicit variable
definitions, data collection training for observers, and use of objective variables limiting scope for inter-observer error
35 / Dalgic 2009 / Review of retrospective case series / 266 patients
with TBI and systemic trauma in 2003 and 2004 / Comparison of APACHE II and GCS in predicting mortality in TBI plus systemic trauma / Clinical demographics, mortality / APACHE II is superior to GCS for prediction of mortality in multitrauma patients, / Retrospective data; single center; used initial GCS on ICU admission and approximated Verbal scores in intubated patients.
Possible selection and treatment bias
36 / Ivascu, 2008 / Retrospective case series review / 109 patients, age 50 or greater, with CT evidence of
ICH, while taking ASA, clopidogrel, or both, over August 1999 to November 2004 / Analysis of interaction of antiplatelet agents with
traumatic ICH on mortality, and assess effect of treatment. / Clinical demographics,
mechanism of injury,
(GCS), grading of head
CT scans, and outcomes / ASA and clopidogrel are both associated with high mortality, secondary to initial hemorrhage and comorbidities rather than extension of bleeding – in contrast to warfarin.
The initial GCS and hemorrhage on
CT scan are most predictive of death.
Progression of hemorrhage after admission
is unusual / Retrospective data; single center. Possible selection and treatment bias, especially with regard to use of platelet transfusion
37 / Zafonte 2001 / Retrospective review of case series / 27 patients > 16 years, comatose with severe TBI admitted to
level I trauma center within 24 hours of injury, between December 1989 and April
1999 / Effect of inpatient rehabilitation on
outcome, after
GSW to the head. / Functional Independence Measure, Disability
Rating Scale, and length of stay / Rehabilitation was associated with improved outcome in survivors. / Small number. Retrospective data; single center. Non-randomized study with possible selection and treatment bias
38 / Arboix 1999 / Secondary analysis of prospective Barcelona Stroke Registry / 184 patients with non-traumatic SAH admitted to two Spanish hospitals between 1977 and 1993. / Analysis of clinical predictors of in-hospital mortality within the first 72 hours of admission / Clinical demographics. mortality / Fischer score, progressive neurological deficit and limb weakness were independent predictive factors of death within 72 hours in patients with non-traumatic SAH / Retrospective data with patients accrued over a 16 year period = small yearly number.
39 / Zhang 2011 / Retrospective review of case series / 155 patients > 18 years, admitted to hospital within 7 days of SAH, from January 2006 to 2009, / Assessment of possible risk factors for fever and its influence on in-hospital mortality / Clinical demographics. in-hospital mortality / Poor Hunt–Hess grade, presence of IVH and older age were independent predictors of fever. In-hospital mortality was associated with fever,unconsciousness on admission
and older age. / Single center. Retrospective data. Possible treatment bias.
40 / Arboix 2000 / Secondary analysis of 1840 patients in stroke registry. / 393 patients with diabetes in larger stroke registry, between January 1986 and December 1997. / Analysis of the effect of diabetes on ischemic stroke and determining predictors of in-hospital mortality. / Clinical demographics. in-hospital mortality / Higher incidence of atherothrombotic stroke and lacunar infarction in patients with diabetes. No significant effect of diabetes on mortality.Patient's age, decreased consciousness, chronic nephropathy, congestive heart failure and atrial fibrillation were all predictors of in-hospital mortality. / Retrospective review. Limited follow-up of outcome.