Neurocritical Care

, Volume 18, Issue 1, pp 131–142

Protocol Management of Severe Traumatic Brain Injury in Intensive Care Units: A Systematic Review

Authors

  • Shane W. English
    • Department of Medicine (Critical Care)The Ottawa Hospital
  • Alexis F. Turgeon
    • Department of Anesthesia (Critical Care)L’Enfant-Jésus
  • Elliott Owen
    • Department of Medicine (Critical Care)The Ottawa Hospital
  • Steve Doucette
    • Clinical Epidemiology ProgramThe Ottawa Hospital Research Institute
  • Giuseppe Pagliarello
    • Department of Surgery (Critical Care)The Ottawa Hospital
    • Department of Medicine (Critical Care)The Ottawa Hospital, The Ottawa Hospital Research Institute
Review Article

DOI: 10.1007/s12028-012-9748-3

Cite this article as:
English, S.W., Turgeon, A.F., Owen, E. et al. Neurocrit Care (2013) 18: 131. doi:10.1007/s12028-012-9748-3

Abstract

To examine clinical trials and observational studies that compared use of management protocols (MPs) versus usual care for adult intensive care unit (ICU) patients with acute severe traumatic brain injury (TBI) on 6-month neurologic outcome (Glasgow Outcome Scale, GOS) and mortality, major electronic databases were searched from 1950 to April 18, 2011. Abstracts from major international meetings were searched to identify gray literature. A total of 6,151 articles were identified; 488 were reviewed in full and 13 studies were included. Data on patient and MP characteristics, outcomes and methodological quality were extracted. All 13 included studies were observational. A random effects model showed that use of MPs was associated with a favorable neurologic outcome (GOS 4 or 5) at 6 months (odds ratio [OR] and 95 % confidence interval [CI] 3.84 (2.47–5.96)) but not 12 months (OR, 95 % CI 0.87 (0.56–1.36)). Use of MPs was associated with reduced mortality at hospital discharge and 6 months (OR and 95 % CI 0.72 (0.45–1.14) and 0.33 (0.13–0.82) respectively), but not 12 months (OR, 95 % CI 0.79 (0.5–1.24)). Sources of heterogeneity included variation in study design, methodological quality, MP design, MP neurophysiologic endpoints, and type of ICU. MPs for severe TBI were associated with reductions in death and improved neurologic outcome. Although no definitive conclusions about the efficacy of MPs for severe TBI can be drawn from our study, these results should encourage the conduct of randomized controlled trials to more rigorously examine the efficacy of MPs for severe TBI.

Keywords

Clinical protocolsBrain injuriesIntensive care unitsCritical careEvidence-based medicine

Introduction

Severe traumatic brain injury (TBI) is a significant and potentially life-threatening event that affects more than 1.4 million Americans every year [13], leaving 50,000 dead and approximately 7 % permanently disabled [3]. The rate of TBI has declined over the last decade, however, affecting more than 16,000 Canadians in 2003/2004 with an incidence of 90.1/100,000 population in those over 60 years of age [4].

The care of this patient population most often occurs in an intensive care unit (ICU) for the most severely afflicted. Multiple treatment intervention strategies, including intracranial pressure (ICP) monitoring, hyperosmolar therapy, therapeutic hypothermia, hyperventilation, and barbiturate coma, among others, were proposed over time, all of which having varying levels of evidence to support their use. In 1996, the Brain Trauma Foundation guidelines were published, updated in 2000 and 2007, in an attempt to standardize the management of patients with severe TBI and improve outcomes [57]. Despite the publication of these guidelines, there appears to be a wide variation in the management practice of patients with severe TBI [810].

The use of management protocols (MPs) or care pathways have been proposed to guide patient care, incorporate the early use of the best evidence available, and coordinate a multidisciplinary approach to a patient [11, 12]. However, the efficacy of these MPs and guidelines is not very well known. We conducted a systematic review of controlled trials involving adult patients suffering from severe TBI who were managed with the use of a TBI MP as compared to those managed by usual care on neurologic outcome and mortality.

Materials and Methods

Study Population

We aimed to identify all observational and interventional studies in the published and unpublished literature that compared the ICU medical management of adult patients with severe TBI by means of an established protocol (descriptive or algorithmic) versus a control based on standard care at the time. We searched for randomized controlled trials (RCTs) and observational studies including cohort studies (prospective or retrospective) that had a control group. Studies were selected if they included adult patients (age ≥18 years) with severe TBI (post-resuscitation GCS ≤8). The interventional arm must have included protocol-based care and be compared with a control group (managed by non-protocolized usual care). The protocols must have been based on the medical management (with or without decompressive craniotomy) of these patients in an intensive/acute care unit setting. A protocol was categorized as algorithmic if its organization was a structured or a step-based flow diagram or had a step-wise algorithm to follow. A protocol using only descriptors, order sets, or described phases was categorized as descriptive.

Primary and Secondary Outcomes

Our primary outcome measure was a neurologic outcome as defined by the Glasgow Outcome Scale (GOS) score at ≥6 months. A favorable outcome was defined as a GOS 4 or 5; unfavorable outcome was defined as a GOS of 1, 2, or 3. Six-month GOS scoring is considered an acceptable time period for assessment in head-injured patients, allowing enough time for recovery from the neurologic injury and minimizing losses to followup [13]. Secondary outcomes included mortality measured at the end of the study period (discharge, ≥6 months, ≥12 months), length of stay (LOS) in ICU and hospital, duration of mechanical ventilation, and resource utilization. We also examined the effect of protocol use on ICP/CPP control and on adverse events including ventilator associated pneumonia and CNS infection rates and seizure occurrence.

Study Identification

An OVID search strategy was used to search MEDILINE (1950–March week 4, 2009), EMBASE (1980–March week 4, 2009), EBM Reviews (includes Cochrane database of systematic reviews), ACP Journal Club, DARE (Database of Abstracts of Reviews of Effects), CCRT (Cochrane Central Registry of Controlled Trials), CMR (Cochrane Methodology Register), HTA (Health Technology Assessment), and NHSEED (NHS Economic Evaluation Database). In addition, an EBSCOhost search strategy was used to search the CINAHL database for additional articles in this primarily nursing literature database. Our search strategy included a combination of MeSH terms and keywords: (1) TBI, (2) intensive/critical care, and (3) clinical protocol. The OVID search strategy for Medline® is appended in Appendix 1 of Supplementary material. Searches of the gray literature included http://www.guidelines.gov, Cochrane Proceedings and Google Scholar as well as a hand search of the published abstracts from the conference proceedings of the Society of Critical Care Medicine (2004–2008), the International Symposium on Intensive Care and Emergency Medicine (2000–2009), and the Neurocritical Care Society Annual Meeting (2004–2008). Hand searches of the reference lists of the identified trials and review articles were also undertaken. There were no language restrictions. Authors of the included articles were contacted when information was unclear or not available. The original search was updated April 18, 2011 using Medline® In-Process and Other Non-Indexed Citations, Medline® and EMBASE. The former was selected as it includes pre-indexed citations as well as those indexed for PubMED and not Medline®. Conference proceedings (including those listed above) are now also indexed in EMBASE. The same search strategy as outlined in Appendix 1 of Supplementary material was used for the update.

Data Extraction of Primary Studies

Citations were initially reviewed by an examination of the titles and key words by one reviewer (SE). All subsequent citations were reviewed independently by two reviewers (SE, EO) first with screening of the abstracts and then by screening the full article if they met our pre-defined selection criteria.

Data extraction was performed independently by two reviewers (SE, EO), and any disagreements were resolved through consensus or with the assistance of a senior reviewer (LM). Extracted information included data on study design, patient population, MPs, and clinical outcomes including physiologic measures, adverse events, and morbidity and mortality measures.

Methodological Quality and Risk of Bias Assessment

Methodological quality assessment of the included studies was completed independently by two reviewers (SE, EO) by means of a modified checklist as presented by Downs and Black [14]. This tool was selected because it allows simultaneous assessment of both RCTs and observational studies, has been validated and rated as high quality in a review of greater than 150 tools [15], and is one of the few recommended tools by the Cochrane Collaboration [16]. This checklist includes 27 items divided into five categories for a total score of 32 (Appendix 3 of Supplementary material); the higher the score, the stronger is the methodological quality of the study. No formal cutoffs for poor versus good methodological quality has been suggested (written correspondence with author). The categories of quality measurement focus on reporting external and internal validity and study power. As a part of the assessment for possible bias within internal validity, compliance was measured based on whether the study included a measure of adherence to the protocol.

Data Analysis

We pooled results for the primary and secondary outcomes of neurologic outcome and mortality with the use of random effect models by odds ratios (ORs) and 95 % confidence intervals (95 % CIs) by means of Comprehensive Meta Analysis version 2.2 (NJ, USA). An OR less than 1 suggested a reduced risk of unfavorable neurologic outcome or death with the use of a protocol. An OR greater than 1 suggested improved odds of favorable neurologic outcome. Forest plots were generated for favorable neurologic outcomes (based on the GOS), as well as mortality (hospital, 3-month, 6-month). Standard errors of the mean for ICU and hospital LOS, as well as duration of mechanical ventilation were converted to standard deviations for ease in comparison. Given the heterogeneity in the measure of these outcomes (e.g., median with interquartile range and means with standard deviations), pooling of these data were not completed. Statistical heterogeneity was assessed by means of Cochrane Q statistic/I2 [16]. Clinical heterogeneity was examined focusing on patient population factors including the continuous variables of age, injury severity score, and APACHE score, the categorical variable presenting GCS and dichotomous variables: the presence of hypotension on admission and the presence of hypoxemia on admission. Other factors considered were the following categorical variables: type of ICU (medical/surgical ICU, highly specialized neuro ICU, and community ICU), the use of ICP monitoring devices (dichotomous variable), and year of publication (continuous variable). Additional sensitivity analyses were planned a priori based on the risk of bias, the use of co-interventions, and protocols that were designed in accordance with the Brain Trauma Foundation management guidelines [57].

We also aimed to examine if ICP and/or CPP were targeted by the protocol and if maintaining targets was improved by the use of the protocol. This was to be measured by occurrences of sustained over-(ICP) or under-(CPP) shooting the predefined target as set out in the protocol. In one study in which this data was captured, a polynomial best-fit trend was calculated by means of the average ICP data gathered during their preliminary study [17]. Finally, a measure of resource utilization associated with the institution of a clinical protocol was sought to establish if costs were decreased as a function of protocolized care.

Results

Study Selection and Characteristics

Our search strategy found 6,151 citations. Of these, 12 full publications and one abstract (Fig. 1) met our inclusion criteria (Tables 1, 2). All studies retrieved were observational studies; no RCTs were found. All were published in the English language. Six studies were European in origin, five originated from the United States, and the final study from a center in Saudi Arabia. All the included articles were published after 1997.
https://static-content.springer.com/image/art%3A10.1007%2Fs12028-012-9748-3/MediaObjects/12028_2012_9748_Fig1_HTML.gif
Fig. 1

Study selection

Table 1

Characteristics of included studies

Author (year)

Protocol management group

Control group

Interventions

CPP targeted

ICP targeted

Type of ICU

Primary outcome

Protocol group

Control group

Protocol group

Control group

Protocol group

Control group

Eker et al. (1998) [22]

Prospective n = 53

Historic n = 38

Descriptive protocol targeting brain volume regulation and microcirculation

Usual conventional care

Indirect target (>50)

Not described

Indirect target

Not described

General ICU

GOS at ≥6 months

McKinley et al. (1999) [17]

Prospective n = 12

Retrospective historic n = 12

ICP/CPP-targeted algorithm

Pre-printed orders and usual care

Yes (>70)

At treating MD discretion

Yes (<20)

At treating MD discretion

ICU NOS

Maintenance of ICP/CPP goals

Vukic et al. (1999) [24]

Retrospective n = 23

Retrospective n = 16

Descriptive protocol

Usual care

Yes (>70)

Not monitored

Yes (<15)

Not monitored

ICU NOS

GOS at 6 months

McIlvoy et al. (2001) [21]

Prospective n = 159

Retrospective historic n = 43

ICP/CPP-targeted descriptive algorithm within four phase plan

Usual care

Yes (>70)

At treating MD discretion

Yes (<20)

At treating MD discretion

Level I Trauma Center ICU

ICU and hospital LOS

Palmer et al. (2001) [27]

Retrospective/prospective n = 56

Retrospective/prospective n = 37

Descriptive algorithm

Usual care

Yes (>70)

Not described

Yes (<20)

Yes (<20)

Community ICU

GOS at 6 months

Vitaz et al. (2001) [23]

Prospective n = 119

Retrospective n = 43

ICP/CPP-targeted descriptive algorithm within four phase plan

Not described

Yes (>70)

Not described

Yes (<20)

Not described

ICU NOS

ICU and hospital LOS

Elf et al. (2002) [20]

Prospective n = 154

Retrospective historic n = 121b

Descriptive algorithm

Usual care

Yes (>60)

At treating MD discretion

Yes (<20)

At treating MD discretion

NICU

eGOS at ≥6 months

Patel et al. (2002) [25]

Retrospective n = 129

Retrospective n = 53

ICP/CPP-targeted algorithm

Not described

Yes (>70)

Not described

Yes (<20)

Not described

NICU

GOS at 6 months

Clayton et al. (2004) [28]

Retrospective n = 353

Retrospective n = 316

CPP-targeted algorithm

Not described

Yes (>70)

Not described

Only if CPP target not met

Not described

General ICU

ICU and hospital mortality

Fakhry et al. (2004) [19]

Prospective n = 611

Retrospective n = 219

Descriptive algorithm and order set

Not described

Yes (>70)

Not described

Yes (<20)

Not described

Level I Trauma Center ICU

Mortality and LOS

Cremer et al. (2005) [18]

Retrospective/prospective outcome assessment n = 211

Retrospective/prospective outcome assessment n = 122

ICP/CPP-targeted algorithm

Usual care

Yes (>70)

Not monitored

Yes (<20)

Not monitored

Level I Trauma Center ICUs

eGOS at ≥12 months

Bader et al. (2008)a [26]

Retrospective/prospective n = 205

Retrospective/prospective n = 37

Descriptive algorithm

Usual care

Yes (>70)

Not described

Yes (<20)

Yes (<20)

Community ICU

GOS at 6 months

Arabi et al. (2010) [29]

Retrospective/prospective = 362

Retrospective/prospective n = 72

Preprinted order form

Not described

Yes (>70)

At treating MD discretion

Yes (<20)

At treating MD discretion

M/S ICUTC Hospital

Hospital mortality

NOS not otherwise stated, NICU neuro ICU, M/S ICU medical-surgical ICU, TC hospital tertiary care hospital, eGOS extended Glasgow Outcome Scale

aAbstract data

bPreviously published

Table 2

Study methodological quality

Article

Reporting

External validity

Internal validity-bias

Internal validity-confounding

Power

Total

/11

/3

/7

/6

/5

/32

Eker et al. [22]

5

3

5

2

0

15

McKinley et al. [17]

6

1

4

2

0

13

Vukic et al. [24]

5

1

5

2

0

13

McIlvoy et al. [21]

6

1

3

1

0

11

Palmer et al. [27]

8

2

4

2

0

16

Vitaz et al. [23]

4

1

3

1

0

9

Elf et al. [20]

7

2

3

2

0

14

Patel et al. [25]

7

2

3

1

0

13

Clayton et al. [28]

6

3

4

2

0

15

Fakhry et al. [19]

6

3

5

2

0

16

Cremer et al. [18]

10

3

3

2

0

18

Bader et al.a [26]

3

1

4

1

0

9

Arabi et al. [29]

9

3

5

3

0

20

aAbstract publication

Eleven of the 12 studies were single centered. One study compared two different centers in two regions of the Netherlands [18]. Six of the 12 studies prospectively studied the protocol group [17, 1923]; the other six studied the protocol group retrospectively. All 12 studies used a control group, only one of which was not historic [18]. In addition to the implementation of a management protocol, two studies included the use of ICP monitors with the intervention group [18, 24] (which had not been used in the controls) and one simultaneously opened a dedicated Neurocritical Care Unit within their institution [25]. The sample sizes of all of the studies ranged from 24 to 830 patients.

Three studies had three separate groups [1921]. Two of the studies each had two treatment groups and one study had two control groups, each of which were collapsed into one for our analysis. In the first instance, the treatment groups differed in time and degree of compliance with the protocol (1995–1996 with reported 50 % compliance and 1997–2000 with 88 % compliance reported in 1997) [19]. In the other, the two treatment groups differed in time and the protocol used (1995–1997 and 1997–1998) [21]. Finally, in the last study, the two control groups differed in time with the creation of a Neurosurgical Intensive Care (NIC) Unit for the latter group (1980–1981 pre-NIC and 1987–1988 post-NIC) [20].

One of the included studies was a published abstract [26]. It is an extension of the data presented by another included article [27] with a longer followup, and thus a larger sample size in the intervention arm, but used the same historic control (written communication with author). Given the overlap in some of the patient data, and since we were unable to assess the full study from the abstract, these data were not used in the analyses. Two single center studies with different intervention sample sizes, reported out of the University of Louisville Hospital [21, 23], were included in this study.

Methodological quality and risk of bias varied among the included studies with scores ranging from nine to 20 on the 32 point Downs and Black scale (Table 2). Given that no study was a randomized trial, two points were automatically lost for their observational nature with the majority (>70 %) of remaining deficiencies lying in power adequacy, confounding risk and study reporting. With the exception of the previously mentioned study [19], protocol adherence was not measured, and no report on the adherence measures for the duration of the study. This was also reflected in the methodological quality assessment scores.

Patient Characteristics

A summary of the patient characteristics from the included studies is shown in Table 3. Severe TBI patients were included in all the studies: one study included both moderate and severe TBI patients [25]; one included mild, moderate, and severe TBI patients [28]; and one included patients who deteriorated to severe TBI within 24 h [18]. The mean admission GCS ranged from 3.5 to 6.9 and was similar among the control and intervention groups (Table 3). The mean age ranged from 29.5 to 41.4 years and was similar between the intervention and control groups. Men represented ≥64 % of the included patients in those trials reporting gender. Seven studies [1821, 23, 25, 29] report Injury Severity Scores [30] with most reporting mean/median ≥24, suggesting a multiple trauma population.
Table 3

Patient baseline characteristics

Author (year)

Total n

Mean age (SD)

Male sex (%)

Mean admission GCS (SD)

Mean ISS (SD)

Multiple injuries (mean)

Cx

Tx

Cx

Tx

Cx

Tx

Cx

Tx

Cx

Tx

Eker et al. (1998) [22]

91

          

McKinley et al. (1999) [17]

24

    

5.67

6.08

  

50

58

Vukic et al. (1999) [24]

39

36

39

69

78

5.3

5.7

    

McIlvoy et al. (2001) [21]

161

32.9 (14.8)

33.9 (16.1)

  

6.3 (1)c

6.2 (1.3)c

26.8 (8.6)

25.5 (7.8)

74

46

Palmer et al. (2001) [27]

93

41.4 (3.7)

38.1 (2.5)

78

84

6.43 (0.67)

6.88 (0.50)

32.82 (2.35)

28.32 (1.20)

  

Vitaz et al. (2001) [23]

162

40.2 (20.1)

34.7 (18.2)

  

5.4 (1.6)c

6.2 (1.1)c

24.5 (8.9)

27.7 (8.1)

  

Elf et al. (2002) [20]

275

36.4

41

  

41.2d

92d

  

18

36

Patel et al. (2002) [25]

182

33.8 (13.5)

33.8 (13.6)

64b

168b

5a

5a

25a

25a

  

Clayton et al. (2004) [28]

669

28 (15–45)a

32 (18–51)a

74

75

5 (3–8)a

6 (3–9)a

14 (10–19)a, f

18 (14–22)a, f

  

Fakhry et al. (2004) [19]

830

33.8

35.1

73.5

75.3

4

3.5

25.2

24.6

40

42.6

Cremer et al. (2005) [18]

333

39 (24–64)a

38 (25–54)a

67

68

3–5: 24e

6–8: 30e

9–12: 19e

13–15: 6e

3–5: 23e

6–8: 25e

9–12: 16e

13–15: 4e

25 (17–29)a

24 (17–29)a

  

Arabi et al. (2010) [29]

434

31.8 (14.3)

29.5 (14.0)

96

95

5.3 (1.9)

4.9 (1.8)

33.1 (11.5)

31.6 (11.5)

  

aReported as median (interquartile range)

bIncludes data from moderate and severe TBI patients

cCalculated at 24 h

dReported as GCSm (motor) ≥4, %

eReported as GCS range: %

fAPACHE II score

MP Characteristics

Several study MPs [18, 19, 21, 23, 2729] were designed by means of the original (1996) Brain Trauma Foundation guidelines [5] for the management of severe TBI or the European equivalent [31] as a guide. In all but one study [22] the intervention group enrollment period included time after the publication of these guidelines. However, the majority of the control group enrollment periods predated it [17, 1923, 25] and thus care for the control or usual care preceded the existence of management guidelines in 7 studies. Palmer et al. [27], Clayton et al. [28], and Arabi et al. [29] had control groups the enrollment periods of which extended into or occurred after 1996, but then introduced protocolized care plans based on the BTF guidelines.

All the protocols were designed for physician and/or nursing care. Ten studies published a written description of their protocols [1724, 27, 29], two published protocol diagrams or a flow chart [25, 28], and two combined a description with a flow chart [17, 27]. Four of the descriptive protocols included preprinted order sets or standardized operative procedures [19, 20, 23, 29]. Three of the protocols included care outside the acute ICU management of TBI (e.g., ER management and rehabilitation) [21, 23, 27]. All the studies managed TBI patients on the principles endorsed by the BTF guidelines [57] except one [22] the intervention group of which was treated based on the Lund Concept [32]. Ten of the protocols targeted specific ICP and CPP goals concomitantly [1721, 2325, 27, 29]; CPP goals were targeted solely in one study [28]. ICP targeting was indirect in the unique protocol instituted in the study by Eker et al. [22], where the treatments targeted cerebral hypoxia and transcapillary filtration.

Primary Outcome

Neurologic outcome defined by the GOS was reported in seven studies: one at discharge [19], five at 6 months [22, 24, 27] or ≥6 months [20, 25], and one at ≥12 months [18]. The pooled OR for favorable GOS at ≥6 month by means of a random effects model was 3.84 (95 % CI 2.47–5.96) in favor of a MP (Fig. 2). There was no evidence of statistical heterogeneity with an I2 of 33 % and Q statistic p = 0.20.
https://static-content.springer.com/image/art%3A10.1007%2Fs12028-012-9748-3/MediaObjects/12028_2012_9748_Fig2_HTML.gif
Fig. 2

Forest plot of favorable neurologic outcome (GOS 4 or 5)

Secondary Outcomes

The use of protocolized care as compared to control was not associated with a favorable neurologic outcome at hospital discharge and at ≥12 months (Fig. 2). All the studies reported mortality data even though this data were not included in the outcome analyses of two individual trials [21, 23]. Two studies reported a reduction in both ICU and hospital mortality associated with the use of the MP [28, 29]. Eight studies reported mortality at discharge yielding a statistically insignificant pooled OR [1721, 23, 28, 29] (OR 0.74 and 95 % CI 0.47–1.15) (Fig. 3). Four trials reported mortality at six [22, 24, 27] or ≥6 months [25] and found a statistically significant reduction associated with the use of a protocol with a pooled OR of 0.33 (95 % CI 0.13–0.82). Only one study reported mortality at ≥12 months and found statistically insignificant results [18]. Significant heterogeneity was observed in the mortality data with I2 of 81 % (p < 0.00001) and 68 % (p = 0.02) at discharge and 6 months, respectively.
https://static-content.springer.com/image/art%3A10.1007%2Fs12028-012-9748-3/MediaObjects/12028_2012_9748_Fig3_HTML.gif
Fig. 3

Forest plot of mortality

ICU and hospital LOS were reported in seven [18, 19, 21, 23, 2729] and six [19, 21, 23, 2729] studies, respectively. The mean ICU LOS among protocol-managed groups was similar to control groups. ICU LOS in the intervention groups ranged from a median of 3 days to a mean of 24 days, while the control group LOS ranged from a median of 3 days to a mean of 21 days. Mean hospital LOS for protocol-managed groups as compared to controls also appeared similar. Among both LOS measures, the trend appears to be a shortened LOS in the intervention group; the exception being the results presented by Cremer et al. [18] (Table 4). With the same exception, time spent on mechanical ventilation did not appear to be lengthened in the five studies reporting the outcome [18, 21, 23, 27, 29].
Table 4

ICU and hospital LOS and length of mechanical ventilation

Author

Control

Treatment

n

Mean, days (±SD)

n

Mean, days (±SD)

ICU

 Fakhry et al. [19]

219

9.7

611

8.1

 McIlvoy et al. [21]

43

21.2 (9.3)

118

16.6 (5.9)

 Palmer et al. [27]

37

21.0 (16.4)

56

22.0 (12.0)

 Vitaz et al. [23]

43

21.2 (9.3)

119

16.8 (7.4)

 Clayton et al.a [28]

391

3 (1–40)d

452

3 (1–33)d

 Cremer et al. [18]

 

8 (4–14)d

 

14 (8–23)d

 Bader et al.b [26]

37

21.0 (16.4)

205

24.3 (14.9)

 Arabi et al.c [29]

52

11.5 (7.3)

294

11.9 (7.9)

Hospital

 Fakhry et al. [19]

219

21.2

611

16.1

 McIlvoy et al. [21]

43

31 (19.7)

118

22.5 (8.5)

 Palmer et al. [27]

37

24.4 (19.5)

56

25.4 (14.2)

 Vitaz et al. [23]

43

31.0 (19.7)

119

22.5 (8.5)

 Clayton et al. [28]

391

10 (1–351)d

452

10 (1–354)d

 Bader et al.b [26]

37

24.4 (19.5)

205

30.2 (21.6)

 Arabi et al.c [29]

52

82.3 (78.9)

294

71.4 (79.1)

Mechanical ventilation

 McIlvoy et al. [21]

43

14.4 (7.5)

118

11.3 (5.9)

 Palmer et al. [27]

37

17.5 (13.4)

56

19.2 (12.0)

 Vitaz et al. [23]

43

14.4 (7.5)

119

11.5 (5.8)

 Cremer et al. [18]

 

5 (2–9)d

 

12 (7–9)d

 Bader et al.b [26]

37

17.5 (13.4)

205

19.2 (10.5)

 Arabi et al.c [29]

52

10.4 (6.9)

294

11.2 (7.4)

aIncludes admission GCS >8

bAbstract data

cAmong survivors

dMedian (range)

One study [17] reported the effect of managed therapy on ICP/CPP control. Protocol-managed patients had a reduction in duration of average ICP >20 mmHg from 4.7 to 2.3 days and an overall increase in mean CPP from 65 mmHg (±24 SE) to 72 mmHg (±14 SE) (target CPP during this study was >70 mmHg) corresponding to less variation in the protocol group (p < 0.001) (data not shown).

Resource utilization was reported in three studies [19, 21, 27]. Palmer et al. [27] reported a mean charge increase in the MP group of more than $95,000 (Table 5). Unpublished followup data from the same group [26] that examined their 10 year experience showed that these costs had risen to greater than $200,000, but it is unclear if adjustment for time and inflation were made. Fahkry et al. [19] reported a small reduction in the mean cost per patient managed by means of a protocol (calculated in 1997 dollars and did not include physician billings, diagnostic services, or medications). Similar findings were reported by McIlvoy et al. [21] who used an estimated cost reduction analysis accounting for major costs incurred by a typical TBI patient in 1996 dollars with adjustment for inflation.
Table 5

Cost analysis

Author

Control

Treatment

n

Mean, $ (±SD)

N

Mean, $ (±SD)

Fakhry et al. [19]

219

36694

423

28987

McIlvoy et al. [21]

43

72426

118

60927

Palmer et al. [27]

37

196128 ± 140913

56

293065 ± 149360

Bader et al.a [26]

37

196128 ± 140913

205

444539 ± 293673

aAbstract data

Inconsistent and lack of reporting in the literature prevents analysis of adverse events related to the use of a MP.

Sensitivity Analyses

As previously described, heterogeneity also existed in study design, methodology, and quality. Clinical heterogeneity existed in the MP design, the presence of co-interventions, differing neurophysiologic endpoints, and type of ICU used. Executing the a priori sensitivity analysis plan was limited given the lack of RCTs identified and the relatively few and overall poor methodological quality observational studies included. The quality score based on the modified Downs and Black checklist did not explain the heterogeneity observed (dichotomized poor and good). Similarly, sensitivity analyses examining the effect of removing potentially overlapping data [21, 23] or including only those protocols based on the Brain Trauma Foundation guidelines (or equivalent) [18, 19, 21, 23, 2729] had very little effect on results. Removing the three studies with a clear co-intervention [18, 24, 25] increased the effect of protocol management on 6 month favorable neurologic outcome, OR 5.06 (95 % CI 3.36–7.61), and 6 month mortality, OR 0.18 (95 % CI 0.11–0.31), with resulting I2 of 0 % for both. Owing to the minimum number of studies meeting our inclusion criteria, the performance of a meta-regression analysis to explain heterogeneity based on our a priori hypotheses was not feasible.

Discussion

In our study, we observed that MPs used for the treatment of severe TBI patients were associated with a significant increase in favorable neurologic outcomes at ≥6 months as well as a decrease in mortality at discharge and at ≥6 months. The institution of these protocols did not appear to be associated with an increased LOS (ICU or hospital) nor a significant increase in duration of mechanical ventilation. However, their use may be associated with an increase in resource utilization.

MPs exist across many disciplines. Their use has improved the outcomes seen in patients suffering from stroke [33] or undergoing surgical procedures [34]. In the intensive care setting, MPs have been successful in improving the management of patients with adult respiratory distress syndrome [35], weaning from mechanical ventilation [36], insulin therapy for hyperglycemia [37], optimizing sedation and analgesia [38], and blood product transfusion [39]. Not only has their introduction been associated with a decrease in health care costs but also the improvement of health care consistency and quality [40, 41]. However, MPs are not without their pitfalls or controversy. The nature of ICU patients by virtue of illness severity, varying medical co-morbidities, and disease complexities, make performing high-quality gold-standard practice-setting research challenging. Further, the generalizability of such research is difficult for the same reasons. As such, recommendations or directions within a MP may be based on poor or lacking evidence, subjecting its utility to debate. For example, of the numerous recommendations set out by the BTF guidelines for the management of severe TBI [7], the only Level I recommendation (i.e., based on Class I evidence or good quality RCT data) is against the use of steroids. Protocols have also been met by resistance by some practitioners as they are viewed as “mindless” medicine that do not allow for the adaptation to any patient-specific need that falls out of the realm of the ideal patient for which the protocol was designed [41]. The findings from our study may suggest otherwise.

The treatment of the severe TBI patient can be complex. A recent review on their management in the ICU setting advocated for the use of MPs [42]. The BTF have credited, in part, evidence-based protocols with the reduction in mortality seen over the years among patients with TBI [7]. In measuring adherence to these guidelines, Hesdorffer and Ghajar [1] found that the use of treatment protocols were associated with better adherence. In addition to improving outcomes, treatment protocols have also been shown to improve collaboration and communication across disciplines, and better coordinate care [43].

There are limitations to our study that must be considered when interpreting the results. First, we are limited by the absence of RCTs examining the use of protocols in managing patients with severe TBI. The very nature of non-randomized studies using historic controls portend to bias and confounding [22]. Second, there is no clear gold standard or globally accepted means of assessing quality of observational studies used in systematic reviews [44]. The more commonly known MOOSE guidelines [45] were designed for the reporting of the meta-analyses of observational studies rather than for assessing their quality [44]. We used one of the tools [14] suggested by the Cochrane Collaboration [16]; but like others, there are no pre-defined cutoffs for “high” or “low” quality studies using these checklists (author communication October 25, 2010). Finally, among the observational studies included in this review, clinical heterogeneity existed, given differences in the design, makeup, and scope of the protocols. Furthermore, no study completely measured protocol adherence during the intervention. Although attempts were made to explore statistical heterogeneity that existed among the studies, the degree of the clinical differences in study design, patient population, and protocol characteristics limits our ability to draw conclusions from these data. The strengths of this paper lie in the rigorous methodology followed. Our a priori design included a search strategy broad and all inclusive with selection of articles, data extraction, and quality assessment done in duplicate. Our study summarizes this body of literature and represents all of the available published data.

Conclusions

In conclusion, the use of MPs for severe TBI was associated with improved neurologic outcome and a lower death rate. Although no definitive conclusion about MPs for severe TBI can be drawn from this review, these data underscore the need for a better understanding of the efficacy and safety of MPs for severe TBI.

Acknowledgments

This research was completed without external and/or industry funding and the authors have no conflicts of interest to declare. The authors would like to thank Ms. Risa Shorr, librarian at the Ottawa Hospital, for her assistance with deriving the electronic search strategy used in this review. They would also like to thank Dr. Nataliya Milman and Dr. Antonio Caycedo for their assistance with translation.

Supplementary material

12028_2012_9748_MOESM1_ESM.docx (14 kb)
Supplementary material 1 (DOCX 14 kb)
12028_2012_9748_MOESM2_ESM.docx (26 kb)
Supplementary material 2 (DOCX 25 kb)
12028_2012_9748_MOESM3_ESM.docx (32 kb)
Supplementary material 3 (DOCX 32 kb)

Copyright information

© Springer Science+Business Media, LLC 2012