Prolonged Therapeutic Hypothermia After Traumatic Brain Injury in Adults: Title and subTitle BreakA Systematic Review FREE

Traumatic brain injury (TBI) is defined as any brain injury that has been sustained secondary to externally inflicted trauma. This injury can result in death or a lifelong impairment of physical, cognitive, and psychosocial functioning.^{1 - 2} It is an injury predominantly limited to the younger population. Some of the main causes of TBI include motor vehicle crashes, falls, acts of violence, and sports injuries.² Although preventive measures have reduced the overall death rate from this devastating injury,¹ no therapy has been shown to reduce mortality and improve neurologic outcome once the injury occurs.

Therapeutic hypothermia has been investigated as a possible neuroprotective strategy for the prevention or reduction of brain injury due to several causes. The exact mechanism by which it reduces the deleterious effects of brain injury is a matter of debate.³ Possible mechanisms include reductions in cerebral metabolic rate, increased intracranial pressure and cerebral edema formation,^{3 - 6} and attenuation in the opening of the blood-brain barrier.⁷ Hypothermia also inhibits the inflammatory response and the release of glutamate, nitric oxide, and free radicals associated with TBI.^{3 ,8 - 11}

Therapeutic hypothermia is used as a primary preventive therapy for neuroprotection in cardiac surgical patient populations¹² and is currently being investigated for its potential therapeutic benefit in ischemic stroke.^{13 - 14} Recently, 2 randomized controlled trials (RCTs) have suggested that therapeutic hypothermia reduces mortality¹⁵ and improves neurologic outcome^{15 - 16} after cardiac arrest. It was first introduced for the treatment of TBI by Fay in 1945.¹⁷ Results of early experimental models of brain injury,^{18 - 19} as well as uncontrolled case series,^{17 ,20 - 23} suggested that hypothermia may improve clinical outcomes. However, deep and prolonged hypothermia also was thought to be associated with the development of arrhythmias, coagulopathies, and pulmonary infections.^{24 - 26} These adverse effects, as well as perceived benefits of alternative therapies in critical care, were the likely reasons it largely fell out of favor for the treatment of TBI in the 1960s.²⁷ Research interest in this therapy resurged in the 1980s because investigators found that moderate hypothermia (30°C-33°C) in experimental models of TBI reduced neuronal loss and improved behavioral outcomes.^{28 - 31} In addition, a number of small RCTs have suggested potential benefits.^{12 ,32 - 33}

Recent laboratory and clinical research suggests that time from injury to initiation of hypothermia may have a differential impact on the effects of secondary brain injury, and hence, clinical outcome.^{3 ,34} Similar findings have been reported for depth,^{18 ,35} duration,³³ and rate of rewarming after discontinuation of hypothermia.^{36 - 37} Although 2 previous reviews on hypothermia for TBI have been published,^{38 - 39} neither have explored these factors in detail. Therefore, we conducted a systematic review to explore the effects of depth, duration, and rate of rewarming after discontinuation of hypothermia on mortality and neurologic outcome in adults after TBI.

Published and unpublished RCTs of therapeutic hypothermia in adults with TBI were identified using both electronic and manual search strategies. MEDLINE (OVID) database was searched from January 1, 1966, to week 1 of September 2002, using a combination of the following medical subject heading (MeSH) terms: "induced hypothermia," "hypothermia," and "cranio-cerebral trauma"; and the following text words with truncation where appropriate: "refrigeration anesthesia," "cryoanesthesia," "artificial hybernation," "hypothermia," "head trauma," "head injury," and "brain concussion."

An RCT filter and a systematic review filter was applied to the MEDLINE search strategy to identify RCTs and systematic reviews.⁴⁰ The electronic search was supplemented by a search of Current Contents from week 27 of year 1993 to week 40 of year 2002, EMBASE and the Cochrane Library, which contain the CENTRAL Database of Controlled Trials, as well as the Cochrane Database of Systematic Reviews and the Database of Abstracts of Review Effectiveness.⁴¹ We also conducted a review of the conference proceedings from the Cochrane review.³⁸ These proceedings included the International Conference on Recent Advances in Neurotraumatology, Rimi-Riccione, Italy, September 1996; the 2nd International Neurotrauma Symposium Glasgow, Scotland, 1993; the 3rd International Neurotrauma Symposium, Toronto, Ontario, Canada, July 1995; the 4th International Neurotrauma Symposium, Seoul, South Korea, August 1997; the 27th Meeting of the Society for Critical Care Medicine, San Antonio, Tx, February 1998; and the 10th International Symposium on Intracranial Pressure, Williamsburg, Va, May 1997. The search was not restricted by language or type of publication. The bibliographies of all identified RCTs and review articles also were reviewed. Authors of the RCTs were contacted to identify unpublished studies and abstracts, and for their knowledge of any unpublished or ongoing clinical trials.

Each RCT was evaluated on the basis of 4 inclusion criteria: study design (RCT), target population (adults with TBI), therapeutic intervention and comparison (at least 24 hours of therapeutic hypothermia at any time after sustaining TBI vs normothermia), and outcome (all-cause mortality, which was the primary outcome measure). Figure 1 outlines how studies were screened and retrieved for inclusion.

Two reviewers (L.A.M. and J.S.H.) independently extracted data from the primary studies, and any disagreements were resolved through consensus. Extracted information included demographic data, information on the pattern and extent of injuries, presence of comorbid illnesses, and a detailed description of the hypothermia intervention (ie, depth and duration of cooling and rate of rewarming after discontinuation of hypothermia), as well as other cointerventions and physiologic goals for brain injury management through the cooling period (ie, therapies to control intracranial pressure and cerebral perfusion pressure).

The primary outcome measure was all-cause mortality at the end of the trial follow-up period. The secondary outcome measure was neurologic outcome at the end of the trial follow-up. Poor neurologic outcome was defined as a Glasgow Outcome Scale score of severe disability, vegetative state, or death. When necessary, authors of the articles were contacted for clarification and for additional information.

Two reviewers (L.A.M. and J.S.H.) assessed methodological quality of the individual studies and disagreements were resolved through consensus. Similar to methods used in the Cochrane review,³⁸ the 2 items used to assess methodological quality were allocation concealment and blinding of the outcome assessment. Both items are potentially associated with bias in estimation of the true treatment effect.^{38 ,42 - 43} We included blinding of neurologic outcome assessment as a measure of quality for these RCTs because unlike mortality, reported neurologic outcome is subjective and thus another potential source of bias. Studies were considered to be of high methodological quality if allocation concealment was present and the outcome assessment was blinded; of moderate methodological quality if allocation concealment was unclear but the outcome assessment was blinded; and of low methodological quality if allocation concealment was unclear and the outcome assessment was not blinded. Studies that reported losses to follow-up and/or missing outcome assessments also were documented.

Heterogeneity is a major threat to the interpretation and validity of meta-analyses and can be because of differences in methods, study populations, interventions, outcomes, or chance.⁴⁴ We deliberately sought to identify significant differences in the hypothermia interventions between trials that may have caused a differential effect on mortality and neurologic outcome. Sources of heterogeneity in these hypothermia interventions included an examination of different depths (32°C-33°C vs 33.5°C-34.5°C); durations (24 hours, 48 hours, and >48 hours) and rates of rewarming (≤24 hours vs > 24 hours) after discontinuation of hypothermia. Definitions for mild and moderate hypothermia varied among the individual studies. Therefore, we dichotomized depth of hypothermia as 32°C to 33°C and 33.5°C to 34.5°C.

Sensitivity analyses included an examination for the presence of methodological heterogeneity (high vs moderate vs low quality) between the individual trials for the mortality and the neurologic outcome. We also assessed the effect of the mortality outcome by removing the largest RCT.

Data from all 12 studies were combined to estimate the pooled relative risk (RR) and 95% confidence intervals (CIs) with a random-effects model (Meta.Analyst.977, Lau J, Chalmers TC; 1990). An RR less than 1.0 suggests a reduced risk of death or poor neurologic outcome while an RR greater than 1.0 suggests an increased risk of death or poor neurologic outcome, for the hypothermia group compared with the normothermia group. The prespecified analyses for depth, duration, and rate of rewarming following discontinuation of hypothermia and sensitivity analyses were conducted in a similar fashion. Evidence of publication bias was evaluated with an inverted funnel plot.⁴⁵ Presence of statistical heterogeneity between studies was evaluated using the Cochran Q test of homogeneity.⁴⁶

A total of 867 citations were identified by our search strategy. Of these, 13 RCTs met our initial inclusion criteria.^{11 - 12 ,32 - 34 ,47 - 54} One non–English-language citation was identified as potentially relevant but was not retrievable despite a thorough search by our librarian as well as a search from the Canadian national science library (Canada Institute for Scientific and Technical Information).⁵⁴ It is unclear if this study was an RCT. Thus, a total of 12 studies were included in this review. All but 3 citations were identified by the electronic search strategy. These citations were identified through a search of the Cochrane Controlled Trials Register^{52 - 53} and 1 was a published systematic review.⁵¹ One unpublished study was identified from the Cochrane review.³⁸

Twelve RCTs comprising a total of 1069 patients were identified: 543 patients in the therapeutic hypothermia group and 526 patients in the control (normothermia) group. Ages of patients ranged from 16 years to 81 years. In the 7 studies that reported the proportion of males to females, males accounted for more than 70% of the sample.^{11 ,33 ,40 ,47 - 48 ,50 ,52} Nine studies included patients with severe brain injury defined by a Glasgow Coma Score of 8 or less^{12 ,33 - 34 ,47 - 50 ,52 - 53} and 3 with a Glasgow Coma Score of 7 or less.^{11 ,32 ,51} Five studies included the time from injury to initiation of hypothermia as part of the inclusion/exclusion criteria.^{11 ,32 ,34 ,50 ,52} Eight studies initiated cooling within 6 hours of injury or admission to hospital,^{11 - 12 ,34 ,47 ,50 - 52} and this was unclear in 4 studies.^{33 ,48 - 49 ,53} Seven studies excluded patients if major injuries to other organs were present^{12 ,32 ,34 ,47 ,49 - 50 ,52} and 3 studies excluded patients if they had sustained hypotension or hypoxia.^{11 ,32 ,34} One study provided a severity of illness score validated for trauma patients³⁴ and no study reported on presence of comorbid illnesses.

Differences between hypothermia interventions were found among studies (Table 1). Two studies induced cooling for 24 hours,^{11 - 12} 5 for 48 hours,^{32 ,34 ,47 ,49 ,51} and 5 for more than 48 hours.^{33 ,48 ,50 ,52 - 53} Three studies induced hypothermia to target a temperature between 33.5°C and 34.5°C^{33 ,47 ,49} and 5 studies to target a temperature between 32°C and 33°C.^{11 ,32 ,50 - 51 ,53} Four studies could not be categorized according to target temperatures.^{12 ,34 ,48 ,52} The rewarming period after discontinuation of hypothermia was 24 hours or less for 6 studies^{11 - 12 ,32 ,34 ,48 ,51} and more than 24 hours in 4 studies.^{33 ,47 ,49 - 50} Two studies did not report the rate of rewarming after discontinuation of hypothermia.^{52 - 53}

Nine studies outlined physiologic goals for managing intracranial pressure^{11 - 12 ,32 - 34 ,47 - 49 ,51} and 5 outlined physiologic goals for managing both intracranial pressure and cerebral perfusion pressure (Table 2).^{11 ,34 ,47 - 49} Cointerventions instituted to achieve these physiologic goals included drainage of ventricular cerebrospinal fluid and hyperventilation, as well as the use of osmotic agents, diuretics, barbiturates, fluid therapy, and vasoactive agents. Four studies described intensity of therapy prescribed for control of intracranial pressure and/or cerebral perfusion pressure.^{32 ,34 ,47 ,51} No study described in adequate detail treatment protocols to explain how each cointervention was used in the individual studies, and no study described how often hypothermia goals were achieved during the study period. Baseline characteristics of the study patients also are summarized in Table 2.

Four studies reported mortality and neurologic outcome at 3 months,^{12 ,32 ,47 ,51} 4 studies at 6 months,^{33 - 34 ,49 - 50} 1 at 1 year,⁴⁸ and 1 at all these time points.¹¹ Two studies did not report follow-up times for mortality.^{52 - 53} Mortality in the normothermic group ranged from 0% to 82%, with 2 studies reporting mortality rates between 80% and 90%,^{33 ,52} 3 between 40% and 50%,^{48 ,51 ,53} 1 between 30% and 40%,³² 3 between 20% and 30%,^{11 ,34 ,50} 2 between 10% and 20%,^{12 ,47} and 1 with no deaths.⁴⁹ Ten of the 12 studies reported neurologic outcomes.^{11 - 12 ,32 - 34 ,47 - 51}

Two studies were graded as high methodological quality,^{11 ,34} 3 as moderate quality,^{32 ,48 ,50} and 5 as low quality(Table 2).^{12 ,33 ,47 ,49 ,51} Three studies reported losses to follow-up or missing information pertaining to the outcome assessment for a total of 5 patients in the hypothermia group and 21 patients in the normothermia group.^{11 ,32 ,47}

The pooled RR of death was 0.81 (95% CI, 0.69-0.96; Cochran Q = 8.12), conferring a significant protective effect of the therapeutic hypothermia relative to normothermia (Figure 2). Although statistical heterogeneity was not detected in the pooled results, we were still concerned about the possibility of heterogeneity in the hypothermia interventions. Indeed, cooling induced for greater than 48 hours was associated with an even greater reduction in the risk of death (RR, 0.70; 95% CI, 0.56-0.87) (Figure 3). This effect remained significant after removal of 1 study³³ that cooled only a subset of patients (30%) for more than 48 hours (RR, 0.72; 95% CI, 0.57-0.91). None of the other subanalyses were statistically significant. Analyses exploring methodological quality did not detect evidence of bias in estimation of the treatment effect for mortality. Reduction in the risk of death remained significant after removal of the largest clinical trial³⁴ (RR, 0.75; 95% CI, 0.62-0.91). An inverted funnel plot did not suggest evidence of publication bias.

The pooled RR for a poor neurological outcome was 0.78 (95% CI, 0.63-0.98; Cochran Q = 16.05) (Figure 4). However, the analysis of heterogeneity did show statistical significance (Figure 5). The benefit also was present when hypothermia was induced for more than 48 hours (RR, 0.65; 95% CI, 0.48-0.89) or for 24 hours (RR, 0.61; 95% CI, 0.39-0.97). Studies that included both induced moderate hypothermia and rewarming within 24 hours after discontinuation of hypothermia also were associated with a reduction in the risk of a poor neurologic outcome (RR, 0.61; 95% CI, 0.45-0.83 and RR, 0.79; 95% CI 0.63-0.98, respectively). No other comparisons between hypothermia interventions were statistically significant. Analyses according to methodological quality did not reveal evidence of bias in estimation of the treatment effect.

We observed a significant reduction in the risk of death and poor neurological outcome in favor of therapeutic hypothermia compared with normothermia. However, we hypothesized that heterogeneity related to different hypothermia interventions (ie, depth and duration of cooling and rate of rewarming after discontinuation of hypothermia) could have a differential effect outcome. Indeed, we found that hypothermia induced for 24 hours was associated with a significant reduction in the risk of a poor neurologic outcome, and if induced for more than 48 hours, the reduction in risk was significant both for neurologic outcome and death. The risk of a poor neurologic outcome also was reduced with the induction of hypothermia to a target temperature between 32°C and 33°C, and when rewarming occurred within 24 hours after discontinuation of hypothermia.

Although therapeutic hypothermia appears to show a benefit, it also is important to be aware of its potential complications: the risk of arrhythmias, coagulopathies, and infections.^{24 - 26} With hypothermia in the range of 32°C to 34.5°C, infections, and especially pulmonary infections, are the most clinically relevant. Indeed, the Cochrane review found an increased odds of pulmonary infection in the hypothermia group compared with the normothermia group.³⁸ Because no additional studies in our review reported this complication, we did not perform this analysis.

Both laboratory and clinical studies have suggested that time to initiate cooling, as well as duration and depth of cooling, and rate of rewarming after discontinuation of hypothermia may cause different physiological and biochemical effects, and hence, a potential for varying effects on secondary brain injury and clinical outcome. A randomized controlled experimental study of brain injury in a rat model found that neurologic deficits were reduced to a greater extent when hypothermia was induced within 60 minutes of the injury vs after 90 minutes or more.³ A secondary analysis of data from a large multicenter RCT of hypothermia and head injury also supported this finding.³⁴ In a subset of 81 patients who were hypothermic on admission (<35°C) and younger than 45 years, a trend toward a better neurological outcome at 6 months was noted in comparison with patients randomized to the normothermia group (RR, 0.7; 95% CI, 0.5-1.0).³⁴ However, the studies that we reviewed did not provide sufficient detail to analyze the time from injury to initiation of hypothermia.

Prolonged duration of hypothermia also may reduce the effects of secondary brain injury and thus lead to an improvement in clinical outcome. In an experimental model of brain injury, Markgraf et al³ reported that cerebral edema remains elevated at 4 days, and possibly as many as 7 days after the injury. These findings suggest that mechanisms of secondary brain injury are mediated for several days after the initial injury and based on the results of this experimental study, we believe it is plausible that cooling for longer than 48 hours may maximize clinical outcome.

Laboratory studies have suggested that depth of hypothermia may have a varying effect on clinical outcomes. Different depths of hypothermia may have varying protective effects due to differing effects on metabolism and on inhibition of molecular mechanisms of secondary brain injury. Our analysis, according to depth of hypothermia, found that the risk of a poor neurologic outcome in the hypothermia group was reduced with induction of hypothermia to target temperatures of 32°C to 33°C as compared with the normothermia group.

Rate of rewarming after discontinuation of hypothermia also has the potential to affect outcome. In a laboratory study, Matsushita et al³⁶ induced brain injury followed by a hypoxic insult in rats, followed by either hypothermia or normothermia. Rats with hypothermia then were rewarmed over a period of either 15 minutes or 2 hours. In the group that was rewarmed for 2 hours, the volume of contused brain was significantly lower than matched normothermic control brains. However, the hypothermia group rewarmed within 15 minutes showed no benefit, suggesting that rate of rewarming may be important in improving clinical outcome, particularly after sustaining a hypoxic cerebral insult.³⁷ Our analysis, however, did not detect a protective effect for slower rewarming. Indeed, our results suggest that rewarming within 24 hours may reduce the risk of a poor neurologic outcome. Possible explanations for differences in results between laboratory data and our analysis include different mechanisms of brain injury, different baseline characteristics between the 2 treatment groups, and importantly, possible loss of information because of an arbitrary choice of a time threshold of 24 hours in our analysis.

Experience in treating patients with TBI, as well as quality of care,⁵⁵ detailed treatment protocols for hypothermia, and cointerventions for TBI^{56 - 58} are 4 other important potential sources of heterogeneity in these studies. We did not explore these important aspects of clinical care in any depth given the limitation of information reported in the studies we reviewed.

Two other systematic reviews have evaluated the effect of therapeutic hypothermia on outcomes from TBI.^{38 - 39} We believe that our review adds novel information in our more detailed analysis of hypothermia interventions. We also included 2 more trials than the Cochrane review³⁸ and 6 more trials than Harris et al.³⁹ Finally, we searched the non–English-language literature and included 1 non–English-lanuage study.⁵³

Our results suggest that the greatest clinical benefit may be derived when patients are cooled to a target temperature of 32°C to 33°C, with a duration of greater than 48 hours, and then rewarmed within 24 hours after discontinuation of hypothermia. However, our results should be interpreted with some caution because our analysis included a small number of high-quality trials, thus perhaps biasing our results in favor of the treatment effect.⁴² In addition, 1 large multicenter trial demonstrated no benefit associated with hypothermia. However, the inclusion or exclusion of this trial did not affect the overall interpretation of the results. Finally, other potential sources of heterogeneity not explored in this review, such as the intensity and quality of care delivered, time from injury to initiation of hypothermia, and achievement and adherence to hypothermia interventions and cointerventions, may have affected the clinical outcome assessments. For these reasons, our results should not influence clinical practice at this time. Instead, we believe this review provides valuable information to assist in the design of future clinical trials.

Additional studies are now underway to examine the role of hypothermia for the treatment of TBI. Clifton is currently conducting an RCT for patients who are between 16 years and 45 years of age and have hypothermia on arrival to hospital (<35°C) (information obtained from the metaRegister of Controlled Trials). Currently in progress are 2 RCTs in children (oral and written communication with J. S. Hutchison, MD, and P. D. Adelson, MD, November 6, 2002). In light of the results of our analysis, it may be important to conduct a clinical trial that induces hypothermia to a target temperature of 32°C to 33°C, with duration greater than 48 hours, and a rewarming period of 24 hours or less, to see if benefit from this therapeutic intervention can be maximized.