Criteria abstracted from The Users' Guide to Medical Literature, from the Health Information Research Unit and Clinical Epidemiology and Biostatistics, McMaster University

Highlighted lines and questions below provide links to the pertinent description of criteria in The EBM User's Guide, now available at the Canadian Centres for Health Evidence

Article Reviewed:

Please visit the new Evidence Based Journal Club Reviews

Prolonged Therapeutic Hypothermia After Traumatic Brain Injury in Adults: A Systematic Review.

McIntyre LA, Fergusson DA, Hebert PC, Moher D, Hutchison JS.

JAMA 2003;289 2992-2999 [abstract]

Reviewed by Nadeem I. Shafi MD, and M. Michele Mariscalco MD, Texas Children's Hospital, Baylor College of Medicine, Houston TX

Review posted March 13, 2005

---

I. Are the results of the study valid?

A. Primary questions:

1. Did the overview address a focused clinical question?

Yes. This meta-analysis focused on the effects of hypothermia on mortality and neurological outcome in adults after traumatic brain injury (TBI.) The authors also sought a priori to determine if the depth, duration, and rate of re-warming from hypothermia influenced mortality and neurological outcome. Thus, these authors sought to distinguish their systematic review from others performed on this subject by focusing on specific aspects of the hypothermia intervention.

2. Were the criteria used to select articles for inclusion appropriate?

Yes. Inclusion criteria specified study design (RCT), target population (adults with TBI), therapeutic intervention and comparison (≥ 24 hours of therapeutic hypothermia at any time after sustaining TBI vs. normothermia), and primary and secondary outcomes (all-cause mortality and neurological outcome). These criteria are sufficiently general such that author bias should not be inadvertently introduced into the selection of studies.

B. Secondary questions:

3. Is it unlikely that important, relevant studies were missed?

Yes, it is unlikely. The selection process was thorough. Appropriate MeSH terms and text words were used to perform electronic searches in MEDLINE, and additional searches were performed using EMBASE, Current Contents, and the Cochrane Library. Subspecialty conference proceedings, bibliographies of all RCTs and review articles on the subject were reviewed by hand. Authors of RCTs were queried for unpublished studies and abstracts, as well as unpublished and ongoing trials. Searches were not restricted to the English language.

4. Was the validity of the included studies appraised?

Yes. The authors appraised the methodologies of the included studies individually by focusing on whether or not the trials performed "allocation concealment" and "blinding of the outcome assessment." The blinding assessment here was of neurological outcome. Methodological quality was stratified into "high" (allocation concealed, outcome blinded), "moderate" (allocation concealment unclear, outcome blinded), and "low" (allocation concealment unclear, outcome not blinded.) Of 12 included trials, 2 were of high grade, 3 of moderate grade, 5 of low grade, and 2 trials were not assigned a grade because they did not assess neurological outcome. Subgroup analyses performed according to trial grade did not reveal bias in estimation of the treatment effect.

The authors evaluated their included studies for publication bias using an inverted funnel plot. They report this revealed no bias, although neither the plot nor the regression analysis results were published.

5. Were assessments of studies reproducible?

Yes. Two reviewers appraised the methodologies of the trials and accepted them by consensus.

6. Were the results similar from study to study?

Relative risks for mortality fell in favor of hypothermia in 6 of 12 trials, in favor of normothermia in 1 trial, and nearly upon the line of no difference in 5 trials; the 95% confidence intervals for all trials touched or crossed the line of no difference, however (visual inspection of scatter-plot.) Relative risks for poor neurological outcome fell in favor of hypothermia in 7 of 10 trials (2 did not evaluate neurological outcome), in favor of normothermia in 2 trials, and nearly upon the line of no difference in 1 trial; again, the 95% CIs touched or crossed the line of no difference.

The authors developed hypotheses a priori and examined specific clinical sources of heterogeneity, namely, the depth and duration of interventional hypothermia and the rate of re-warming from it. They also identified other potential underlying differences in the trials, which they discuss and summarize in tabular form.

The authors performed a sensitivity analysis to examine whether methodological heterogeneity among the trials (high vs. moderate vs. low quality) influenced measured mortality and neurological outcome. The presence of statistical homogeneity was evaluated using the Cochran Q test. Trial data with respect to mortality were homogeneous (and thus combinable) whereas data with respect to neurological outcome were inhomogeneous (and thus not combinable.) P-values are not provided. While the authors went on to combine neurological outcome for completeness, technically this should not have been done. Readers must recognize this and accept results reflecting neurological outcomes with appropriate caution (or not at all.)

II. What are the results?

1. What are the overall results of the review?

12 RCTs were included in the meta-analysis, representing a total of 1,069 patients - 543 patients in the therapeutic hypothermia group, and 526 patients in the normothermia or control group. Results are presented in scatter plots.

In addition to pooled relative risks, the authors examined RR according to (1) depth of cooling (33.5-34.5 vs. 32-33°C,) (2) duration of cooling (24 hrs vs. 48 hrs vs. > 48 hrs,) (3) rates of re-warming (≤ 24 hrs vs. > 24 hrs,) and (4) trial quality (concealment of outcome and randomization scheme).

The pooled RR of death was 0.81 (95% CI 0.69-0.96,) suggesting a statistically significant protective effect of hypothermia. The RR of mortality was significant in patients cooled for > 48 hrs (0.79 ; 95% CI 0.57-0.91); shorter cooling times did not demonstrate an improvement in outcome. All depths of hypothermia and rates of re-warming favored hypothermia, but RRs were not statistically significant (95% CIs crossed the line of no difference.) Analyses according to trial quality did not detect significant differences in outcome; an examination of the figures, however, demonstrates that the studies of highest quality had a RR of death close to 1, while studies of moderate or poor quality had RR in the 0.7-0.8 range.

Trials were inhomogeneous with respect to neurological outcome, and thus not combinable. In all patients, hypothermia improved neurological outcome, however, with a pooled RR for a poor neurological outcome of 0.78 (95% CI 0.63-0.98). All sub-group analyses favored hypothermia as well. Cooling to 32-33°C had a RR of 0.65 (95% CI 0.3-0.85). Hypothermia improved neurological outcome regardless of the duration of cooling. Statistically significant reductions in RR were seen with rewarming in < 24 hrs. Methodological quality did not influence neurological outcome in a statistically significant way but all studies favored improvement (i.e., RRs were < 1, but CIs crossed the line of no difference.)

The authors conclude, "our results suggest that the greatest clinical benefit may be derived when patients are cooled to a target temperature of 32 to 33°C, with a duration of greater than 48 hours, and then re-warming within 24 hours after discontinuation of hypothermia." Appropriate cautions that should accompany this conclusion are discussed by the authors. It is also important to note that patients were excluded from all trials if they had multiple injuries.

2. How precise were the results?

Confidence intervals were as noted above.

III. Will the results help me in caring for my patients?

1. Can the results be applied to my patient care?

This meta-analysis included trials that enrolled adult patients, and found a protective effect of hypothermia after TBI. The developing (i.e., pediatric) brain, however, may be distinct in its susceptibility to hypothermia. Hypothermia in infants and children could be more neuroprotective or it could cause more harm. It is impossible to say without an RCT that focuses on this question. Such a trial is reportedly in progress.

2. Were all clinically important outcomes considered?

Mortality and neurological outcome (severe vs. mild-to-moderate) are the most important outcomes to be considered after severe traumatic brain injury. More subtle neurological sequelae such as learning disabilities, behavioral disorders, etc. are worthy of consideration in milder head injuries; hypothermia trials have focused on severe TBI, however, making these outcomes less relevant.

3. Are the benefits worth the harms and costs?

This meta-analysis finds that systemic hypothermia after TBI may significantly reduce the risk of death and poor neurological outcome. The authors do well to provide sufficient data so that numbers-needed-to-treat can be calculated. In doing so, we find that 20 patients would need to be treated with systemic hypothermia to prevent 1 death and 12.5 to prevent one poor neurological outcome (calculations using patient totals). These NNTs are quite good when compared to other accepted interventions in medicine.

These authors did not examine "harms" of hypothermia, as Henderson et al. (1) did.

Given the uncertainties that are inherent to meta-analyses, the fact that a number of variables remain to be considered (time to initiation of hypothermia, how long patients truly spend at the desired temperature, co-interventions with respect to ICP management, co-morbid illnesses, and others), McIntyre et al. point out that "our results should not influence clinical practice at this time."

Other comments

McIntyre et al. identify potential underlying differences in the included trials that may confound the results of this meta-analysis. Eight studies initiated cooling within 6 hours of injury or admission to the hospital, whereas time to initiation was unclear in 4. No trials reported co-morbid illnesses. The uniformity of ICP management strategies across trials (such as use of neuromuscular blockade and sedation, osmotic agents, diuretics, barbiturates, fluid therapy, and vasoactive agents) was difficult to discern. Studies varied with respect to the time points at which they reported outcomes (3 mos, 6 mos, 1 yr, or all of these). The authors also point out that the included trials did not mention how frequently target temperatures were achieved.

This meta-analysis by McIntyre et al. is published in the same year that another meta-analysis on the same topic was reported by Henderson et al. (1) (please see accompanying appraisal.) Henderson found no significant reduction in mortality nor improvement in neurological outcome due to induced hypothermia. McIntyre et al. did not limit their searches to the English language, allowing them to discover two additional RCTs published in Chinese journals. McIntyre et al. identified 12 trials which met their inclusion criteria; only 8 of these trials were included in the analysis by Henderson. Thus, McIntyre et al. included 4 trials which were not included in Henderson's review. Two of these trials were assessed to be of low methodological quality. The other two did not have random allocation of patients, and thus were also of low methodologic quality; since they did not examine neurologic outcome, however, they were not scored in this aspect. While the addition of these 4 trials allowed McIntyre et al. to discern a treatment effect, and while McIntyre's analysis found that trial quality did not influence outcomes in a statistically significant way, the inclusion of these 4 studies (totaling about 1/3rd of all patients) should make us take more care in accepting these results. Both meta-analyses could have used a "weighting" method when pooling data, giving more weight to well-designed studies and less weight to those that were not.

The article appraised here also sheds light on another important utility of meta-analyses. By performing subgroup analyses a priori to examine the clinical heterogeneity of treatment , McIntyre et al. were able to discern that deeper cooling (32-33°C) of longer duration (> 48 hrs) and with rapid re-warming (< 24 hrs) had the most effect in the setting of TBI. This observation is the equivalent of a "dose" effect, and in itself supports the notion that hypothermia deserves further study. While the observation must be validated in an RCT, this treatment effect could only have been found by pooling patients from several clinical trials, that is, by meta-analysis. A confounder is that none of the studies described how long the patients actually remained at their target temperatures. This will need to be addressed in the experimental design of future studies as it is possible that the noted "dose effect" may be greater (or less) than what is found in this analysis.

Interestingly, the results of this meta-analysis also suggest that while therapeutic hypothermia may reduce mortality after TBI, its role may be even more significant in reducing poor neurological outcomes among survivors. McIntyre et al. acknowledge that the impact of certain variables such as co-morbid illnesses cannot be appreciated here. Since mortality from TBI is a function of the severity of injury, and since other patient characteristics may also contribute to mortality, these findings suggest that the niche for therapeutic hypothermia may be in improving outcomes among survivors, be that from severe or from mild-to-moderate TBI.

References:

1. Henderson WR, Dhingra VK, Chittock DR, Fenwick JC, Ronco J. Hypothermia in the management of traumatic brain injury: A systematic review and meta--analysis. Intensive Care Med (2003) 29: 1637-1644. [abstract]; [PedsCCM EB Journal Club Review]

Comments

Back to the EB Journal Club Index