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

Hypothermia in the management of traumatic brain injury. A systematic review and meta-analysis.

Henderson WR, Dhingra VK, Chittock DR, Fenwick JC, Ronco JJ.

Intensive Care Med. 2003 Oct;29(10):1637-44. [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?

The authors did not present a "clinical question," per se; instead, they cited as their primary objective, "to appraise critically and summarize all published and peer-reviewed, randomized, controlled trials of the use of hypothermia in traumatic brain injury." Primary and secondary outcome measures were death and neurological impairment, respectively. They also analyzed the risk of pneumonia associated with hypothermia.

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

The inclusion criteria, assembled here, were trials that (1) were controlled and "properly randomized;" (2) enrolled patients with traumatic brain injury; (3) had an intervention arm consisting of iatrogenic hypothermia; (4) examined at least one clinically relevant outcome (i.e., mortality and/or Glasgow Outcome Score); (5) were published; and (6) were published in English. Population characteristics (adult vs. pediatric) were not specified. Data about rates of pneumonia were extracted from trials when available.

B. Secondary questions:

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

It is possible that relevant studies were missed. Appropriate MeSH terms and text words were used to perform electronic searches in MEDLINE, EMBASE, Science Citation, Dissertation Abstracts, Current Contents, and the Cochrane Database. Abstracts from subspecialty conference proceedings were hand searched. However, searches were limited to published trials and to those in the English language, and authors of RCTs were not queried for on-going or unpublished trials.

Interestingly, McIntyre et al (1) published a systematic review on the same topic in the same year (see accompanying appraisal.) Four studies were included in the McIntyre review that were not in the Henderson review. Of these four, two studies were intentionally excluded by Henderson due to design issues and two studies were not identified by Henderson's search strategy (2,3).

4. Was the validity of the included studies appraised?

Though searches were limited to published trials, no assessment of publication bias was performed. Publication bias refers to the greater tendency for studies demonstrating a positive effect to be published. Inclusion of only published data in a meta-analysis thereby increases the risk of over-estimating a treatment effect. Since Henderson et al. did not ask investigators about unpublished data, publication bias may have been introduced.

An objective strategy for assessing trial quality is not described. The authors discuss and tabulate differences in the inclusion and exclusion criteria of the selected trials as well as differences in the onset, degree and length of hypothermia. However, neither the contribution of each trial to the pooled result nor the pooled result after systematic exclusion of each trial (sensitivity analysis) is described. Other important potential confounding variables are also discussed (please see Other Comments below.)

5. Were assessments of studies reproducible?

Yes. While the number of reviewers is not mentioned, "inter-reviewer disagreements regarding inclusion were resolved by consensus." However, data was extracted from the studies by one author.

6. Were the results similar from study to study?

Heterogeneity of the included trials was assessed using the Breslow-Day test. The trials were statistically non-heterogeneous with respect to mortality and thus combinable (chi-square = 8.78, p = 0.19), but they were heterogeneous with respect to neurological outcome in survivors and thus not combinable (chi-square = 16.56, p = 0.020.) One trial did not randomize patients with refractory ICP > 25 mm Hg (thinking that the potentially beneficial effect of iatrogenic hypothermia on lowering elevated ICP may have been masked;) after excluding this trial, Breslow-Day testing yielded similar results. No testing for heterogeneity was performed with respect to rates of pneumonia.

Unfortunately, the authors do not present a scatter plot for easy inspection. Results are presented in tables as odds ratios (e.g. odds of dying in the hypothermia groups vs. odds of dying in the control groups). Of eight trials included in this systematic review, the OR of mortality favored hypothermia in two trials; however, the 95% confidence intervals (CIs) of both trials crossed the line of no difference. The remaining trials did not demonstrate significant benefit or harm. With respect to the "effect of hypothermia on death or severe disability" (GOS 1, 2, or 3,) the OR favored hypothermia significantly in two trials (with CIs < 1.0), and the remaining trials found no benefit or harm.

II. What are the results?

1. What are the overall results of the review?

8 RCTs were included in this systematic review, representing a total of 748 patients - the number of patients each in the treatment and control groups is not discernable. From meta-analysis, the pooled odds ratio for mortality in the hypothermic group was 0.83 (95% CI 0.59-1.13, p = 0.22), demonstrating a statistically insignificant benefit from hypothermia given that the CI crosses 1.0. Despite finding the neurological outcome data to be heterogeneous and thus unable to be combined, a pooled OR for death or poor neurological outcome of 0.75 (95% CI 0.56-1.01, p = 0.06) is reported.

In one sensitivity analysis, the authors excluded a trial that did not randomize patients with refractory ICP > 25 mm Hg. Pooled ORs remained statistically insignificant for mortality alone, and 0.67 (95% CI 0.49-0.92, p = 0.01) for death or severe neurological outcome (although, again, the trials were found to be statistically heterogeneous in the latter.)

In a second sensitivity analysis, the authors excluded 3 trials in which the duration of hypothermia was prolonged or unknown. Pooled OR remained statistically insignificant for mortality (0.9, 95% CI 0.62-1.32, p = 0.1) and for poor neurological outcome (0.7, 95% CI 0.54-1.08, p = 0.13.)

Five of eight trials reported pneumonia rates. Although no testing for heterogeneity was performed on this data, the pooled OR for pneumonia in the normothermic group was 0.42 (95% CI 0.25-0.70, p = 0.001), i.e. the control groups had an overall lower rate of pneumonia that was statistically significant.

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?

From this systematic review and meta-analysis, we must conclude that hypothermia does not improve mortality or neurological outcome from TBI in adults. This conclusion, however, comes with the uncertainties inherent to meta-analyses. If the conclusions from McIntyre et al. (1) are also taken into consideration (who find a statistically significant protective effect with respect to mortality and neurological outcome), we can say that perhaps we have yet to learn the best way of utilizing hypothermia after TBI. This is supported by the sub-group analyses of McIntyre et al., whose findings suggest that deeper cooling (32-33 °C) of longer duration (> 48 hrs) and with quick re-warming (< 24 hrs) may be the best strategy for therapeutic hypothermia.

Both meta-analyses include trials that enrolled adult patients. The developing (i.e., pediatric) brain, however, may be distinct in its susceptibility to hypothermia. Thus, hypothermia in infants and children could afford more neuroprotection or do more harm. It is impossible to say without an RCT that focuses on this question.

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. Henderson et al. chose to look at rates of pneumonia in hypothermia vs. control groups as well. Pneumonia is an oft-observed association in hypothermia trials; the significant association of pneumonia with hypothermia found in this meta-analysis is useful information. Other known complications of hypothermia (cardiac arrhythmias, coagulopathy) are generally not seen at the depths of hypothermia achieved in these trials.

3. Are the benefits worth the harms and costs?

At this time, we can say that no significant benefit has been demonstrated with utilization of hypothermia after TBI. Thus, while no harm is demonstrated here, currently the costs seem to outweigh the benefits of adopting therapeutic hypothermia after TBI into routine care. Certainly more work needs to be done in pediatric populations first.

Other comments

Henderson et al. appropriately discuss several important potential confounders that may have dampened the treatment effect. Three of the included trials which showed the greatest treatment effects did not discuss how temperature was managed in the control groups, i.e., if hyperthermic patients were actively cooled or if hypothermic patients were actively warmed. Thus, if hyperthermia in control patients was not appropriately managed, these trials may have demonstrated a "deleterious effect of hyperthermia rather than a beneficial effect of hypothermia." Variation in the practice of actively rewarming control patients may have confounded outcomes in a similar fashion. The authors state that it was not possible to determine the uniformity of ICP management strategies across trials (such as the use of neuromuscular blockade or sedation.) Three of the trials used ICP normalization as an endpoint for therapeutic hypothermia, resulting in longer hypothermia times. Consistency in the diagnostic criteria of pneumonia across trials is not known.

Several factors could have made this systematic review stronger. Search criteria could have been broadened to include unpublished and ongoing trials, as well as trials published in non-English languages. Publication bias and trial quality could have been assessed objectively. Data regarding rates of pneumonia could have been tested for combinability. To get a feeling for how much individual included trials affected the pooled results, a sensitivity analysis (i.e., pooled result after systematic exclusion of each trial) may have been useful.

Finally, comparing the Henderson and McIntyre (1) reviews highlights the point that study selection and inclusion can dramatically affect the results of meta-analyses. Henderson chose to exclude two studies due to experimental design issues that McIntyre included. McIntyre found 2 additional studies that were not published in English. McIntyre et al. found an improvement in outcome with therapeutic hypothermia whereas Henderson at al. did not. Both meta-analyses could have used a "weighting" method when pooling the data, giving more weight to well-designed studies and less weight to those that were not.

References:

1. McIntyre LA, Fergusson DA, Moher D, Hutchison JS. Prolonged therapeutic hypothermia after traumatic brain injury in adults: A systematic review. JAMA; 2003 289 (22): 2992-2998. [abstract]; [PedsCCM EB Journal Club Review]
2. Zhang K and Wang JX. Comparative study on mild hypothermia in patients with severe head injury and the most severe head injury. Inner Mongol Med Journal 2000; 32: 4-6.
3. Yan Y and Tang W. Changes of evoked potentials and evaluation of mild hypothermia for treatment of severe brain injury. Chinese J Traumatology 2001; 4: 8-13. [abstract ]

Comments

Back to the EB Journal Club Index