This was a complex review that included TBI and stroke, and also cardiac arrest for temperature reduction data, and cardiac arrest and neonatal HIE for adverse effects of methods and devices.

Impact of head cooling on functional outcome

The searches found 46 studies of non-invasive head cooling in TBI, stroke and brain injury, of which three were assessed as RCTs with good allocation concealment (see Figure 1). Good-quality RCTs with blinded outcome assessment were prespecified for inclusion in analysis of functional outcome, but none of the three met this criterion and was suitable for this purpose.

Temperature reduction with head cooling

For assessing the effect of head cooling on temperature, studies and trials in cardiac arrest were eligible in addition to those in TBI and stroke. Twelve studies had useable data, including four RCTs and four in cardiac arrest (see Table 1). There was considerable heterogeneity (of patients, reasons for cooling and interventions) in these studies, making it difficult to summarise the data.

Two studies showed no effect,46,48 but in both cases this seems likely to be because of the methods used. In one of these, ambient temperature nasal airflow delivered to intubated, ventilated patients to replicate normal nose breathing showed no effect on intracranial temperature at a distance from the nasopharynx.46 With this method the temperature gradient between the patient and the airflow was relatively small. In the other, ice bags on the head and neck for 5–30 minutes did not further reduce temperature in patients who were already on average hypothermic.48 This method does not actively remove heat by coolant flow.

However, in broad terms the data indicate that liquid head- and neck-cooling devices and the Rhinochill intranasal cooling device can reduce intracranial and/or core trunk temperature by around 1 °C or more, within 1 hour in some studies (see Table 1). (These methods create a relatively steep temperature gradient between the patient and the coolant, and actively remove heat by coolant flow.) This is promising and, in particular, suggests that there may be a role for liquid head-cooling devices for induction and maintenance of modest temperature reduction in TBI and stroke (the Rhinochill device is not designed for prolonged use). A small observational study51 showed that it was possible to successfully treat fever refractory to standard management (paracetamol, metamizole, alcohol washing and ice packs) in this way (see Table 1). It is noteworthy that, even in the presence of active body warming, intracranial temperature was reduced with a liquid head-cooling device and could be reduced below core trunk temperature.45,50

This is in contrast with mathematical modelling studies of head cooling, which are sometimes cited to support the view that external head cooling with various devices, including liquid cooling helmets, has a very limited effect on intracranial temperature. The modelling data, even when incoming carotid temperature is varied (which is more realistic than models that treat it as fixed at 37 °C) suggest that these devices reduce brain temperature only superficially, up to about 18 mm below the parenchymal surface.107,108 Even if the usual site of parenchymal temperature measurement – at 1 cm below the brain surface – is considered too superficial to provide valid information on whether or not deeper brain is cooled, the data in this review include examples of ventricular temperature and core body temperature reduction with head cooling (see Table 1). It is hard to conceive that core trunk temperature would be reduced with head cooling, for example with the Rhinochill intranasal cooling device, in the absence of some deeper brain temperature reduction (see Appendix 1).

Head cooling compared with systemic cooling

The reason commonly given for using head cooling is that there may be fewer side effects than with systemic hypothermia.109 Some investigators simply assume that cooling the head and keeping the body warm will minimise systemic complications of hypothermia.45,50 Some head-cooling device providers have also made that assumption. The TraumaTec website states: ‘Selective brain cooling with the Neuro-Wrap™ avoids the complications seen with full body cooling … Complications of systemic hypothermia do not occur as systemic normothermia is maintained’ (; accessed 28 April 2011). Also, on the Benechill website, regarding the Rhinochill device: ‘it is core temperature reduction that causes problems in cooling – not brain temperature reduction’ (; accessed 28 April 2011).

Harris and colleagues45 were unable to achieve an intracranial temperature of 33 °C with head cooling while maintaining bladder temperature at 36 °C with active warming (see Table 1), although whether or not such a large temperature gradient is necessary or desirable remains to be determined. Because a statistically significant intracranial body temperature was not achieved, they concluded that their head-cooling device was not useful in management of TBI. Nevertheless, mean intracranial temperature was reduced below body temperature by 0.67 °C, and Wang and colleagues,50 also in the presence of active body warming, achieved a mean 1.6 °C reduction of intracranial temperature below body temperature. This is a reversal of the norm, in which intracranial temperature is usually higher than body temperature,110 and could well be considered clinically relevant for that reason, although whether there is therapeutic benefit or otherwise is not known. It is difficult to measure intracranial temperature gradients in humans but, in animals, head cooling in the presence of body warming to normothermia has been shown to significantly increase intracranial temperature gradients compared with systemic normothermia and hypothermia, although, again, it is not known if this is harmful.111,112

This review found no high-quality RCT evidence on the relative complications and benefits of head compared with systemic cooling in adults and there are no RCTs making that comparison in neonatal HIE. However, there is some circumstantial evidence from other sources that is relevant to the question of whether or not a hypothermic brain and relatively warmer body may produce fewer complications than systemic hypothermia.

The side effects of systemic hypothermia at temperatures of 33–35 °C include pulmonary oedema, rebound increases in ICP on rewarming, higher temperatures post hypothermia, coagulation abnormalities, metabolic effects and immune suppression.113 However, the reporting and definition of complications in clinical trials of systemic hypothermia is variable, which makes assessing their impact difficult. Nevertheless, systematic reviews of trials of systemic hypothermia after brain injury have shown a non-significant increase in occurrence of infections with cooling therapies (i.e. not only with hypothermic therapy) in stroke14 and of pneumonia in hypothermia for TBI.15 But whether or not head cooling results in fewer complications than systemic hypothermia is likely to depend on the mechanisms by which the complications are caused, how the cooling and warming of the blood as it circulates through a cooler brain and relatively warmer body affects these, and how extreme the temperature gradients are.

With regard to infection, because immune response is modulated by the brain114,115 it seems unwise to assume that brain cooling, with the body relatively warmer, will cause less immune depression and infection than systemic cooling. The primary hormonal pathway for brain– immune system interactions is the hypothalamic–pituitary–adrenal axis115 and, consequently, it is thought that brain cooling does suppress immune function because the pituitary is cooled.116 Furthermore, if immune defence is accelerated and more efficient at increased temperatures,117 reducing brain temperature even to the normothermic range might increase morbidity and mortality from infection.

Another undesirable effect of cooling is shivering, with the requirement for sedation to prevent it. Shivering and stress response to cold may occur even if brain temperature alone is reduced below the ‘set-point’, as cooling of the preoptic area of the hypothalamus is sufficient to cause heat production and retention responses.118,119 Lim120 controlled brain temperature and core trunk temperature independently in anaesthetised dogs using bilateral carotid antegrade cerebral perfusion with independent control of body temperature. Cool brain–warm body and warm brain–cool body conditions both produced shivering. Therefore, brain cooling, even if the body is warm, may not prevent shivering.

If the question of whether or not head cooling does produce fewer complications than systemic cooling is to be answered then a good-quality RCT is needed. There is a small safety and efficacy study ongoing in stroke. The Cerebral Hypothermia in Ischaemic Lesion (CHIL) trial has a head-cooling arm in China, a systemic hypothermia arm in Australia and a normothermic control group, with blinded outcome assessment of the National Institutes of Health Stroke Scale (NIHSS), modified Rankin Scale (mRS) and Barthel Index (BI) at 90 days (see Appendix 6, Characteristics of ongoing studies).

Head-cooling terminology and search terms

There is no agreed terminology to describe cooling that is directed specifically at the head and brain. In the absence of this we have previously suggested the term direct brain cooling and developed the classification of head-cooling methods used in this review.23 This was in part an attempt to provide an alternative to the term selective brain cooling, which has been commonly and erroneously used in clinical papers to describe head-cooling interventions. Selective brain cooling is a natural thermoregulatory mechanism in which brain temperature is reduced below carotid blood, defined in the Glossary of terms for thermal physiology.121 Although applying cooling to the head can reduce brain temperature below body temperature (see Table 1), this is not physiological selective brain cooling.

Because the purpose of therapeutic cooling is to reduce brain temperature, investigators may use head- and brain-related terms to describe cooling, even when they have used systemic methods. Hayashi’s group in Japan for example refer to ‘brain hypothermia therapy’ and ‘cerebral hypothermia’ but they use whole-body cooling.116,122

The lack of standard terminology makes literature searching more difficult. Key words are variable, if used at all, in relation to cooling method. Medical subject headings (MeSH) do not specifically help in searches for cooling interventions or cooling targeted at particular organs. Gastric hypothermia is the only named method in the MeSH tree structure for therapeutic cooling and brain cooling crosses a number of subject areas. We used MeSH terms within our searches but it would refine indexing and aid searching if cooling interventions were incorporated, at least on the basic level of whether they are invasive or non-invasive, brain directed or systemic.

A helpful consensus has recently been reached on a number of factors related to targeted temperature management in critical care123 and this could usefully be extended to agreeing terminology for cooling methods.

Poor reporting of methods and temperature data

Poor reporting of study methodology and/or temperature data were the main reasons why studies were excluded (see Appendix 6, Characteristics of excluded studies). Poor reporting has ethical implications as well as being frustrating for readers and reviewers. It is a recognised problem that is being actively addressed by the CONSORT (Consolidated Standards of Reporting Trials) group ( with some success.124

If the studies found for this review, even if not randomised, had reported temperatures satisfactorily there would have been more information on proof of concept of head cooling with regard to temperature reduction. But many did not adequately report the cooling interventions, the temperature outcomes, where temperature was measured or temperature management in control groups (see Appendix 6).

A consensus report from a meeting of five international critical care societies has recently been published, which includes criteria for reporting studies of targeted temperature management in critical care. The therapeutic effect, safety and reproducibility of temperature management should be reported just as with a drug, including:

Accurate reporting of the indication for temperature management, the interval between disease onset and cooling, the management profile, including the rates of decrement and increment as well as the temperatures achieved, and a comprehensive description of the effects on each body system. (p. 1114).123

Therefore, in addition to rigour in reporting study design,125 information reported in head-cooling studies should include sufficient detail on the cooling method(s) to allow replication, the temperature measurement sites, actual temperatures in intervention and control groups at baseline and with cooling and during rewarming, temperature management strategy (e.g. normothermia or no intervention) and temperature in control groups, and complications/adverse effects from cooling. Providing this information is particularly important because head-cooling research is still largely at the explanatory stage of whether and to what extent different methods reduce temperature.

Chinese studies of head cooling

Twenty-six of the studies on head cooling found for this review were Chinese. The reports were generally relatively short and, unfortunately, none gave sufficient detail on methods to allow trial quality to be adequately assessed or sufficient information on temperature. Poor reporting was also found in a review of leading Chinese medical journals by the Chinese Cochrane Centre.126 It may be partly explained by lack of formal training in research methods and failure of journals to adopt reporting criteria such as CONSORT,126,127 but is not limited to Chinese trials124 (and see Appendix 6).

Information on the method of randomisation was missing or scanty (e.g. ‘computerised’, ‘number method’) and sample size calculation and whether or not the analysis was on an intention-to-treat basis were not reported. It was unclear if analysis was prespecified and sometimes the time scales of result reporting were ambiguous (days on which blood tests were carried out, for example), which suggested that positive results may have been selected for reporting (e.g. see Xu and colleagues128). None of the Chinese studies reported on blinding of treatment allocation, analysis or outcome assessment. In cooling studies it is not necessarily feasible to blind investigators to treatment allocation but we considered that blinded outcome assessment was important and therefore prespecified in the review protocol that studies in which this was not undertaken would be excluded from the formal analysis.

Recently the Chinese Cochrane Centre conducted a study to assess the adequacy of randomisation of peer-reviewed trials published in Chinese that purported to be RCTs.129 Trial investigators were interviewed on the telephone and of 2235 studies only 207 (6.8%; 95% CI 5.9% to 7.7%) were found to be authentic RCTs. Most of those interviewed (85.6%) did not fully understand randomisation when they claimed that their trials were randomised. However, 5.1% did understand randomisation and still claimed that their trials were properly randomised when they were not. Although we had limited success in contacting Chinese authors for this review, the corresponding author of one study who did respond said that the trial was not randomised, although in the paper it was reported to be randomised using a randomisation table.75

Selection bias owing to inadequate randomisation may be one reason for the relatively high proportion of positive results that has been noted in Chinese trials.129 Those found in the searches for this review were all positive and, although the inadequate reporting of methods has made it impossible to assess their quality, it is plausible that selection bias and bias from unblinded outcome assessment and analysis were contributing factors.

Typically, if hypothermia was the aim, only the target temperature for head cooling was reported or, for reduction of fever, for example, the temperature at which the cooling device was set. The actual temperatures prior to cooling and during induction, maintenance and reversion in the intervention groups were not reported nor was the site of temperature measurement always specified [when it was, this is noted under characteristics of included/excluded studies (see Appendix 6)]. General information on the time to reach target temperature was sometimes provided, for example Yang and colleagues130 noted that it took 30–60 minutes to achieve a brain temperature of 35 °C and 3–4 hours to reach 32–35 °C. The implication is that large reductions in brain temperature can be achieved rapidly with non-invasive head cooling, which could be important and clinically relevant if more detailed information was available. There was also little information on temperature management in control groups, for example normothermia. Most papers simply stated that groups were treated the same with the exception of cooling.

A number of the Chinese stroke studies assessed outcome with the neurological deficiency score (NDS) (see Appendix 6). This is not an assessment of functional outcome and is not considered well validated by the Cochrane Stroke Group. It was not one of our prespecified assessment tools (; accessed 24 April 2011).

Cerebral oedema volume was used as an outcome measure in several Chinese stroke studies and a Japanese stroke study (see Appendix 6), but the methods used were not adequately explained. This was not one of our prespecified outcomes because the validity of cerebral oedema volume as a measure of improvement in brain injury is not established and there is no agreed method of measuring it (e.g. see Degos and colleagues131 and Lescot and colleagues132).

Ethical review and informed consent was not always reported in the Chinese studies. Medical ethics in China has been described as ‘anaemic’, a state of affairs attributed to there having been no equivalent to the Nuremberg Code because Second World War atrocities were not addressed, as they were in Europe by the Nuremberg Trials.133 However, there have been considerable developments in medical research ethics in China since the 1990s, which include the requirement for ethical review.134 But there are philosophical differences between the principle of individual autonomy on which ethics in Europe and North America are based, and the traditional Chinese focus on ‘social harmony over individual interests’ (p. 1867).134 No single approach necessarily has the monopoly on ethical ‘correctness’ and sensitivity to cultural differences is important.135

Medical treatment is not free in China.136 Hospitals receive little government subsidy and therefore have to sell services. Drugs and medical consumables in particular are relatively expensive and may be subject to corrupt pricing,137 and corrupt purchasing and prescribing in hospitals.138 The issue of cost was touched on in some of the Chinese study reports and has implications for trial validity. In one study, patients who were allocated to head cooling but could not afford the head-cooling device were cooled with ice packs.139 Ou and colleagues140 investigated different durations of head cooling and noted that longer cooling increased the cost for patients. This also meant that the timing of patients’ discharge from hospital was not necessarily dictated by their condition but by their ability to pay for continued care, which is a potentially confounding factor in studies where functional or neurological outcome after head cooling was followed up on hospital discharge (e.g. see Dong and colleagues141).

This is not intended to single out Chinese studies for special criticism: it just happens that they formed a large proportion of the studies found by searches for this review. In summary, what we found with regard to reporting quality in the Chinese studies is consistent with a recent systematic review of the quality of Chinese RCTs142 and a review of reporting quality in leading Chinese medical journals,126 both undertaken by the Chinese Cochrane Centre. There are initiatives to improve trial conduct and reporting in China, and such initiatives are already having an effect elsewhere, although reporting is still not all that it should be.143

Potential biases in the review process

This systematic review addresses clear research questions and used predefined inclusion criteria to select and appraise studies. We conducted extensive and sensitive searches but the possibility of publication bias remains. There may be, for example, more trials published in Chinese journals than we found. The majority of the trials found were small and/or, on the basis of the reports, of low methodological quality. If the trial reports did not reflect the true quality of the trials then it is possible that there are excluded trials that should have been included.

Agreements or disagreements with other reviews

Four previous reviews of head cooling were found: two, published in Chinese, in patients with cerebral haemorrhage,144,145 one that included human and animal studies in TBI146 and one that included human and animal studies in TBI, stroke, cardiac arrest and neonatal HIE.23 None were systematic.

No head-cooling studies were included in the most recent Cochrane systematic reviews on hypothermia for traumatic head injury,15 modest cooling therapies (35 °C to 37.5 °C) for TBI,13 and cooling therapy for acute stroke (ischaemic or haemorrhagic).14 Sydenham and colleagues15 excluded two head-cooling studies, one75 because the cooling intervention was of < 12 consecutive hours’ duration, the other147 was not a RCT, another study76 was awaiting assessment. Saxena and colleagues13 excluded five head-cooling studies46,47,50,76,147 for physiological end points46,47, methodological reasons76 and target temperature outside the review scope,50,147 respectively. den Hertog and colleagues14 excluded three head-cooling studies50,148,149 because the outcome measures were not relevant to the review (relevant outcome measures were functional outcome, mortality, mean temperature 24 hours after start of cooling, cerebral haemorrhage, complications).

Image ch4f1