Introduction
Traumatic brain injury (TBI) is a great challenge to public health; more than 50 million people suffer from TBI every year worldwide [
1]. TBI can cause swelling (oedema) in the brain, can increase intracranial hypertension (ICP) and can worsen the injury. Cell death can occur minutes to hours after the injury, and the harmful effects can last for 72 h or longer [
2]. Therapeutic hypothermia (TH) can reduce ICP [
3] and, to some extent, play the role of a neuroprotective agent, thereby protecting the function of neurons, improving the prognosis of patients and achieving the goal of reducing mortality [
4].
To date, TH in patients with TBI remains controversial. The results of a large number of animal experiments support TH management [
5], and numerous studies have shown that TH can improve neurological outcomes and reduce mortality [
6,
7]. However, in recent years, some studies have considered that TH, compared with the control condition, did not ameliorate outcomes among patients with severe TBI [
8,
9]. Moreover, a large multicentre trial showed that TH played a negative role in the mortality rate and functional outcome [
2]. From this, we can see that the TH strategies remain controversial in patients with TBI.
Systematic reviews have also reported conflicting results [
10‐
12]. A large meta-analysis reported a benefit of TH, but this may be due to the influence of a large number of low-quality studies [
12]. However, a recent meta-analysis suggested that TH could cause more mortality and poor outcomes in high-quality studies [
13]. The aim of this meta-analysis is to use RCTs to update the evidence according to when and who administered TH to patients with TBI by analysing 6-month mortality rates, functional outcome, and pneumonia morbidity.
Methods
This meta-analysis was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses: the PRISMA statement [
14]. The review was registered with the PROSPERO International prospective register of systematic reviews (registration number CRD42019121207).
Eligibility criteria
All studies included in our meta-analysis met the following criteria:
1.
Type of research: Clinical randomised controlled trial
2.
Population: Patients with TBI
3.
Intervention: TH management
4.
Control: Normothermia management or fever control
a)
Primary outcomes: 6-month mortality, unfavourable functional outcome [Glasgow Outcome Scale (GOS) score 1–3: 1, death; 2, a vegetative state; 3, severe disability. Or Glasgow Outcome Scale-Extended (GOS-E) score 1–4: 1, death; 2, vegetative state; 3–4, severe disability]
b)
Secondary outcome: Pneumonia morbidity
Search strategy
We searched the PubMed, Web of Science, Embase, Cochrane (Central) and Clinical Trials databases from inception to January 17, 2019, for studies discussing TH management in patients with TBI. All of the studies we included were independently screened and read by two authors. By reading the abstracts and topics, we excluded clearly unrelated literature, and by reading the full texts, we included only articles that fully met the requirements. When there was a disagreement about a study, the third author arbitrated discussions until a decision was reached. All of the included studies were limited to English articles that could be retrieved. In addition, we manually reviewed the relevant journals that were available.
Data were collected using an author-created information extraction form. The two authors independently extracted the required content by screening the literature. When there was a dispute about a study, the two authors reached a consensus through discussion. If no consensus could be reached, the third author arbitrated until a final decision was reached. The data extracted from each trial included the following: first author, publication date, sample content, inclusion criteria, exclusion criteria, Glasgow Coma Score (GCS) on admission, outcome data for the GOS or GOSE, induction time, target temperature of the hypothermia group, hypothermia duration, rewarming rate, follow-up time and study results.
Bias risk assessment
The Cochrane Collaboration’s tool for assessing the risk of bias was used. The items assessed were random sequence generation, allocation concealment, blinding of the participants and personnel, blinding of the outcome assessment, incomplete outcome data, selective reporting and other biases (Additional file
1: Figure S1). In order to quantify the quality of the articles, we performed a subgroup analysis of quality assessment according to the modified Jadad score (Additional file
13: Table S1). In addition, we also conducted a subgroup analysis with reference to the quality evaluation method of Watson et al. [
13] (Additional file
14: Table S2).
Trial sequential analysis
To prevent the constantly updated meta-analysis from increasing the risk of type I errors, we conducted a TSA that could also estimate the amount of information required for such research, thereby stopping similar research in time and preventing the waste of medical resources. We performed a one-sided TSA to summarise and analyse the data of the included studies for the functional outcome with 5% risk of type I error and 80% power.
Statistical analysis
All statistical aspects of the meta-analysis were performed using Review Manager 5.3 software. All our outcomes comprised dichotomous data, and the pooled risk ratios and 95% confidence intervals of these data were calculated. In terms of statistical heterogeneity, a quantitative analysis was performed using the Mantel-Haenszel (MH) chi-square test and the I-square test; when p was < 0.05 for the MH chi-square test or I2 was > 50% for the I-square test, there was obvious heterogeneity. To evaluate the publication bias, we created funnel plot charts. In addition, we conducted a sensitivity analysis using STATA version 15.1 to determine whether any single study incurred undue weight in the analysis.
Discussion
TH management remains controversial for patients with TBI [
11,
35]. Despite extensive research, there is no high-quality evidence that hypothermia is beneficial to TBI patients, as is to cardiac arrest [
36]. Similar to a recent meta-analysis, our meta-analysis demonstrated that TH could cause more mortality in the subgroup of high-quality studies. And TH initiated within 24 h could reduce mortality in patients with TBI [
13]. Furthermore, we also find TBI patients benefit from TH when hypothermia is used for therapy instead of prophylaxis. Additionally, post-craniectomy TBI patients may benefit more from TH than patients who have not received a craniectomy. In terms of functional outcome, our meta-analysis is consistent with previous meta-results [
2,
12]. Patients with TBI can show improved neurological outcomes with TH within 72 h of injury [
8].
Some RCTs suggest that elevated ICP is associated with worsening outcomes in patients with TBI [
37,
38]. Elevated ICP may result in decreased cerebral perfusion pressure and cerebral blood flow, which may further lead to hypoxic-ischaemic brain damage [
39]. Many studies have shown that TH can play a role in neuroprotection in many ways, mainly because hypothermia can reduce ICP, reduce the brain metabolic rate, reduce the blood flow in the brain, change the release of neurotransmitters and maintain the function of the blood-brain barrier [
40,
41]. Moreover, TH can reduce the inflammatory response and biochemical cascade that is activated early after TBI [
42], thereby limiting secondary brain injury [
43,
44]. A study by Roman et al. showed that TH can improve the functional prognosis of GOS (4–5) by reducing ICP [
7]. Our study also found that hypothermia can improve the patient’s functional outcomes.
However, some studies have shown that hypothermia sometimes plays an adverse role. Several recent multicentre large RCT studies have shown that TH not only failed to reduce patient 6-month mortality but may also be harmful to patients with lesser damage [
2,
8,
9,
32]. Long-term hypothermia is considered a form of immunosuppression that increases the infection rate of pneumonia and sepsis [
45]. In addition, it has been reported in the literature that hypothermia can cause propofol infusion syndrome because propofol can reduce liver metabolism; this may be an important cause of fatal symptoms at low temperatures [
46]. It has also been reported that low temperatures can affect the metabolism of certain drugs, including muscle relaxants such as atracurium, which may further affect 6-month mortality [
47]. The CRASH study, a large multicentre trial, also found a lower 6-month survival rate in the hypothermia group and a higher 2-week mortality rate in patients treated with methylprednisolone [
48]. It has been shown that the use of steroidal antipyretics may also be one of the important factors that influence mortality. The results of our subgroup analysis also showed that high-quality studies suggested that TH can cause an increase in mortality.
We observed smaller studies may note some “benefits” from hypothermia while the more structured large RCTs have failed; we believe it may be because the smaller studies are mostly with small sample size and single-centre. Through the sample size-bias curve, we found that the sample size and the bias score showed a significant positive correlation after the abnormal point was removed (Additional file
11: Figure S11), that is, as the sample size increased, the bias gradually decreased. However, the study of Zhi et al. has a large sample size but a low bias score, and different to the results of other studies with large sample size, we suspect that it may invite bias into the results. The column chart about research centres and bias also support our conjecture (Additional file
12: Figure S12). But since all the included studies meet the inclusion criteria for our meta-analysis, there is no reason to remove any RCT study, which is a limitation of our research. So we recommend more large multicentre RCTs to continue this research.
Why is there controversy regarding TH management for patients with TBI? First, it seems that the induction time of TH is the key point. Our meta-analysis found that both survival rates and functional outcomes will benefit if TH is administered within 24 h after TBI. In animal experiments, the most obvious link between intracranial temperature changes and nerve injury occurred within the first 24 h [
49,
50]. When TH is applied within 24 h after TBI, it may be possible to control the increase of ICP earlier, thereby reducing intracranial nerve injury and improving the functional prognosis of patients. There are data indicating that hypothermia may regulate both the JNK signalling cascade via XIAP and the preconditioning pathways that activate caspases. Thus, hypothermia mediates TNFR1 responses via early activation of the JNK signalling pathway and caspase-3, leading to endogenous neuroprotective events [
51]. Recent studies by Watson et al. also supported early hypothermia in patients with TBI [
13]. At present, there is no consensus regarding when TH management should be used after TBI. After the occurrence of TBI, the severity of numerous destructive biochemical cascades plays a decisive role in the survival of nerve cells [
43]. TH is an effective protective mechanism to inhibit these reactions. We believe that TH management within 24 h is conducive to maximally limiting the infinite expansion of these cascades in a short period of time, thereby avoiding risks, and when the TH time is later (more than 24 h), patients may have more serious damage, and ICP may be more difficult to control; such destructive reactions have been irreversible.
A subgroup analysis of TH for prevention or treatment suggests that TH may be more effective in reducing mortality when used for therapeutic purposes. We believe that the hypothermia applied in patients with TBI after cerebral oedema, increased intracranial pressure or craniotomy is defined as a therapeutic effect, and the application of hypothermia as soon as possible without relevant complications is a preventive effect. These two concepts have not been clearly defined internationally, and we need to recognise the subjectivity of this subgroup analysis. Moreover, the subgroup analysis of post-craniotomy showed that TH after surgery had a tendency to reduce mortality. Previous studies have reported that the use of mild hypothermia as a preventive application of neuroprotective agents has also failed; prophylactic hypothermia is not recommended to improve final outcomes [
52]. Clifton et al. also found that, compared with diffuse brain injury, TH may play a better effect in those with surgically evacuated haematomas [
31]. It has also been confirmed at the experimental level that intra-ischaemic hypothermia after haematoma removal is associated with improved outcomes [
53]. Through the analysis of the above two subgroups, we hypothesise that we can use hypothermia as a treatment for TBI patients while evacuating the haematoma and after cranietomy, which can effectively reduce the mortality rate.
We need to discuss some of the limitations of our work. First, although we performed a comprehensive database search and a manual search and made a funnel plot, and the funnel plot had symmetry, we did not search the grey literature or contact authors to confirm whether there were any unpublished studies. Therefore, we still cannot rule out the existence of a publication bias. Second, the inclusion criteria for each group of trials included in this study were not completely consistent, which may have led to heterogeneity in the observations. The forest plot shows that the difference in weight is relatively large, which may affect the final result to some extent. Finally, we found substantial heterogeneity in some of the outcomes. We tried to reduce clinical and methodological heterogeneity through different subgroup analyses; however, some analyses did not have an obvious effect, and the heterogeneity was still high. Therefore, we used a random effects model instead of a fixed effects model to address the observed heterogeneity. Despite such differences, our sensitivity analysis identified no outlier studies, hinting that our results were relatively reliable.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.