Herbal Medicine for Traumatic Brain Injury: A Systematic Review and Meta-Analysis of Randomized Controlled Trials and Limitations

Background: This systematic review aimed to evaluate the effectiveness (functional outcomes and clinical symptoms) and safety (incidence of adverse events) of herbal medicine (HM) as monotherapy or adjunctive therapy to conventional treatment (CT) for traumatic brain injury (TBI). Methods: We comprehensively searched 14 databases from their inception until July 2019. Randomized controlled trials (RCTs) using HM as monotherapy or adjunctive therapy to treat TBI patients were included. The primary outcome was functional outcomes, consciousness state, morbidity, and mortality. Meta-analysis was performed to calculate a risk ratio (RR) or mean difference (MD) with 95% confidence intervals (CIs), when appropriate data were available. Methodological quality of RCTs and the strength of evidence were also assessed. Results: Thirty-seven RCTs with 3,374 participants were included. According to meta-analysis, HM as a monotherapy (RR 1.29, 95% CI: 1.21–1.37) or an adjunctive therapy to CT (RR 1.21, 95% CI: 1.16–1.27) showed significantly better total effective rate based on clinical symptoms, compared to CT alone. Subgroup analysis showed that HM had significantly improved post-concussion syndrome, dizziness, headache, epilepsy, and mild TBI, but not traumatic brain edema, compared to CT. Moreover, HM combined with CT had significantly improved post-concussion syndrome, mental disorder, headache, epilepsy, and mild TBI-like symptoms, but not cognitive dysfunction and posttraumatic hydrocephalus, compared to CT alone. When HM was combined with CT, functional outcomes such as activities of daily living and neurological function were significantly better than in patients treated using CT alone. In terms of the incidence of adverse events, HM did not differ from either CT (RR 0.88, 95% CI: 0.33–2.30) or placebo (RR 2.29, 95% CI: 0.83–6.32). However, HM combined with CT showed better safety profile than CT alone (RR 0.64, 95% CI: 0.44–0.93). Most studies had a high risk of performance bias, and the quality of evidence was mostly rated “very low” to “moderate,” mostly because the included studies had a high risk of bias and imprecise quantitative synthesis results. Conclusion: The current evidence suggests that there is insufficient evidence for recommending HM for TBI in clinical practice. Therefore, further larger, high-quality, rigorous RCTs should be conducted.

According to the CDC report (3), nearly half of patients with moderate-to-severe TBI undergoing inpatient rehabilitation experience pathological changes in their cognitive function between 1 and 5 years after injury (11). Therefore, to prevent long-term negative consequences and improve QoL, TBI requires long-term management as well as acute, post-injury treatment.
Complementary and integrative medicine (CIM) approaches, including acupuncture and herbal medicine (HM), are often used to supplement the limitations of conventional medicine (12,13), improve effectiveness, and sometimes reduce side effects, even in the management of TBI (14,15). In particular, HM has been used to manage brain trauma such as hemorrhage-related hydrocephalus (16), as well as long-term neurological diseases such as stroke (17), cerebral palsy (18), Parkinson's disease (19), vascular dementia (20), and Alzheimer's disease (21). In the field of brain trauma, common HMs such as Goreisan have been shown to prevent chronic subdural hematoma recurrence (22,23), and the mechanism may involve the regulation of aquaporin, a water channel (24)(25)(26). Similarly, some HMs such as Yokukansan (27) and Xuefu Zhuyu decoction (28) have beneficial effects on TBI-related behavioral changes or cognitive impairment. In the management of TBI, HMs may have beneficial effects through complex mechanisms; they may reduce tumor necrosis factor-α or nitric oxide expression, improve blood-brain-barrier permeability, and reduce brain water content (29). However, no studies have yet synthesized all the clinical evidence for the effectiveness and safety of HM as an adjunctive or alternative therapy for various outcomes of TBI, including functional outcomes (mobility and global disability), mortality, quality of life, global clinical improvement, and adverse events. The present systematic review aimed to evaluate the effectiveness and safety of HM on these outcomes in TBI compared to placebo, no treatment, and conventional treatment (CT), to inform clinicians, policy makers, and patients in how to manage this disease.

Study Registration
The protocol of this systematic review has been published and registered in PROSPERO (registration number, CRD42018116559) (30), and the study was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (31) and the Cochrane Handbook for Systematic Reviews of Interventions (32).

Data Sources and Search Strategy
As previously described, the following 14  , Wanfang Data, and VIP), and one Japanese database (CiNii). The initial search date was December 2, 2018 and we conducted an updated search on July 27, 2019 to retrieve more up-to-date and comprehensive evidence. Additionally, we searched the reference lists of the relevant articles and performed a manual search on Google Scholar to identify further eligible studies. We also included "gray literature, " such as degree theses and conference proceedings, as well as the literature published in journals. There was no restriction on language, publication date, or publication status. The search strategies for all databases are available in Supplemental Digital Content 1.

Types of Studies
We included randomized controlled trials (RCTs) and excluded quasi-RCTs that used an inappropriate randomization method such as alternate allocation or allocation by birth date. Studies were excluded if they used the term "randomization" (随机) but failed to detail the randomization methods used. We included both parallel and crossover studies. Other study designs, such as in vivo, in vitro, case reports, and retrospective studies were excluded.

Types of Participants
We included studies involving patients diagnosed with TBI through medical or radiological examination, regardless of target symptoms, disease severity, sex, age, or race. We included all studies involving TBI patients, even if the diagnostic method of TBI was not clearly stated. We excluded studies that included participants with drug allergies or other serious medical conditions, such as cancer, liver disease, or kidney disease.

Types of Interventions
We included studies that used HM as a treatment intervention, regardless of which formulation of HM was used (e.g., decoction, tablets, capsules, pills, powders, and extracts); however, we only included studies in which HM was administered orally. We excluded studies that failed to detail the composition of the HM used, except when patent medicines were used whose composition could be found by searching the Internet. Studies comparing different types of HM were excluded. As control interventions, we included placebo, no treatment, and CT including surgery, medication, rehabilitation treatment, and psychotherapy for acute management and rehabilitation, which are baseline treatments for TBI. In the present study, acute management was defined as any treatment administered to stabilize the patients immediately after the injury (within 1 month). Rehabilitation was defined as any treatment of longterm impairments that aimed to restore to their previous level of health and was administered more than 1 month after injury (33). We included studies that combined HM with other therapies if the other therapies were used equally in both the treatment and control groups.

Types of Outcome Measures
The primary outcome measure was functional outcome, measured using the following validated scales: Barthel index (BI) (34), functional independence measurement (35), Fugl-Meyer assessment (36), and Glasgow Outcome Scale (GOS) (37). We also analyzed consciousness state measured using validated scales such as the GCS (38), with morbidity and mortality as primary outcome measures.
The secondary outcome measures were QoL, measured using validated assessment tools such as the 36-Item Short Form Health Survey (SF-36) (39), and adverse events (AEs), measured using the Treatment Emergent Symptom Scale (TESS) (40) or the incidence. We also analyzed the total effective rate (TER) as a secondary outcome; this is a non-validated outcome measure that is processed secondarily using certain evaluation criteria, such as improvement in clinical symptoms based on clinician ratings. In TER assessment, participants are generally classified as "cured" (痊 愈), "markedly improved" (顯 效), "improved" (有效), or "non-responsive" (無 效) after treatment. The TER is calculated using the following formula: TER = N1 + N2 + N3/N, where N1, N2, N3, are the number of patients who are cured, markedly improved, and improved, respectively, while N is the total sample size. This outcome was considered a secondary outcome in this review as it lacks a unified standard and can be potentially heterogeneous.

Study Selection
As previously reported, two researchers (B. Lee and C-Y Kwon) independently selected the studies according to the above inclusion criteria. After removing duplicates, we screened the titles and abstracts of the retrieved studies for relevance; we then evaluated the full texts of the selected studies for final inclusion. Any disagreement was resolved through discussion with the other authors.

Data Extraction
Using a standardized data collection form in Excel 2007 (Microsoft, Redmond, WA, USA), two researchers (B. Lee and C-Y Kwon) independently extracted and double-checked the data from the included studies. Discrepancies were resolved through discussion with the other authors.
Using a predefined data collection form, we extracted information regarding the first author's name, publication year, country, institutional review board (IRB), informed consent, sample size, and number of dropouts, diagnostic criteria, participant details, intervention, comparisons, duration of intervention and follow-up, outcome measures, outcomes, and AEs. We also extracted details of the HM used, including the name, source, dosage form, and dosage of each medical substance, as well as the principles, rationale, and interpretation of the intervention in terms of the Consolidated Standards of Reporting Trials Extension for Chinese Herbal Medicine Formulas 2017 (41). If the data were insufficient or ambiguous, we contacted the corresponding authors of the included studies via e-mail to request additional information.

Quality Assessment
As previously reported, two researchers (B. Lee and C-Y Kwon) independently evaluated the risk of bias of the included studies and the quality of evidence of the main findings. We resolved discrepancies through discussion with other researchers.
We assessed the methodological quality of the included studies using the Cochrane Collaboration's risk of bias tool (42). The following items were evaluated as either "low risk, " "unclear, " or "high risk": (1) random sequence generation, (2) allocation concealment, (3) blinding of participants and personnel, (4) blinding of outcome assessment, (5) completeness of outcome data, (5) selective reporting, and (6) other biases. In particular, we assessed other bias categories with an emphasis on baseline imbalance between the treatment and control groups in terms of participant characteristics such as mean age, sex, or disease severity, because baseline imbalance in factors that are strongly related to outcome measures can cause bias when estimating the intervention effect.
The quality of evidence for each main finding was assessed using the Grading of Recommendations Assessment, Development, and Evaluation approach (43), which uses the online program GRADEpro (https://gradepro.org/). The following items were evaluated as either "very low, " "low, " "moderate, " or "high": risk of bias, inconsistency, indirectness, and imprecision of the results, and probability of publication bias.

Data Synthesis and Analysis
As previously described, we conducted descriptive analyses of the participants' details, interventions, and outcomes for all included studies. Using Review Manager version 5.3 software (Cochrane, London, UK), a meta-analysis was performed across studies that used the same types of intervention, comparison, and outcome measure. We pooled the dichotomous data using the risk ratio (RR) with 95% confidence intervals (CIs) and the continuous data using the mean difference (MD) with 95% CIs. We assessed clinical heterogeneity by comparing the distribution of important participant factors, such as age, sex, disease severity, and specific types of TBI, and we compared intervention factors such as co-interventions and control interventions among the included studies. Furthermore, statistical heterogeneity between the studies was assessed using both the chi-squared test and the I 2 statistic; I 2 ≥ 50% indicated substantial heterogeneity, while those ≥75% indicated high heterogeneity. In the meta-analyses, a random-effects model was used when the heterogeneity was significant (I 2 ≥ 50%), while a fixed-effects model was used when the heterogeneity was not significant or when the number of studies included in the meta-analysis was <5, where estimates of inter-study variance have poor accuracy (44,45). If the necessary data were available, we performed subgroup analyses to explain the heterogeneity or to assess whether the treatment effects varied between subgroups categorized according to the following criteria: (1) objective of interventions, such as acute management or rehabilitation, assessed in terms of time frame following injury; (2) severity of TBI, and (3) target symptoms, such as headache, dizziness, cognitive disorder, or mental disorder. To ascertain the robustness of the meta-analysis result, we conducted a sensitivity analyses by excluding (1) studies with a high risk of bias and (2) outliers that were numerically distant from the rest of the data.

Reporting Bias
We assessed reporting biases, such as publication bias, using funnel plots if more than 10 studies were included in the meta-analysis.

Study Description
We identified 27,258 studies through database searching and one study from the references of the relevant studies. After removing duplicated studies, we considered 626 studies relevant after screening of the titles and abstracts. Among these, we finally included 37 studies with 3,374 participants  in the qualitative synthesis, and 33 studies with 3,000 participants (46-48, 50, 51, 53-59, 61-74, 76-82) in meta-analysis after screening of the full-text articles (Figure 1).

Risk of Bias
All the included studies reported appropriate random sequence generation methods; however, only two used a sealed opaque envelope (79) or independent allocation manager (46) to conceal allocation. Only one study (46) appropriately blinded both the participants and personnel, and two studies (68, 69) used placebo drugs as a control intervention but did not report appropriate blinding of personnel. None of the included studies reported blinding of the outcome assessor. Two studies (51, 68) that performed per-protocol analysis were assessed as having a high risk of attrition bias, while two (50, 51) that reported only TER, a secondary processed outcome without the raw data, were assessed as having a high risk of reporting bias. Thirty-five studies (46-51, 53-60, 62-82) reported no significant baseline difference in demographic data between the two groups, and were rated as having low risk of bias in the other potential sources of bias domains (Figures 2, 3).
In one study involving traumatic brain edema (58), the groups did not differ in terms of functional outcome, as measured using the GOS, after 1 month of post-intervention follow-up (MD: 0.10, 95% CI: −0.13 to 0.33), nor did they differ in terms of consciousness state, measured using the GCS after 14 days of intervention (MD: 0.05, 95% CI: −0.12-0.22). In addition, the two groups did not differ in terms of intracranial pressure or neurological function, measured using the China stroke scale after treatment. However, in 10 studies, the TER based on clinical symptoms was significantly improved in the HM group (RR: 1.29, 95% CI: 1.21-1.37, I 2 = 0%). In a subgroup analysis based on the target symptoms of TBI, the HM group showed significantly better outcomes in patients with PCS, dizziness, headache, epilepsy, and mild TBI of all causes except traumatic brain edema (Table 2; Figure 4) (Supplemental Digital Content 3). In a study by Xu et al. (59), when HM was administered to patients with PCS, the symptom improvement time and hospitalization time were significantly shorter than in the CT group (P < 0.05, all). Wang and Tian (67) reported that, when HM was administered to patients of epilepsy, the number of seizures was significantly lower than in the CT group (P < 0.05).

Safety
Three studies reported AEs during the intervention, and a metaanalysis of these showed no difference in the incidence of AEs between the two groups (RR: 0.88, 95% CI: 0.33-2.30; Table 2) (Supplemental Digital Content 3).

HM vs. Placebo
Efficacy Three studies (46,68,69) compared HM with a placebo. Two of these (46,68) were conducted on patients with cognitive dysfunction, while the other one (69) did not include participants with specific symptoms. Collectively, the functional outcomes showed inconsistent results between studies, and there was no significant difference in QoL between two groups. However, memory impairment was improved more in the HM group.
In a study by Wang (69), the HM group showed improved functional outcomes, as assessed using the Fugl-Meyer assessment (MD: 9.63, 95% CI: 8.21-11.05) and modified BI (MD: 18.54, 95% CI: 17.27-19.81), after 8 weeks of treatment. Additionally, hand function in the HM group was significantly better than in the placebo group (P < 0.01). After patients with cognitive dysfunction were treated ifor 6 months (46), physical disability was measured using the GOS and QoL measured by the QoL after brain injury scale showed no significant differences between the two groups (GOS: MD, 0.00; 95% CI: −4.17 to 4.17; QoL after brain injury scale: MD, 1.91; 95% CI: −9.58 to 13.40; Table 2) (Supplemental Digital Content 3). In addition, after intervention, there were no significant differences between the groups in terms of neurobehavioral sequelae, mood, or fatigue. However, complex attention and executive functioning in the HM group were significantly better than in the placebo group (P < 0.05). In a study by Wang et al. (68) involving patients with memory impairment, the HM group showed significantly better memory quotient, measured using the Wechsler Memory Scale, than the placebo group after 4 weeks of treatment (P < 0.01). The results of sensitivity analysis by excluding low quality studies (that had 4 or less low risk of bias on the seven domains of the risk of bias tool) were consistent in GOS and QoL (Supplemental Digital Content 4).

Safety
Two studies (46,68) recruiting patients with cognitive dysfunction reported AEs during the treatment period. There was no difference in the incidence of AEs between the two groups (RR: 2.29, 95% CI: 0.83-6.32, and I 2 = 79%; Table 2; Figure 4) (Supplemental Digital Content 3), nor was there any difference between the two groups in a sensitivity analysis that excluded studies with a high risk of bias (Supplemental Digital Content 4).
Huang and Li (82) conducted 4 weeks of treatment in patients with PCS; they found that activities of daily living were significantly better in the HM plus CT group than in the CT alone group (MD: −3.30, 95% CI: −5.04 to −1.56). Ping (77) conducted 15 days of treatment in patients with posttraumatic hydrocephalus; their results showed that functional outcomes, as measured using BI, were significantly better in the HM group (MD: 11.14, 95% CI: 5.43-16.85) ( Table 2) (Supplemental Digital Content 3). When HM was added to the CT, there was a significant difference in neurological function after treatment compared to that with CT alone, as measured using the National Institute of Health Stroke Scale (NIHSS) (P < 0.01), and degree of hydrocephalus differed significantly between the groups after 1 month of post-intervention follow-up (P < 0.05) (77). Two studies (64,79) reported the QoL of patients after treatment. One (79) showed that patients with PCS treated using HM had significantly better mental component summary score, as measured using the SF-36 scale, than the CT alone group after 6 weeks of treatment (MD: 36.51, 95% CI: 13.76-59.26). However, there was no difference in physical component summary score (MD: 3.84, 95% CI: −13.27-20.95). Another study (64) treated patients with mental disorder for 8 weeks.
The HM group showed significantly better scores in the areas of physical health, psychological health, and social functional status domain, measured using the generic QoL inventory 74.
However, there was no difference between the groups in terms of living condition (physical health: MD, 11.68, 95% CI, 9.11-14.25 (Figure 5). In a subgroup analysis according to target symptoms of TBI, there were significant differences in PCS, mental disorder, headache, epilepsy, and mild TBI-like symptoms, but not in cognitive dysfunction or post-traumatic hydrocephalus ( Table 2) (Supplemental Digital Content 3). However, a sensitivity analysis that excluded studies with a high risk of bias showed no difference in TER based on clinical symptoms between the two groups (Supplemental Digital Content 4).
When HM plus CT was administered to treat patients with PCS, neurological function, as measured using the NIHSS, was better than when CT alone was used (P < 0.05) (60), and cure time was significant shorter in the combination group (P < 0.05) (51). In patients with mental disorder after TBI, symptoms of depression (53), anxiety (53), and schizophrenia (62,64,66) were significantly better in the combination group than in the CT alone group (P < 0.05 in all cases). Furthermore, when HM plus CT was administered, cognitive function, as measured using the mini-mental state examination, was significantly improved (P < 0.05) (61), and the recurrence rate of headache was significantly lower than in the CT group (P < 0.05 in all cases) (50,56). Two studies showed that clinical symptom relief time was significantly shorter in the combination group (P < 0.05 in all cases) (52,73).

Quality of Evidence
In the studies that compared HM with CT, the quality of evidence was graded as "very low" or "low" ( Table 2). Additionally, the quality of evidence was graded as "very low" to "moderate" in studies that compared HM with a placebo, as well as in those that compared HM plus CT with CT alone ( Table 2). The main reason for these low grades was the high risk of bias of the included RCTs. Furthermore, most findings had low precision because they did not fulfill the optimal sample size and had wide CIs. Indirect outcome measures also lowered the quality of evidence, especially in studies that measured TER as an outcome.

Publication Bias
No evidence of publication bias emerged from the funnel plots of TER based on clinical symptoms in studies that compared the effectiveness of HM with that of CT, or in studies that compared the effectiveness of HM plus CT with that of CT alone. Furthermore, the funnel plot comparing AE incidence between the HM plus CT group and the CT alone group was also symmetrical (Figure 7).

DISCUSSION
This review aimed to assess the effectiveness and safety of HM as a monotherapy or adjunctive therapy to conventional treatment for TBI. We conducted a comprehensive and systematic search of English, Korean, Chinese, and Japanese-language databases and retrieved a total of 37 RCTs .
In summary, when comparing HM with CT, there was no conclusive evidence in functional outcome or consciousness state in patients with traumatic brain edema because there was only one study. However, the function measured by Fugl-Meyer assessment, BI, and NIHSS was significantly improved when HM was added to CT in studies that focused on symptomatic treatment or rehabilitation. Results regarding QoL were inconsistent between the two groups after treatment. The present meta-analysis showed that the TER of various symptoms showed significantly better results in the HM group in all comparisons. However, TER is a non-validated outcome measure that is secondarily processed, and thus, assertions regarding HM's effectiveness cannot be made confidently. Regarding the safety of HM, none of the study participants showed obvious abnormalities in electrocardiogram examinations or laboratory tests, such as the blood routine, urine routine, fecal routine, and liver and kidney function tests. There was no difference in the incidence of AEs between the two groups when HM monotherapy was compared with CT or placebo. Conversely, the incidence of AEs and TESS was significantly better in the HM plus CT group than in the CT alone group. However, the risk of bias in the included studies was generally high, whereas the quality of evidence of the main findings was generally low; thus, only limited confidence can be placed in the estimate of the effect, that is, the true effect may be different from the estimate.
Interestingly, pattern identification based on blood stasis was most frequently used in the included studies. In addition, the most commonly used HM was Xuefuzhuyu decoction, and the commonly used single herbs comprising the HM were Cnidii Rhizoma, Angelicae Gigantis Radix, Persicae Semen, Carthami Flos, and Paeoniae Radix Rubra, which improve blood stasis (84,85). In East-Asian traditional medicine, blood stasis is considered the main pathology in traumatic injury (84). According to this pathological concept, blood stasis-removing therapy is widely used to treat TBI in clinical practice, and some clinical evidence has shown that blood stasis-removing HM is effective in the treatment of TBI (86,87). Our review does not prove that blood stasis-removing HM is effective in improving TBI, but suggests that this type of herbal medicine is promising in the field of research for TBI treatment in the future.
Many studies have tried to explain the mechanism through which HM functions in TBI, showing that HM decreases neuronal injury by increasing superoxide dismutase and catalase activities, as well as by suppressing the expression of interleukin (IL)-1, IL-6, nuclear factor kappa B, and glial fibrillary acidic protein (88). Another study showed that HM protected a rat model of TBI, possibly via immune-promoting, antiinflammatory, and neuroprotective effects (89). However, the underlying mechanism of HM in the treatment of TBI is still not fully understood; future studies should address this question to help establish an optimal management strategy for BI. Our review had the following limitations. Firstly, although we conducted a systematic and comprehensive search in English, Korean, Chinese, and Japanese databases, most studies were conducted and published in China. This may have resulted in reporting biases, such as language and location bias. In addition, many studies assessed TER, which is a secondarily processed outcome measure according to certain criteria, and the meta-analysis showed significant results suggesting better outcomes in the HM group. However, this nonstandardized outcome measure may have caused outcome reporting bias, and the results may not have been reliable. Secondly, most of the included studies were not of high quality. In particular, many had a high risk of performance bias. Therefore, our confidence in the effect estimate, as assessed using GRADE methodology, was low. Thirdly, we attempted to perform subgroup analysis in terms of either the objective of intervention (acute management or rehabilitation) or the TBI severity, as described in the study protocol (30). However, few studies clearly specified the objective of intervention or the severity of TBI in a subgroup analysis. Finally, although we performed subgroup analysis according to different target symptoms of TBI to address heterogeneity, we could not resolve clinical heterogeneity because the participants had diverse clinical characteristics and a wide range of interventions were used in the included studies. Relatedly, because the studies showed clinical heterogeneity, we performed only a few quantitative syntheses.
The following recommendations may be considered in future studies. To evaluate the effectiveness of HM in PCS, participants should be enrolled using standardized diagnostic criteria, such as the international statistical classification of diseases and related health problems or the diagnostic and statistical manual of mental disorders. In addition, the multicompound, multi-target nature of HM may improve a wide range of symptoms after TBI, such as PCS; therefore, the underlying molecular mechanism of HM should be studied. Particularly, priority should be given to HM and/or herb, which are especially known for ameliorating blood stasis, in further HM researches on TBI. To optimize the use of HM during treatment of TBI and to resolve the clinical heterogeneity, future studies should characterize the participants in detail, with particular focus on TBI severity and target symptoms after TBI, such as headache, mental disorder, and cognitive dysfunction, and on the objectives of HM, such as acute management or rehabilitation. In PCS, validated disease specific tools should be adopted to evaluate the effect of HM on various symptoms and deficits; these may include the Rivermead Postconcussion Symptoms Questionnaire, the World Health Organization Disability Assessment Schedule 2.0, and the British Columbia Post-concussion Symptom Inventory-Short Form (90). Finally, only three of the retrieved studies compared HM with a placebo and these showed marked clinical heterogeneity, and thus, we could not draw a definite conclusion about the efficacy of HM. Blinding of participants and personnel using placebo with the same taste, flavor, and formulation should be conducted to avoid performance bias. In future, rigorously conducted, placebo-controlled trials to evaluate the efficacy of HM in TBI should be performed considering the above implications.

CONCLUSION
The current evidence suggests that there is insufficient evidence for recommending HM for TBI in clinical practice. Although some RCTs reported that HM as an adjuvant therapy to CT may have benefits for some functional outcomes of TBI, the low quality of evidence significantly limited its reliability. Therefore, further rigorous, well-designed, high quality, placebo-controlled RCTs should be conducted to confirm these results.