Herbal Medicine for Cervicogenic Dizziness: A Systematic Review and Meta- Analysis


 Background: Herbal medicines (HMs) have been widely used in the treatment of cervicogenic dizziness (CGD) based on their empirical effectiveness and safety. Herein, we reviewed and evaluated the clinical evidence on the efficacy and safety of HM for CGD. Methods: Among the relevant studies published up to December 2019 in 11 electronic databases, only randomised controlled trials (RCTs) were included. The studies’ methodological quality was assessed using the revised Cochrane risk-of-bias tool for randomised trials, and the strength of evidence for the main findings was evaluated using the Grading of Recommendations Assessment, Development, and Evaluation approach. Results: All 17 included RCTs with 1,797 participants were conducted with six types of modified HM prescriptions and three types of active controls. More than half of the included studies were of low quality because of the high risk of bias due to deviations from the intended intervention. HMs plus active controls were more effective in CGD treatment than active controls alone. HMs plus antivertigo drugs, HMs plus manual therapy, and HMs plus acupuncture therapy were all effective in CGD treatment, with HMs plus antivertigo drugs showing the most reliable effect. All HM prescriptions were effective for specific patterns of CGD when administered with active controls, with Banxia Baizhu Tianma tang and Dingxuan tang demonstrating the most reliable effect. No serious adverse events were reported in all included studies. Conclusions: The current evidence suggests that HMs may enhance the treatment effect on CGD when combined with other treatments without serious adverse events. Further high-quality evidence is needed to draw a definite conclusion.Systematic review registration: PROSPERO (registration number: CRD42020199222), and the Research Registry (Review Registry Unique Identifying Number: reviewregistry1036)

names. Dichotomous outcomes were pooled using a risk ratio (RR), and continuous outcomes were pooled using a mean difference (MD) or standardised mean difference (SMD) with 95% con dence intervals (CIs).
Statistical heterogeneity among studies was assessed by computing I 2 statistics. Data were pooled using a random-effects model if the included studies had signi cant heterogeneity (I 2 values ≥ 50% indicated substantial heterogeneity and those ≥ 75% indicated considerable heterogeneity; both were considered as signi cant). Otherwise, a xed-effects model was applied (31). Sensitivity analysis was performed to increase the robustness of the results by excluding studies with high risks of bias and outliers. If the number of studies was su cient (n ≥ 10), visual inspection of the funnel plot was performed to assess for publication bias. Data on the safety of HMs for CGD were described qualitatively.

Study selection
A total of 7,735 studies were identi ed through the database searches, and one additional study was identi ed through other sources. After removing 305 duplicates, 6,765 studies were excluded by screening the titles and abstracts. Through review of the full-texts, 649 studies were further excluded: 17 studies with unavailable full-texts, 31 non-clinical studies, 21 case reports, 164 comparative studies, 5 nonrandomized controlled trials, 258 studies not related to CGD, 48 studies not related to eligible intervention, and 105 studies not related to the clinical question. Finally, 17 RCTs with 1,797 participants were included in the analysis (Fig. 1).

Study characteristics
All included studies were RCTs conducted in China. They were classi ed according to the comparison types as follows: 1) studies comparing HMs plus antivertigo drugs with antivertigo drugs (n = 6), subdivided into studies prescribing unarizine (n = 3) and those prescribing betahistine (n = 3); 2) studies comparing HMs plus manual therapy with manual therapy alone (n = 5); and 3) studies comparing HMs plus acupuncture therapy with acupuncture therapy alone (n = 6). There was no study which assessed the e cacy of HM as a monotherapy for CGD.
In addition, 10 types of outcome measurements were identi ed: two studies evaluated overall function scores, 11 studies evaluated simple scores, seven studies assessed the average blood ow velocity in the vertebral arteries, eight studies assessed the average blood ow velocity in the basilar artery, 16 studies evaluated the total effective rate, two studies measured the ET and CGRP levels, and three studies measured the Fib and TC levels. The incidence of adverse events was reported in only one study. The study characteristics and the main results are summarised in Table 1.

Risk-of-bias assessment
For the bias arising from the randomisation process, seven studies were evaluated as 'low-risk' because the randomisation process for the allocation sequence was clearly described. The remaining 10 studies were determined as 'some concerns' because the relevant information was provided insu ciently. For the bias due to deviations from intended interventions, 11 studies, most of which were studies with manual or acupuncture therapy as active controls, were judged as 'high-risk' because it was unclear whether blinding of participants and trial personnel had been su ciently performed by using sham-massage or sham-acupuncture. The remaining six studies were determined as 'some concerns'. For the bias due to missing outcome data, 13 studies were evaluated as 'low-risk' and one study was judged as 'high-risk' because there was missing data, but only the result of per protocol analysis was reported. The remaining three studies were determined as 'some concerns' because of insu cient provision of the relevant information. For the bias in the measurement of the outcome, eight studies were evaluated as 'low-risk' and the remaining nine studies were determined as 'some concerns' because it was di cult to judge whether outcome indicators used in the studies were affected by the awareness of outcome assessors. For the bias in the selection of the reported results, all studies were evaluated as 'some concerns' because there was no basis for judging, such as study protocols. Finally, for the overall risk of bias, 11 studies were evaluated as low-quality studies because they were assessed as 'high-risk', and the remaining six studies were evaluated 'some concerns' (Fig. 2). Summarising the results of the subanalysis according to HM prescription names, BBTT, DXT, and YCT showed signi cant treatment effects in various primary and secondary outcomes, and had relatively more clinical evidence compared to the remaining HM prescriptions. Each of BYT, GGT, and GJT were investigated in only one clinical RCT. In the GGT study, one primary outcome (change in the simple scores) and one secondary outcome (total effective rate) were statistically signi cant. However, in the BYT and GJT studies, only one primary outcome (change in the simple scores) was signi cant, whereas the other outcomes were not statistically signi cant. The results of the total analysis and the subanalyses on the e cacy of HMs are shown in Table 3. If the evidence of more than two studies showed MD < 4 for the change in the blood ow velocity in the vertebrobasilar artery or RR > 2 for the total effective rate, it was considered that there was a strong association for a treatment effect. If the evidence of more than two studies showed MD < 4 for the change in the blood ow velocity in the vertebrobasilar artery or RR > 2 for the total effective rate, it was considered that there was a strong association for a treatment effect. If the evidence of more than two studies showed MD < 4 for the change in the blood ow velocity in the vertebrobasilar artery or RR > 2 for the total effective rate, it was considered that there was a strong association for a treatment effect. If the evidence of more than two studies showed MD < 4 for the change in the blood ow velocity in the vertebrobasilar artery or RR > 2 for the total effective rate, it was considered that there was a strong association for a treatment effect. If the evidence of more than two studies showed MD < 4 for the change in the blood ow velocity in the vertebrobasilar artery or RR > 2 for the total effective rate, it was considered that there was a strong association for a treatment effect. If the evidence of more than two studies showed MD < 4 for the change in the blood ow velocity in the vertebrobasilar artery or RR > 2 for the total effective rate, it was considered that there was a strong association for a treatment effect.  If the evidence of more than two studies showed MD < 4 for the change in the blood ow velocity in the vertebrobasilar artery or RR > 2 for the total effective rate, it was considered that there was a strong association for a treatment effect. If the evidence of more than two studies showed MD < 4 for the change in the blood ow velocity in the vertebrobasilar artery or RR > 2 for the total effective rate, it was considered that there was a strong association for a treatment effect.  If the evidence of more than two studies showed MD < 4 for the change in the blood ow velocity in the vertebrobasilar artery or RR > 2 for the total effective rate, it was considered that there was a strong association for a treatment effect. If the evidence of more than two studies showed MD < 4 for the change in the blood ow velocity in the vertebrobasilar artery or RR > 2 for the total effective rate, it was considered that there was a strong association for a treatment effect.

Safety
Of the 17 studies, only one study reported an adverse event; one case of gastrointestinal discomfort, evaluated as not serious, was reported in the BBTT plus manual therapy group ( Table 1).

Quality of evidence
In the comparison of HMs plus active controls with active controls alone, the quality of evidence for the primary outcomes ranged from 'very low' to 'moderate'. For the secondary outcomes, the quality of evidence for the total effective rate was graded as 'moderate', but that for the other outcomes was graded as 'very low'. In other words, the overall quality of evidence in the total analysis was rated 'low'.
In the comparison of HMs plus antivertigo drugs with antivertigo drugs alone, the quality of evidence for all outcomes, except for the simple scores, was graded as 'moderate'. In the comparison of HMs plus manual therapy with manual therapy alone, the quality of evidence for most outcomes was graded as 'moderate'. but that for the simple scores and the blood ow velocity in the basilar artery was graded as 'very low'. In the comparison of HMs plus acupuncture therapy with acupuncture therapy alone, the quality of evidence for the overall functional scores and the total effective rate was graded as 'moderate', but that for the other outcomes ranged from 'very low' to 'low'. In summary, in the subanalysis according to the comparison types, the overall quality of evidence was higher, but decreasing in the order of HMs plus antivertigo drugs, HMs plus manual therapy, and HMs plus acupuncture therapy.
In the comparison of BBTT plus active controls with active controls alone, the quality of evidence for most outcomes, except for the blood ow velocity in the right vertebral and basilar arteries, was graded as 'moderate'. Similarly, in the comparison of DXT plus active controls with active controls alone, the quality of evidence for most outcomes, except for the simple scores and the blood ow velocity in the basilar artery, was graded as 'moderate'. In the comparison of GGT plus active controls with active controls alone, the quality of evidence for the simple scores and the total effective rate was also graded as 'moderate'. In the comparison of GJT plus active controls with active controls alone, the quality of evidence for the simple scores was graded as 'moderate', but that for the total effective rate was graded as 'low'. In the comparison of YCT plus active controls with active controls alone, the quality of evidence for outcomes ranged from 'very low' to 'high'. Finally, in the comparison of BYT plus active controls with active controls alone, the quality of evidence for the outcomes ranged from 'very low' to 'moderate'. In summary, in the subanalysis according to HM prescription names, the overall quality of evidence was highest for BBTT, DXT, and GGT, and lowest for BYT. The main reason for the downgrade was the high risk of bias for the included studies, and the inconsistency of the results due to high heterogeneity between them (Table 3).

Publication bias
For two outcomes included in more than 10 studies, we checked for publication bias through funnel plot analysis. In the comparison of HMs plus active controls with active controls alone, the overall functional scores showed asymmetry, suggesting the existence of potential publication bias (Fig. 3); thus, there may be negative results not published in the literature. Conversely, the funnel plot for the total effective rate was symmetrical (Fig. 4).

Discussion
In this study, we reviewed and evaluated the available clinical evidence on the e cacy and safety of HMs as monotherapy or adjunctive therapy for the treatment of CGD to promote evidence-based decision-making in clinical practice. Since none of the included 17 RCTs (32-48) could assess the e cacy of HMs for CGD as monotherapy, we evaluated only their e cacy as adjunctive therapy used with other active controls. The included studies were conducted with six types of modi ed HM prescriptions and three types of active controls. In the risk-of-bias assessment, more than half of the included studies were evaluated to be of low quality because of the high risk of bias due to deviations from the intended intervention. The results of the e cacy analyses for HMs plus active controls indicated the following. First, HMs plus active controls were more effective in CGD treatment than active controls alone (administration duration varied from 10 days to 8 weeks). Second, HMs plus antivertigo drugs ( unarizine and betahistine), HMs plus manual therapy, and HMs plus acupuncture therapy were all effective in CGD treatment. Among all, HMs plus antivertigo drugs showed the most reliable effect. Third, BBTT, BYT, DXT, GGT, GJT, and YCT were effective for speci c patterns in patients with CGD, when administered with active controls. Among the HM prescriptions, BBTT and DXT had the most reliable effect when combined with active controls. Regarding the safety of HMs plus active controls in the treatment of CGD, no serious adverse events were reported in all included studies.
In traditional Chinese medicine, HMs are prescribed to match the speci c pattern of patients' signs and symptoms. It is reasonable to select and prescribe the most appropriate HM for a speci c pattern in each patient with CGD, as opposed to consistently prescribing one HM prescription to all patients with CGD, even if it is statistically the most evidenced prescription for CGD. Thus, although BBTT and DXT had the highest level of clinical evidence for the treatment effect on CGD in this review, it may be more effective to use other HM prescriptions for certain patterns in some patients with CGD. BBTT, which has the effect of dispelling pathogenic wind and eliminating phlegm, is used for the wind-phlegm type. Several studies have reported quantitative clinical evidence for the use of BBTT with antivertigo drugs (38,43,47). DXT has the effect of removing a pathogenic mass as the original prescription, and it can be prescribed for both de ciency and excess syndromes by modi cation of the original prescription. HMs can be modi ed for better e cacy and fewer side effects (49). In cases of combined excess and de ciency syndromes, such as spleen de ciency and dampness type, or qi de ciency and blood stasis type, DXT was modi ed by addition of herbs that have effects on invigorating the qi and spleen (Codonopsis Pilosulae Radix, and Atractylodis Rhizoma Alba), regulating qi-owing (Citri Reticulatae Pericarpium), enriching the blood (Angelicae Gigantis Radix), and soothing the nerves (Fossilia Ossis Mastodi), but with subtraction of other herbs from the original prescription, which have effects on suppressing hyperactive liver for calming endogenous wind (Uncariae Ramulus cum Uncus, and Scorpio) and promoting blood circulation with removing blood stasis (Salviae Miltiorrhizae Radix) (34,45). On the other hand, in cases of only excess syndrome, such as hyperactivity of liver yang type, DXT was modi ed by adding Puerariae Radix, which has the effect of dispelling wind-heat (32,48). For the combination of DXT and other treatments, quantitative clinical evidence has been reported for the use of DXT with manual therapy (32,45,48). Both YCT (33,37,(39)(40)(41) and BYT (36) are used for the qi and blood de ciency type. However, YCT has a higher quality clinical evidence than BYT, because BYT showed low precision for outcomes. For the combination of YCT and other treatments, the most quantitatively reported clinical evidence was for the use of YCT with acupuncture therapy (37,(39)(40)(41). Additionally, GGT was used with manual therapy for the wind type with disharmony between ying and wei (35), and GJT was used for the collateral stasis type with betahistine (44). Through this review, we obtained a clue about the corresponding relation of patterns in patients with CGD and HM prescriptions; however, it remains unknown which HM prescription is the most effective for a certain pattern in these patients, because all included studies used only one HM prescription with one pattern of patients with CGD. Further studies are needed to con rm which HM prescription is the most effective in a certain pattern in these patients.
In this review, we identi ed Fib, ET, TC, and CGRP as haematological indicators used in clinical studies on CGD. ET and CGRP were used as indicators to verify the e cacy of HMs plus manual therapy compared with manual therapy alone. Fib and TC were used to verify the e cacy of HMs plus acupuncture therapy compared with acupuncture therapy alone. ET is an endogenous vasoconstrictor that reduces the perfusion of brain tissues by constricting the blood vessels in the brain (50,51). CGRP is a vasodilator mainly distributed in the central nervous system (52). In a previous study, ET and CGRP were reported as important factors affecting the development of CGD with vertebrobasilar arteriospasm (53). Furthermore, Fib is able to promote the formation of artherosclerotic plaques (54), and TC is able to accelerate atherosclerosis and cause lipid metabolism disorders (55). In summary, control of the ET and CGRP levels improves the prognosis of CGD, and evaluation of the Fib and TC levels help predict CGD progression (38). Therefore, it can be recommended to use them as outcomes when conducting further clinical trials for CGD.
This review has some limitations. First, the quality of the included RCTs was generally poor, in particular, because of the high risk of bias due to deviations from the intended intervention. Second, most meta-analyses showed high heterogeneity among studies. Third, potential publication bias could not be ruled out, because the assessment of publication bias was not conducted in the meta-analyses in which the number of included studies was less than 10, and all RCTs were conducted in China and published in Chinese. Fourth, there is a possibility of attrition bias, because few studies presented dropout or withdrawal statistics. Fifth, it was unknown whether the treatment effect of HMs plus active controls maintained after completion of the intervention, because most studies did not perform follow-up assessments. Finally, the safety of HMs for CGD is still unclear, because few studies had clearly reported that there were no adverse events.
However, this research is valuable as the rst systematic review that comprehensively evaluated the e cacy and safety of HMs for treating CGD and will aid clinicians to select and prescribe HMs suitable for speci c patterns of patients with CGD based on evidence-based decision-making. Furthermore, it provides knowledge of which treatments will be effective in combination with HMs. Thereby, this research may contribute for the development of effective strategies in the treatment and management of the increasing number of patients with CGD due to population ageing. Nonetheless, further high-quality evidence from rigorously conducted clinical studies, preferably conducted outside of China, is required to support the clinical recommendations regarding the use of HMs for CGD. In addition, placebo-controlled RCTs are needed to evaluate the e cacy of HMs for CGD as monotherapy. Furthermore, experimental studies on the mechanism of action and the dose-response relationship of HMs are necessary to determine the optimal dosage.

Conclusions
Current evidence suggests that HMs may have the potential to enhance the treatment effect on CGD when combined with other treatments without serious adverse events. Since the overall quality of the studies included in this review was generally low, additional high-quality evidence is needed to draw a de nite conclusion. Risk of bias summary for all included studies. The risk of bias was evaluated as "low," "high," or "some concerns", represented with the following symbols, respectively: "+," "-," and "?". Abbreviations: D, bias due to deviations from intended interventions; Me, bias in measurement of the outcome; Mi, bias due to missing outcome data; O, overall risk of bias; R, bias arising from the randomisation process; S, bias in the selection of the reported result