The Effect of Different Acupuncture Therapies on Neurological Recovery in Spinal Cord Injury: A Systematic Review and Network Meta-Analysis of Randomized Controlled Trials

Many acupuncture therapies were used to treat spinal cord injury (SCI) and its complications. The difference in efficacy among these therapies has not been assessed. To compare the efficacy of different acupuncture therapies for SCI, we searched databases (PubMed, Embase, Cochrane Library, CNKI, and WanFang) for relevant RCTs in both English and Chinese before June 2019 that reported the association between acupuncture therapies and SCI. The RCTs were categorized according to the location of the acupoints used in them. The neural function was assessed by American Spinal Injury Association (ASIA) motor score, and daily living ability was accessed by Modified Barthel Index (MBI) after SCI. In total, 22 trials involving 1644 participants were included. The pairwise meta-analysis and random effects model network meta-analysis were conducted. The results indicated that exercise combined with electro-acupuncture (EA) is superior to exercise without acupuncture in improving the ASIA motor score. EA was associated with a significantly higher improvement in the MBI score than exercise alone, except for EA of head + limbs and limbs. Additionally, EA on the head + back and back + front (chest and abdomen) rank the top in both increasing the ASIA motor score and the MBI score. Acupuncture can significantly increase motor function and daily living ability of individuals who suffer from SCI, especially acupuncture of the back + front or the head + back. The evidence supports acupuncture of the back + front or the head + back as an effective treatment for SCI.


Introduction
SCI and its secondary complications have always been a great social and economic burden for the patients' families and health care system. It is estimated that the annual occurrence of SCI is roughly 17,000 new cases each year or approximately 54 cases per million population in the US. Take patient with high tetraplegia (C1-C4), for example, the average yearly expenses (heath care and living expenses) cost an average of 1,065,980 dollars in the first year and 185,111 dollars in the each subsequent year which did not even include any indirect costs such as losses in wages, fringe benefits, and productivity which cost average 72,047 dollars per year in 2015 in the US [1].
People with SCI almost always show some degree of motor and sensory function recovery below the injury level, but spontaneous recovery in patients with complete SCI is quite limited. [2] Although there have been reports of positive outcomes following treatment options such as surgery [3], pharmacological interventions [4], rehabilitation [5], and alternative methods such as acupuncture [6][7][8][9], reviews by a number of researchers concluded that use of acupuncture or EA in acute SCI can significantly improve functional recovery in long-term and help the management of chronic pain associated with these injuries, with a risk that could be considered minimum [6,7,[10][11][12].
At present, a standard acupuncture therapy for people with SCI is not yet available. So, the question remains that which one of these existing therapies is most effective? us, a comparison of efficacy of different acupuncture therapies is necessary.
However, a meta-analysis of direct comparisons between different acupuncture therapies is unavailable because such analyses are limited by comparators and insufficient studies which probably due to the fact that acupuncture is not yet widely used throughout the world.
Network meta-analysis overcomes this limitation by creating indirect comparisons and allowing data synthesis, which could help identify the most effective interventions. erefore, we performed this Bayesian network metaanalysis to compare efficacy of the different acupuncture therapies, which include both direct and indirect comparisons of the changes in both the ASIA motor score and the MBI score of these methods.

Methods
is study was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Network Meta-Analysis (PRISMA-NMA) checklist [13].

Study Selection. Two authors (Feng Xiong and Chenying
Fu) independently identified the eligible studies, firstly by titles and abstracts, and then full texts were scanned. Any disagreements regarding study inclusion were resolved by discussion with a third author (Lihong Peng or Zejun Liang).

Eligibility Criteria.
Randomized controlled trials that met following criteria were included: (1) participants diagnosed with SCI; (2) trials comparing different acupuncture/EA therapies, or comparing acupuncture/EA to regular rehabilitation training or drugs; (3) trials reporting one of following outcomes: ASIA motor score and MBI score. e inclusion of studies was limited to RCTs reporting the prospective use of the acupuncture interventions in study participants who suffered from spinal cord injury (SCI). All trials selected were those that focused on functional recovery, while those that only reported complications such as bladder dysfunction or pain were excluded. is review included parallel-group RCTs that assessed the efficacy of acupuncture regardless of blinding or the type of control comparison.
Our study mainly considered needle acupuncture as the intervention of interest which included manual acupuncture, electro-acupuncture, auricular acupuncture, and acupoint injection. Laser acupuncture, acupressure, and moxibustion were excluded. Studies on complications, prospective nonrandomized studies, duplicate publications, or studies without extractable data for analysis were also excluded.

Data Sources and Searches.
We performed searches of the following databases: PubMed, Embase, Cochrane Library, CNKI, and WanFang database. e search keywords included SCI patients, acupuncture, EA, and randomized controlled clinical trials. e following keywords were used in combination with both MeSH terms and text words: SCI, acupuncture, EA, and randomized controlled trials. No restriction was placed on publication status or language. In addition, to guarantee the saturation of literature, we also scanned relevant trials included by previous systematic reviews and meta-analyses for publication as well. e study selection process is shown in Figure 1.

Data Collection Process.
Four authors extracted and assessed the data (data extraction: Feng Xiong and Chenying Fu; assessment: Zejun Liang and Lihong Peng) using a standard spread sheet (performed by Excel; Version 2010) consisting of four sheets, which included (1) general information (i.e., study design, intervention types, and information about data extractor); (2) study characteristic (i.e., patients, intervention, frequency, duration, follow-up, and outcomes); (3) a risk of bias assessment (ROB: randomization, allocation, blinding of participants and personnel, blinding of outcome assessor, incomplete data, and selective reporting); and (4) a summary of outcome data (ASIA motor score and MBI score).

Categorizing Strategy.
Acupuncturists use different combinations of acupoints to treat different conditions, according to the function of each meridian and acupoint. Since up to 20 meridians are used in acupuncture (acupoints used in ear acupuncture not included), theoretically there can be 1,048,576 kinds of meridian combinations. If we judge the therapies in combinations of acupoints, the number would only be greater. It would be impossible to compare all those different combinations this way. So, we settled for comparing combinations of four main regions. e meridians in the same region are thought to be interconnected and functionally related. As for the scalp and ear acupuncture, it would be ideal to classify them as different therapies, but due to limited number of RCTs, they were classified into the same group, the head acupuncture. e regions include the back (governor meridian, Jiaji acupoints, bladder meridian, and Yangming meridian), the front (Ren channel, kidney meridian, and stomach meridian), the head (scalp acupoints, ear acupoints), and the limb (three yin meridians of hand, three yin meridians of foot, three yang meridians of hand, and three yang meridians of foot). In this way, each combination would have more RCTs to compare with. e included studies each used one or a combination of these regions for acupuncture in treating SCI patients and many produced positive outcomes.
For example, some studies may only use acupoints from the back and the head (back + head), and some others may only use acupoints from the back and the limb (back + limb).

Statistical Analysis.
e pairwise meta-analysis was performed to synthesize studies that compared the same interventions with random effect models (direct comparison) using the R software (version 3.5.1; http://www.rproject.org). en, to determine comparative effectiveness, a random effects model network meta-analysis (a combination of direct and indirect comparison) was conducted in a Bayesian framework by introducing the WinBUGS package and the Jags package (http://www.r-project.org). e direct and indirect comparisons for each specific pair of treatments were combined by modeling the outcomes in every treatment group of included studies. e ASIA motor and MBI score were reported as a standard mean difference (SMD) with 95% confidence intervals (CI) for direct comparisons or 95% credible intervals (CrI) for indirect comparisons. e Brooks-Gelman-Rubin method was then used to assess the convergence between direct and indirect variances. To show the relationship between different acupuncture treatments, we used network plots to show direct comparisons between arms (see Figure 2). e effectiveness of each treatment among all existing treatments was ranked in order by calculating the SMD; probability density functions were generated. e I 2 test was conducted to quantitatively assess heterogeneity, and an I 2 value greater than 75% was considered significant.  Evidence-Based Complementary and Alternative Medicine e clinical, statistical, and methodological differences may attribute to heterogeneity. e statistical codes were shown in Supplement 1 (available here).

Study Selection.
e search was performed and updated up to date (June 5th, 2019), and 191 references were identified. After duplicate studies were removed, other 113 records were excluded by reading the titles and abstracts, and the full texts of 78 articles were then assessed for eligibility. A total of 22 studies were selected covering 8 groups which include exercise, 7 different combinations of exercise, and acupuncture of different regions (the back, the limbs, the head, and the front).

Sensitivity Analysis and Network
Assumption. e heterogeneity in the pairwise meta-analysis was high in five comparisons: back vs. back + front, back vs. head + back, exercise vs. back + limb + front, exercise vs. head + limb, limbs vs. head + limbs; so, we chose the random effect model to cancel the statistical heterogeneity. As for clinical heterogeneity, meta regression was not done due to limited number of RCTs included in this article.
In total, 3 loops were in the network meta-analysis of the ASIA motor score, and 1 loop was in that of the MBI score. e pairwise meta-analysis and the network meta-analysis results did not significantly differ.

Discussion
e purpose of this network meta-analysis is to determine the effectiveness of various acupuncture therapies for SCI in improving motor function and daily living ability. Since rehabilitation training and neurotrophic drugs are combined for clinical routine treatment of SCI, routine rehabilitation training and medication are also included and analyzed as routine treatment. An assessment of the ROB indicates that most of the included studies were medium ROB. Preliminary analysis showed that except for head + limbs and limbs therapy, all other therapies improved patients' muscle strength and daily living ability comparing with conventional rehabilitation training (and drugs). e main results of the ranking chart showed that the acupuncture on the back + front, the back + head is most effective in improving muscle strength, followed by   [17] L H U L U Guo X. J. [23] L U U L L Guan W. [21] L H L L U Jiang H. Y. [30] L H L L U Wu K. X. [27] U H U L L Ou Y. P. [32] L U U L U Li X. N. [25] L U U L L Deng N. [20] L H L L U Deng C. [14] L U U L U Qian Y. [15] L U U L U Sun S. B. [16] U H U L U Huo H. X. [24] U H L L L Xiao H. [34] U H L L L Xie J. D. [33] U U U L L Guo Y. H. [29] L H U L L Liang T. Y. [31] L H L L L Du Y. P. [28] L U L L L Chen D. [19] L U U U L Guo J. [22] L H U U L Zhao L. S. [18] L U U U L Wong A. M. [12] L L H L L L: low risk of bias; H: high risk of bias; U: unclear. Risk of bias assessment. 8 Evidence-Based Complementary and Alternative Medicine the back and back + limbs, head + limbs, limbs, back + front + limbs, and simple rehabilitation training (and drugs); while the back + front is most effective in improving the ability of daily life, followed by the back + head, back + limbs, back + front + limbs, back, rehabilitation training, limbs, and head + limbs. In addition, ancient mystical clinical devices or means may have a stronger placebo effect than drugs; therefore, acupuncture could be associated with greater placebo effects than drugs or rehabilitation. In terms of safety analysis, no serious adverse events related to acupuncture were reported in the selected studies, which may be associated with an overall low rate of serious AE in acupuncture. Longer time treatments appear to be more effective than short-term treatments. ese results can be used as a reference for the efficacy and safety of clinical acupuncture in the treatment of SCI. In short, the back + front, and the back + head strategy are the most effective ones. e best strategies would involve the acupoints in the back and the head is hardly a surprise, but what does the acupoints in the front have to do with the spinal cord, many would ask that question. e answer may lie in the way that acupuncturist sees the body. ey see the body like a computer with keyboards or many buttons. Most of the time this computer runs on itself, but occasionally some error appears and it needs to be fixed. e acupuncturist knows where to push, and they use a needle to push it. ose buttons were called acupoints. We do not know who designed the computer or the keyboard, but some times the button that needs to be pushed lies not in the proximity of the illness or injury site, and may not even be linked to the injury site by the nerve system or the circulating system or any other systems, except by the meridians. ere is currently not a satisfying explanation in modern science concerning this phenomenon. And, in the case of spinal cord injury, some of the buttons that needs to be pushed may lie in areas that is far away from the injured spinal cord. A skilled acupuncture finds such buttons by following the ancient ways. As for the deeper reasons for how and why these buttons work, may just be the real lesson that the modern medicine has to learn from Chinese traditional medicine.
Acupuncture or EA has not been used as a routine treatment for SCI, except in several Asian countries and regions. Acupuncture or EA can regulate immune function, such as cholinergic anti-inflammatory pathways, and participate in the inhibition of inflammatory response, while EA may promote nervous system regeneration by stimulating the nervous system. Most of the existing literature focused on acupuncture treatment of SCI complications such as neurogenic bladder, while less attention is paid to motor function or daily living ability. In contrast, acupuncture or EA treatment of SCI has a variety of specific methods, which only make it much more difficult to choose among various treatment methods.
To the authors' knowledge, all previous meta-analyses on acupuncture treatment of SCI laid emphasis on the effectiveness of acupuncture [6,9,35], but none recognized that acupuncture itself contains many different intervention strategies. is article analyzed the efficacy of acupuncture performed on different body parts to treat the same condition, SCI. And, judging from the results of this article, there are differences between these therapies.
Acupuncture and its derivative therapies mainly include needle acupuncture, moxibustion, EA, needle implantation, and even meridian massage. Even only within the category of needle acupuncture, there are differences such as the Admittedly, the categorizing strategy in this article is not the ideal way, but it is the first attempt to investigate the difference between many acupuncture therapies. Due to the limited number of relevant studies, this analysis has not been able to further classify and evaluate the selection of acupoints and meridians and their efficacy. We hope the number of related studies would increase in the future, and allow a more accurate classification and analysis of these acupuncture therapies. e strengths of this study are as follows. We used Bayesian frameworks to compare various acupuncture therapies and rehabilitation (and medicine), and the results showed that acupuncture may be an effective and safe treatment, and there are differences between existing acupuncture therapies for SCI. And, its ranking chart may be useful for clinical practice and further research. In terms of data processing, we used scores at the end of treatment to arrive at clinically valuable conclusions.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
e authors declare that they have no conflicts of interest.