Effectiveness of Acupuncture and Electroacupuncture for Chronic Neck Pain: A Systematic Review and Meta-Analysis

Transcription

1 The American Journal of Chinese Medicine, Vol. 45, No. 8, World Scientific Publishing Company Institute for Advanced Research in Asian Science and Medicine DOI: /S X Effectiveness of Acupuncture and Electroacupuncture for Chronic Neck Pain: A Systematic Review and Meta-Analysis See Yoon Seo,* Ki-Beom Lee, Joon-Shik Shin, Jinho Lee, Me-Riong Kim, In-Hyuk Ha, Youme Ko and Yoon Jae Lee *Department of Korean Medicine, College of Korean Medicine Sangji University, Wonju, Republic of Korea Jaseng Spine and Joint Research Institute Jaseng Medical Foundation, Seoul, Republic of Korea Department of Korean Preventive Medicine Graduate School, Kyung Hee University Seoul, Republic of Korea Published 10 November 2017 Abstract: The aim of this systematic review was to assess evidence from randomized controlled trials (RCTs) on the effectiveness and safety of acupuncture and electroacupuncture in patients with chronic neck pain. We searched nine databases including Chinese, Japanese and Korean databases through 30 July The participants were adults with chronic neck pain and were treated with acupuncture or electroacupuncture. Eligible trials were those with intervention groups receiving acupuncture and electroacupuncture with or without active control, and control groups receiving other conventional treatments such as physical therapy or medication. Outcomes included pain intensity, disability, quality of life (QoL) and adverse effects. For statistical pooling, the standardized mean difference (SMD) and its 95% confidence interval (CI) were calculated using a fixed-effects model. Sixteen RCTs were selected. The comparison of the sole acupuncture group and the active control group did not come out with a significant difference in pain (SMD 0.24, 95% CI ), disability (SMD 0.51, 95% CI ), or QoL (SMD 0.37, 95% CI ), showing a similar effectiveness of acupuncture with active control. When acupuncture was added into the control group, the acupuncture add-on group showed significantly higher relief of pain in studies with unclear allocation concealment (SMD 1.78, 95% CI ), but did not show significant relief of pain in studies with good allocation concealment (SMD 0.07, 95% CI ). Significant relief of pain was observed when the sole electroacupuncture group Correspondence to: Dr. Yoon Jae Lee, Jaseng Spine and Joint Research Institute, Jaseng Medical Foundation, 858 Eonju-ro, Gangnam-gu, Seoul, Republic of Korea. Tel: (þ82) , Fax: (þ82) , goodsmile8119@gmail.com 1

2 2 S.Y. SEO et al. was compared to the control group or electroacupuncture was added onto the active control group, but a lot of the results were evaluated to have low level of evidence, making it difficult to draw clear conclusions. In the result reporting adverse effects, no serious outcome of adverse event was confirmed. Acupuncture and conventional medicine for chronic neck pain have similar effectiveness on pain and disability when compared solely between the two of them. When acupuncture was added onto conventional treatment it relieved pain better, and electroacupuncture relieved pain even more. It is difficult to draw conclusion because the included studies have a high risk of bias and imprecision. Therefore better designed largescale studies are needed in the future. Keywords: Effectiveness; Safety; Acupuncture; Electroacupuncture; Systematic Review; Meta-Analysis; Review. Introduction Neck pain is one of the three major musculoskeletal diseases (Hogg-Johnson et al., 2008), causing considerable discomfort and economic loss (Hoy et al., 2010). Most people experience neck pain at least once in a lifetime (Haldeman et al., 2010), and the lifetime prevalence reaches 26% 71% (Hogg-Johnson et al., 2008; Peloso et al., 2007). The symptom can be experienced incidentally throughout a lifetime along with back pain (Bot et al., 2005; Guzman et al., 2008). The definition of neck pain varies with different studies, and covers pain of a wide range including the neck, shoulders, as well as the upper chest (Bot et al., 2005). The natural history of neck pain is unclear, and most patients complaining of neck pain are not known to have a particular cause such as tumor, infection, fracture, or inflammation (Dabbs and Lauretti, 1995). Although neck pain is prone to chronicization and the definition of chronic pain differs by study, the Cochrane back review group defines chronic neck pain as neck pain of 12 weeks or longer duration (Furlan et al., 2009). The general treatment of neck pain includes oral medication, injection, physical therapy, etc. The first-line of oral medication is NSAIDS, conventionally used for musculoskeletal diseases. According to Dabbs et al., the use of NSAIDs is reported to have severe adverse effects such as gastritis, ulcer, myocardial infarction, etc. (Somerville et al., 1986; Graham et al., 1988; Dabbs and Lauretti, 1995; Peloso et al., 2007; White et al., 2011). Injection therapy includes trigger point injections, corticosteroid injections for anti-inflammation, and muscle relaxants. For trigger point injections and muscle relaxants there is a lack of evidence for concrete effects (Peloso et al., 2007; Scott et al., 2009), and corticosteroid injections are reported to have complications such as spinal cord, nerve injury, infection, epidural hemorrhage, blood vessel penetration, subarachnoid penetration, etc. (Manchikanti et al., 2015). In Gross et al. s systematic review of physical therapy, multiple cervical manipulation sessions are reported to show a better effect on short-term and long-term pain reduction and functional improvement, but to have adverse events (Peloso et al., 2007; Gross et al., 2015). Higher quality studies are warranted. Therefore, many people, including patients who do not improve with existing medication treatment, are interested in complementary and alternative medicine (CAM).

3 SYSTEMATIC REVIEW FOR ACUPUNCTURE EFFECTS ON CHRONIC NECK PAIN 3 According to Burke et al. (2006) 8.19 million Americans are reported to have used acupuncture treatment. According to a 2014 study done in Korea the major treatment modalities used in Korean medicine were acupuncture (59.2%), herbal medicine in decoction form (27.6%), herbal medicine in powder form (4.9%), and physical therapy (4.6%) in order, showing that acupuncture takes up more than half of the total Korean medicine treatment, especially in the treatment of musculoskeletal diseases. (Korea Health Industry Development Institute, 2014). Acupuncture is one of the main treatment modalities of traditional Chinese medicine (TCM) and has been used for more than 3000 years in the East Asian region along with herbal medicine and tuina therapy. It is known to activate the endogenous pain control systems (Coutaux, 2017). However, because it is impossible to perform double-blinding studies and set up appropriate control, an effective evaluation of acupuncture is not easy (Coutaux, 2017), and well-designed studies that can validate the clinical efficacy of acupuncture are lacking. For clinicians, comparison studies using active controls generally used in clinics may be more meaningful than inactive control studies done with sham acupuncture or placebo. Active control groups generally receive alternative treatment or medication with proven effects normally provided for the given condition. Active controls may also be referred to as positive controls or be taken to indicate standard treatment (Guideline, 2000). Yuan et al. (2015) reviewed the efficacy of acupuncture in an existing systematic review, and concluded that acupuncture is effective compared with inactive control, but did not draw a conclusion when comparing active control due to the small number of studies. Therefore in this study we compared acupuncture and electroacupuncture for neck pain with active controls and analyzed references evaluating the improvement of pain and quality of life (QoL). We included randomized controlled trials (RCTs) published recently and performed systematic review and meta-analysis. Methods Literature Search Strategy We searched Cochrane Central Register of Controlled Trials (CENTRAL), OVID-MED- LINE (1946 to July 2016), OVID-EMBASE (1980 to July 2016), OVID-AMED (1985 to July 2016) and one Chinese database (China National Knowledge Infrastructure (CNKI)) (search date: 26 July 2016) to find studies published up to 30 July The search words were neck pain (e.g. Neck pain, cervical spondylosis, cervical radiculopathy, cervical disc herniation, and myofascial pain syndrome), acupuncture (e.g. Needling, acupuncture, acupressure), and electroacupuncture (e.g. Electric acupuncture, electroacupuncture). The references for relevant publication (e.g. clinical guidelines for spinal diseases, reviews, etc.) were hand-searched. There was no language limitation. Inclusion and Exclusion Criteria Types of Studies. All RCTs of acupuncture and electroacupuncture for neck pain were included. Cross-over studies were not included unless first phase data were usable.

4 4 S.Y. SEO et al. Non-randomized studies, quasi-trials, and observational studies were excluded, as well as animal studies, qualitative studies, letters, etc. Types of Participants. Clinical studies of adult participants with mechanical neck disorders, myofascial pain syndrome, cervical spondylosis, cervical spine diseases accompanying radiating pain, and myalgia were included. Studies of patients complaining of myelopathy, or headache and dizziness without neck pain were excluded. Whiplash injury and external cause of injury were excluded, for they could have a different natural history. Types of Interventions. All types of acupuncture and electroacupuncture for neck pain were included. Studies using interventions other than acupuncture and electroacupuncture that are applied with the same method in the acupuncture group and control group were included. However, dry needling not based on oriental medicine and meridian theory was excluded. Types of Control Groups. Conventional therapy generally used for neck pain such as usual care, medication, and physical therapy modalities done by the general physician were included. However, herbal medicine as the control group could not be seen as a conventional therapy, and thus was excluded. Types of Outcome Measures. In this study we analyzed pain intensity, functional disability of the neck, and QoL to evaluate the efficacy of acupuncture. The outcome measures that were analyzed followed the NASS guideline and were determined by clinician counseling. The pain intensity was measured using Visual Analogue Scale (VAS) or McGill Pain Questionnaire, and functional disability was measured using the Neck Disability Index (NDI). The measurement of QoL was done by using general measurement tools assessing the QoL such as SF-36 or EQ5D. However, study results reported improvement rates that do not use objective evaluation tools and standards or that use tools which are not validated for reliability were excluded. Data Extraction Two independent reviewers reviewed the title and abstract of the selected reference and included or excluded studies according to the eligibility criteria. Later we subscribed the full text of the selected reference to determine the final selection. Disagreement was resolved by discussing with a third reviewer. Two reviewers independently read the full text of all of the studies and extracted data using a predefined form. Any disagreement was resolved by consultation between the two reviewers and the final data were examined by another reviewer. Assessment for Risk of Bias Two reviewers independently evaluated the risk of bias of the final selected studies using the risk of bias tool of Cochrane Collaboration. This tool consists of seven domains, which are selection bias (random sequence generation and allocation concealment), performance bias (blinding of participants and personnel), detection bias (blinding of outcome assessment), attrition bias (incomplete outcome data), reporting bias (selective outcome

5 SYSTEMATIC REVIEW FOR ACUPUNCTURE EFFECTS ON CHRONIC NECK PAIN 5 reporting), and other source of bias. Each study was evaluated as High, Low, or Unclear risk of bias on all of the domains, and the assessment criteria were based on the Cochrane handbook. (Higgins and Green, 2011). Disagreement between the two reviewers was settled by discussion. Analysis Statistics analysis was done using the Review Manager program (Version 5.3 Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). All the studies were grouped and analyzed considering the outcome variable of each study and the characteristics of intervention. All the outcome variables are continuous data and were presented as standardized mean differences (SMD) using inverse variance analysis. The number of studies included in each analysis was small and thus, was analyzed using a fixed-effects model. Heterogeneity between studies was evaluated by using 2 (chi-squared) test with p-value of p < 0:10 and I 2 statistic. In case of substantial heterogeneity the cause of heterogeneity was identified by analyzing subgroups. Publication bias was not done in cases of less than 10 studies in a group. Level of Evidence The level of evidence was evaluated according to the Grading of Recommendations, Assessment, Development and Evaluation (GRADE), on each outcome. The level of evidence was classified as high, moderate, low or very low. The evaluation of the level of evidence was done on the following domains; risk of bias, imprecision, inconsistency, indirectness, publication bias, large magnitude of effect, dose-response, and confounding. The domains were evaluated based on the criteria of the GRADE group (Balshem et al., 2011). Results Search Results A total of 4513 studies were searched, 337 from Ovid-Medline, 568 from Ovid-EMBASE, 300 from Ovid-AMED, 471 from the Cochrane Library, 1322 from CNKI (China), 743 from OASIS (Korea), 731 from NDSL (Korea), 17 from ICHUSHI (Japan), and 24 from J-stage (Japan) studies remained after eliminating overlapping studies, and 3028 studies were eliminated from this after primary exclusion based on the title and abstract. After secondary exclusion of the 708 remaining studies, 72 were eliminated for duplication, 35 were eliminated for not meeting the participant criteria, 354 were eliminated for not meeting the outcome measurement criteria, 58 were eliminated for not meeting the control criteria, 81 were eliminated for not meeting the intervention criteria, 91 were eliminated because they were not RCTs, and one was eliminated for not being able to access the original article. Sixteen total RCTs were included in the qualitative analysis and fourteen were included in the quantitative analysis (meta-analysis). Two studies were presented in

6 6 S.Y. SEO et al. Figure 1. PRISMA flow diagram of literature searching and article selection process. RCTs, randomized controlled trials; PRISMA, preferred reporting items for systematic reviews and meta-analyses. median value, not in mean value, and therefore were not included from the quantitative analysis (Giles and Muller, 1999, 2003). The flow chart of the analysis is presented in Fig. 1. Included Studies and Characteristics Ten of the sixteen studies were on patients with chronic neck pain and six of them were on patients with neck pain that accompanies radiculopathy. Nation-wise, one study was done in Korea, one in Brazil, one in the United States, two in the United Kingdom, two in Australia, and nine in China. Intervention-wise, nine studies were done using acupuncture and seven using electroacupuncture. In detail, seven of the acupuncture studies performed acupuncture with de qi, one study performed shallow-needling without de qi, and one study used Bo s abdominal acupuncture. In the electroacupuncture studies, the frequency used in five of the studies were Hz, 2 Hz, 2 4 Hz, Hz, and Hz, while two studies did not report the frequency used. The acupuncture points used more than twice in the acupuncture studies were GB21 and BL11. The acupuncture methods were diverse in each study and various acupuncture points including a shi points were identified. The most commonly used

7 SYSTEMATIC REVIEW FOR ACUPUNCTURE EFFECTS ON CHRONIC NECK PAIN 7 Figure 2. Risk of bias assessed using the Cochrane Risk of bias tool. þ, high risk of bias;?, unclear risk of bias; -, low risk of bias. acupuncture point in electroacupuncture studies was EX-B2. In the acupuncture studies, NSAIDs and usual care were most commonly compared with acupuncture as controls, while physical therapy, traction, medium frequency pulsation electrotherapy, and Methycobal Tab were also used as control groups. The number and period of treatment sessions varied in each study. A summary of the included studies in more detail is presented in Table 1. Assessment for Risk of Bias Random Sequence. Of the 16 studies 10 (Zhang and Jin, 2003; Salter et al., 2006; Franca et al., 2008; Wang et al., 2008; Hu, 2011; Tao and Cheng, 2011; Cho et al., 2014;

8 8 S.Y. SEO et al. Table 1. Characteristics of the Included Studies Study Country (Period) Disease Number of Participants (Age: Mean SD) Intervention Comparison Acupuncture Points Treatment Period Follow up Period Acupuncture vs. active control (Medication) Cho et al. (2014)* Korea (11/ / 2011) Giles and Muller (2003) Franca et al. (2008)* Australia (02/ / 2001) Brazil (03/ / 2007) Chronic neck pain IG: 15 ( ) CG: 15 ( ) Chronic spinal pain syndromes 13 weeks Tension neck syndrome with chronic pain IG: 34 (median: 39) CG: 40 (median: 39) IG: 15 (38.0) CG: 15 (33.0) Acupuncture (de qi) Acupuncture (painful needling) Acupuncture (TCM, de qi) NSAIDs (zaltoprofen 80 mg) NSAIDs (Celebrex mg/day or Vioxx mg/ day or paracetamol up to 4 g/day, depending on physician s discretion) Physiotherapy (therapeutic exercise) SI9, SI10, SI11, SI12, SI14, BL11, BL12, TE14, TE15, TE16, TE17, GB21 ND (Local tender points and distant acupuncture points depending on the condition being treated) LR3, ST36, LI4, SP6, KI3 IG: Total 9 sessions, 3 times a week for 3 weeks CG: 3 times a day for 3 weeks IG: 20 min each, 2 times per week until recovery (maximum 9 weeks) CG: until recovery (maximum 9 weeks) IG, CG: 1 or 2 times a week over 10 weeks Week 0(baseline), 1, 3 (primary endpoint), 7 Initial visit, week 2, 5, 9 after the initial treatment Before treatment, after 10 weeks, after 6 months Acupuncture with control vs. control Birch and Jamison (1998) Cho et al. (2014)* Franca et al. (2008)* MacPherson (2016) US (NR) Myofascial neck pain lasting 6 months Korea (11/ / 2011) Brazil (03/ / 2007) UK (03/ /2013) IG: 15 (40.9) CG: 15 (38.6) Chronic neck pain IG: 15 ( ) CG: 15 ( ) Tension neck syndrome with chronic pain Chronic nonspecific neck pain (median duration: 6 years) IG: 16 (30.0) CG: 15 (33.0) IG: 173 ( ) CG: 172 ( ) Acupuncture (shallow needling: no de qi) þ NSAIDs (Trilisate 500 mg) Acupuncture (de qi) þ NSAIDs (zaltoprofen 80 mg) Acupuncture (TCM, de qi) þ physiotherapy Acupuncture (TCM, de qi) þ usual care NSAIDs (Trilisate 500 mg) NSAIDs (Zaltoprofen 80 mg) Physiotherapy (therapeutic exercise) Usual care (consultants, medications, other routine health care) SI3, BL62, GB41, TE5, GB20, GB21, GB12, BL10, BL11, GV14 SI9, SI10, SI11, SI12, SI14, BL11, BL12, TE14, TE15, TE16, TE17, GB21 LR3, ST36, LI4, SP6, KI3 ND (specific to acupuncture diagnosis) IG: Total 14 sessions, 2 times a week for 4 weeks, once a week for 4 weeks, every other week for 2 weeks, CG: 500 mg a day for 4 weeks IG: Total 9 sessions, 3 times a week for 3 weeks, CG: daily for 3 weeks IG, CG: 1 or 2 times a week over 10 weeks IG: Total 12 sessions (600 min), 50min each, 1 2 times a week CG: ND Immediate post treatment, 3 months after treatment Week 0 (baseline), 1, 3 (primary endpoint), 7 Before treatment, after 10 weeks, after 6 months Baseline, 3, 6, 12 months Relevant Outcomes VAS, NDI, SF-36, BDI, EQ-5D Oswestry Questionnaire, NDI, SF-36, VAS, pain frequency, CROM VAS pain, VAS muscle tension NDI C- CFT CPEQ, SFMPQ, PIR, SF-36, SCL-90-R, MQS, belief and helpfulness measures, Physiological measures VAS, NDI, SF-36, BDI, EQ-5D VAS pain, VAS muscle tension NDI, C- CFT NPQ, SF-12 ver. 2, Chronic Pain Self- Efficacy Scale pain intensity, adverse effects

9 SYSTEMATIC REVIEW FOR ACUPUNCTURE EFFECTS ON CHRONIC NECK PAIN 9 Table 1. (Continued) Study Country (Period) Disease Number of Participants (Age: Mean SD) Intervention Comparison Acupuncture Points Treatment Period Follow up Period Salter et al. (2006) Hu (2011) China (02/ / 2010) Tao and Cheng (2011) Xu et al. (2015) UK (ND) Neck pain in the previous 12 months China (05/ / 2010) Cervical spondylosis of nerve root type Cervical spondylosis of neck type China (ND) Cervical spondylosis of neck type IG: 10 ( ) CG: 14 ( ) IG: 41 (41.96) CG: 41 (42.53) IG: 36 ( ) CG: 39 ( ) IG: 33 ( ) CG: 32 ( ) Acupuncture (TCM: Five Element Approach, de qi) þ usual GP care Acupuncture (abdominal acupuncture, de qi) þ traction Acupuncture (de qi) þ methycobal Tab Acupuncture (Bo s abdominal acupuncture) Usual GP care (medication, massage, exercise, etc.) Traction (cervical traction, fixed-point) ND (specific to acupuncture diagnosis) CV12, CV4, KI17, ST24, KI18 IG: Up to 10 treatments over 3 months (average: 7.9 sessions) GP: ND IG, CG: Total 18 sessions, once a day, 6 times a week for 3 weeks Methycobal Tab. GB20, GV14, Ashi point IG: Once a day for 1 week, CG: 3 tablets a day for 1 week Medium frequency pulsation electrotherapy (frequency range: Hz, maximum output current: 100 ma) CV12, CV4, KI17, ST24 IG: Total 10 sessions, 25 min a day, 5 times a week, CG: Total 10 sessions, 20 min a day, 5 times a week Baseline, 1 month 3 months Immediate post treatment Immediate post treatment After first and last treatment. Electroacupuncture vs. physiotherapy (sole) Chen et al. (2015) Hu et al. (2015) China (06/ / 2013) China (06/ / 2013) Cervical Spondylotic Radiculopathy Cervical spondylosis of nerve root type IG: 32 ( ) CG: 32 ( ) EG: 32 ( ) PT: 30 ( ) Electroacupuncture (frequency range: Hz, intensity: patient s endurance capacity) Electroacupuncture (frequency range: Hz, intensity: patient s endurance capacity) Traction (force: N, patient s endurance capacity) Physical therapy (traction þ microwave) (traction: microcomputer retraction, maximum force: 20 kg) EX-B2, BL10, EX-HN 15 EX-B2, LI11, LI4, HT3, LI10 IG, CG: Total 20 sessions, daily, rest 3 days between 10 days IG, CG: Total 10 sessions, daily, rest 2 days between 5 day Before and after treatment Before and after treatment Relevant Outcomes NPQ, medication use (baseline), SF-36, EQ-5D SF-MPQ MPQ, VAS Efficacy evaluation criteria, VAS, NDI, Trapezius semg Chinese medicine diagnosis criteria, MPQ Chinese medicine diagnosis criteria, SF- MPQ

10 10 S.Y. SEO et al. Study Country (Period) Lu and Jiao (2016) China (06/ / 2014) Zhang and Jin (2003) China (08/ / 2000) Giles and Muller (1999) Australia (07/ / 1998) Table 1. (Continued) Disease Number of Participants (Age: Mean SD) Intervention Comparison Acupuncture Points Treatment Period Follow up Period Cervical spondylosis of neck type Cervical spondylosis (nerve root type) Chronic Spinal Pain Syndromes 13 weeks IG: 36 (22 61**) CG: 36 (23 59**) IG: 60 (24 72**) CG: 60 (26 75**) IG: 20 (median: 46.5) CG: 21 (median: 35.0) Electroacupuncture (frequency range: NR, intensity: patient s endurance capacity) Electroacupuncture (frequency range: 2 4 Hz, intensity: patient s endurance capacity) Electroacupuncture (low-voltage) Physical therapy (computerized intermediate frequency therapy apparatus, frequency range: Hz, maximum output current: 100 ma) Traction (average force: 7.5 kg, patient s endurance capacity) EX-B2, GB20, GB21, SI3 BL10, EX-HN15, BL11 External wind, cold, and dampness: GV14, GB20, BL12, GB21, TE5 /Phlegm-induced stagnation of collateral: LI11, BL20, ST40, BL17/Stagnation of qi and blood: 17, LI15, LI11, SI15, SI14/ Deficiency of qi and blood: BL18, BL20, ST36/Deficiency of liver and kidney: SI6, BL18, BL23, KI3 Manipulation ND (Local tender points and distant acupuncture points depending on the condition being treated) IG, CG: Total 10 sessions, 5 times a week for 2 weeks IG, CG: Total 45 sessions, 30 min each, rest 2 days between 15 sessions. IG, CG: Total 6 sessions, min each, over 3 4 weeks Before and after treatment, 3 months Before and after 15 sessions of treatment Before and after treatment Relevant Outcomes Chinese medicine diagnosis criteria, effective rate, NPQ, recurrence rate SF-MPQ OBPDI, NDI, VAS of local pain

11 SYSTEMATIC REVIEW FOR ACUPUNCTURE EFFECTS ON CHRONIC NECK PAIN 11 Table 1. (Continued) Study Country (Period) Disease Number of Participants (Age: Mean SD) Intervention Comparison Acupuncture Points Treatment Period Follow up Period Relevant Outcomes Electroacupunctureþusual care vs. usual care (add-on) Wang et al. (2008) Yu et al. (2009) China (09/ / 2008) Hu et al. (2015) China (06/ / 2013) China (NR) Cervical spondylosis of nerve root type Cervical spondylosis of nerve root type Cervical spondylosis of nerve root type IG: 66 ( ) CG: 66 ( ) Total: 64 ( ) IG: 32 CG: 32 IG: 31 ( ) CG: 30 ( ) Electroacupuncture (frequency range: 1 W 300 Hz, intensity: patient s endurance capacity) þ usual care Electroacupuncture (frequency range: 2 Hz, intensity: 0.5 2mA) þ usual care Electroacupuncture (frequency range: Hz, intensity: patient s endurance capacity) þ physical therapy Usual care (traction þ tuina þ TDP) Usual care (traction þ manipulation) Physical therapy (traction þ microwave) (traction: microcomputer retraction, maximum force: 20 kg) EX-B2, GB20, GB21, LI15, SI9, TE5, SI3 IG, CG: Total 10 sessions, 20 min each (IG: 40 min, CG: 20 min), everyday EX-B2 IG, CG: Once a day for 20 days EX-B2, LI11, LI4, HT3, LI10 IG, CG: Total 10 sessions, daily, rest 2 days between 5 days Before and after treatment Before and after treatment Before and after treatment Chinese medicine diagnosis criteria, Effective rate, VAS, SF-MPQ, SEP Effective rate VAS Chinese medicine diagnosis criteria, SF- MPQ Note: IG: Intervention Group, CG: Control Group, VAS: Visual Analogue Scale, NDI: Neck Disability Index, BDI: Beck s Depression Inventory, EQ-5D: Euroqol 5-D health utility, SF-36: 36-Item Short Form Health Survey, CPEQ: Comprehensive Pain Evaluation Questionnaire, MPQ: McGill Pain Questionnaire, SF-MPQ: Short Form McGill Pain Questionnaire, PIR: Pain Intensity Ratings, SCL-90-R: Symptom Checklist 90, MQS: Medication Quantification Scale, MPQ: Multiple Pain Quality, VAS: Visual Analogue Scale, semg: surface Electromyography, PRI: Pain rating index, PPI: Present Pain Intensity, NPQ: Neck Pain Questionnaire, SEP: Somatosensory Evoked Potentials, OBPDI: Oswestry Back Pain Disability Index, CROM: Cervical Range of Motion, ND: Not Defined, SD: Standard deviation, C-CFT: Cranio-Cervical Flexion Test. *: Three-armed study, **: Presented as range.

12 12 S.Y. SEO et al. Hu et al., 2015; Xu et al., 2015; MacPherson, 2016) used computer-programmed random sequencing or a random number table, and were thus evaluated as low risk of bias. Four of the remaining studies (Birch and Jamison, 1998; Yu et al., 2009; Chen et al., 2015; Lu and Jiao, 2016) did not mention the method of random sequencing, and two studies (Giles and Muller, 1999, 2003) used the sealed-envelop method, but the actual number of participants and the number of envelops were not matched. Thus, this was evaluated as an unclear risk of bias (Fig. 2). Allocation. In four studies, an independent investigator performed allocation concealment so that the risk of bias was evaluated as low (Salter et al., 2006; Franca et al., 2008; Cho et al., 2014; MacPherson, 2016). Giles and Muller (1999, 2003) used sealed-envelops, and thus were evaluated as having a low risk of bias (Giles and Muller, 1999, 2003). The rest of the studies did not mention allocation concealment, and how the risk of bias was evaluated is unclear (Birch and Jamison, 1998; Zhang and Jin, 2003; Wang et al., 2008; Yu et al., 2009; Hu, 2011; Tao and Cheng, 2011; Chen et al., 2015; Hu et al., 2015; Xu et al., 2015; Lu and Jiao, 2016). Blinding. Due to the nature of the active control and acupuncture, most of the studies did not perform blinding. Only in the study of Cho et al. (2014) was the outcome assessment blinded, resulting in a low risk of bias. The rest were evaluated as high risk of bias. Incomplete Outcome Data. In four studies the number of drop-out was high resulting in high risk of bias (Giles and Muller, 1999, 2003; Hu et al., 2015; Lu and Jiao, 2016). The rest of the studies were evaluated as low risk of bias. Selective Reporting. Cho et al. (2014); MacPherson (2016); Salter et al. (2006) and Franca et al. (2008) either published or registered the protocol of the study, or reported all the results of the variables, thus evaluated as low risk of bias. The rest of the studies did not register the protocol, and many did not mention adverse events, thus evaluated as unclear risk of bias. Other potential Sources of Bias. The study by Giles and Muller (1999) was found not to have randomized the preplanned number of participants because of a shortage of funding and because the termination of study could have influenced the results of the study. It was difficult to clearly assess the risk of bias (Giles and Muller, 1999). Analysis. Since the RCTs included in this study vary in study designs, they need to be categorized by the types of interventions, outcome measurement scales, and study designs. Studies were categorized for analysis by the types of intervention (acupuncture or electroacupuncture), by the measurement scales (VAS, NDI, and SF-36) and by study designs ( Sole when the intervention group used acupuncture only and Add-on when acupuncture was accompanied by other treatments.) Electroacupuncture was also analyzed in the same way. Acupuncture vs. Active Control (Sole) Pain (VAS). There were two studies adequate for meta-analysis of VAS after sole treatment of acupuncture and other treatment (Franca et al., 2008; Cho et al., 2014). Analysis based

13 SYSTEMATIC REVIEW FOR ACUPUNCTURE EFFECTS ON CHRONIC NECK PAIN 13 on the characteristics of the control group showed a SMD of 0.23 (95% CI 0.95, 0.48, p ¼ 0:52) in a study that compared acupuncture with NSAIDs, and a SMD of 0.73 (95% CI 0.00, 1.46, p ¼ 0:05) in a study that compared acupuncture with physical therapy, which were both not significant. The meta-analysis also showed a SMD of 0.24 (95% CI 0.27, 0.75) which is statistically non-significant (p ¼ 0:36) (Fig. 3). In the study of Giles and Muller (2003), which was not included in the meta-analysis because they presented median in the results, acupuncture decreased the pain VAS to 50%, which is better than the control group of medication, which was 0%, and of manipulation, which was 42% (Giles and Muller, 1999, 2003). In the study of Giles and Muller (1999), acupuncture decreased the median VAS by 1.0, while manipulation decreased it by 1.5 and medication by 0.5, meaning that acupuncture is more effective than medication, but is not as effective as manipulation (Giles and Muller, 1999). Disability (NDI). Two studies included for NDI analysis of acupuncture sole treatment showed SMD of 0.51 (95% CI 0.01, 1.02 p ¼ 0:05), which did not show significant differences in the effects (Fig. 3). In the study of Giles and Muller (1999), which was not included in the meta-analysis, the medication group showed a difference of 0 in the median NDI score, while the acupuncture group showed a difference of 6.0, and the manipulation Figure 3. Forest plot of outcomes in patients with neck pain comparing acupuncture vs. active control (Sole).

14 14 S.Y. SEO et al. group This shows that acupuncture is more effective than medication for disability, but not as effective as manipulation (Giles and Muller, 1999). In the study of Giles and Muller (2003), the medication group showed a difference of 8% in the NDI, while the acupuncture group showed 16%, and the manipulation group 38%, showing that acupuncture is more effective than medication, but not as effective as manipulation (Giles and Muller, 1999, 2003). Acupuncture with Control vs. Control (Add-On) Pain (VAS). There were eight studies that measured pain with VAS score for the intervention group treated with acupuncture and other treatments and the control group. These studies showed SMD of 0.57 (95% CI 0.73, 0.40, p < 0:00001), but due to the high heterogeneity (p < 0:00001, I 2 ¼ 94%), the data were analyzed in subgroups according to the risk of bias, especially the allocation concealment. Four studies with well-performed allocation concealment showed a SMD of 0.07 (95% CI 0.26, 0.12, p ¼ 0:47), which was not significant, and four other studies with poor allocation concealment showed an SMD of 0.57 (95% CI 0.01, 1.02, p < 0:00001) which was significant. But, the heterogeneity was not resolved in spite of the subgroup analysis (Fig. 4). Disability (NDI). Two studies reported the results using NDI (Cho et al., 2014; Franca et al., 2008), with SMD of 0.72 (95% CI 0.20, 1.24, p ¼ 0:007), where the NDI score in the control group was significantly lower (Fig. 4). Quality of Life (SF-36). One study measured the QoL (Cho et al., 2014) with SMD of 1.48 (95% CI 2.31, 0.06, p ¼ 0:0004), which was significantly higher in the control group, but the difference was not a great compared to the baseline (Fig. 4). Electroacupuncture vs. Active Control (Sole): Pain (VAS) There were four studies that measured VAS on intervention group of electroacupuncture in comparison to other treatments. The result showed SMD of 0.51 (95% CI 0.73, 0.28, p < 0:00001), indicating significant effect of electroacupuncture for relieving pain, and also with low heterogeneity (p ¼ 0:54, I 2 ¼ 0%). Subgroup analysis was also performed on other physical therapy and traction, categorized by the type of the control group, where other physical therapy group showed SMD of 0.42 (95% CI 0.77, 0.08, p ¼ 0:02), and traction group showed a SMD of 0.51 (95% CI 0.86, 0.27, p ¼ 0:0002), with no difference in the effects depending on the types of the control group (Fig. 5). Electroacupuncture with Control vs. Control (Add-On): Pain (VAS) There were three studies that measured VAS score for electroacupuncture accompanied by other treatments. The resulting SMD of 1.29 (95% CI 1.57, 1.02, p < 0:00001) indicates that electroacupuncture accompanied by other treatments significantly relieves pain, with the heterogeneity in a tolerable level (p ¼ 0:12, I 2 ¼ 52%) (Fig. 5).

15 SYSTEMATIC REVIEW FOR ACUPUNCTURE EFFECTS ON CHRONIC NECK PAIN 15 Figure 4. Forest plot of outcomes in patients with neck pain comparing acupuncture with control vs. control (Add-On). Figure 5. Forest plot of outcomes in patients with neck pain comparing electroacupuncture vs. active control and electroacupuncture with control vs. control (Sole and Add-on).

16 16 S.Y. SEO et al. Table 2. Meta-Analysis of Outcomes and Level of Evidence Variable Overall Effect Studies (N) Sample Size (N) SMD 95% CI p I 2 p ( 2 test) Statistical Method Level of Evidence Acupuncture vs. active control (Sole) Pain (VAS) vs. medication , Fixed inverse variance Pain (VAS) vs. physiotherapy , Fixed inverse variance Disability (NDI) , Fixed inverse variance QoL (SF-36) , Fixed inverse variance Acupuncture with active control vs. control (add-on) Pain (VAS) with allocation concealment , Fixed inverse variance Pain (VAS) without allocation concealment , 1.48 < 0: Fixed inverse variance Disability (NDI) , Fixed inverse variance QoL (SF-36) , Fixed inverse variance Electroacupuncture vs. active control (Sole) Pain (VAS) , 0.28 < 0: Fixed inverse variance Electroacupuncture with active control vs. active control Pain (VAS) , 1.02 < 0: Fixed inverse variance 1 30 Moderate 1 31 Moderate 2 61 Moderate 1 30 Moderate Low Very low 2 61 Moderate 1 30 Moderate Low Low Note: VAS: Visual analogue scale, NDI: Neck disability index, QoL: Quality of life, SMD: Standardized Mean difference.

17 SYSTEMATIC REVIEW FOR ACUPUNCTURE EFFECTS ON CHRONIC NECK PAIN 17 Adverse Events Eleven studies did not mention of any adverse effects (Birch and Jamison, 1998; Giles and Muller, 1999; Zhang and Jin, 2003; Wang et al., 2008; Yu et al., 2009; Hu, 2011; Tao and Cheng, 2011; Chen et al., 2015; Hu et al., 2015; Xu et al., 2015; Lu and Jiao, 2016). Cho et al. (2014) reported no serious adverse events, with only one case of skin rash in the group treated with acupuncture and NSAIDs. However, this was irrelevant with the intervention. Franca et al. (2008) reported of no occurrence of any adverse effects. Salter et al. (2006) reported a temporary aggravation of symptoms (n ¼ 6), dizziness (n ¼ 6), and tiredness (n ¼ 4) and Giles and Muller (2003) reported adverse effects in 6.1% in the medication group, with no specific mention of symptoms. MacPherson (2016) reported of 80 cases of adverse effects, 30 of which were serious and 50 were non-serious. Of the serious cases, none were related to the interventions. Although serious adverse effects did not result from acupuncture treatment, there needs to be further study since there were many studies that did not deal with adverse effects at all. Level of Evidence The level of evidence for each analysis is in Table 2. A small number of studies with few participants was included, and all the outcome was downgraded of a level in the precision domain. Comparison of acupuncture vs. active control did not result in a significant difference, and the result from the small sample size downgraded the quality of evidence in imprecision, and was evaluated as moderate. Comparative analysis of acupuncture with active control vs. active control did not result in a significant difference in pain within the studies with well-performed allocation concealment. The level of evidence was evaluated as low due to imprecision and inconsistency. In studies with an unclear risk of allocation concealment, the results showed a significant decrease of pain, but were evaluated as very low in terms of evidence due to risk of bias, imprecision, and inconsistency. Disability and QoL showed significant improvement in the control group with a moderate level of evidence due to imprecision. Comparison of pain in the electroacupuncture and active control resulted in a low level of evidence, downgraded for risk of bias and imprecision. Comparison of electroacupuncture with active control vs. active control was evaluated as low in the level of evidence. Discussion This study analyzes 16 RCTs on neck pain that used acupuncture and electroacupuncture for comparison with active controls such as medication and physical therapy. In the comparison of acupuncture vs. active control, pain, disability, and QoL did not show a significant difference, which means that acupuncture exerts a similar amount of effect as the active control. When acupuncture was added to the treatment of the control group, acupuncture showed significantly increased pain-relieving effects, where the significant effect was only identifiable in studies with underperformed allocation concealment. Although the

18 18 S.Y. SEO et al. control group showed a significantly improved functional index and QoL index in comparison to the acupuncture add-on group, further research and studies are needed since this result was the analysis of only 61 patients. Electroacupuncture compared to control or electroacupuncture add-on to the control group both showed significant pain-relieving effects. However, since all the studies were published in China with a risk of bias, there needs to be additional large-scale clinical studies that are well designed before drawing out conclusions. Studies show that electroacupuncture is more effective in relieving neck pain in comparison to acupuncture, but the risk of bias prohibits clear conclusions. Especially since there an inadequate amount of literature for each analysis, and the number of candidates for each study was limited, lowering the credibility of the evidence. Therefore, the level of evidence for some of the outcome variables turned out to be moderate, but there were limits that lower the credibility of the studies to low and very low. Since this study aims to compare acupuncture and active control, there needs to be a large-scale study to confirm the effect. However, most of the studies included were pilot studies for the calculation of sample size, and therefore had very few participants. Adverse effects reported from studies were limited, and within the reported adverse effects, it can be concluded that the adverse effects from acupuncture are not as severe or serious as other active controls. Although acupuncture is free from the risk of serious adverse effects (Ernst and White, 1997), most serious adverse effects can be prevented via careful and hygienic administration and education. The mechanism by which acupuncture relieves neck pain is not clear, but is being confirmed by many studies. From the 1970s, animal studies and human studies have reported the analgesic effects of acupuncture through sympathetic nerve and the endocrine system (Coutaux, 2017). When pain is first detected, the plasticity of the peripheral nervous system and central nervous system increases to accelerate hypersensitivity (Basbaum et al., 2009). Acupuncture regulates the opioid, serotonin, especially the sympathetic nervous system located at the spinal cord level to decrease the hypersensitivity of pain for pain relief (Kim et al., 2005; Cidral-Filho et al., 2011; Pachman et al., 2011; Song et al., 2011). This study separately analyzed studies that used acupuncture and electroacupuncture since there can be difference in their effects on neck pain. Electroacupuncture is a treatment method for the maximization of the effects of acupuncture via the simultaneous administration of acupuncture needling and electric stimulation. In inflammation-induced rat models, stimulation of electroacupuncture on ST36, acupoint located near the hind paw increased the pain threshold (Sekido et al., 2003; Taguchi et al., 2010), which results from the activation of the secretion of endocrine opioids like β-endorphins secretion from lymphocytes, monocytes/macrophages, and granulocytes located at the sites of inflammation (Cabot et al., 1997; Rittner et al., 2001). Also, electroacupuncture is reported to activate the peripheral sympathetic nerve fibers and cannabinoid system to increase the concentration of opioids in the sites of inflammation to relieve pain (Kim et al., 2008; Su et al., 2011). Electroacupuncture inhibits COX-2, thereby increasing opioids in the site of inflammation, which shows similar effect as anti-inflammatory medications (Lee et al., 2006). Through this mechanism, it can be supposed that electroacupuncture is better than acupuncture in relieving pain. Although there is a study that suggests 2 10 Hz

19 SYSTEMATIC REVIEW FOR ACUPUNCTURE EFFECTS ON CHRONIC NECK PAIN 19 electroacupuncture is more effective than 100 Hz in inflammatory and neuropathic pain (Zhang et al., 2014), many studies on electroacupuncture included using various frequencies resulting in no clear difference in effects. There needs to be further studies on the difference in effects upon frequencies. In 2015, a systemic review by Yuan et al. (2015) studied the effect of acupuncture on neck pain. Yuan et al. concluded that acupuncture for chronic neck pain showed evidence of moderate effects in comparison to sham acupuncture. Comparison with active control that compared acupuncture with medication, massage, and traction did not result in a concrete conclusion due to the small number of participants. This study includes more studies published afterwards, and difference in inclusion and exclusion criteria resulted in different analytic results on the effects of acupuncture. Another uniqueness of this study is that acupuncture and electroacupuncture were separately analyzed. Giles and Muller (1999, 2003) included in the systematic review of Yuan et al. (2015) only presented the median value, and therefore the authors excluded it from the meta-analysis and only included it in the qualitative analysis. Of the studies included in the systematic review of Yuan et al., the study of Li et al. (2006) was excluded for the selection of spinal stenosis as a disease and the study of Thomas et al. (1991) was excluded for the inability to extract first-phase data applied with acupuncture although it was a cross-over design. A Cochrane review on acupuncture treatment for neck pain was compared with inactive control (Trinh et al., 2006). There is still controversy over the proper sham control group, and a systemic review of the literature for clinical practice to assist in clinical decision making by comparing it with the active controls used in actual clinical settings is meaningful. Therefore, this study separated the intervention group into acupuncture and electroacupuncture to compare the effects with the active control in neck pain for analysis. Although it was hard to conclude that acupuncture alone for neck pain is better than analgesics or physical therapies, the addition of acupuncture to conventional treatment better controls pain. Even better effects could be expected when electroacupuncture is used. Also, regarding acupoint selection, there is the possibility that selecting acupoints in the neck region may slightly increase the treatment effect. Franca et al. (2008) used distal points such as acupoints in the feet as opposed to proximal points, and the distinct lack of effect in outcomes may be carefully conjectured to be attributable to the difference in needling area. Furthermore, studies exclusively utilizing acupoints in the abdominal region for neck pain include those by Hu (2011) and Xu et al. (2015), and while the effect size was slightly smaller in the study by Hu (2011), it was difficult to accurately ascertain the difference in studies with an uncertain risk of allocation concealment. Moreover, many studies employed individualized acupoints, thus precluding direct comparison and the assessment of the difference in effect by acupoint. Future studies may feel the need to further consider the effects of acupoint selection on the treatment effects in light of the fact that acupuncture needling non-proximal to the site of pain may decrease its effect size. While the authors had originally aimed to investigate the difference in effect size by treatment duration and frequency of the included studies, the majority administered long term treatment or 10 sessions due to the fact that the patient population was chronic neck

20 20 S.Y. SEO et al. pain patients. It was therefore difficult to determine treatment effect by treatment frequency as most studies applied 10 treatment sessions. For example, Cho et al. (2014) employed nine acupuncture sessions, Salter et al. (2006) 10 sessions, and MacPherson (2016) 12 sessions, rendering comparison by treatment frequency irrelevant. While the number of acupuncture sessions in the study by Tao and Cheng (2011) was smaller at about seven sessions, the treatment size did not show a difference between studies. There was an unclear risk of allocation concealment. However, with the diverse designs and qualities of studies included, the level of evidence was not that high. In this analysis, other factors, like QoL in addition to pain, were also analyzed, but due to the small number of studies included, it was hard to conclude that acupuncture improves function and QoL in patients with neck pain. As mentioned above, the relatively many studies on the effects of acupuncture on neck pain were not enough to reach a concrete conclusion. A qualified systemic review can only be established upon well-performed clinical trials. Many bibliographical studies are not enough for drawing a meaningful conclusion if qualified RCTs are not enough (Ernst and White, 1997). Since the level of evidence has been lowered in all the outcome variables from imprecision, studies of larger scales are called for. The effects of acupuncture should be studied afterwards with well-designed clinical trials. It is meaningful enough that acupuncture is as effective as, if not better than, the common treatment strategies for neck pain such as medication and physical therapies. Since there were no serious adverse effects from acupuncture within the studies included in this analysis, acupuncture treatment can be considered for neck pain. Acupuncture, especially electroacupuncture, added onto the conventional treatments may be expected to decrease neck pain even more. However, since the included studies were small and risk of bias was high, further larger scale studies of higher quality design are needed. Conclusion Comparison of the effects of acupuncture and conservative treatments for chronic neck pain revealed that both have similar degrees of effects on pain management and functional improvement. There was greater improvement in pain control when acupuncture was added onto conventional treatment, and the use of electroacupuncture decreased pain into a higher degree. However, with the risk of bias and imprecision of the studies included, a concrete conclusion is difficult to draw. Thus, well-designed, rigorous studies are warranted in the future. Acknowledgments This study was supported by the Traditional Korean Medicine R&D Program funded by the Ministry of Health & Welfare through the Korea Health Industry Development Institute (KHIDI) (HB16C0035). The funder did not play any role in the design, collection, screening, interpretation, writing, and submission for publication of this review.