Cupping therapy versus acupuncture for pain-related conditions: a systematic review of randomized controlled trials and trial sequential analysis

Background Both cupping therapy and acupuncture have been used in China for a long time, and their target indications are pain-related conditions. There is no systematic review comparing the effectiveness of these two therapies. Objectives To compare the beneficial effectiveness and safety between cupping therapy and acupuncture for pain-related conditions to provide evidence for clinical practice. Methods Protocol of this review was registered in PROSPERO (CRD42016050986). We conducted literature search from six electronic databases until 31st March 2017. We included randomized trials comparing cupping therapy with acupuncture on pain-related conditions. Methodological quality of the included studies was evaluated by risk of bias tool. Mean difference, risk ratio, risk difference and their 95% confidence interval were used to report the estimate effect of the pooled results through meta-analysis or the results from each individual study. Trial sequential analysis (TSA) was applied to adjust random errors and calculate the sample size. Results Twenty-three randomized trials with 2845 participants were included covering 12 pain-related conditions. All included studies were of poor methodological quality. Three meta-analyses were conducted, which showed similar clinical beneficial effects of cupping therapy and acupuncture for the rate of symptom improvement in cervical spondylosis (RR 1.13, 95% CI 1.01 to 1.26; n = 646), lateral femoral cutaneous neuritis (RR 1.10, 95% CI 1.00 to 1.22; n = 102) and scapulohumeral periarthritis (RR 1.31, 95% CI 1.15 to 1.51; n = 208). Results from other outcomes (such as visual analogue and numerical rating scale) in each study also showed no statistical significant difference between these two therapies for all included pain-related conditions. The results of TSA for cervical spondylosis demonstrated that the current available data have not reached a powerful conclusion. No serious adverse events related to cupping therapy or acupuncture was found in included studies. Conclusion Cupping therapy and acupuncture are potentially safe, and they have similar effectiveness in relieving pain. However, further rigorous studies investigating relevant pain-related conditions are warranted to establish comparative effectiveness analysis between these two therapies. Cost-effectiveness studies should be considered in the future studies to establish evidence for decision-making in clinical practice.


Background
Traditional Chinese non-pharmaceutical therapies, such as acupuncture, are applied under the guidance of the Traditional Chinese Medicine (TCM) theory of syndrome differentiation. As an important part of TCM, these therapies mainly use manual or technique stimulations at specific body parts (especially acupoints) to dredge the meridian system. Generally, non-pharmaceutical therapies, including acupuncture, cupping therapy, moxibustion, massage (tuina), and guasha, are more likely to be accepted by patients since they have been used in treating numerous diseases or conditions and may have fewer side effects than drugs [1]. Acupuncture, as one of the most popular non-pharmaceutical therapy, has been widely used to treat diseases by regulating the functions of qi (vital energy) and blood of the organs through puncturing certain acupoints of meridians in the body with needles, to strengthen the resistance of the body against diseases [2]. A current clinical guideline issued by the American College of Physicians (ACP) recommends non-pharmacologic treatments (such as acupuncture, massage and superficial heat) as the priority treatments to patients with acute, subacute or chronic low back pain [3]. In addition, systematic reviews have reported that acupuncture was indicated for the treatment of chronic pain, mainly headaches, migraines, cervical pain, back pain, and pain from osteoarthritis [4][5][6][7][8][9].
Cupping therapy also belongs to TCM non-pharmaceutical therapy, which has been used for long time [10]. Cupping practitioners utilize the flaming heating power to achieve suction (minus pressure) inside the glass cups to make them apply on the desired part of the body, and this suction on selected acupoints produces hyperemia or hemostasis, which may result in a therapeutic effect [11]. There are different types of cupping including retained cupping, flash cupping, moving cupping, wet cupping, medicinal cupping, and needling cupping [12]. Since a report about the Olympic swimmer Michael Phelps using cupping therapy to relieve his muscular discomfort, this treatment has become more and more popular outside China. However, although beneficial effects of cupping therapy have been reported in treating various diseases/ conditions, there is lack of high-quality evidence to confirm its efficacy [13]. Our previous systematic reviews on cupping for pain-related conditions also identified no high-quality evidence to prove its effectiveness [14,15].
Both acupuncture and cupping therapy are commonly used in treating similar conditions, especially pain-related conditions. Though the mechanism of acupuncture and cupping therapy may be different, both, therapies employ the meridian and acupoints to activate blood stasis and regulate the flow of qi to relieve pain. Cupping therapy has more advantages than acupuncture, such as a non-invasive therapy with relatively shorter treatment duration and potential less treatment cost. It is worthy to critically review the evidence of the comparison of these two therapies to inform clinical practice. Herein, to the objective of this review is to comprehensively review the evidence from randomized controlled trials (RCTs) comparing cupping therapy with acupuncture for pain-related conditions.

Methods
The protocol of this review was registered in PROSPERO (CRD42016050986) on 15th November 2016 (Achieved at http://www.crd.york.ac.uk/PROSPERO/). Since painrelated conditions were most commonly treated by cupping therapy and acupuncture, we limited the target conditions (such as musculoskeletal pain, tissue pain and neuralgia pain) in this review to reduce the clinical heterogeneity among included studies.

Inclusion criteria
RCTs comparing cupping therapy with acupuncture were included. Pain-related conditions were classified by the type of tissue according to international statistical classification of disease and health related problems by World Health Organization [16], including musculoskeletal system pain (such as spondylopathies, lumbar spondylosis, knee osteoarthritis, acute tissue pain), and neurologic pain (such as lateral femoral cutaneous nerve and herpes zoster pain). Acupuncture is defined as the insertion of fine needles, sometimes in conjunction with electrical stimulus, to influence physiological functioning of the body. In this review, we included both manual acupuncture (including auricular therapy, scalp needle, and abdominal acupuncture) and electro-acupuncture. Cupping is defined that practitioners utilize the flaming heating power to achieve suction (minus pressure) inside the glass cups to make them apply on the desired part of the body. In this review, all types of cupping (i.e. wet cupping, herbal cupping, moving cupping, flash cupping or retained cupping) were included. Primary outcome measures included severity of pain, functional capacity, quality of life (QoL). Secondary outcomes included depression, rate of symptom improvement and adverse effects. There was no limitation on language and publication type.

Data extraction and risk of bias assessment
Two authors (YJ Zhang and XL Li) extracted the data from the included studies independently. The methodological quality of RCTs was assessed according to the criteria from the Cochrane Handbook for Systematic Reviews of Interventions [17]. The quality of studies was categorized into low, unclear, or high risk of bias according to the risk for each important outcome within included studies, including adequacy of generation of the allocation sequence, allocation concealment, blinding, whether there were incomplete outcome data and selected reporting the results. Studies which met all criteria were categorized to low risk of bias, studies which meet none of the criteria were categorized to high risk of bias, and those were categorized to unclear risk of bias if insufficient information acquired to make the judgment.

Data analysis
Data were extracted and calculated for frequency using Microsoft Excel 2007 (American: The Microsoft Corporation, 2007). Binary outcomes were summarized using risk ratio (RR) with 95% confidence intervals (CI) for relative effect and risk difference (RD) with 95% CI for absolute effect. The continuous outcomes were summarized using mean difference (MD) with 95% CI. Revman 5.3 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2016) was used for data analyses. Meta-analysis was used if the studies had similar clinical characteristics (such as study design, participants, interventions, control, and outcome measures) and acceptable statistical heterogeneity. Random-effect model was used for meta-analysis. Statistical heterogeneity was detected by I 2 test, an I 2 > 50% indicates the possibility of statistical heterogeneity among the studies. If I 2 was larger than 75%, which means there was obviously statistical heterogeneity among studies, only results from each single study were present respectively rather than pooling analysis. Funnel plot analysis was planned to be generated to detect publication bias.

Trial sequential analysis (TSA)
TSA can be performed if there are more than 5 included studies in the meta-analysis. We applied TSA version 0.9.5.5 (Copenhagen: The Copenhagen Trial Unit, Center for Clinical Intervention Research, 2016) to calculate the required sample size in a meta-analysis and to detect the robustness of the result. We used the diversity-adjusted required information size estimated from a control event proportion of the included studies and a priori intervention effect of 5%, and the diversity which was estimated in the included studies.

Basic information of included studies
After primary searches from the 6 databases, 3102 studies were identified. After screening title and abstract, the majority of studies were not included due to obvious ineligibility, and full texts of 3079 studies were retrieved. Finally, 23 RCTs  were included in this review, 21 of which were published in Chinese and the remaining 2 [21,40] were published in English. All of the 23 studies were conducted in China (see Fig. 1).

Description of acupoints
In the cupping group, 22 (95.7%) studies used wet cupping on tender points (Ashi), and 1 [38]  In general, the number of acupoints in acupuncture group is more than that in cupping therapy. According to the included studies of this review, it's interesting to find that acupuncturists often selected acupoints according to syndrome differentiation, as well as the tender points (Ashi), while cupping therapy practitioners were more likely to choose the tender points (Ashi) for retaining cupping.
The pain-related conditions included locomotor pain (cervical spondylosis, lumber disk herniation, scapulohumeral periarthritis and knee osteoarthritis); tissue injury (soft tissue contusion, acute ankle joint); muscle pain (muscles fibrositis and fibromyalgia), neuralgia pain (lateral femoral cutaneous neuritis and postherpetic neuralgia) and acute pain (toothache). Relieving pain was the main goal of the treatments for the included trials, and the study purposes of them were to compare the effectiveness on pain reduction between acupuncture and cupping therapy.

Outcome assessment
Twenty studies (86.95%) used the rate of symptom improvement as the primary outcome, which was defined as the rate of the sum of the cured, markedly effective and effective improvement. Symptom improvement was categorized into four grades (i.e. cured, markedly effective, effective, and ineffective) according to the changes or improvement of the symptoms. 'The cured' was the disappearance of the pain symptoms, 'markedly effective' was the almost disappearance of the symptoms, 'effective' was symptoms alleviation, and 'ineffective' was no changes in symptoms. Besides, 8 [18,19,21,26,29,30,32,40]

Risk of bias assessment
According to our pre-defined methodological quality criteria, no study was evaluated as low risk of bias, 8 [19-21, 26, 29, 31, 32, 40] studies unclear risk of bias, and the remaining 15 studies high risk of bias. The sample size in each study varied from 40 to 407 participants with an average of 56 patients per group. None of the studies specified the sample size calculation. Eleven studies described the details of randomization procedure. Among them, 6 studies [19-21, 26, 29, 32] used a random number table to generate the random allocation. One study [27] was declared as 'randomized controlled trial' , however, it used a quasi-randomization method by dividing participants into two groups based on the sequence of the registration order. Only 1 [26] of the above 11 studies reported using sealed envelope to implement allocation concealment. Three studies [31,39,40] mentioned they used blinding method, in which 2 studies [31,40] reported that they blinded outcome assessors, while the other one did not specify who were blinded. Two studies [21,26] reported the number of dropouts, but none of them used intention-to-treat analysis or other appropriate methods to deal with the missing data (Fig. 2). There were 19 (82.6%) studies reported the comparability of baseline, but only 1 [40] study reported the baseline data of both the outcomes and the demographic characteristics of participants. Nine [19,21,26,29,31,32,34,39,40] (39.13%) studies reported the inclusion criteria, and 9 [18,19,21,26,29,31,32,38,40] (39.13%) studies reported the exclusion criteria. Seventeen (73.91%) studies reported the diagnostic criteria (as we mentioned before). Only one protocol of a study [40] was available which was registered in Clinicaltrials.gov, thus we acquired the protocol to supporting the judgment of reporting bias of this study and it was of low risk.

Results from individual trials
Due to the variations in study quality, participants' characteristics and outcome measures of the included RCTs, data from 14 studies could not be synthesized by meta-analysis. Among them, 11 [21, 25, 29-33, 36-38, 40] studies (78.5%) reported that there was no statistical significant difference between cupping and acupuncture on the rate of symptom improvement for specific conditions, including cervical spondylotic radiculopathy, lumber disk herniation, knee osteoarthritis, toothache, muscles fibrositis, soft tissue contusion, acute lumber sprain, postherpetic neuralgia and fibromyalgia. Three [19,26,39] studies reported cupping might have better effects than acupuncture on increasing the total effective rate in acute ankle joint knee osteoarthritis, acute lumbar sprain and the third lumber vertebral transverse process syndrome according to the VAS scores. Table 2 shows the detailed results of the 14 studies.
Funnel plot analysis could not be conducted due to the insufficient number of included trials in each meta-analysis.

Trial sequential analysis
TSA was conducted with the data from 6 [18][19][20][22][23][24] studies which reported the symptom improvement rate for cervical spondylosis. TSA illustrated that the cumulative Z-curve (blue curve) only across the traditional boundary of 5% significance (horizontal red line) but did not cross the monitoring boundaries (red inward sloping curves), which is needed to obtain firm evidence controlling for the risk of random error. This resulted in a required information size of 2847 participants. TSA of all included trials suggests that about 18 high quality RCTs (2201 participants) were required to confirm possible intervention effect (Fig. 4).

Adverse events
Seventeen (73.92%) out of the 23 studies did not report the outcome on adverse event, and other 6 [24-27, 31, 32] (26.08%) studies reported that there was no adverse event in both groups during the treatment.

Summary of main findings
Twenty-three RCTs with 12 pain-related conditions were included in this review. The majority of the included studies had poor methodological quality according to cochrane risk of bias assessment. Manual or electro-acupuncture were compared with cupping therapy (including wet cupping, moving cupping, flash cupping and retaining cupping) in these studies. Due to the obvious clinical or statistical heterogeneity, only data from ten studies could be pooled in three meta-analyses for cervical spondylosis, lateral femoral cutaneous neuritis and scapulohumeral periarthritis. It should be point out that though some of the included trials reported statistical significant differences between the two therapies according to the symptom improvement rate or pain relieve assessed by scales (such as VAS), according to clinical experience these differences have not achieved clinically meaningful difference between cupping therapy and acupuncture since the differences were too small (RR less than 2 or MD of the VAS less than 3 cm) [41] and the result of TSA for cervical spondylosis also indicated the character available data has not reached a powerful conclusion. All of included studies showed similar effects of these two therapies on increasing symptom improvement rate and other pain reduction for the covered diseases/ conditions. No serious adverse events were reported from both groups.

Strengths and weakness
To the best of our knowledge, this review is the first systematic review comparing cupping therapy with acupuncture for pain-related conditions. The findings of our review might provide evidence-based information of the therapeutic effects of the two therapies to support clinical treatment decisions on pain-related conditions.
There are some limitations of this review. Only 23 RCTs covering 12 conditions were included, thus for most of the conditions we got only 1 or 2 small studies for each to compare the effect of cupping therapy and acupuncture. Furthermore, the sample size of the included participants was too small to provide a powerful evidence. In addition, Fig. 4 Trial sequential analysis on cupping therapy versus acupuncture for symptom improvement rate in patients with cervical spondylosis only three meta-analyses could be conducted with 2-6 studies in each. Though funnel plot analysis could not be conducted, most of the included studies were conducted and published in China, and outcomes of the original studies all got 'positive' conclusions, though some of which did not show any clinically meaningful difference between the two therapies. Therefore, the 'uncertainty principle' might be destroyed due to the 'preference' of the researchers and lead to publication bias.

Implication for clinical practice
Regardless the potential bias and limitations of this review, all of the included studies showed that cupping therapy had similar therapeutic effects with acupuncture in treating pain-related conditions (such as cervical pain, back pain, osteoarthritis, and acute soft tissue injury).
For pain-related conditions, cupping therapy and acupuncture have different choice of acupoints. Based on our review, acupuncture often use two basic acupoints which are tenderness point (Ashi) and the meridians passing through the pain area; while for cupping therapy, wet cupping uses special acupoints (such as Haoyi for cervical spondylopathy) for local pain, and flash cupping combined with moving cupping often use the meridians passing through the pain area for widespread pain.
Considering the similar therapeutic effects of these two therapies, clinical practitioners may consider to choose any of them, based on their own clinical experience and the preference of patients.

Implications for future research
The methodological quality of included studies was general poor, therefore more well designed, conducted and reported RCTs are warranted in the future. According to the Consolidated Standards of Reporting Trials (CONSORT) [42], randomization methods need to be clearly described and fully reported. Although blinding of patients and practitioners might be very difficult for cupping therapy and acupuncture, we strongly suggest the blinding of outcome assessors should be attempted to minimize assessment bias. We also suggest future researchers to select outcome measures based on international consensus for relevant pain conditions. The study protocol is strongly suggested to be registered in authoritative registration plat form [43], such as WHO international Clinical Trial Registration Platform (WHO ICTRP) or Clinicaltrial.com. The reporting of sample size calculation, and using intention-to-treat analysis to deal missing data might be considered.
In addition, due to different treatment durations and variations of the cupping therapy and acupuncture, it is worthy developing cost-effectiveness analyses to further comparing these two TCM non-pharmaceutical therapies, in order to find more economic clinical procedure. Clinical equivalence trials with large sample size will be needed to confirm the non-significant difference of effectiveness between acupuncture and cupping therapy. In addition, compliance, satisfaction and preference of participants to treatment would be considered to employ as outcome measurements in future study.

Conclusion
Currently, there are limited evidence shows that cupping therapy and acupuncture has similar effect on relieving pain and improving other symptoms of pain-related conditions. Larger and rigorously designed RCTs are needed to confirm this conclusion. Economic studied might be considered in the future to further compare the costeffectiveness of cupping therapy and acupuncture.