Comparative efficacy of acupuncture-related therapy for postmenopausal osteoporosis: protocol for Bayesian network meta-analysis


Osteoporosis is a systemic skeletal disease characterised by decreasing bone mass and microarchitecture degeneration, resulting in increased bone fragility and fracture risk.1 Due to its high incidence and serious complications, osteoporosis has emerged as a global public health issue as the world’s population ages.2 The most prevalent type of osteoporosis, postmenopausal osteoporosis (PMOP), affects women 5–10 years following menopause. According to reports, osteoporosis affects over 50% of postmenopausal women globally.3–5 Women with PMOP not only have a higher risk of fracture, but also require long-term medication, which adversely affects physical and mental health, and imposes a heavy economic burden on families and society.6 7 Currently, the first-line Western medications are bisphosphonates, denosumab, teriparatide, abaparatide, selective oestrogen receptor modulators, hormone replacement therapy, parathyroid hormone, calcitonin, etc.8 The combination of calcium and vitamin D is the basis of PMOP treatment.9 10 However, in recent years, side effects such as adverse cardiovascular events, thromboembolism events, or even increased risk of cervical cancer, breast cancer and ovarian cancer,11 12 and individual differences in medication use have received increasing attention and research. Also in clinical practice, the patient’s lack of knowledge of the disease and treatment plan, safety issues and inconvenient dosing regimens may lead to poorer adherence to medications, which are associated with an increased risk of fracture. Therefore, clinical practitioners and researchers are continuously exploring effective complementary therapies,13–16 including acupuncture, tai chi, exercise and Chinese herbal medicines.

Acupuncturerelated therapy, as one of the most widely accepted complementary therapies, has achieved good clinical efficacy worldwide due to its clear efficacy, long-lasting effect and low adverse effects. In recent years, many clinical trials and meta-analyses of acupuncture-related therapies for the treatment of osteoporosis have been published.17–19 Through these studies, we understand the mechanism by which acupuncture increases bone density and improves osteoporosis. With bone loss, microstructure destruction, bone mechanical property decline and microfractures, patients may have low back pain, spinal deformation in severe cases, and even osteoporotic fractures and other serious consequences.20 21 Meanwhile, the analgesic effect of acupuncture-related therapy has been internationally recognised.22 23

Previous systematic reviews and meta-analyses have confirmed that a single acupuncture-related treatment can improve bone mineral density (BMD) and relieve pain in patients with osteoporosis,24–27 such as acupuncture, electroacupuncture, moxibustion, warm needling, acupoint catgut embedding, etc. However, only comparisons of acupuncture therapy with other therapies or placebo, other than directly comparing different acupuncture-related treatments, have been reported. Therefore, it is still worth discussing which method can achieve the best efficacy in treating PMOP. In contrast to the traditional pairwise meta-analysis, network meta-analysis (NMA) is able to summarise direct as well as indirect evidence and evaluate the relative efficacy of multiple treatment comparisons. More importantly, NMA can provide rankings of these treatments according to the effectiveness. We will compare the effectiveness of commonly used clinical acupuncture-related therapies with each other by using NMA, and rank the efficacy in the hope of finding the optimal treatment to provide some clinical guidelines in the treatment of PMOP.

Methods and analysis

Study registration

The protocol for this NMA is based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocols,28 and the registration number is CRD42023401003.

Patient and public involvement

No patients or members of the public were involved.

Inclusion criteria

Research types

This study will collect all randomised controlled trials (RCTs) on acupuncture-related therapies for patients with PMOP. There will be no restriction on blindness, publication status and location, but the languages would only be Chinese and English.

Research objects

The trial subjects will be postmenopausal patients with osteoporosis who meet the relevant diagnostic criteria according to the European guidance for the diagnosis and management in 201829 and the 2021 position statement of the North American Menopause Society.30

Intervention measures

The intervention will be conventional Western medicine treatment in the control group, and the treatment group could only be treated with a single acupuncture-related therapy or a single acupuncture-related therapy combined with conventional Western medicine.

Outcome indicators

Primary outcome

The primary indicators will focus on the overall clinical effectiveness rate, BMD as well as a Visual Analogue Scale (VAS) score. The overall clinical effectiveness rate, based on the criteria of Chinese medicine clinical evidence points for the clinical standard used to judge the efficacy of acupuncture-related therapies, will be divided into four levels: (1) clinically cured, (2) markedly effective, (3) effective and (4) invalid. The overall clinical effectiveness rate will be calculated as: the overall clinical effectiveness rate (%)=[(number of patients clinically cured+markedly effective+effective)/number of patients]×100%.31 BMD is the gold standard for assessing the severity of osteoporosis. In 1994, the WHO established the measurement standard of BMD, which is determined by the strength and density of the bone.32 The lower the value, the more severe the osteoporosis is. According to the diagnostic criteria,29 30 the diagnosis of osteoporosis is usually based on the BMD of the spine, femur and femoral neck. We will perform subgroup analyses based on pretreatment and post-treatment changes in BMD at different sites. Besides, the VAS score will be used to estimate the pain measure.

Secondary outcome

The secondary indicator will involve adverse reactions, in order to understand and compare the safety of different acupuncture-related therapies and conventional Western medicines for PMOP.

Retrieval strategy

Electronic databases will be searched for RCTs of acupuncture-related therapies for postmenopausal osteoporosis, and the Chinese literature will be searched in four databases: the China Biomedical Literature Database, China National Knowledge Infrastructure, VIP Database and Wanfang Database. In addition, we will search the English literature through PubMed, Web of Science, Embase as well as Cochrane Library. Search terms will include “acupuncture”, “electroacupuncture”, “moxibustion”, “warm needle acupuncture”, “catgut implantation at acupoint”, “acupoint sticking therapy”, “postmenopausal osteoporosis” and their synonyms. Furthermore, it is anticipated that the search will employ both subject phrases and free words. The search period will take place between 1 January 2002 and 31 December 2022. The search and screening process of literature is detailed in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow chart (figure 1).

Figure 1
Figure 1

Flow diagram of the study selection process. PMOP, postmenopausal osteoporosis.

Literature selection and data collection

All retrieved studies will be imported into NoteExpress; duplicate studies will be eliminated, then followed by initial screening by manually reading the titles and abstracts of the remaining literature, and ultimately downloading the full text for full-text reading to include those that are eligible. For some missing data, we will try to contact the first author or corresponding author of the article by email. To extract material that satisfies the inclusion requirements, which include patient information, the details of the intervention, the outcome indicator and other factors, two researchers will independently conduct the literature screening and then confirm their findings with one another. A third investigator of our team will organise the discussion and decide if there are any doubts or disagreements that cannot be resolved after communication. These data will be extracted and formed into the following: study characteristics (author, publication year, mean age, sample size, etc), intervention and control measures (acupoints, operation, treatment duration and frequency), diagnostic criteria, outcomes, methods of randomisation and blinding to demonstrate the basic characteristics of the included literature. Moreover, NoteExpress and Microsoft Excel V.2019 will each be used to manage all references and extract data, respectively.

Evaluation of literature quality

The high quality of the literature will be evaluated in accordance with the Cochrane Collaboration’s tool of Systematic Review,33 from the seven aspects of random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data and selective reporting and other bias that may potentially affect authenticity. The quality of the literature can be evaluated at three levels: ‘low risk of bias’, ‘unclear risk of bias’ and ‘high risk of bias’. A ‘low risk of bias’ means that the study meets all criteria, ‘unclear risk of bias’ means that the study does not supply enough information to make a judgement and ‘high risk of bias’ indicates that the study does not satisfy any of the criteria. Two researchers will assess the risk of bias independently, and then cross-check to ensure there are no errors. Disagreements will be resolved by consulting a third investigator. Review Manager V.5.4 will be used to make the risk of bias diagram.

Data analysis

Pairwise meta-analysis

We will primarily perform a pairwise meta-analysis using the Review Manager V.5.4 and Stata V.14.0 software. The effect values of 95% CIs will be measured by the Stata V.14.0 software, and it will include the OR for dichotomous data and the mean difference for continuous data. The heterogeneity of each pairwise comparison will be assessed by the Q test and the I² through RevMan V.5.4. Simultaneously, the presence and magnitude of heterogeneity will also determine which effect model to use to merge the data. If the I2<50% and p>0.1, we will choose the fixed-effects model; otherwise, the random-effects model will be selected.

Network meta-analysis

Bayesian NMA will be conducted to compare the effects of different acupuncture-related therapies. Since the outcome indicators we selected include both dichotomous and continuous-type variables, the number of events and the total number of samples will be used as the effect values of dichotomous-type variables for the statistical analyses, reported using OR with 95% CIs. On the other hand, the BMD and VAS scores as continuous-type variables will be statistically analysed using the mean and SD as effect values, and using standardised mean difference with 95% CIs. This process will be run by Stata V.14.0.

Furthermore, we will use Stata V.14.0 for data analysis and graph drawing: first, we will use the command ‘networkplot’ to draw network evidence plots for showing the quantitative relationship between the individual interventions. Second, if the evidence network plot forms a closed loop, then we will conduct the inconsistency test. Third, we will use the command ‘network forest’ to draw a forest plot to obtain the effect values and 95% CIs. Fourth, we will enter the ‘netleague’ command to conduct an NMA and draw an ‘inverted triangle’ diagram based on the results of pairwise comparison. Fifth, we will use the ‘sucra prob’ command to rank the efficacy of different interventions and draw a cumulative probability graph, and use the area under the curve (surface under the cumulative ranking curve) to indicate the superiority or inferiority of each intervention. The larger the value, the better the efficacy of the intervention. In order to make the results more objective, we will further corroborate our results with minimal or partial contextualisation.34–36 Sixth, the funnel plot will be created using the ‘netfunnel’ command to assess the publication bias and small sample of the included literature. We will conduct a narrative review and summarise the evidence, if the available data are not suitable for synthesis.

Subgroup analysis and assessment of heterogeneity

First, we will perform a traditional pairwise meta-analysis of all comparable outcome indicators, examining and evaluating their consistency and heterogeneity. The I2 statistic and p values will be applied to assess the heterogeneity across all individual studies. To obtain more reliable estimates of the effect, I2 greater than or equal to 50% and p<0.1 will be used as thresholds to indicate significant heterogeneity. If the heterogeneity is small, we will choose the fixed-effects model; if not, we will deal with it by the following methods: (1) checking whether the original data are correct and the accuracy of the data extraction method; (2) conducting the source of heterogeneity analysis through subgroup analyses and meta-regression; (3) conducting sensitivity analysis to find which studies generated the heterogeneity. If necessary, subgroup analyses will be performed according to the age of the patients, the different types of Western medicines in the control group and the BMD of the different sites. Meta-regression analyses will be performed according to the sample size and the duration of treatment.

Assessment of publication bias

When more than 10 articles are included, we are going to use Stata V.14.0 to draw the comparison-adjusted funnel plots to assess the publication bias and the small study effects in our NMA. The assessment will be based on whether the funnel plot is symmetrical or not. If the funnel plot is found to be asymmetrical, we will attempt to explain the asymmetry of the funnel plot. If small-study effects are absent, the funnel plot should be symmetrical along the midline around the regression line with a p>0.05; otherwise, researchers will explore this further with appropriate network meta-regression and models.

Sensitivity evaluation

We will explore the effect of certain low-quality methodological studies, or non-blinded studies, on the total effect by excluding them. The results of the NMA will be proved relatively stable or with lower sensitivity, if the results of the NMA do not change at this point.

Grading the quality of evidence

We will conduct GRADE evaluation of the included studies according to their quality of evidence, by applying the GRADEpro system from five aspects of risk of bias, inconsistency, indirectness, inaccuracy and publication bias, which will be divided into four levels: high, medium, low or very low.

Ethics and dissemination

Ethical approval is not required for this protocol as no private information of participants is involved. The results of this study will be disseminated in peer-reviewed journals or conference presentations. Important protocol amendments will be documented and updated in PROSPERO.

This post was originally published on