Analysis of Acupoint Selection and Combinations in Acupuncture Treatment of Endometriosis: A Protocol for Data Mining

Endometriosis is a chronic, inflammatory, estrogen-dependent gynecological disease commonly seen in women of reproductive age. It is characterized by the presence of active endometrial tissue outside the uterine cavity. 1–3 Endometriosis affects approximately 10–15% of reproductive-aged women worldwide. Moreover, the prevalence of disease is even higher among women with unexplained infertility. 4–6 The clinical manifestations of endometriosis are diverse. The most common symptoms include dysmenorrhea, pelvic pain, dyspareunia, and infertility. 4 These symptoms severely impair patients’ quality of life, work capacity, and psychological well-being, and also impose a substantial economic burden on both patients and society. 7 , 8 Currently, pharmacological therapy (such as nonsteroidal anti-inflammatory drugs (NSAIDs), progestins, combined oral contraceptive (COC), and gonadotropin-releasing hormone (GnRH) agonists and antagonists) and laparoscopic surgery are important treatment modalities for endometriosis. 9 However, these approaches all carry certain side effects and risks of recurrence. 4 , 10–15 At the same time, accurate diagnosis is equally important. Recent studies have explored non-invasive markers for endometriosis; Yildiz et al reported that systemic inflammatory markers were associated with endometrioma diagnosis and Douglas infiltration, potentially reducing the need for histopathological examination. 16 These diagnostic advances, together with the limitations of current therapies, underscore the need for safe, non-pharmacological options. Acupuncture is a safe and natural non-pharmacological therapy. A large body of research has demonstrated that acupuncture is significantly effective in relieving endometriosis-associated pain and improving patients’ quality of life. Systematic reviews and meta-analyses have further confirmed that acupuncture is an effective treatment for endometriosis. 17 However, previous systematic reviews and meta-analyses have focused on whether acupuncture as a whole improves symptoms, rather than on how acupoints are selected and combined. Evidence for symptom improvement is therefore not equivalent to evidence for specific acupoint combinations. To our knowledge, no study has yet systematically examined the acupoint selection and combination patterns used in acupuncture treatment of endometriosis. This gap is clinically relevant: in practice, there are significant differences in acupoint selection—different practitioners choose different acupoints based on their own theoretical understanding and clinical experience—and treatment protocols lack unified standards. Even in published clinical studies, the selection of main acupoints, adjunct acupoints, and their combinations differ from one another, which limits the comparability of research results and makes it difficult to establish a standardized clinical acupuncture treatment protocol. Therefore, it is necessary to use existing technologies to analyze the acupoint selection and compatibility patterns in acupuncture treatment for endometriosis, so as to summarize these patterns and provide references for optimizing treatment protocols. Data mining technology is an efficient data analysis method that can extract valuable information hidden behind large amounts of redundant data from massive literature, thereby effectively sorting out the application patterns of acupuncture points and revealing the intrinsic associations between acupoint combination schemes. In recent years, data mining methods have been successfully applied to research on acupuncture treatment for various pain conditions, including piriformis syndrome, 18 migraine, 19 carpal tunnel syndrome, 20 and trigeminal neuralgia. 21 More recently, data-mining approaches have also been used to explore acupoint selection patterns in acupuncture treatment for gynecological conditions, including infertility, 22 and primary dysmenorrhea. 23 However, the application of a combined data-mining approach—integrating frequency analysis, association rules, factor analysis, and decision-tree analysis—to the study of acupoint selection in gynecological disorders remains limited. To address this gap, this study aims to make full use of data mining technology to conduct an in-depth analysis of the acupoint application patterns in acupuncture treatment for endometriosis, thereby providing references for more standardized treatment protocols for the clinical practice of acupuncture for endometriosis.

This study protocol has been registered in INPLASY (registration number: INPLASY202690092). This study will adhere to the PRISMA 2020 statement. 24 , 25 We will systematically search Chinese and English databases, with the search timeframe covering from the inception of each database to September 30, 2026. The literature search has not yet started; it is planned to begin on October 1, 2026, and to be completed by mid-October 2026. The final search date and the complete search results will be reported before data extraction and synthesis. The databases searched include Web of Science (WOS), PubMed, Embase, Cochrane Library, Wanfang, Chongqing VIP database, China National Knowledge Infrastructure (CNKI), and the Chinese Biomedical Literature Database (CBM). Only literature in Chinese and English will be included. The search strategies will be adjusted according to the specific retrieval characteristics of each database. Table 1 presents the detailed search process for the WOS database. Table 1 Search Strategy for Web of Science Database No. Search Terms #1 ((((((TS=(Endometriosis)) OR TS=(Endometriotic)) OR TS=(Endometriosis-associated pain)) OR TS=(Endometrioma)) OR TS=(Ovarian endometriosis)) OR TS=(Secondary dysmenorrhea) #2 ((((((((((((((((((((((((((((((((((((((((((((((((((TS=(Acupuncture)) OR TS=(Acupuncture therapy)) OR TS=(Auricular acupuncture)) OR TS=(Ear acupuncture)) OR TS=(Fire needling)) OR TS=(Scalp acupuncture)) OR TS=(Wrist-Ankle Acupuncture)) OR TS=(Cheek acupuncture)) OR TS=(Abdominal acupuncture)) OR TS=(Navel needling therapy)) OR TS=(Acupotomy)) OR TS=(Moxibustion)) OR TS=(Warm needling)) OR TS=(Medicinal cake moxibustion)) OR TS=(Ginger separated moxibustion)) OR TS=(Moxibustion with salt)) OR TS=(Salt moxibustion)) OR TS=(Separated aconite cake moxibustion)) OR TS=(Medicinal vesiculation)) OR TS=(Tianjiu)) OR TS=(Thunder-fire moxibustion)) OR TS=(Grain-sized moxibustion)) OR TS=(Spreading Moxibustion)) OR TS=(Fire dragon moxibustion)) OR TS=(Huolong moxibustion)) OR TS=(Cupping)) OR TS=(Fire cupping)) OR TS=(Dry cupping)) OR TS=(Wet cupping)) OR TS=(Fire Dragon Cupping)) OR TS=(Bloodletting cupping)) OR TS=(Blood-letting)) OR TS=(Pricking blood)) OR TS=(Pricking bloodletting)) OR TS=(Electroacupuncture)) OR TS=(Electric Stimulation Therapy)) OR TS=(Transcutaneous Electrical Acupoint Stimulation)) OR TS=(Acupoint application)) OR TS=(Acupressure)) OR TS=(Acupoint Injection)) OR TS=(Ear acupressure)) OR TS=(Ear acupressure with seeds)) OR TS=(Scraping therapy)) OR TS=(Guasha)) OR TS=(Midnight-noon and ebb-flow doctrine)) OR TS=(Midnight-noon ebb-flow)) OR TS=(Ziwu liuzhu)) OR TS=(Eight methods of intelligent turtle)) OR TS=(Linggui bafa)) OR TS=(Catgut implantation at acupoint)) OR TS=(Point Embedding Therapy) #3 #1 AND #2 Search Strategy for Web of Science Database Inclusion criteria will be as follows: (1) Patients must be definitively diagnosed with endometriosis. (2) The study should use acupuncture therapy as the primary intervention, which could be applied alone or in combination with other treatment modalities. (3) Studies comparing the efficacy of different acupuncture techniques for the treatment of endometriosis will also be included. (4) Each trial must include more than ten participants. Exclusion criteria will include the following: (1) Non-clinical trial literature such as reviews, systematic reviews and meta-analyses, animal experiments, trial protocols, case reports and so on. (2) Conference papers, dissertations, and other literature that has not undergone comprehensive and rigorous peer review by experts in the field. (3) Literature that presents only laboratory or physiological data. (4) Studies with incomplete acupuncture prescriptions or incomplete access to the literature. (5) Studies with an unclear or unconfirmed diagnosis of endometriosis. We will use NoteExpress 4.3 reference management software to establish a literature database for acupuncture treatment of endometriosis. We will first perform duplicate detection using author, year, title, and journal as duplicate-check fields, and remove duplicates. Subsequently, we will further conduct manual screening to comprehensively eliminate duplicate literature. The above procedures will be carried out independently by Linglin Dai and Yin Zhang, and finally be cross-checked to ensure the uniqueness and accuracy of the included literature. All researchers will receive complete training in systematic evaluation. After duplicate records were removed, Linglin Dai and Yin Zhang will independently review and evaluate the titles and abstracts of all literature retrieved through the search, and exclude articles that clearly do not meet the inclusion criteria (such as review articles, trial protocols, dissertations, etc). The full texts of the remaining literature will be collected and further reviewed to exclude publications with incomplete retrieval, missing acupoint information, or any publications deemed irrelevant. This work will be completed independently by the two researchers. In case of any disagreement, consensus will be reached through discussion with the third author (Lili Wang). Finally, all remaining literature will be re-assessed for applicability according to the aforementioned inclusion and exclusion criteria, to ensure the accuracy of the sources of the acupuncture prescription data. Figure 1 illustrates the literature screening process of the two authors. Figure 1 Flowchart of Literature Screening. A flowchart illustrating the literature screening process for research inclusion in data mining. The flowchart outlines the process of literature screening for research inclusion. It begins with the identification phase, where records are identified from databases such as WOS, PubMed, Embase, Cochrane, VIP, Wanfang, CNKI and CBM, as well as other sources. Duplicate records are removed. In the screening phase, records are screened by title and abstract. Exclusions are made for unrelated literature, studies with fewer than 10 participants, non-clinical trial literature, animal experiments, trial protocols, case reports, conference papers, dissertations and literature lacking comprehensive peer review or presenting only laboratory data. Records are then assessed for eligibility, excluding studies with incomplete acupuncture prescriptions or literature acquisition and unrelated publications. Finally, eligible research is included in data mining. Flowchart of Literature Screening. We will extract data from the remaining literature and enter it into WPS Spreadsheet (version 12.1.0.26375) to establish an acupoint prescription database for acupuncture treatment of endometriosis. Prescriptions containing adjunct acupoints will be structured as effective data combining “one group of primary acupoints with one group of adjunct acupoints.” All meridian acupoints and auricular acupoints will be standardized according to the latest national standards, GB/T 12346–2021 “Nomenclature and Location of Meridian Acupoints” and GB/T 13734–2008 “Nomenclature and Location of Auricular Acupoints,” respectively. In addition, we will supplement basic information such as the anatomical location of the meridian acupoints, the corresponding meridian, specific attributes, and so on. To ensure full data entry and comprehensive statistical analysis, in cases where data are missing, we will contact the corresponding author by Email to obtain the original or supplementary data. If the corresponding author cannot be reached or is unable to provide the requested information, we will conduct the analysis based solely on the data that are available. The Cochrane Risk of Bias Assessment Tool (RoB 2) covers five domains: the randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selective reporting of results. It is an authoritative tool for assessing the risk of bias in specific outcomes of randomized controlled trials (RCTs). 26 Two independent reviewers, Linglin Dai and Yin Zhang, will each evaluate the methodological quality of the included RCTs using the RoB 2 tool and reach conclusions of “low risk,” “high risk,” or “some concerns.” If non-randomized studies of interventions (NRSIs) are included, the ROBINS-I tool will be used to assess their risk of bias. In addition, the STRICTA (Standards for Reporting Interventions in Clinical Trials of Acupuncture) checklist will be used to assess the reporting quality of acupuncture interventions. 27 After the independent assessments are completed, the two reviewers will conduct a consistency check. Any disagreements will be resolved through discussion and consultation with the third author (Lili Wang). We will use WPS Spreadsheets (version 12.1.0.26375) to enter the acupuncture point prescriptions of all included literature. Subsequently, based on the database and its pivot table generation function, we will perform descriptive statistics and quantitative analysis on multiple dimensions, including the frequency of acupoint use, the body region of the acupoints, the meridian to which the acupoints belong, and the attributes of specific acupoints. The results of these analyses will be reported in the future full study after data extraction is completed. According to Price’s law, the identification of high-frequency acupoints will adopt the formula M = 0.749 × (Pmax) 1 / 2 , where Pmax is the highest frequency of use among the acupoints included in the studies. If more than 50 high-frequency acupoint results are obtained, a second method will be applied: calculating the average frequency of use of all acupoints (total frequency/total number of acupoints), and defining acupoints with a frequency ≥ this average value as high-frequency acupoints. 19 We will use bar charts to visualize the high-frequency acupoint results. SPSS Modeler is a powerful data mining and predictive analytics tool launched by IBM. We will use this tool along with the Apriori algorithm to perform association rule analysis on high-frequency acupoints. Support degree indicates the probability that a combination of acupoints appears simultaneously. The confidence degree indicates the conditional probability that the first item of the acupuncture prescription appears in the record, and the second item appears at the same time. The minimum support and minimum confidence thresholds will be determined by a prespecified, objective procedure. A series of candidate thresholds (support: 5%–20%; confidence: 50–90%) will be tested. The final thresholds will be selected as the lowest values at which the number of generated rules reaches a stable plateau (ie, further lowering the threshold no longer substantially increases the number of interpretable rules). The procedure and the final thresholds will be reported in the future full study. Sensitivity analyses will be performed to assess robustness. The association rules generated will be analyzed after data extraction is completed, and the results will be reported in the future full study. Subsequently, we will construct a complex network diagram in a point-line format to visualize the compatibility relationships among acupoints. SPSS Statistics is a comprehensive statistical analysis tool developed by IBM, widely applied in the field of data mining. In this study, this tool will be used to conduct exploratory factor analysis. Prior to the analysis, the Kaiser-Meyer-Olkin (KMO) test and Bartlett’s test of sphericity will be performed on the data. When the KMO value is greater than 0.6 and the P-value from Bartlett’s test of sphericity is less than 0.05, the data are considered to exhibit sufficient intercorrelation and are deemed suitable for exploratory factor analysis. Conversely, the analysis will be deemed inappropriate for application to this dataset. In addition, a minimum subject-to-item ratio of 5:1 will be required, and the total number of acupoint prescriptions should be no fewer than 100, consistent with standard recommendations for exploratory factor analysis. Cluster analysis is an unsupervised machine learning method that, without a predetermined classification standard, automatically groups similar samples into the same category based on the intrinsic characteristics of the data itself. For this analysis, we will continue to use the SPSS Statistics software. To assess model stability, bootstrap resampling (1,000 replications) will be performed. Only clusters that appear in more than 70% of the resamples will be considered stable and retained for interpretation. Later, we will construct a cluster analysis dendrogram to visually display the clustering relationships among acupoints. Decision tree analysis is a supervised machine learning method that simulates the human decision-making process by constructing a tree-like model to perform classification or prediction on data. In this study, SPSS Statistics software and the CHAID algorithm will be used to build a decision tree model for acupoint data. The target (dependent) variable will be defined as the presence of a core acupoint combination (1 = present, 0 = absent), that is, whether a given acupoint appears in a high-frequency combination identified by the association-rule analysis. High-frequency acupoints and their attributes (eg, meridian, body region, specific acupoint category) will serve as predictor variables. This definition ensures that the supervised learning task has a clearly specified outcome. To prevent model overfitting, 10-fold cross-validation will be employed to evaluate model performance. The mean accuracy and standard deviation across the 10 folds will be reported to reflect model stability. Finally, this study will construct a decision tree model to present the hierarchical relationships among acupoints in a visual form, and to screen out potentially relevant acupoint combinations for treating endometriosis.

The results of this study will identify commonly used acupoints, recurring acupoint combination patterns, and candidate protocols for acupuncture treatment in patients with endometriosis. These findings will provide a basis for clinical decision-making and for the selection of acupoints in subsequent well-designed prospective randomized controlled trials, and will help promote more informed choices by both doctors and patients.

Endometriosis, as an estrogen-dependent disease, is a common gynecological condition in women of reproductive age. 28 Its characteristic feature is that ectopic endometrial tissue undergoes repeated bleeding and shedding in response to cyclic fluctuations in estrogen levels, yet is unable to be expelled from the body normally, resulting in the persistence and continuous progression of ectopic lesions. Pain, such as dysmenorrhea and chronic pelvic pain, is the most prominent symptom of endometriosis, affecting approximately 60–80% of patients. 28 The pathogenesis of endometriosis-associated pain is complex. The currently prevailing view holds that local inflammation caused by ectopic lesions, abnormal innervation, hormonal influences, immune dysregulation, and peripheral and central sensitization are all involved, and collectively lead to the generation, amplification, and persistence of pain related to endometriosis. 29 Moreover, even after the original lesions are removed, central sensitization may persist, leading to the continuation of pain. 30 The persistence of pain not only causes physical discomfort but also triggers a range of issues such as anxiety, depression, and sleep disturbances, which severely interfere with the patient’s physical and mental health as well as daily life. 7 , 31 Currently, mainstream treatment options include NSAIDs, hormonal preparations, and surgical resection. NSAIDs can only provide symptomatic relief of pain and cannot delay the progression of endometriosis. Moreover, long-term use of NSAIDs increases the risk of myocardial infarction, stroke, hypertension, gastrointestinal perforation, and other adverse events. 10 , 11 Hormonal preparations can also effectively relieve pain. However, the adverse reactions such as breakthrough bleeding, mood disorders, weight gain, and decreased libido still warrant attention. 12 , 13 In addition, GnRH analogues, as second-line therapy for endometriosis, are unsuitable for patients with fertility desires due to their ovulation-suppressing properties. And long-term use is associated with side effects, including decreased bone mineral density, hot flashes, and sleep disturbances. 32 As for laparoscopic surgery, although it provides an intervention option for patients refractory to medical therapy, it is associated with a relatively high postoperative recurrence rate and potential damage to ovarian reserve function. 14 , 15 Therefore, in the comprehensive management of endometriosis, a chronic disease requiring long-term care, non-drug, safe, natural treatment strategies with minimal adverse effects are demonstrating increasingly important clinical value. Acupuncture originates from China. As an important component of traditional Chinese medicine, it is widely applied in the field of pain treatment and management. 20 It has demonstrated significant efficacy in relieving endometriosis-associated pain, which has been confirmed in multiple randomized controlled trials and systematic reviews. 6 , 33 , 34 Studies have shown that acupuncture can effectively reduce the levels of VEGF, NGF, and BDNF, downregulate the serum inflammatory factors IL-6, TNF-α, and PGE2, suppress peripheral inflammatory responses, and simultaneously improve endometrial hemodynamics, thereby exerting analgesic effects. 35 Furthermore, pre-clinical mechanistic evidence from an animal study suggests that the combined use of acupuncture and medication can downregulate the expression of CLEC5A mRNA and protein in both ectopic and eutopic endometrial tissues, reduce the protein levels of IL-1β and NLRP3, thereby significantly shrinking endometriotic cysts, raising the pain threshold, and consequently alleviating pain. 36 The mechanisms underlying acupuncture’s therapeutic effects are characterized by multiple targets and multiple dimensions. Additionally, acupuncture may also exert its effects through the regulation of targets such as dynorphin, β-endorphin, and prostaglandins, as well as through central pain processing pathways. 37 , 38 However, acupuncture treatment protocols—particularly acupoint selection and compatibility—exhibit significant variation across different studies. This heterogeneity directly compromises the reproducibility of research findings, undermines the comparability of treatments between studies, and largely limits the standardized promotion of acupuncture in the clinical management of endometriosis. Thus, in order to explore the rules of acupoint selection and compatibility, this study plans to adopt data mining technology. This technology, which integrates multiple disciplinary methods including statistics, machine learning, database technology, and visualization, can extract implicit and valuable effective information from massive and disordered published clinical literature. Through descriptive statistics, we can preliminarily mine the high-frequency acupoints, key meridians, specific acupoints, and core body regions most commonly used in acupuncture treatment of endometriosis. Association rule analysis was proposed by Agrawal et al, 39 Through a rule-based unsupervised machine learning algorithm, this technology can mine the associations among acupoints from large-scale acupuncture prescriptions, thereby helping to reveal the compatibility relationships among different acupoints, screen out high-frequency compatibility combinations, and provide data support for clinical acupoint selection in acupuncture treatment. The Apriori Algorithm, one of the typical algorithms for association rule analysis, can mine frequently occurring strong-association acupoint combinations in the database through an iterative strategy of layer-by-layer search. 40 The concept of exploratory factor analysis was first proposed by Charles Spearman in 1904. 41 This analysis is a multivariate statistical method that relies on the correlation coefficient matrix among variables. By reducing the dimensionality of variables, it identifies the latent dimensions (acupoint group factors) underlying a large number of acupoints. Cluster analysis is a classic pattern recognition method that relies on the distance or similarity between samples. By classifying samples, it groups similar acupoint prescriptions into distinct clusters. This technique helps simplify the complexity of data and extract meaningful groups from vast amounts of information, which holds significant application value in multi-parameter data analysis. Decision tree analysis is a knowledge-based data mining method. Framed within a tree structure and employing a recursive tree-shaped classification model, this analysis extracts combination rules from large-scale, redu ndant clinical acupuncture data. This method is particularly suited for making informed decisions based on existing knowledge. In this model, high-frequency acupoints serve as index factors. The root node represents the starting point of the decision process. And each child node corresponds to a classification task. Then, the data are partitioned into two or more branches according to the importance hierarchy of acupoints. Each branch can be further iteratively divided until the leaf nodes yield the final prescription categories. Each path from the root node to a leaf node represents a single acupoint compatibility rule. 42 , 43 Ultimately, the decision tree model presented in a visualized form will help clinicians identify core acupoints, understand the hierarchical relationships among acupoints, and quickly screen out potentially revelant acupoint compatibility protocols in relevant fields. Given that different data mining methods have different analytical perspectives and emphases, their combined application can provide multi-faceted, multi-dimensional references for clinical acupoint selection and compatibility in acupuncture treatment. Meanwhile, this combined approach can avoid the result bias caused by a single method and enhance the credibility of the overall research conclusions. Therefore, this study will comprehensively apply the aforementioned methods, aiming to identify frequently used acupoints and potentially relevant acupoint combinations for acupuncture treatment of endometriosis, provide data support for clinical decision-making in acupuncture for endometriosis, and offer references for acupoint selection and protocol optimization in future prospective randomized controlled trials.

This study has several limitations. First, as a data-mining study based on published literature, it is inevitably affected by publication bias, since studies reporting positive results are more likely to be published than those reporting negative or neutral results. Second, acupuncture techniques vary considerably across studies—including differences in needle type, depth, manipulation, stimulation frequency, and treatment duration—which may introduce heterogeneity that cannot be fully controlled in the analysis. Third, the comparators and outcome measures used in the included trials differ substantially, which limits the comparability of acupoint prescriptions across studies. Fourth, the frequency and co-occurrence of acupoints reflect patterns in the published literature and cannot establish comparative efficacy; acupoints that are used more frequently are not necessarily more effective. Fifth, acupuncture trials may include heterogeneous populations with endometriosis, adenomyosis, or both, and adenomyosis is a distinct and potentially confounding gynecological condition that may influence pain severity and treatment response. 44 This should be considered when extracting and interpreting clinical characteristics across the included studies. Finally, because this is a protocol, the findings have not yet been empirically validated and should be interpreted as hypothesis-generating rather than definitive.