Acupuncture for Major Depressive Disorder: A Data Mining-Based Literature Study

Major depressive disorder (MDD) is highly prevalent (at least 3.8% globally), recurrent, and associated with significant functional impairment and disease burden. Standard treatments include antidepressants and adjunctive therapies (psychological therapy, cognitive therapy, transcranial magnetic stimulation, light therapy), but medications carry side effects and limitations, prompting interest in complementary therapies. Acupuncture is increasingly used worldwide and may enhance antidepressant efficacy, reduce medication use, and improve quality of life, and is recommended as a complementary option in clinical guidelines. Given wide variation in acupoint selection, stimulation methods, and treatment courses across studies, optimal acupoint prescriptions for MDD are unclear. Data mining offers tools to synthesize acupoint information. This study applies association rule mining, network analysis, and hierarchical cluster analysis to summarize acupoint prescriptions and identify characteristics and patterns of acupuncture for MDD to inform research and clinical practice.

Design: Data mining-based literature study of clinical trials (RCTs and CCTs) of acupuncture for MDD. Search strategy: Eight databases (PubMed, Embase, Web of Science Core Collection, Cochrane Library, CBM, CNKI, Wanfang, CQVIP) were searched from inception to May 28, 2022, using terms related to acupuncture and depression (English and Chinese). Reference lists were screened. Eligibility criteria: Included clinical trials of acupuncture alone or combined with basic treatments (e.g., Western medicine, psychotherapy) for MDD with clinical efficacy outcomes (HAMD, MADRS, SDS). Excluded: reviews, systematic reviews, meta-analyses, case reports, commentaries, clinical protocols, animal studies; non-needling techniques (e.g., Tuina, laser); unconventional acupuncture types (head acupuncture, ear acupuncture, wrist–ankle, abdominal); trials comparing different acupuncture protocols; studies with sample size <10. Only studies reporting positive results were included for acupoint analysis. Study selection and data extraction: Two reviewers (SX, XW) de-duplicated, screened titles/abstracts/full texts, and extracted data using a preset form; disagreements were resolved by a third reviewer (XS). Extracted data included acupoints, meridians, acupuncture methods, outcomes, and adverse effects. Acupoint nomenclature and locations followed GB/T 12346-2021. Data processing and analyses: Frequencies of acupoints and meridians were summarized, including body-part distribution and specific acupoint categories. Association rule mining (ARM) with the Apriori algorithm (Python 3.8) computed support, confidence, and lift for acupoint pairs/triads; co-occurrence matrices were visualized (Gephi 0.9.7), network built with Fruchterman–Reingold layout, and core nodes identified via k-core analysis. Hierarchical clustering: From studies using pattern identification, a matrix of pattern–acupoint weights (frequency normalized by total acupoints per text) was constructed; Ward’s method clustering was performed (scipy 1.8.0), and dendrograms plotted (matplotlib 3.5.1). Safety data (AEs) and treatment parameters (method, frequency per week, duration) were summarized across intervention groups.

- Search and inclusion: 10,738 records identified; after screening, 311 records (292 Chinese, 19 English) yielded 664 acupuncture prescriptions for analysis. - Acupoint frequency: 147 acupoints used (138 meridian points, 9 extraordinary); top 5: GV20, LR3, PC6, SP6, GV29. Among meridians, GV, BL, LR, SP, and ST were most frequent. - Yin/Yang distribution: Of 138 meridian acupoints, 59.42% were Yang meridian points and 40.58% Yin. - Specific acupoints: 102/147 (69.39%) were specific-point categories; most used were five-shu points (23.13%). Crossing points, yuan-primary points, and back-shu points were also frequently used. - Body-region distribution: Highest usage frequency in lower limbs (30.36%; 33 acupoints, 896 uses). The head/face/neck had the greatest number of distinct acupoints (42; 867 uses). - Association rules (top pairs): GV29→GV20 had support 56.41% and confidence 97.78% (lift 1.20). Other frequent associations included GV20 with PC6, LR3 with GV20, SP6 with GV20, HT7 with GV20; SP6→PC6 (support 40.38%, confidence 79.75%, lift 1.40). A triplet example: SP6→GV20,PC6 (support 37.50%, confidence 74.05%, lift 1.42). - Network analysis: ARM produced a network with 30 nodes and 207 edges; average degree 13.8; network diameter 2; average path length 1.524. Three acupoints had degree ≥25: GV20, PC6, and SP36 (reported as core nodes by k-core ≥14). - Clustering (pattern identification): Five clusters were identified. Examples: Cluster 1 (ST36, BL15, SP6, BL20) – strengthening heart and spleen; Cluster 2 (LR3, LR14) – soothing liver and regulating Qi; Cluster 3 (KI3, BL18, BL23) – Yin deficiency of liver and kidney; Cluster 4 (LR2, HT7, PC6, LI4, CV17, GB34, GV20, GV29, SJ6) – tonifying Shen and Qi; others in Cluster 5. - Acupuncture methods and treatment parameters: 330 intervention groups included: manual acupuncture (MA) 210, electroacupuncture (EA) 112, warm acupuncture (WA) 8. Treatment frequency most often 7 times/week (30.55%); duration most commonly 42 days (36.53%). Other frequent durations: 56 days (21.67%) and 28 days (14.55%). - Outcomes by scales: To improve HAMD, GV20, LR3, PC6 were most used; for SDS, PC6, GV20, GV29; for MADRS, Ren meridian points (CV6, CV12, CV4, CV10) were used. To reduce antidepressant use: ST36, SP6, LR3, HT7, GV20, PC6, etc. - Safety: 84 publications reported AEs. Intervention groups had more types of AEs (e.g., needling pain, dizziness, mild burns, subcutaneous hematoma), but for the same AE types, incidence rates were higher in control groups. Reported AEs were mild and self-limiting.

The analysis reveals consistent patterns that can guide standardized acupoint prescriptions for MDD. Frequent use of Governor Vessel and Bladder meridian points, and overall predominance of Yang meridian acupoints, align with TCM theory that MDD is a yin disorder benefitting from yang regulation. Concentration of acupoints on the head/face/neck supports targeting the confluence of Yang meridians and the brain ("capital of the mind"). GV20 emerged as a central node and, with GV29, formed the most common pair; evidence suggests modulation of affective and default mode networks via GV20. Combinations with PC6 and LR3 (both Jueyin meridian) support regulation of yin–yang through distal–proximal pairing. Five-shu points and other specific categories (crossing, yuan-primary, back-shu) were prominently used, consistent with multi-meridian involvement in MDD. Cluster analysis based on pattern identification delineated therapeutically coherent sets matching TCM focuses (e.g., draining liver Qi, tonifying Qi/blood, nourishing liver–kidney Yin, calming Shen). Methodologically, MA was used more than EA, with typical regimens of daily or near-daily sessions over approximately 6 weeks. Safety data suggest acupuncture-related AEs are mild, and incidence of shared AEs may be lower than controls, potentially reflecting reduced medication side effects when combined with acupuncture. Overall, the findings address the research question by extracting principled acupoint selections, combinations, and treatment parameters that can inform protocol optimization and future RCT designs.

Data mining of clinical trials identified key characteristics of acupuncture prescriptions for MDD. The most frequently used acupoints were GV20, LR3, PC6, SP6, and GV29. Yang meridian points, particularly on the Governor Vessel, predominated. Specific acupoints—especially five-shu—were heavily utilized. GV29 combined with GV20 was the most frequent pair; core acupoints were GV20, PC6, and SP36. Hierarchical clustering yielded five acupoint sets with distinct therapeutic focuses aligned with TCM pattern identification. Manual acupuncture was the most common method, with typical treatment frequency of 7 sessions per week and duration of about 42 days. These results can aid standardization of acupuncture prescriptions for MDD. Further high-quality clinical and experimental studies are needed to validate efficacy, refine syndrome-specific prescriptions, and clarify mechanisms.

- Heterogeneity of included studies in outcome measures and acupuncture stimulation methods may confound attribution of effects to specific acupoints. - Inclusion of non-RCT clinical trials with variable and sometimes unclear quality may affect objectivity and generalizability. - Only studies reporting positive results were included for acupoint analysis, introducing publication/reporting bias. - Cluster analysis relied on existing databases; acupoint specificity for MDD syndromes was not explored in depth. Future studies should examine syndrome-specific acupoint efficacy. - Exclusion of unconventional acupuncture types limits generalizability to standard body acupuncture protocols.