Major Depressive Disorder (MDD) is a prevalent mental health disorder with a high relapse rate, impacting at least 3.8% of the global population and disproportionately affecting women. Current treatments primarily involve antidepressants, which often have side effects. Complementary therapies like acupuncture are gaining traction, with some studies suggesting it improves depressive symptoms and quality of life, even when combined with medication. Acupuncture involves stimulating acupoints, making acupoint selection and combination crucial for effective treatment. However, the varied acupuncture approaches in existing studies hinder the identification of optimal acupoint combinations for MDD. This study leverages data mining techniques, specifically association rule mining (ARM), network analysis, and hierarchical cluster analysis, to analyze clinical trial data and identify patterns in acupuncture prescriptions for MDD, aiming to inform future research and clinical practice. Data mining offers a robust approach to analyze large amounts of potentially imperfect data to extract meaningful insights, overcoming challenges posed by variations in acupuncture treatment approaches in previous studies.

The introduction section cites several studies highlighting the prevalence, recurrence, and gender disparities in MDD. It also mentions the limitations of antidepressant medication and the growing interest in complementary therapies like acupuncture. A meta-analysis is referenced, demonstrating the potential benefits of acupuncture in conjunction with medication for MDD patients, leading to improved symptom reduction, reduced medication reliance, and enhanced quality of life. The American College of Physicians' inclusion of acupuncture as a complementary therapy in their clinical guidelines is also mentioned, further strengthening the rationale for the current study. The introduction emphasizes the lack of standardization in acupuncture treatments for MDD, a crucial issue motivating the present research to use data mining techniques to find patterns and potentially useful combinations of treatments for this condition.

The study followed a rigorous methodology involving a comprehensive literature search across eight electronic databases (PubMed, Embase, Web of Science, Cochrane Library, CBM, CNKI, Wanfang Data, and CQVIP) from inception to May 28th, 2022. Two independent reviewers conducted the search using keywords related to acupuncture and depression. The inclusion criteria focused on clinical trials (RCTs and CCTs) of acupuncture for MDD with a sample size of at least ten participants and outcomes measured using scales like HAMD, MADRS, and SDS. Reviews, meta-analyses, case reports, and animal studies were excluded. Interventions included acupuncture alone or combined with basic treatments. Data extraction involved two reviewers using a standardized form, with discrepancies resolved by a third reviewer. Missing data were obtained by contacting authors. The study utilized three data mining techniques. Association Rule Mining (ARM) employed the Apriori algorithm in Python 3.8 to analyze acupoint combinations, assessing support, confidence, and lift. Network analysis, using Gephi 0.9.7, visualized acupoint co-occurrence, identifying core acupoints. Hierarchical cluster analysis, using the "dendrogram" package with scipy software, clustered acupoints based on their usage in pattern identification from the literature. The study included data on acupuncture methods (manual, electroacupuncture, warm acupuncture), treatment frequency, and duration.

The study analyzed 664 acupuncture prescriptions from 311 records (292 Chinese, 19 English). GV20 (Baihui), LR3 (Taichong), PC6 (Neiguan), SP6 (Sanyinjiao), and GV29 (Yintang) were the top five most frequently used acupoints. Yang meridian acupoints were used more than Yin meridian acupoints, with the Governor Vessel (GV) showing the highest frequency. Specific acupoints constituted 69.39% of the prescriptions, with five-shu points being the most common type. Lower limbs had the highest frequency of acupoint usage, while the head, face, and neck regions had the most acupoints used. The most frequent acupoint combination was GV29 and GV20. Network analysis identified GV20, PC6 (Neiguan), and SP36 (Zusanli) as core acupoints. Hierarchical cluster analysis grouped acupoints into five clusters based on their usage in different pattern identifications in the literature, suggesting specific therapeutic focuses for each cluster. Manual acupuncture was the predominant technique, with 7 treatments per week and a 42-day duration as the most common treatment course. Analysis of various outcomes indicated different acupoint combinations were associated with improvements in specific symptoms. Finally the study also notes a few adverse events (AEs) which were generally mild and related to the acupuncture procedure itself.

The findings demonstrate that MDD acupuncture treatments are not confined to a single meridian but involve multiple meridians and organs, aligning with Traditional Chinese Medicine (TCM) principles. The prevalence of Yang meridian acupoints, particularly those on the Governor Vessel, suggests a focus on regulating Yang energy, which is consistent with TCM's view of MDD as a Yin deficiency condition. The central role of GV20 (Baihui), along with frequent combinations involving GV29, PC6, and LR3, suggests these acupoints are crucial for MDD treatment, potentially due to their impact on brain regions involved in mood regulation. The frequent use of distal acupoints on the lower limbs supports the TCM concept of distal healing. The predominance of five-shu points highlights their importance. The hierarchical cluster analysis reveals distinct acupoint clusters reflecting various therapeutic approaches for different MDD patterns, aligning with TCM principles. The observed mild adverse events primarily related to the acupuncture procedure itself suggest acupuncture has a relatively favorable safety profile compared to pharmacologic treatments.

This data-mining study revealed patterns and characteristics of acupoint selection in MDD acupuncture treatments. GV20, LR3, PC6, SP6, and GV29 were identified as frequently used acupoints, with a bias toward Yang meridian acupoints and the use of specific acupoints, especially five-shu points. Hierarchical cluster analysis identified distinct acupoint clusters potentially suited to specific MDD patterns. Manual acupuncture was the most common method. Further research is needed to validate these findings and to explore the underlying mechanisms of action.

The study acknowledges several limitations. The heterogeneity of outcome measures and acupuncture stimulation methods across the included studies might have influenced the assessment of the acupoints' therapeutic effects. The quality of some included non-RCT studies was challenging to assess. The study only included publications reporting positive results, which may bias the findings. Finally, while the study explored TCM pattern differentiation, it did not delve into the specific characteristics of acupoints for each MDD syndrome, leaving room for future research.