Analysis of Acupoint Selection and Combinations in Acupuncture Treatment of Trigeminal Neuralgia: A Protocol for Data Mining [Response to Letter]

Dear editor

Thank you for your thoughtful letter¹ regarding our protocol entitled “Analysis of Acupoint Selection and Combinations in Acupuncture Treatment of Trigeminal Neuralgia: A Protocol for Data Mining.”² We greatly appreciate your insights, which will help strengthen the methodological rigor of our study. We have carefully considered each of your suggestions and would like to respond as follows:

Regarding the management of temporal and terminological heterogeneity, first of all, what we included are not studies from the past 30 years, but studies from the establishment of the database to May 2025 (approximately the past 70 years). However, we agree with your insights. Due to changes in diagnostic criteria, such bias may indeed arise. In fact, many studies (including meta-analyses and similar data mining studies^3–7) have adopted this approach, aiming to cover a broader range of data. Even so, your opinions are indeed worthy of careful consideration. In future studies, we may include studies separately before and after the changes in diagnostic criteria, conduct subgroup analyses, and compare their similarities and differences. In addition, we will include the management of acupoint aliases in the complete study, and elaborate on the alias mapping rules of acupoints to improve reproducibility.

Regarding “Uncontrolled Confounding from Synergistic Therapies”, it must be acknowledged that, on the basis of comprehensive analysis, further conducting stratified subgroup analysis between pure acupuncture and acupuncture combined with other therapies such as drugs is of significance, as this can reduce bias. In fact, this also represents a shift in research trends in similar articles over the past year. Clinically, acupuncture is mostly used in combination with drugs; thus, stratified subgroup analysis can more effectively demonstrate the association between drugs and acupoint selection. Therefore, given that the present study is the most fundamental one in this research field, stratified subgroup analysis according to intervention methods will be further conducted after the completion of this study.

Regarding the KMO threshold for exploratory factor analysis, first, it is true that a KMO value between 0.50 and 0.59 indicates that the data is “barely suitable” for factor analysis. However, numerous studies have shown that in certain specific research fields, factor analysis remains feasible when the KMO value is slightly lower than 0.6, supplemented by other tests (such as Bartlett’s test of sphericity). If the KMO value is slightly low (0.5–0.6) but the Bartlett test is significant (P<0.05), factor analysis can still be conducted cautiously to reject the hypothesis of variable independence.^3,8–11 Nevertheless, as you mentioned, adopting the dual-threshold approach will indeed yield more reliable results. Therefore, in future studies, we will perform principal component analysis on data with KMO values between 0.50 and 0.59.

Once again, thank you for your valuable feedback. These suggestions will significantly enhance the robustness and validity of our study. We welcome further discussions to refine our protocol.