The specificity of acupoint indication (i.e., reverse inference-diseases for which an acupoint could be used) might differ from the specificity of acupoint selection (i.e., forward inference-acupoints used for a disease). In this study, we explore acupoint specificity through reverse inferences from the dataset of prescribed acupoints for a certain disease in clinical trials. We searched acupuncture treatment regimens in randomized controlled trials included in the Cochrane Database of Systematic Reviews. For forward inference, the acupoints prescribed for each disease were quantified. For reverse inference, diseases for each acupoint were quantified. Data were normalized using Z-scores. Bayes factor correction was performed to adjust for the prior probability of diseases. The specificities of acupoint selections in 30 diseases were determined using forward inference. The specificities of acupoint indications regarding 49 acupoints were identified using reverse inference and then subjected to Bayes factor correction. Two types of acupoint indications were identified for 24 acupoints: regional and distal. Our approach suggests that the specificity of acupoint indication can be inferred from clinical data using reverse inference. Acupoint indication will improve our understanding of acupoint specificity and will lead to the establishment of a new model of analysis and educational resources for acupoint characteristics.
Keywords: Bayesian modeling; acupoint indication; acupoint specificity; clinical trials; data mining.