Computational linguistics and sentiment analysis of misconceptions in pharmacotherapy of osteoarthritis pain

Abstract

Background. Osteoarthritis (OA) is one of the leading causes of chronic pain in adults, wherein half of the cases is coupled to a neuropathic component. Agents with chondroprotective properties such as chondroitin sulfate (CS) and glucosamine sulfate (GS) have been successfully used in the treatment of OA-related pain. CS/GS exhibit diverse analgesic, anti-inflammatory, antioxidant and chondroregenerative effects that contribute to the restoration of cartilage tissue.Objective: to analyze the misconceptions associated with the medical terminology used for CS/HS in the treatment of OA-related pain, approaches to standardize the quantitative and qualitative composition of CS/HS extracts.Material and methods. Expert analysis was performed along with computational linguistics methods (sentiment analysis, i.e. analysis of text-related emotional modality). Sentiment analysis was carried out using the topological theory of data analysis and algorithms, with 90% accuracy allowing to classify texts into 16 classes of sentiment (manipulative constructs, research without positive results, propaganda, data falsification, etc.). This technique was tested earlier on 20 million publications retrieved from PubMed/MEDLINE database.Results. In recent years, the use of highly dubious terms such as “symptomatic slow acting drug for osteoarthritis, SYSADOA”, etc., has been extensively promoted at certain international conferences. The introduction of such barely scientific terms is not justified neither by the results of basic research nor clinical practice. Using the methods of computational linguistics and data mining of the biomedical literature, we have shown that some misconceptions actively promoted at the so-called "grand conferences" and "international congresses" virtually lack in real-world published scientific literature. Such misconceptions, logically contradicting the entire system of other medical terms, confuse scientific terminology. Moreover, texts promoting this misconceptions are easily recognized as manipulative not only by experts in the analysis of medical literature, but also by artificial intelligence algorithms.Conclusion. A number of misconceptions associated with inadequate interpretation of data obtained during basic and clinical studies of CS/GS has been explored. Specific examples show how practitioners can distinguish between manipulative propaganda and a balanced presentation of research data.