Ayurveda in Knee Osteoarthritis—Secondary Analyses of a Randomized Controlled Trial

Abstract

Background: Ayurveda is widely practiced in South Asia in the treatment of osteoarthritis (OA). The aim of these secondary data analyses were to identify the most relevant variables for treatment response and group differences between Ayurvedic therapy compared to conventional therapy in knee OA patients. Methods: A total of 151 patients (Ayurveda n = 77, conventional care n = 74) were analyzed according to the intention-to-treat principle in a randomized controlled trial. Different statistical approaches including generalized linear models, a radial basis function (RBF) network, exhausted CHAID, classification and regression trees (CART), and C5.0 with adaptive boosting were applied. Results: The RBF network implicated that the therapy arm and the baseline values of the WOMAC Index subscales might be the most important variables for the significant between-group differences of the WOMAC Index from baseline to 12 weeks in favor of Ayurveda. The intake of nutritional supplements in the Ayurveda group did not seem to be a significant factor in changes in the WOMAC Index. Ayurveda patients with functional limitations > 60 points and pain > 25 points at baseline showed the greatest improvements in the WOMAC Index from baseline to 12 weeks (mean value 107.8 ± 27.4). A C5.0 model with nine predictors had a predictive accuracy of 89.4% for a change in the WOMAC Index after 12 weeks > 10. With adaptive boosting, the accuracy rose to 98%. Conclusions: These secondary analyses suggested that therapeutic effects cannot be explained by the therapies themselves alone, although they were the most important factors in the applied models.