The two-stage therapeutic effect of posture biofeedback training on back pain and the associated mechanism: A retrospective cohort study

Introduction: Back pain is an extremely common symptom experienced by people of all ages and the number one cause of disability worldwide. Poor posture has been identified as one of the factors leading to back pain. Digital biofeedback technology demonstrates the promising therapeutic ability in pain management through posture training. One common goal of such an approach is to increase users' posture awareness with associated movement correction. However, we lack a deep understanding of the biofeedback therapeutic mechanisms and the temporal dynamics of efficacy.Objective: This study investigates the temporal dynamics of the biofeedback learning process and associated outcomes in daily life settings, testing the mechanism of the biofeedback-associated pain reduction. Methods: This retrospective real-world evidence study followed 981 users who used the UpRight posture biofeedback platform. Piecewise mixed models were used for modeling the two-stage trajectory of pain levels, perceived posture quality, and weekly training duration following an 8-week biofeedback training. Also, the mediation effect of perceived posture quality on the analgesic effect of training duration was tested using Monte Carlo simulations based on lagged effect mixed models. Results: The analysis revealed significant pain level reduction (p <.0001) and posture quality improvement (p <.0001) during the first 4 weeks of the training, maintaining similar pain levels and perceived posture quality during the next 4 weeks. In addition, weekly training duration demonstrated an increase during the first 3 weeks (p <.001) and decreased during the next 5 weeks (p <.001). Moreover, training duration predicted following-week perceived posture quality (p <.001) and in turn perceived posture quality predicted following-week pain (p <.001) (p = 0.30). Finally, perceived posture quality mediated the effect of weekly training duration on the pain levels in 2 weeks (p <.0001). Conclusion: Our findings provide a better understanding of the therapeutic dynamic during digital biofeedback intervention targeting pain, modeling the associated two-stage process. Moreover, the study sheds light on the biofeedback mechanism and may assist in developing a better therapeutic approach targeting perceived posture quality.