Development and internal validation of a therapeutic effect predictive model for myofascial pain syndrome

Comprehensive Summary

The study develops and evaluates a machine learning predictive model to estimate treatment efficacy in patients with myofascial pain syndrome (MPS) using a combination of clinical, psychological, and inflammatory variables. Researchers retrospectively analyzed data from 340 patients who underwent standardized 8-week treatment and applied statistical methods alongside popular machine learning algorithms (including random forest, support vector machine, and K-nearest neighbors) to identify key predictors and construct the model. Results identified six significant factors as independent predictors of poor treatment response: disease duration, baseline pain intensity, depression (PHQ-9), pain catastrophizing, interleukin-6, and high-sensitivity C-reactive protein. These factors are combined within a logistics regression model, specifically a support vector machine model, where these 6 factors are individually weighted to produce an aggregate score that predicts the probability of poor treatment responses. The support vector machine model demonstrated the best performance, achieving strong discrimination (AUC ~0.87–0.90), good calibration, and meaningful clinical utility. Overall, the study highlights that the utility of machine learning models, when integrated with biopsychosocial data, can effectively predict treatment outcomes in MPS.

Outcomes and Implications

This study highlights the potential of machine learning to streamline individualized treatment strategies in myofascial pain syndrome (MPS) by integrating clinical, psychological, and inflammatory data into a single predictive framework. By identifying key predictors (such as depression severity, pain catastrophizing, and inflammatory markers), the research demonstrates to providers the importance of a biopsychosocial approach to chronic pain management and how these diverse factors can be used to improve clinical decision-making. The model, especially the support vector machine, illustrates how AI tools can help clinicians estimate treatment response before intervention, allowing for earlier optimization of therapy, which reduces resource usage and improves clinical outcomes. The study informs researchers and clinicians about the importance of psychological and inflammatory contributors to treatment outcomes, highlighting areas where interventions (e.g., mental health or anti-inflammatory strategies) may be intensified. However, before AI models can be widely implemented into clinical practice, challenges, including lack of external validation, potential confounding variables, and limited generalizability across diverse populations, must be addressed. While the study demonstrates strong potential for AI-assisted personalization in chronic pain treatment, gaps remain in validation, implementation, and real-world integration.