Pain catastrophizing predicts tDCS response patterns in knee osteoarthritis pain

OBJECTIVES: We planned to identifyied key predictors of the heterogeneity of treatment effects of transcranial direct current stimulation (tDCS) in individuals with knee osteoarthritis (KOA). METHODS: This is a secondary analysis of a randomized clinical trial involving 60 participants who underwent 15 daily sessions of 2-mA tDCS over 3 weeks. We applied group-based trajectory modeling to classify participants into distinct subgroups based on longitudinal KOA pain and symptom patterns from baseline to 3 months postintervention to examine differential responses to tDCS. Four learning-based classifiers-multilayer perceptron, ElasticNet, random forest, and gradient boosting decision trees-were then trained to predict the trajectory subgroups using demographic, clinical, and quantitative sensory testing data collected at baseline. Feature selection methods-f-regression, mutual information, and SHapley Additive exPlanations (SHAP)-were employed to identify the influential features. In addition, SHAP was used to analyze the correlation and impact of each feature on classification. RESULTS: Participants exhibited distinct response patterns to tDCS: high responders (low initial symptoms with significant improvement, n = 28) and low responders (high initial symptoms with minimal improvement, n = 32). The influential features included pain catastrophizing, conditioned pain modulation (CPM), and pressure pain thresholds (PPTh) at the trapezius. SHAP revealed that pain catastrophizing was the most influential feature. Greater pain catastrophizing, lower CPM, and lower PPTh were associated with a higher likelihood of being classified as low responders. CONCLUSION: Baseline assessments of pain catastrophizing, CPM, and PPTh may be used to stratify participants, optimize treatment allocation, or tailor stimulation parameters for individuals less likely to respond to tDCS protocols.