Predicting Treatment Response to Pulsed Radiofrequency in Postherpetic Neuralgia Using an Explainable Machine Learning Model.

Objective: Postherpetic neuralgia (PHN) is a chronic neuropathic pain disorder with substantial heterogeneity in response to pulsed radiofrequency (PRF). This study aimed to develop and validate an explainable machine learning (ML) model for predicting PRF efficacy in PHN patients. Methods: This single-center retrospective study included 404 PHN patients treated with PRF between January 2022 and December 2024. Patients were randomly assigned to a training set (n=283) and a validation set (n=121). Treatment efficacy was assessed 6 months after discharge. LASSO and the Boruta algorithms were used for feature selection. Six ML algorithms were developed and compared. Model performance was evaluated by area under the receiver operating characteristic curve (AUC), calibration, and decision curve analysis. SHapley Additive exPlanations (SHAP) were used to interpret the optimal model. Results: Of the 404 patients, 255 (63.12%) were classified as effective and 149 (36.88%) as ineffective. Six predictors were finally selected: age, duration of pain, diabetes mellitus (DM), tumor necrosis factor-α (TNF-α), interleukin-6 (IL-6), and brain-derived neurotrophic factor (BDNF). Among the candidate models, the random forest (RF) model showed the best overall predictive performance, with an AUC of 0.878 (95% CI: 0.806-0.949) in the validation set. SHAP analysis indicated that older age, longer pain duration, DM, and higher TNF-α and IL-6 were associated with poor response, whereas higher BDNF was associated with better outcome. Conclusion: An explainable ML model for predicting PRF efficacy in PHN patients was developed, and the RF model showed the best predictive performance. This model may assist individualized risk stratification and support clinical decision-making, but further prospective multicenter validation is required.