Clarifying What the Model Predicts After Pulsed Radiofrequency for Postherpetic Neuralgia [Letter]
Journal of Pain Research · 2026-08 · Volume 19 · 1-3
DOI: 10.2147/jpr.s640690 ↗Kurz gefasst
Leserbrief in der Zeitschrift Journal of Pain Research, 2026-08. Verfasst von Mingyang Wang und 2 weiteren Autor:innen. Beteiligte Einrichtung: Clinical Medicine Grade 2022, School of Clinical Medicine, Hangzhou Normal University, Hangzhou, 311121, People's Republic of China. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 3.0).
Abstract (Originalsprache)
We read with interest the article by Li et al, who developed an explainable machine-learning model in 404 patients with postherpetic neuralgia treated with pulsed radiofrequency (PRF). 1 The random forest model achieved an area under the curve (AUC) of 0.878 in the internal validation set (n=121).The model combines routine clinical variables and biomarkers and uses SHapley Additive exPlanations (SHAP) to interpret its predictions.Knowing which patients are likely to have a poor outcome after PRF would be useful in clinical practice.Our concern is narrower: the operational definition of the outcome leaves uncertainty about what the model predicts.First, the two outcome-related Methods subsections do not state the same classification rule. 1 Under "Efficacy Assessment and Grouping", ineffective treatment included a numerical rating scale (NRS) reduction of less than 50% at 6 months after discharge, pain-related anxiety or depression requiring pharmacological intervention, or pain-related sleep disturbance.Under "Postoperative Analgesic and Neuropathic Pain Medications", the definition also included escalation of neuropathic pain medication because of inadequate pain control and stated that additional analgesic use was considered when efficacy was determined.This subsection appeared to require pharmacological intervention for sleep disturbance, whereas the preceding subsection did not.It remains unclear whether additional analgesic use alone triggered ineffective classification and whether a transient event at any point during follow-up was sufficient or the criterion had to be present 6 months after discharge.The prevalence and co-occurrence of the outcome components were also not reported.Without these details, the ineffective label cannot be reproduced from the published description, and the clinical meaning of the model output remains uncertain.The authors should provide the chart abstraction protocol, the final operational grouping rule, and the prevalence and co-occurrence of each outcome component.Second, the psychological and sleep components are clinically relevant but not reproducibly defined.The article does not indicate whether anxiety, depression, and sleep disturbance were identified using structured instruments, clinical diagnoses, medication records, or unstructured chart entries. 1 Without these details, readers cannot determine whether the same criteria were applied across patients or whether documentation and treatment decisions influenced group assignment.The authors should report the ascertainment source, timing, and decision rule for each component.Third, baseline and 6-month NRS scores were recorded, yet neither Table 1 nor Table 2 reports them.The Data Collection subsection does not list baseline NRS among the candidate predictors, and Supplementary Table 1 describes only the six predictors retained in the final model. 1 Because baseline NRS enters the percentage-reduction calculation, the article should clarify whether it was considered during feature selection and justify its inclusion or exclusion.The distributions of baseline NRS scores and NRS scores collected 6 months after discharge should also be reported.The TRIPOD+AI (Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis) statement specifically recommends clear reporting of the outcome definition, time horizon, assessment method, and initial predictor set. 2 The absence of these
Autor:innen und Institutionen
- Clinical Medicine Grade 2022, School of Clinical Medicine, Hangzhou Normal University, Hangzhou, 311121, People's Republic of China.
- Stomatology Grade 2023, School of Clinical Medicine, Hangzhou Normal University, Hangzhou, 311121, People's Republic of China.
- Nursing Department, Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou, 310016, People's Republic of China.
Zitieren
APA
Wang, M., Yang, X., & Yang, L. (2026). Clarifying What the Model Predicts After Pulsed Radiofrequency for Postherpetic Neuralgia [Letter]. Journal of Pain Research, Volume 19, 1-3. https://doi.org/10.2147/jpr.s640690
Vancouver
Wang M, Yang X, Yang L. Clarifying What the Model Predicts After Pulsed Radiofrequency for Postherpetic Neuralgia [Letter]. Journal of Pain Research. 2026;Volume 19:1-3. doi:10.2147/jpr.s640690
BibTeX
@article{Wang2026,
author = {Wang, Mingyang and Yang, Xiaoyu and Yang, Li},
title = {Clarifying What the Model Predicts After Pulsed Radiofrequency for Postherpetic Neuralgia [Letter]},
journal = {Journal of Pain Research},
year = {2026},
volume = {Volume 19},
pages = {1--3},
doi = {10.2147/jpr.s640690}
}
EUVIMED