Improved Efficacy of a Predictive Model for Swallowing-Induced Breakthrough Pain Based on a Redefined Delineation...

The objective of this study was to explore the performance of a predictive model for swallowing-induced breakthrough pain established using a redefined delineation method based on the common occurrence sites of radiation-induced oral mucositis (RIOM) in locally advanced nasopharyngeal carcinoma (NPC).

A total of 208 patients with locally advanced NPC were included in the study cohort, and the test cohort consisted of 88 patients. The oral mucosa structure was contoured using oral cavity contour (OCC), mucosal surface contour (MSC), and oral-pharyngeal mucosa (OPM) methods, and relevant dosimetric parameters were collected. Assessment of the severity of RIOM was made with the National Cancer Institute's Common Terminology Criteria for Adverse Events, version 4.0. The random forest classification method was chosen to establish and validate the predictive models based on 3 contouring methods.

The area under the curve of the OPM-based model was higher than that of the OCC- and MSC-based models in both the validation cohort and the test cohort (0.800, 0.739, and 0.750; 0.670, 0.605, and 0.609, respectively). Better predictive performance could also be observed under the OPM method than the OCC and MSC methods in terms of accuracy. The OPM-based model showed high specificity (greater than 90%) in both the validation cohort and the test cohort. According to the mean decrease in the Gini index, the maximum dose was the most important predictor of severe oral mucositis in the OPM-based model.

We redefined a delineation method for oral mucosa structure based on the common occurrence sites of RIOM in locally advanced NPC. The model for swallowing-induced breakthrough pain constructed based on this method demonstrated good predictive performance. New parameters were found as predictors of severe swallowing-induced breakthrough pain in locally advanced NPC.