Purpose: To determine the values of two normal tissue complication probability (NTCP) models for pharyngeal constrictors regarding the endpoint of patient reported severity of dysphagia based on the EAT10 scoring system 12 months after radiotherapy for oropharynx cancer. Additionally, to identify risk thresholds.
Methods: 151 patients with HPV-associated oropharyngeal squamous cell carcinoma were treated with 60Gy IMRT. Dysphagia was assessed based on the EAT10 score as the top 5% percentile of net difference from baseline at 12 months post-RT. The DVHs of the superior (SPC) and combined pharyngeal constrictors were exported for each patient. The clinical data was fitted by the Lyman-Kutcher-Burman (LKB) and Relative Seriality (RS) NTCP models.
Results: Dysphagia was observed in 5.3% of the patients. The values of the D50, m and n parameters of the LKB model were 69.1Gy, 0.11 and 1.0 for SPC and 65.4Gy, 0.10 and 0.22 for the combined constrictors. Similarly, the values of the D50, γ and s parameters of the RS model were 74.6Gy, 2.1 and 0.0001 for SPC and 65.6Gy, 6.7 and 1.0 for the combined constrictors. The AUC values of the correlations ranged between 0.75-0.77. Statistically significant Odds Ratios (ORs) of 5.5 (95%CI: 1.3-23.6) and 6.5 (1.5-28.3) were found for the SPC for the two NTCP models for a gEUD threshold of 58Gy. The respective ORs for the combined constrictors were 9.2 (1.5-57.5) and 11.6 (1.8-76.2) for the LKB and RS models for a gEUD threshold of 58Gy and 63Gy, respectively.
Conclusion: The LKB and RS NTCP models could fit the clinical data and determine the dose-response curve of SPC and combined constrictors regarding dysphagia based on the EAT10 score at 12 months post-RT. gEUD < 58Gy to SPC or combined constrictors was found to significantly reduce by 5-12 times the risk for patient reported dysphagia.
NTCP, Dose Response, ROC Analysis
TH- Response Assessment: Modeling: other than machine learning