Voxel Dose Pattern for Patient-Reported Dysphagia Among Head and Neck Cancer Patients Receiving Definitive Radiotherapy

Abstract

Acute dysphagia following head and neck cancer (HNC) chemoradiation (CRT) is a significant complication. Since swallow movement involves a wide range of upper aerodigestive tract structures, the impact of radiation injury to each swallowing structure is not fully understood. As patient-reported outcomes (PRO) offer the potential for more accurate assessments, we sought to explore whether the spatial dose pattern in swallow and salivary related structures can help us better model acute PRO-dysphagia in HNC patients treated with CRT. HNC patients treated with intensity-modulated-CRT from 2015-2018 in our institution were evaluated. Our primary outcome was Sydney Swallow Questionnaire total score≥500 at 3-6months post-RT. Variables includes patient clinical data (age, gender, race, TNM staging, HPV status, chemotherapy, surgery, tumor location, and baseline dysphagia status), and voxel-based dose data. Twenty-four swallow related regions of interest (ROIs) were identified, including the ipsi/contralateral parotid glands, ipsi/contralateral submandibular glands, base of tongue, epiglottis, larynx, soft palate, and the constrictor, anterior/posterior digastric, ipsi/contralateral masticatory, cricopharyngeus, geniohyoid, hyoglossus, hyoid, and mylohyoid muscles. Each patient’s CT was spatially normalized to a common coordinate system (CCS) with nonrigid registration. The obtained deformation fields were used to map each patient’s dose to the CCS. The dose map was sampled in a raster pattern across the composite region of all of the ROIs, for a total of 3132 points. Ridge logistic regression with regularization methods were used to investigate the influence of the dose voxel patterns in each ROI on PRO-dysphagia. Among the 150 patients identified, 37(24.6%) reported severe dysphagia (≥500). The results show that the following structures, listed in decreasing order of importance, were the most influential for predicting PRO-dysphagia: ipsi- and contralateral parotid glands, the contralateral masticatory muscle, floor of mouth muscles, epiglottis, cricopharyngeus, and constrictor muscles. The area under the receiver operating characteristic curve (AUC) from 5-fold cross-validation was 0.71±0.06. We evaluated the spatial dose relationship of swallow-related ROIs and PRO-dysphagia. Our analysis demonstrated the feasibility of applying a voxel-based, spatial dose analysis with an atlas of swallow-related structures. More importantly, we demonstrated that important irradiated regions for PRO-dysphagia include the ipsi/contralateral parotids, epiglottis, and masticatory, floor of mouth, cricopharyngeus, and constrictor muscles. These results also support the intuitive relationship between salivary function and PRO-dysphagia, through a spatial description of the pattern of influence. Future directions incorporating more refined measures of dysphagia may help improve the prediction.