Investigation of Longitudinal Dose-weighted FDG-Positron Emission Tomography Metabolic Imaging Biomarkers (PET MIBs) of Radiation-associated Dysphagia in OPC Cohort

Abstract

Head and neck cancer radiotherapy (RT) is associated with injury to dysphagia/aspiration at risk structures (DARS) with subsequent radiation-associated dysphagia (RAD). We sought to characterize the kinetics of standardized uptake values (SUV) derived from routine pre- and post-RT FDG-PET in DARS. We further sought to examine correlations between our proposed PET MIBs and Dynamic Imaging Grade of Swallowing Toxicity (DIGEST), a validated CTCAE-compatible modified barium swallow (MBS) grading scale of pharyngeal-stage dysphagia, in a subgroup of patients with standardly acquired MBS. Data for patients with oropharyngeal cancer (OPC) treated with RT (2005-2016) with available pre- & post-RT PET-CT scans, and DICOM-RT data, were retrospectively collected after an IRB approval. Analyzed DARS included soft palate (SP), genioglossus (GG), mylogeniohyoid complex (MGH), superior, middle and inferior constrictor muscles (SPC, MPC, and IPC) as well as ipsilateral (ipsi) and contralateral (contra) anterior bellies of digastric (DA) muscles. After CT-CT deformable image co-registration using commercial software (Velocity AI), DARS were sub segmented into smaller subvolumes based on the dose they received in 5 Gy increments, i.e. 0-5 Gy, 5-10 Gy subvolumes, etc. Volumetric as well as median dose-weighted SUV (calculated as: SUV * Planned DARS dose) were extracted from these whole and sub DARS volumes pre- & post-RT then compared using in-house algorithm. A corollary analysis was then performed on a subset of patients with prospectively collected DIGEST scores at 3-6-month post-RT (DIGEST 3-6ms) using Kruskal-Wallis tests. 904 DARS of 113 RT-treated OPC patients, were sub segmented & analyzed. DIGEST 3-6ms was reported for 39 patients as: grade 0-1, i.e. No or mild dysphagia (59%) or grade 2-3, i.e. moderate/severe (41%). Volumes and SUVmedian increased significantly after RT (p<0.0001) for all muscular DARS, including SPC, MPC, IPC, both DA, GG, and MGH complex, unlike soft palate that showed a decline in SUVmedian post-RT (p<0.0001). A Spearman rank correlation test showed a rising magnitude of delta-changes with higher DARS Dmean (p<0.05), e.g. IPC volume; ρ = 0.2, contra DA SUVmedian; ρ = 0.2, and MGH complex SUVmedian; ρ = 0.27. There was an increased likelihood of more severe dysphagia among patients with higher mean planned RT dose for all DARS (except MGH complex), more advanced cancer stage, those who received induction chemotherapy or reported some degree of dysphagia at presentation (p<0.05). Using multiple logistic regression to model DIGEST 3-6ms, SP Dmean, post-RT SP volume, post-RT, IPC SUVmedian, along with induction chemotherapy could adequately predict RAD severity; AUC = 0.86 (95% CI: 0.6-0.94; p<0.022). We sought to quantify pre- and post-RT FDG-PET metrics of DARS. Longitudinal kinetics of PET-derived metrics manifest dose-dependent changes that can potentially serve as MIBs of RAD measures using DIGEST score.