Room: AAPM ePoster Library
Purpose: Patients with head and neck cancer generally exhibit anatomical changes (e.g., tumor shrinkage and weight loss) while on treatment that may affect the temporal accuracy of VMAT plans. We investigate the effect of anatomical/dosimetric changes and other patient characteristics on patient reported quality of life (QoL) outcomes.
Methods: 156 head and neck cancer patients treated with radical (chemo)radiotherapy completed the Xerostomia Questionnaire (XQ), and MD Anderson Symptom Inventory and Dysphagia Inventory surveys when attending follow-up clinics from June-October 2019. We identified patients with planned doses exceeding planning objective criteria and stratified the cohort into corresponding “acceptable” and “violation” groups. ANOVA statistical tests indicated which QoL responses were statistically significantly different (p<0.05) between the patient subgroups. This process was repeated using delivered doses. Multiple testing corrections controlled the false discovery rate. Higher scores denote poorer QoL.
Results: Delivered doses were associated with a greater number of statistically significant differences in QoL compared to planned doses. Delivered contralateral parotid gland Dmean=26 Gy had significantly worse scores (mean=2.1/5 vs. 2.9/5; p<0.01) for the question: “People ask me, ‘Why can’t you eat that?” Patients with delivered brainstem D0.03cc=54 Gy had significantly greater difficulty in talking (mean=2.3/10 vs. 5.3/10; p=0.03) and greater frequency of sleeping problems (mean=2.2/10 vs. 5.0/10; p=0.05) due to mouth dryness. Brainstem hotspot does not directly cause swallowing or dry-mouth toxicities, rather it likely indicates underlying correlations exist that are not captured by conventional statistical techniques. Patients with greater low-dose CTV volume loss had significantly better XQ scores; nasopharyngeal cancer patients had significantly poorer overall QoL.
Conclusion: QoL is an important means for informing quality of care. While conventional statistical methods indicate that QoL responses are more strongly associated with delivered dose compared to planned dose, artificial intelligence approaches may aid in better understanding complex associations in the dataset.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by the Natural Sciences and Engineering Research Council of Canada. The study team has no relevant financial disclosures or conflicts of interest to declare.