Big data analysis of associations between patient reported outcomes, observer reported toxicities, and overall quality of life in head and neck cancer patients treated with radiation therapy.

PURPOSE: To assess associations between a variety of patient-reported outcomes (PROs), observer reported toxicities (ORTs), and patient-reported overall quality of life (QOL) for head and neck cancer patients treated with radiotherapy, in order to identify important items for inclusion in prospective patient reporting in the clinic.
METHODS: 612 patients completed 27 PRO items from three questionnaires at 1273 follow-up visits, and clinicians provided ORTs according to CTCAE criteria. Using a big data approach, we measured associations among all PROs, between all PROs and ORTs, and between PROs/ORTs and QOL with Pearson (ρ) and Kendall (τ) correlation coefficients, and a novel analysis method based on receiver operating characteristic (ROC) curves used to detect thresholds in response levels demonstrating strong interactions.
RESULTS: PROs most strongly associated with QOL were recreation/entertainment, activity, and fatigue, with ρ = 0.51-0.60. Several PROs assessing a common functional outcome (eg. xerostomia) were highly associated with each other (PRO-PRO), with maximum ρ = 0.84. Maximum ORT-PRO correlations were ρ = 0.61 (dysgeusia versus taste), and ρ = 0.5 for ORT-QOL (dry mouth - day). The ROC method identified response thresholds with high area under the curve (AUC) scores for many ORT-PRO associations with maximum AUCavg = 0.78.
CONCLUSIONS: PRO associations identified activity, lifestyle and fatigue as items for strong consideration for inclusion in questionnaires in the clinic, and suggest that outcome information can be captured in fewer items than the 27 in this study. The ability of clinicians to assess patient toxicities is highest with more severe toxicities, underscoring the need for PRO collection in patient visits to understand and address patient symptoms.