Can patient reported quality of life predict locoregional recurrence in oropharyngeal cancer?

Purpose/Objective(s): The excellent prognosis of p16+ oropharyngeal squamous cell carcinoma (OPSCC) invites efforts to reduce surveillance burden after treatment. Current surveillance guidelines suggest frequent clinical examination, demanding time, costs, and discomfort with nasopharyngoscopy. Recent publications demonstrated low rates of asymptomatic locoregional recurrence (LRR) detection in these frequent visits, but no alternatives exist. We sought to assess correlation of changes in patient reported quality of life (QOL) forms with LRR to evaluate feasibility of using such forms in an individualized surveillance approach. Materials/Methods: Patients with QOL forms treated on a single institution randomized trial and patients on a prospective registry were evaluated (74 and 102 patients respectively). Patients completed EORTC C30 and HN35 at baseline, 1, 3, 6, 9, 12, 18, 24, and 36 mos after end of treatment. Patients were grouped by time to LRR, and QOL forms from the two time points prior to LRR were evaluated for changes in the following prespecified EORTC QOL subscales: from C30, physical functioning, role functioning, fatigue, pain; from HN35, pain, swallowing, social eating, feeling ill. A minimal clinically important difference (MCID) of 10 for each subscale was considered significant, and area under the curve (AUC), sensitivity (sens) and specificity (spec) were calculated for each subscale for each group of failures. Results: With a median follow up of 12.4 mos, 176 patients experienced 29 failures. 71.1% of patients had OPSCC. Baseline forms were completed by 157 patients, with increasing attrition over time. Combining all eight subscales led to a prediction tool with average sensitivity of 80% for detection of LRR, with failures from 3-6 mos (AUC 0.69, sens 83%, spec 34%) and 6-12 mos (AUC 0.78, sens 100%, spec 49%) most significant. Failure within 1-3 mos was best predicted by MCID changes in C30 role functioning (AUC 0.85, sens 100%, spec 62%). Failure within 3-6 mos was best predicted by MCID changes in HN35 pain (AUC 0.71, sens 67%, spec 84%). Failure within 6-12 mos was best predicted by MCID changes in C30 pain (AUC 0.90, sens 80%, spec 82%), HN35 swallowing (AUC 0.80, sens 60%, spec 90%), HN35 social eating (AUC 0.72, sens 40%, spec 88%), and HN35 feeling ill (AUC 0.71, sens 40%, spec 91%). Assessment from 12-24 mos was complicated by low numbers of events. Conclusion: MCID changes are present in 8 subscales of EORTC C30 and HN35 QOL forms prior to clinical presentation of LRR, and the implementation of a rule assessing MCID changes in any of these subscales led to a tool with 83% sensitivity of detecting LRR. Utilization of QOL forms to prompt closer surveillance in integration with a novel less burdensome monitoring paradigm for good-prognosis OPSCC could represent a new option for individualized surveillance. Further validation and study is needed.