Early Response Assessment on Mid-treatment Computed Tomography Predicts for Locoregional Recurrence in Oropharyngeal Cancer Patients Treated With Definitive Radiation Therapy

Purpose: To evaluate whether a response assessment using mid-treatment computed tomography (CT) scans during definitive radiation therapy (RT) for oropharyngeal head and neck cancer can predict for locoregional recurrence (LRR). Methods and Materials: Head and neck cancer patients who receive RT at our institution undergo CT repeat scans at the 15th fraction, with treatment replanning in the case of an inadequate dose to gross disease or an increased dose to organs at risk. A retrospective cohort analysis was performed of 96 consecutive patients with oropharyngeal cancer treated from 2007 to 2015 with mid-treatment repeat CT scans available. The primary disease volume and involved lymph node volume were delineated on the pre- and mid-treatment CT scans. Univariable and multivariable Cox proportional hazards regression analyses were used to evaluate the efficacy of the mid-treatment reduction in tumor volume as a predictor of LRR. Risk stratification was performed by dichotomizing the patients into high- and low-risk groups according to the mid-treatment response and p16 status and smoking history. Results: With a median follow-up of 34 months, 14 patients experienced LRR. The median reduction in the total tumor volume was 18.7% (interquartile range 8.4%-30.9%). A reduction in total tumor volume greater than the median was an independent predictor of LRR (hazard ratio 0.22, 95% confidence interval 0.05-0.89; P =.020). The reduction in primary tumor volume was an even stronger predictor of LRR (hazard ratio 0.11, 95% confidence interval 0.02-0.57; P =.002). Stratifying patients into a high-risk group for those with a reduction in the total tumor volume at mid-treatment at or less than the median, p16 negative status, and smoking status of >10 pack-years and a low-risk group for those without these factors, we found a clear separation in Kaplan-Meier curves, with actuarial 3-year locoregional control, progression-free survival, and overall survival rates for the high-risk patients of 45.7%, 38.2%, and 71.8% compared with 90.7%, 70.6%, and 89.8% for low-risk patients, respectively (P ≤.021 for all). Conclusions: Our results have shown that the treatment response from an early assessment using mid-treatment CT scans is an independent predictor of LRR and can be used to effectively distinguish high- and low-risk patients, allowing for risk-adaptive treatment stratification at the midway point.