Long Term Outcomes of a Positron Emission Tomography (PET) -Based Adaptive Radiation Therapy Protocol in Locally Advanced Head and Neck Cancer.

Abstract

To report the long-term results of our PET-based adaptive head and neck cancer (HNC) protocol and to confirm outcomes/toxicities are not inferior to historic controls. In addition, we also report PET response parameters that are predictive of locoregional control (LRC). Patients with node positive HNC were eligible for the study. Patients underwent concurrent cisplatin or cetuximab treatment with radiation therapy (RT). The prescription dose was 70 Gy to the gross disease and 60 Gy to uninvolved nodal basins using a concomitant boost in 35 fractions. Simulation employed a thermoplastic mask on a PET-CT simulator and was repeated following fractions 10 and 22. Additionally, weekly PET-CT was obtained for response assessment. Plans were modified at those points taking into account reduction in disease burden and anatomical changes secondary to weight loss. The above protocol patients were matched 1:1 to non-protocol patients based on primary site, performance status, clinical T and N stage, smoking status, treatment dose/technique, chemotherapy use, and P16 status. The matched pairs were then analyzed to compare outcomes and toxicities between the two groups. In addition, the initial and weekly PET scans were analyzed for predictive correlation on LRC using 18FDG retention index (RI) and dose response matrix (DRM). The DRM represents a tumor’s SUV voxel response matrix to treatment. This can map out heterogeneity between and within individual neoplastic sites (either primary site or nodal disease) that may be more or less sensitive/resistant to treatment. A total of 37 patients were included on the adaptive protocol. Baseline characteristics were as follows: The median age was 63 years, 76% were oropharynx, 62% were cT1-2, 81% were cN2, 68% were smokers, and P16 was positive in 65% of patients. 3-year LRC, distant metastasis free survival, and overall survival were 81%, 75%, and 87%. The rates of chronic grade 3 toxicities were 8.1% for dysphagia and esophageal stricture and 2.7% for hoarseness and dental problems. There were 23 matched pairs generated and no significant difference in clinical disease outcomes was noted between the two groups. There was a significant decrease in chronic grade 2 xerostomia in the adaptive group vs the non-adaptive group (27% vs 70% p=0.017). Of interest, mean initial RI and the volume of DRM above certain thresholds were significant predictors of LRC on univariate analysis (p<0.05). Adaptive treatment for patients with locally advanced HNC resulted in significantly reduced rates of xerostomia relative to a matched cohort treated with non-adaptive treatment, and it did so without compromising disease control rates. RI and volume of DRM above certain thresholds are predictive for LRC and may be important factors in clarifying the optimal adaptive RT strategy for HNC in terms of both reducing treatment toxicity and potentially as a mid-treatment response parameter for adaptive dose painting in resistant tumors.