Assessing dosimetric benefit from daily online adaptive radiation therapy for esophageal cancer

A cohort of 19 esophageal cancer patients treated at our institution were analyzed to assess the clinical feasibility and dosimetric benefit from daily adaptive radiation therapy (ART). An Ethos-planning template was developed to generate the initial ethos reference plan and daily adaptive plans using 9-field IMRT for reduced on-treatment optimization time. The template included relevant OAR goals and used an institutionally validated esophagus RapidPlan model. Clinical planning margins were used. Ethos-generated plans were validated against the clinically-approved plans which used 2-arc VMAT. Weekly on-treatment CBCT images were used to simulate doses from daily nonadaptive and adaptive techniques using the Ethos 2.0 Emulator. Timing data was recorded for each fraction. Dose metrics from institutional guidelines were compared between standard-of-care (SOC) and adaptive plans using a Wilcoxon signed rank test; these included mean dose for heart, lungs and liver, lungs V20Gy, and max dose for heart, spinal canal, and stomach. All Ethos template-generated plans were found equivalent to clinical plans. The daily adaptive workflow required 14.1 ± 6.8 min. Dosimetric improvements were variable by patient. Some patients experienced large metric reductions while others saw very minor benefits if any. For a sub cohort of patients that received high benefit, statistically significant (p < 0.05) reductions in the lung mean, lung V20, heart mean, and heart V30 and V40 were observed. From these results we conclude that Ethos ART is feasible in terms of plan quality and on-treatment time for esophageal cancer. ART can produce significant normal tissue dose reductions; however, not all patients benefited equally. Thus, identifying high-benefit patients prior to treatment is necessary. Preliminary modeling results suggest this can be done prospectively, but these models are internally validated and require larger datasets to fully develop.