Clinical implementation of an automated VMAT treatment planning script for head and neck cancer patients: three-year experience

This is the authors' abstract. We don't add key points for this paper.

Purpose To assess the impact of implementing an in-house automated volumetric modulated arc therapy (VMAT) planning script for patients with head and neck (HN) cancer. Methods The automated planning script was implemented at our institution in April 2020. During validation, 10 auto-plans were compared with 10 corresponding manual plans for dosimetric indices, Five radiation oncologists blindly reviewed these plans for clinical acceptability and treatment preference. For clinical evaluation, dosimetric indices from 1000 HN patients consecutively treated between 2017 and 2023 (500 manual pre-implementation, 500 automated post-implementation) were compared using t-tests (p<0.05). Results In validation testing, 10 auto-plans maintained PTV D 95% prescription coverage and similar global D max to corresponding manual plans (107.4% vs 107.3%) while significantly reducing maximum doses D max to the brainstem (-5.1 Gy) and spinal cord (-2.9 Gy), both p < 0.03, and mean doses to the ipsilateral parotid (-4.8 Gy), esophagus (-3.9 Gy), cochleae (-3.8 Gy), contralateral submandibular gland (-3.6 Gy), contralateral parotid (-2.2 Gy), and pharynx (-2.0 Gy), all p < 0.05. In blinded review by physicians, 94% of automated plans and 86% of manual plans were rated clinically acceptable and most physicians preferred 7 auto-plans. In the clinical evaluation phase following three years of clinical use, 500 auto-plans achieved PTV D 95% prescription coverage and similar global D max to manual plans while significantly reducing maximum doses to the brainstem (-3.6 Gy) and spinal cord (-2.1 Gy), both p < 0.001. In addition, auto-plans demonstrated significant mean dose reductions for the contralateral submandibular gland (-4.1 Gy), ipsilateral parotid (-3.9 Gy), oral cavity (-2.5 Gy), cochleae (-2.4 Gy), larynx (-2.0 Gy), contralateral parotid (-1.5 Gy), and maximum doses to the mandible (-2.9 Gy) and lips (-2.3 Gy), all p < 0.04. Conclusions Automated planning improved organ-at-risk sparing without compromising target coverage or dose homogeneity, with high clinical acceptability.