Cross-prescription validation of knowledge-based planning in prostate VMAT using a structured validation framework

Knowledge-based planning (KBP) has been widely adopted to improve plan quality, consistency, and efficiency in prostate volumetric-modulated arc therapy (VMAT), but adaptation to different prescription regimens often requires rebuilding separate models, which may be impractical in routine clinical settings. This study investigated the practical feasibility of reusing a KBP model across prescription regimens with nonequivalent dose-constraint systems in prostate VMAT. A model trained for conventional fractionation (78 Gy in 39 fractions) was evaluated using a 3-step framework consisting of closed-loop validation for reproducibility, open-loop validation for generalizability to previously unseen patients, and cross-prescription validation for reuse in moderate hypofractionation (60 Gy in 20 fractions) planned under a different clinical constraint framework. Model-generated plans were compared with manual plans using target dose metrics, organ-at-risk dose-volume indices, and plan quality metric scores. The model reproduced the dosimetric characteristics of the training dataset under standardized conditions and maintained clinically acceptable dosimetric performance in independent cases. When reused for a different prescription regimen, the model achieved comparable overall plan quality and acceptable organ-at-risk sparing, despite slight reductions in target dose homogeneity. These results suggest that KBP model reuse may reduce the need for repeated model rebuilding and provide a practical approach to efficient and standardized prostate VMAT planning, particularly in institutions with limited case numbers or limited resources for model development and maintenance.