Tightening the gradient: ten years of dosimetric refinement in head and neck reirradiation

This study aimed to quantify changes in plan quality for head and neck reirradiation (reRT) on a constant C‑arm platform.

This retrospective study included 775 patients treated on a TrueBeam STx (Varian Medical Systems, Palo Alto, USA) between 2015 and 2025. Delivery transitioned from fixed-field IMRT during 2015-2019 to VMAT during 2020-2025. Planning was performed in Eclipse, with one upgrade from version 11.0.31 to 15.6.06. Endpoints included PTV D95, D98, D50, D2; different gradient indices; and doses to the spinal cord, brainstem, carotids, mandible, oral cavity, and DARS. Year-wise differences were tested using ANOVA with Bonferroni-adjusted pairwise comparisons and nonparametric sensitivity analyses. Linear models included calendar year and the log-transformed PTV volume.

Mean PTV D95 increased from 96.4 ± 0.5% to 99.6 ± 0.2%, while D2 decreased from 108.5 ± 1.6% to 104.0 ± 0.1%. Gradient indices decreased over the study period, indicating reduced dose spillage and steeper dose fall-off. Spinal cord and brainstem doses also declined substantially, with additional reductions in oral cavity and DARS doses. All four primary target endpoints showed significant year effects after multiplicity adjustment and nonparametric sensitivity testing. The PTV volume decreased from 700.2 ± 16.7 cc to 250.4 ± 14.7 cc, representing an approximately 64% reduction, and likely contributed to the observed improvements in gradient indices and OAR sparing. Temporal associations persisted after adjustment for the log-transformed PTV volume.

Dosimetric outcomes improved over the study period, but these changes coincided with reduced target volume, treatment technique transition, TPS upgrade, and workflow maturation. The findings should be interpreted as multifactorial rather than as evidence of an isolated technological effect.