Density override-based VMAT optimization to improve robustness at tissue-air interfaces in head and neck radiotherapy.

BACKGROUND: Volumetric Modulated Arc Therapy (VMAT) is integral to head and neck (HN) radiation therapy; however, plan robustness may be compromised at air-tissue interfaces. Highly modulated fluence directed at these areas can produce significant unexpected high-dose regions in the event of minor variations in tissue density. PURPOSE: This study characterizes an instability in VMAT planning for HN radiation therapy (RT) occurring at the interface between target volume (e.g. mucosal primary tumor) and internal air spaces (e.g. pharyngeal lumen, sinuses, nasal cavity, etc.). A density-override based planning technique is presented and validated to mitigate this instability. METHODS: Treatment plans for fifteen HN patients receiving VMAT RT at our institution between 2023 and 2024 were retrospectively analyzed. All patients received 60-69.96 Gy in 30-33 daily fractions using 6 MV photons. To model swelling or tumor growth near the air-tumor interface, verification plans were generated by overriding internal air within 3 mm of the planning target volume (PTV) to soft tissue density (HU = 0). A robustness-enhancing technique was evaluated, wherein optimization was performed with luminal air overridden to HU = -300, followed by recalculation and renormalization using the original CT HU values. Robustness was then reassessed after overriding luminal air to HU = 0. RESULTS: Baseline plans demonstrated controlled hotspot magnitudes within the BODY, with a median D0.1cc of 109.0% [Interquartile range: 107.8-110.4]. Simulated tissue-air interface changes produced substantially elevated hotspots in verification plans, increasing to a median D0.1cc of 115.6% [111.9-126.9], corresponding to a median paired increase of 9.0% (Wilcoxon p = 1.22 × 10- 4). Robust plans generated using the proposed density-override optimization maintained comparable baseline dosimetry (median D0.1cc 108.1% [107.4-108.7]) while demonstrating stable hotspot behavior under simulated density perturbations (median D0.1cc 108.1% [107.1-108.6], p = 0.18). Target coverage remained preserved, with robust verification plan D95 values ranging from 99.7% to 100.5% (p = 0.26). CONCLUSION: A density-override optimization approach mitigates hotspot formation at air-tumor interfaces, producing VMAT plans that remain dosimetrically stable despite variation at internal air-tumor interfaces.