ESTRO 2024 - Abstract Book

S2853

Interdisciplinary - Health economics & health services research

ESTRO 2024

Purpose/Objective:

Adaptive treatment planning offers the potential to tailor dose distributions to anatomical changes, reducing required PTV margins or setup- robustness settings, consequently enhancing proton therapy’s healthy organs -at-risk (OARs) sparing [1]. Various offline and online plan adaptation strategies have been investigated, with varying levels of dosimetric benefit. However, the associated increments on additional workload and fraction duration remain unknown. This study aims to provide insights on the effects of five adaptive strategies on workload, fraction duration and toxicity risk reduction, to aid making well-informed decisions regarding the introduction of alternative adaptive strategies. Furthermore, considering the rapid advancement of treatment planning technology, a scenario analysis is performed to estimate the impact of automation on workload and treatment time.

Material/Methods:

Five adaptive strategies were considered: 1) Offline trigger, our current offline adaptive schedule: an initial treatment plan is generated on the planning-CT using 3 mm setup-robustness, with replanning based on clinical evaluations. 2) Offline weekly : initial treatment plan using the planning-CT with 2 mm setup-robustness, which is weekly updated using repeat-CTs. 3) PL static : a Plan Library (PL) approach [2], where five treatment plans that vary in setup robustness (0 to 5 mm) are generated using the planning-CT. Daily, the plan with the smallest setup-robustness settings that meets target constraints on a repeat-CT is selected. 4) PL progressive : follows the same concept as PL static , but the PL is weekly extended using a repeat-CT. 5) Daily Online dose Refinement (OnlineRefinement): initial plans are generated on the planning-CT with 1 mm setup-robustness and daily automatically adjusted. For each adaptive strategy, the required activities in the adaptation workflow were identified for our institute. Time costs for activities were estimated using a combination of questionnaires, time measurements, scheduled time, and literature. Activities that were expected to be further automated (Technology Readiness Level ≥8 [3, 4]) within the near-future (<5 years) were identified through discussions among the authors (MO, SB, MG, SH, ZP, BH, MH). Activities for which more than 80% consensus was achieved regarding their potential to be automated (fully or partially) were considered as automated steps within the workflow. Time cost was converted into hands-on workload through multiplication by the number of people performing the activity. Dosimetric performance was assessed by simulating 25 treatments of 35 fractions for each of 15 head-and-neck cancer patients treated with 70/54.25 Gy. Target coverage and risk of toxicities of xerostomia and dysphagia ≥ grade II [5] were compared. Accumulated target coverage was similar for every strategy. Time costs of Offline trigger was estimated to currently be 5.7 hours per patient for the full treatment course, while this was 41, 30, and 152 hours for Offline weekly , PL static and PL progressive (Figure 1). As OnlineRefinement is not possible within our current clinical systems, no estimation was obtained. Considering the anticipated automation, the respective time costs for the strategies are expected to be 1.0, 6.0, 4.2, 22, and 26 hours for Offline trigger , Offline weekly , PL static , PL progressive and OnlineRefinement (Figure 1), prolonging fraction time with 0, 4.0, 5.5, and 27 minutes compared to Offline trigger , respectively (Figure 2, left). Implementing the alternative strategies would result in a reduction of the risk of xerostomia ≥ grade II of 1.9% - point, 1.2%-point, 2.7%-point, and 4.6%-point for on average for Offline weekly , PL static , PL progressive and OnlineRefinement compared to our current strategy Offline trigger (Figure 2). Similarly, the risk of dysphagia ≥ grade II can be reduced by 1.9%-point, 0.7%-point, 2.3%-point and 4.2%-point, respectively. Results: