Purpose: To determine the extent to which the dosimetric consequences of organ-at-risk (OAR) delineation variability (DV) are washed-out by inherent random and systematic patient setup variability (SV).
Methods: For (N=5) head-and-neck patients, VMAT treatment plans were generated for multiple alternative OAR structure sets to meet clinical plan quality metrics (PQMs). These alternative structure sets were generated with simulated DV by randomly perturbing clinical OARs with standard deviations of Ïƒ_c=2,5,10 mm. Five alternatives were generated for each Ïƒ_c. Each plan was cross-evaluated using the clinical OARs and the four other OAR alternatives with the same Ïƒ_c.The effect of SV was evaluated by simulating 1000 treatment course deliveries for each cross-evaluation and simulating per-fraction random (Ïƒ_s) and per-treatment-course systematic (Î£_s) setup errors at three levels: Î£_s=Ïƒ_s=2,4,10 mm, quantifying the probable dosimetric variability due to SV. Combined inter-comparisons with alternative OARs and with SV simulations quantified the wash-out of the effect of DV by SV. The relative impact of DV was assessed by examining changes in probabilities of achieving PQMs (PPQM) with and without DV given a fixed level of SV.
Results: In general, the impact of DV decreased with increasing SV. Without SV, simulated DV with Ïƒ_c=2,5,10 mm reduced PPQM by an average of 0.9%, 3.3%, and 8.3%, respectively. Average PPQM reductions decreased to 0.1%, 1.5%, and 5.4% with SV=2mm and 0.4%, 1.1%, and 3.4% with SV=4mm. However, while average PPQM reductions were low, some OARs showed larger changes (95th percentile PPQM reductions were 6.1%, 9.8%, and 24.5% with SV=4mm).
Conclusion: DV and SV both reduced the probabilities of achieving clinical PQMs that were met in static plan dosimetry. The relative impact of DV decreased when SV was considered. While DV was largely washed-out on average, its effect was still significant for some individual OARs.
Funding Support, Disclosures, and Conflict of Interest: Funded by NIH RO1CA222216.