(P 311) ROBUST VMAT TREATMENT PLANNING FOR EXTREMITY SOFT TISSUE SARCOMAS

Objective: Volumetric modulated arc therapy (VMAT) is a frequently employed radiotherapy (RT) modality for extremity soft tissue sarcomas (eSTS). Prior studies have shown that significant tumor volume changes during treatment result in loss of target volume coverage with highly conformal techniques such as VMAT, but few solutions exist to these issues aside from adaptive replanning. Here, we describe a related but novel phenomenon in which relatively minor changes in surface volume contour can result in unexpected subcutaneous hotspots. This is of significant clinical concern given the known association between skin dose and major wound complications during preoperative RT for eSTS. By evaluating daily cone-beam CT (CBCT) images from thirteen eSTS patients treated with VMAT RT, we identify daily surface contour changes (range: 2 mm-15 mm, median: 8 mm) which are frequently below conventional adaptive replanning thresholds. When applied under experimental conditions, these external contour changes did not have major impacts on target volume coverage (range: 30.2%-91.2%, mean: 72.5%) but did result in unexpected hotspots of 125.8% on average (range: 110.0%-142.2%) in the subcutaneous tissues. To mitigate this issue, we develop a methodology for VMAT treatment planning using flash PTV and virtual bolus (VB) to produce robust treatment plans that are more resistant to target volume changes, surface contour changes, and setup uncertainties than conventional planning methods. With this methodology, robust plans were equivalent to standard plans at baseline, but, after incorporation of surface volume changes, both maintained target volume coverage (p < 0.001) and prevented development of subcutaneous hotspots (p < 0.001) better than standard plans. This method may facilitate development of robust VMAT treatment plans that minimize development of subcutaneous hotspots and preserve target volume coverage in the context of routine volumetric changes during pre-operative RT.

Methods: Thirteen eSTS patients with PTV's extending close to the skin surface were simulated supine with their affected lower extremity immobilized in a vac-loc bag. Organs at risk included ipsilateral bones, subcutaneous strip, joints, skin (5 mm rind), and uninvolved body areas.
Treatment cone-beam CT (CBCT) images were evaluated and the largest external volume change from the plan CT was measured. For a representative example patient, we developed verification plans with varying degrees of PTV cropping from the external contour (3 mm and 5 mm) and degrees of uniform artificial tissue expansion (0 mm, 5 mm, 10 mm, and 15 mm) to explore these parameters. For the entire study population, an evaluation PTV (PTV_3mm_Eval) for optimization was cropped 3 mm from the surface. A 7 mm expansion was created from PTV_3mm_Eval and was cropped within the 5 mm skin rind (PTV_7mm). A 2 cm bolus was created for planning (Bolus_2cm), which covered PTV_7mm at the surface and was assigned a Hounsfield Unit (HU) value of -400. A 1.5 cm expansion from PTV_7mm (PTV_Flash) was created to provide a region of flash around the evaluation PTV and was cropped 1 mm outside PTV_3mm_Eval (Figure 1).
Eclipse with Acuros XB version 16.1.0 was used with VMAT on a Varian Truebeam linear accelerator. Plans used the arc angles, collimator angles, and field sizes from each original treatment plan, with adjustments made only to accommodate the VB and PTV_Flash as needed (Figure 2). Plans were normalized so that 100% of the prescription dose covered 95% of the PTV.
Initial plans were calculated on the original planning CT scan without any changes to external body contour. To create the verification plan dataset, the external body contour was expanded by 1.5 cm in the areas where the PTV_7mm was in the skin contour (External_15mm). Parts of External_15mm extending outside the original external contour were isolated (Res_Override_Tis) and assigned an HU value of –101, to model the expansion of the affected limb during treatment. A 5 mm skin rind was created from External_15mm for dose evaluation (Skin_New_5mm, Figure 3).
Initial and verification plans were created with virtual bolus (VB) and with no virtual bolus (NB). Initial plans were re-optimized using the optimization objectives from each patient’s original plan. Virtual bolus planning objectives included the addition of a lower objective asking for 75% of the prescription dose to PTV_Flash to control hot spots upon tissue expansion. Plans with VB were optimized with the bolus linked in Eclipse. After optimization, the VB was removed from the plan, and the dose was re-calculated with the same MLC sequence and monitor units (MUs) to create the VB plan. The final VB plan was evaluated to ensure that the dose remained clinically acceptable after removing the bolus. After the initial VB and NB plans were created, the plans were recalculated using the verification dataset to see the change in dose after volume changes had occured.

Results: The maximum external contour change observed (15 mm) was applied to the study.
Using a single patient from the cohort, with a PTV cropped 5 mm from the skin surface, we found no appreciable difference in hotspots regardless of external contour change (range: 107.8%-109.2%; Table 1). For a PTV cropped 3 mm from the surface, an increase in hotspot was observed after external contour change with apparent relationships to the degree of external contour change and the location in which hotspots were measured. The greatest hotspot (116.0%) occurred with 15 mm of external contour change and within a 15 mm external rind. These differences generally increased as external contour change increased and were more pronounced at depths closest to the original skin location; however, increased hotspots were observed with external contour changes as small as 5 mm (Table 1).
We investigated these effects in the overall cohort. Plans with VB better maintained V100% PTV coverage after external contour changes compared to NB (p < 0.001; Table 2). Without VB, external contour changes comparable to that seen on CBCT (15 mm) resulted in a mean V100% PTV coverage of 72.5% (range: 30.2%-91.2%), while those with VB and flash PTV resulted a mean V100% PTV coverage of 93.1% (range: 85.5%-97.4%) on verifications plans (Table 2).
Plans with VB were more robust with regard to hotspots after external contour changes compared to NB (p < 0.001; Table 3). Without VB, extremity contour changes resulted in the development of a mean difference in hotspot of 14.6% (range: –0.5%-31.0%) on verifications plans when compared to initial plans (Table 3). With VB and flash PTV there was no appreciable hotspot development on verification plans (mean: -0.47%, range: -1.6%-1.8%; Table 3; Figure 4). This increase in robustness was achieved without increasing hotspots on the initial plans (initial NB plan hotspot mean: 111.2%, range: 108.7%-114.1%; initial VB plan hotspot mean: 111.7%, range: 109.2%-114.1%; Table 4).

Conclusion: This study demonstrates, to our knowledge for the first time in eSTS, that use of virtual bolus and flash PTV not only preserves target coverage with the relatively large changes in extremity contour that are often observed in eSTS, but also mitigates hotspots that can develop with relatively minor changes in external contour. Limitations of this study include its small size and single-institution nature. It is also important to note that we also used artificially expanded limb contours as the basis for presented calculations. These expansions were chosen to reflect the range of external contour changes actually observed on daily CBCT in the study populations, and the controlled nature of these artificial expansions allowed detailed investigation into the effect of various changes on the relevant plans. Overall, however, the results of this study suggest an important but heretofore under-recognized possible pitfall in VMAT planning in extremity sarcomas, as well as a feasible methodology to overcome it.