If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Comparing the robustness of different skin flash approaches using wide tangents, manual flash VMAT, and simulated organ motion robust optimization VMAT in breast and nodal radiotherapy
Compare the robustness of wide tangents (WT) and volumetric modulated arc therapy (VMAT) using different skin flash approaches in breast and nodal radiotherapy. Ten patients treated with WT using 2-cm flash were replanned with VMAT using no flash (NF), manual 2-cm flash (MF), and robust optimization (RO). Plan robustness was assessed for target coverage and organs at risk (OAR) by recalculating on 5 deformed CT scans (SOM1-5), daily cone beam (CBCT), and by shifting the isocenter 5 mm. VMAT NF gave poor coverage of CTVp with its smallest change of −3.2% for V38Gy on CBCT. VMAT RO plans showed the least variations in target coverage loss compared to WT and VMAT MF which dropped as anatomical swelling increased. CTVp D0.5cc decreased on CBCT and increased most for VMAT MF plans (case max increase +3.3 Gy), whereas VMAT RO plans were relatively stable (case max increase +1.2 Gy). OAR dose changed little with anatomical changes (isocenter shifts more important with medial, posterior, and inferior increasing dose). Nodal coverage was superior for VMAT which led to the WT being less robust for coverage toward both geometric and anatomical uncertainties. All techniques except NF plans gave high levels of coverage under minor uncertainties. VMAT RO was highly robust for target coverage for anatomical changes. Manually editing control points on VMAT plans was time-consuming and less predictable. CBCT anatomical changes were modest resulting in small delivered dose changes. OAR dose changes were small with no significant differences between techniques.
Breast cancer has now overtaken lung cancer as the world's most commonly diagnosed cancer, according to statistics released by the International Agency for Research on Cancer in 2020 with 2.26 million new cases.
Breast radiotherapy plays an important part of the multimodality treatment of breast cancer and a vital role in maximizing local disease control, enabling safe breast conservation and contributing to increased survival.
The most common technique in the UK and others used for these patients tends to be a wide tangents (WT) using wedges and or some segments to achieve a uniform dose.
Implementation of Royal College of Radiologists Consensus Statements and National Institute for Health and Care Excellence Guidance: breast radiotherapy practice in the UK.
For the axillary nodes, an anterior field is often applied with or without a posterior field. This type of technique is well established but can often be accompanied by inability to achieve optimal coverage of nodes while maintaining low acceptable levels of dose to organs at risk (OAR).
Matching tangents to nodal fields can also be problematic which may give rise to hot or cold regions between field junctions. In efforts to address the above, volumetric modulated arc therapy (VMAT) is becoming an increasingly used solution that can achieve high target coverage and spare OARs while using only one isocenter. VMAT has been demonstrated to give improved target homogeneity, conformity, and reduce high dose spillage to OARs.
In order to ensure adequate coverage of the clinical target volume (CTV) breast/chest wall, an adequate skin flash margin is required to account for geometrical and anatomical uncertainties such as motion and breast oedema. This ensures the dosimetry is robust to account for on treatment changes. Seppälä et al. showed that during breast radiotherapy that an outer 8 mm margin is needed to account for over 95% of contour changes during treatment. An extreme of 15 mm outer contour change was found across 93 patients on 1731 cone beam computer tomography (CBCT) scans.
Traditionally, the WT techniques use a manual flash (MF) approach retracting the jaws/leaves beyond the patient into air by ∼ 2 cm to ensure that in presence of anatomical changes/motion the CTV will still receive adequate dose. This inherently allows for a level of plan robustness to maintain coverage without increasing normal tissue dose as the field is expanded into air. For rotational intensity modulated radiotherapy (IMRT), the task is more difficult but there are various methods to achieve adequate flash such as manually moving jaws/multileaf collimators (MLCs), using a bolus override structure of some sort outside the body and robust optimization (RO) strategies such as simulated organ motion (SOM) CT scans.
Different planning systems can offer different flash tools to achieve the same goal. Some previous studies have looked at robustness for 3D and VMAT plans using CBCT, simulating tissue expansions/contractions, and the shifting isocenter.
How these different approaches compare with one another and to the traditional tangential flash approach is of interest to ensure an acceptable level of dosimetry is maintained during treatment in the presence of different anatomical deformations and geometrical displacements. This current study will look at 4 different techniques including a manual VMAT flash approach which is generally unheard of in the literature. This may be of interest to those where SOM RO tools are not available. Whereas previous works have looked at different flash approaches, there is less data on specifically SOM RO and how this compares to traditional WT and a manual VMAT flash. This study will aim to include not just anatomical changes via deformable image registration simulations but also estimated delivered doses using CBCT data and also look at isocenter shift effects for both primary and nodal targets as well as OARs.
Phase 3 multicenter randomized controlled trial “PARABLE” (Proton beAm the RApy in patients with Breast cancer: Evaluating early and Late-Effects) study in the UK will randomize between protons and photons where robust VMAT plans will be the standard technique for the photon arm.
It is vital to ensure that as VMAT techniques are adopted nationally that we can be confident in delivered dose in the presence of uncertainties and that is quantified against traditional WT techniques. In our center, we are transitioning to a SOM robustly optimized VMAT technique for these nodal patients including IMN and with such aim to compare and analyze its robustness to anatomical changes and isocenter shifts with WT and two other VMAT flash approaches (VMAT no flash (NF) and VMAT MF).
Material and Methods
Sample and volumes
Ten patients with breast cancer who were treated with WT to their breast/chest wall and axillary nodes including IMN were retrospectively replanned by an experienced dosimetrist with 3 VMAT flash approaches. Treatment and patient characteristics are summarized in Table 1. Patients underwent a planning deep inspiration breath hold (DIBH) CT scan with 3 mm thick slices. All were positioned supine with arms raised up above head on a standard breast board. Outlining of the nodal CTVs was done according to ESTRO guidelines.
The whole breast/chest wall planning target volume (PTVp) was a field-based volume based on the tangential fields from the WT plans. This was clipped away from the body and lung by 5 mm. This approach has been previously described in the FAST FORWARD trial planning pack.
Hypofractionated breast radiotherapy for 1 week versus 3 weeks (FAST-Forward): 5-year efficacy and late normal tissue effects results from a multicentre, non-inferiority, randomised, phase 3 trial.
All nodal CTVs were grown by 5 mm to create their respective PTV volumes (see Table 2 for dose-volume constraints and OARs used). All planning was in Raystation V10A (Stockholm, Sweden) using a 2-mm dose grid and on an Elekta Agility linear accelerator (Elekta Oncology Systems Ltd., Crawley, UK).
Dx, dose in Gy to an x volume; OAR, organ at risk; OAR, organ at risk; PTVp_xxxx, planning target volume primary for x dose in cGy; Vx Gy, Volume receiving x Gy; n_LN, Lymph node PTV.
WT plans consisted of a mono isocentric step and shoot IMRT technique of opposed tangential fields to treat the breast/chest wall and a single direct anterior field for the upper axillary nodes (see Fig. 1). The posterior border of the tangents was aligned straight and additional tangential segments (∼ 4 to 8) were added by inverse optimization to produce a homogenous dose distribution. The plan had two main “open” tangent beams (6 or 10 MV) which delivered approximately 75% of the dose with the remainder by the segments (6 MV). The open tangential beams had a flash margin of 2 cm from the body while the segments did not require such. This flash margin was based on historic practice for a wide range of breast patients and situations. For the direct anterior nodal field (10 MV) which is above the isocenter, the monitors units were increased until a max dose of ∼ 43.9 Gy. The max number of segments was set at 8, the minimum segment area set at 9 cm², and the minimum monitor units (MU) was 5. These WT plans were clinically checked and delivered.
Fig. 1Example beams eye view of the 4 techniques (WT, VMAT NF, VMAT MF, and VMAT RO). PTVp is red (WT field based), body is green and PTVn_LN orange. (Color version of figure is available online.)
These plans consisted of a 6 MV dual arc rotating from gantry angle 350° to 179.9° for left-sided and 10° to 181° for right-sided. The VMAT maximum dose rate was 550 MU/min with MLC leaf width of 5 mm, a constrained leaf motion of 0.6 cm/degree and gantry angle spacing of 4°. Collimator angles were rotated by 10° and revered for the second arc. Each VMAT plan consisted of the first arc rotating from medial to lateral and then the second opposite direction. PTV coverage was priority and was only compromised to achieve OAR mandatory constraints. VMAT plans were optimized by the same planner using additional optimization rings around the breast and nodal PTVs to ensure conformity and minimize hot spots. OAR doses were kept to a minimal by optimizing their Max EUD and Max DVH optimizations functions. Functions optimized used for the PTVs were min dose, max dose, and uniform dose. Three VMAT plans were created with different flash approaches (see Fig. 1 and appendix for details).
•
VMAT NF: First a standard VMAT plan was produced as described above without any additional flash margin for the breast (PTVp which was clipped 5 mm from the body contour).
•
VMAT MF: The NF plans MLCs were retracted on all allowable control points where possible to ensure a 2 cm flash margin from the skin. This consisted of firstly creating a flash contour which was 2.5 cm expansion of the PTVp minus the body contour (see appendix). The MLCs positions were manually edited if needed and retracted to the flash contour ensuring a 2 cm margin from the body contour. Only MLC that were not inside the PTVp is where the MLCs were withdrawn back to the flash contour. This ensured that the planned dosimetry to the target was largely unaffected. Further manual optimization of the plan after MLC adjusting was sometimes required to ensure optimal homogeneity in case of new hot spots.
•
VMAT RO: The PTVp was contacted by 5 mm (except anterior and lateral) to create a CTVp. This CTVp was then used as a motion organ to simulate a max motion of 1.5 cm in the anterior and lateral direction to create 3 additional CT datasets (SOM1, 2, and 5) which were used in the plan optimization. This 15 mm margin was based on more recent literature where most extreme swellings were typically within 15 mm and all these patients would be in DIBH.
The Ipsi_Lung was kept as a fixed organ for this purpose. This process creates CT datasets and deforms the contours on to them from the planning CT so they can be used during the optimization. This ensures the motion uncertainty is accounted for when creating the control points and fluence. The NF plan was copied and reset in which then the RO plan was then created. This method has been described previously by Dunlop et al.
It consists of a minimax optimization where the optimization functions selected (PTV min dose, max dose) are considered in the worst-case scenario. See supplementary data for a detailed procedure.
Robustness evaluations
1.
SOM CT Scenarios: In addition to the 3 CT scans created in the VMAT RO process, another 2 were created which were intermediate amounts of the max 1.5 cm tissue deformations. The WT and VMAT plans were then recomputed on these 5 CT scans (SOM 1 to 5) and the resultant dose statistics were recorded. A CTVp_SP structure was also created which consisted of the most superficial 1 cm to capture the impact of the superficial component of the CTVp. This resulted in a total of 200 dose calculations (4 techniques × 5 CT scenarios × 10 patients). The 5 SOM scenarios consisted of deformation of the CTVp by various amounts (cm) per direction: SOM1: (1.5 cm lateral), SOM2: (1.5 cm anterior), SOM3: (0.75 cm lateral), SOM4: (0.75 cm anterior), SOM5: (0.79 cm lateral and 1.28 cm anterior).
2.
CBCT calculations: WT patients had daily CBCT imaging with online image matching. The CBCT scans were imported into Raystation and fused in the treatment position to the planning CT for every fraction of each plan. To calculate daily doses, additional CBCT structures were created such as a field of view (FOV), body, body minus FOV, and Ipsi_Lung (given density 1.0 g/cc), Ipsi_Lung minus FOV inside the body (given density of 0.26 g/cc). Where CBCT data were missing inferiorly due to scan lengths not capturing some inferior components of the PTVp, missing body contours were created and given a density of 1 g/cc. Each fractional dose was then computed on the CBCT using the auto image value to density table (IVDT) in Raystation and deformed back to the planning CT for summation. Estimated total delivered doses could then be assessed. This resulted in a total of 600 dose calculations (4 techniques × 15 fractions × 10 patients).
3.
Isocenter Shifts: Each plans isocenter was shifted by 5 mm in the anterior, posterior, medial, lateral, superior, and inferior direction and recalculated to assess dosimetric impact on CTV coverage and OAR doses. This resulted in a total of 240 dose calculations (4 techniques × 6 shift scenarios × 10 patients).
Statistical analysis
Plan comparisons and statistical analysis were carried between the techniques and also for dose differences compared against its plan for the various scenarios. This consisted of performing one-way ANOVA with all pairwise comparisons tested using Bonferroni multiple comparisons test, with adjusted p-values ≤ 0.05 indications statistical significance.
Results
WT and VMAT plan doses
Tables 1 and 2 in the supplementary data show the planned doses for WT and VMAT plans. Note the similar high level of coverage for PTVp. For nodal coverage, the VMAT plans are higher and more homogenous than WT. There were no significant differences between the RO VMAT plans and NF VMAT plans.
CTV dose
1
SOM 5 CT Scenarios: Fig. 2 shows for CTVp V38Gy, the NF plans were the least robust with RO plans the most becoming more superior as anatomical changes increased. WT and MF plans showed generally similar levels of robustness. The nodal CTVs V36Gy did not change significantly with anatomical for VMAT plans (< 2%) but WT demonstrated slightly larger reductions being mostly within 5%. The CTVp V42.8Gy changes was relatively stable and low for VMAT RO and NF plans and tended to be higher with anatomical changes for VMAT MF.
Fig. 2Mean CTV dose differences for the 4 techniques (WT, VMAT NF, VMAT MF, and VMAT RO) for isocenter shifts and anatomical changes. (Color version of figure is available online.)
Fig. 3Mean OAR dose differences for the 4 techniques (WT, VMAT NF, VMAT MF, and VMAT RO) for isocenter shifts and anatomical changes. (Color version of figure is available online.)
CBCT calculations: Table 3 and Fig. 2 shows anatomical changes were minimal in this cohort and reflected by its small changes in coverage across all plans. The CTVp V38Gy reduced a maximum of only 3.2% (NF plans). The WT shows more drop in coverage than VMAT plans but on average was within 5%.
3
Isocenter shifts (5 mm): The CTVp V38Gy reduced by a max of 8.4% (NF postshift). All other directions and plans had small changes for CTVp (< 5%). For the CTVn_IMN, VMAT plans showed little changes for coverage (within 2%). WT CTVn_IMN V36Gy, however, reduced by an average of 7.6%, 12.8%, and 8.1% for shift directions lateral, anterior, and superior, respectively. For CTVn_LN only the anterior and superior isocenter shift tended to give reduced nodal V36Gy but within 3% for WT. Max dose increases to CTVp were more likely for MF plans with a mean max absolute increase in D0.5cc of 0.5Gy (SOM2, -0.65 - 2.2 Gy). RO VMAT plans appeared to be most stable with the smallest changes for both D0.5cc and V42.8Gy (Fig. 2).
OAR dose
1
SOM CT scenarios: Fig. 3 and Table 4 show OAR dose changes due to anatomical deformations were very small across all techniques and typically < 0.5 Gy mean dose or < 1% change in volume receiving a dose.
2
CBCT calculations: Changes were minimal (typically < 0.3 Gy mean dose or < 1% change in volume receiving a dose). There were no obvious differences between techniques.
3
Isocenter Shifts (5 mm): Larger dosimetric changes were seen for isocentric shifts than for anatomical changes but still remained on average within 1.5 Gy OAR mean dose. The OAR dose increased for all techniques for medial, posterior, and inferior isocenter shifts. Comparing techniques showed little difference across plans except for Heart V5Gy which tended to be less sensitive to change for the WT plans.
Discussion
The aim of this study was to assess and compare the robustness of the traditional WT technique with VMAT plans using NF, MF, and an organ simulated motion robust optimized plan. To our knowledge, this is the first study that compares robustness using treatment data and simulations using a manual VMAT flash approach. All the techniques could produce clinically acceptable plans. VMAT tended to give superior nodal coverage and was more homogeneous than WT (see supplementary data). VMAT had slightly higher mean OAR doses to WT which may be due to increased coverage and low dose spillage. In terms of coverage, all techniques did not experience a large reduction in CTVp V95% on CBCT. The NF plans did demonstrate the lowest coverage as expected and this became more pronounced as anatomical swelling was simulated for SOM1-5. Studies have demonstrated anatomical changes during treatment in the range of 2.4 ± 2.1 mm with a max of 15 mm.
Although this sample did not have extremely large CBCT changes, it is clear that some flash margin is necessary to ensure coverage and reduce the need for frequent replanning during treatment.
Dunlop et al. similarly showed improved robustness to anatomical changes with the same robust VMAT approach compared to nonrobust plans on CBCT. For the breast CTV, 92% of the robust plans achieved the optimal D98% > 95% clinical goal as compared to 71% of the nonrobust plans (p < 0.01).
The current study has shown that for simulations of swelling at 7.5 mm anterior or lateral (SOM3 and 4) that the robust VMAT plans were most stable for both target coverage and hotness. All 10 patients had CTVp V38Gy > 95% for SOM3 and 4 and 9/10 patients for CTVp_SP. The RO results also demonstrate that despite optimizing over the nominal CT, SOM1, 2, and 5 CT scans that this does not assume that the plan will give adequate coverage in these scenarios. This is seen for SOM1, 2, and 5 where the RO plans CTVp V38Gy dropped by a mean of 4.4%, 5.7%, and 11.9%. Therefore this needs to be considered when evaluating these plans during planning and delivery.
Jensen et al. also looked at geometrically robust VMAT versus 3DCRT but instead of using daily CBCT, used weekly treatment offsets to calculate the perturbed doses. They found that 3D was less robust overall versus VMAT for both breast and OARs.
The current work also showed small changes for isocentric shifts for WT and VMAT RO CTVp but CTVn_IMN coverage changed significantly more in WT than VMAT. OARs in both studies changed very little with Jensen showing perhaps less change with VMAT than 3D, but this current work did not demonstrate an obvious difference based on technique except for Heart V5Gy where medial shifts gave less change for WT than VMAT.
A different study by Van der Veen et al. compared a virtual bolus robustly optimized VMAT technique compared against VMAT and 3D which was similar to the WT in this current study.
The robust analyses were on CBCT and simulating tissue expansions/contractions and set up errors. Like the current study, they showed large drops of V95% CTV breast coverage for NF plans with a few mm of expansion (∼ 20% at 5 to 7 mm). They showed V95% > 92% for expansions up to 5 mm, whereas I showed even better coverage for RO VMAT of > 94% (SOM1 to 4, CBCT). This higher coverage may be due to using a 1.5 cm SOM margin compared to and the 1 cm virtual bolus they used. Similarly they showed little change to axilla CTV V95% with anatomy. Regarding isocenter shifts, like the current work, lateral shifts decreased OAR doses and medial increased them. Although this study only used 5 mm isocenter shifts, they used up to 1 cm to show that CTV V95% drops a lot for a lateral shift for nonrobust VMAT plans (∼ 18% at 1 cm shift). This work showed highest sensitivity for a 5 mm posterior shift on the NF plans with a reduction of 8.4% for CTVp V38Gy.
The WT and MF plans are still a practical method to achieve some robustness of plans over minor anatomical deviations as shown by the CBCT and SOM3 and 4. They tend to lose their value as swelling increases to larger simulated amounts. Additionally, the WT is more sensitive to nodal coverage changes with isocenter shifts possibly due to field edges closer to the PTV in efforts to achieve OAR constraints. Crama et al. compared RO SOM IMRT tangents to manual skin flash IMRT plans calculating dose changes on repeat or synthetic CT scans where > 8 mm swelling existed.
Like the current work, changes were generally within 5% for V95% for this level of swelling with the 8/10 RO plans being more robust than the manual IMRT flash plans. A max dose drop of V95% was about 3.5% in the MF plan.
Liang et al. compared robustness of at step and shoot IMRT MF with an expanded PTV and SOM RO using 1.5 cm.
The assessment was based on recalculating on simulated CT scans with 0, 5, and 10 mm expansions. Like the current study, the nodal targets and OARs did not show any significant findings with regards to changes in dose or differences across techniques. RO was highly robust for D95% and least for MF. Their PTV max dose was highest for MF and did agree with this work showing D0.5cc highest increases on average for the MF VMAT plans showing the increased likelihood of manually creating complicated segments after non-RO. The MF VMAT approach may result in a higher chance of unpredictable hot and cold spots and is quite time-consuming, prone to error and planner variability making it more difficult to be confident in its robustness. Manually altering multiple VMAT segments also creates the need to recompute the plan and renormalize which may then need further plan editing to achieve constraints. This may be reflected by the statistically significantly higher D50% and D2% in VMAT MF compared with VMAT RO (supplementary data).
The RO VMAT plans compared with the NF, and MF plans were not statistically significantly inferior (see appendix). The benefits of RO over some other solutions include that it does not require additional outlining or density overrides. The simulated tissue deformations are included within the optimization process. This does however increase some of the planning time as more CT datasets scenarios are considered. Despite this, RO VMAT has been successfully introduced into our department and the process consists of optimizing and creating a nonrobust plan followed by a script that copied the plan and creates the robust plan so no additional tasks are required for the planner to do. This script code details can be found in the supplementary data.
This study is limited by the lack of a more thorough CBCT dose calculation accuracy validation. However, previous works have shown that the auto IVDT used in Raystation gives reasonable results.
Also, each CBCT was reviewed for any gross artifacts and IVDT altered for that patient for all its calculations. For a separate patient not in this study, a CBCT and CT scan taken on the same day were used to compare the dosimetric accuracy (as both scans where within a few hours of each other, anatomical deformations between them were negligible after rigid registration). Dose calculations demonstrated the dose difference to be largely within 2% to 5% for targets. Moreover, the same calculation table was used across all techniques for all patients so any small uncertainties will be reproduced allowing for comparisons to remain valid. Regions outside the FOV were overridden to densities such as 1 g/cc where tissue existed or 0.26 g/cc for lung. Analyzing the estimated CBCT dose to OARs outside the FOV, such as the contra breast, contra lung, and some of the heart muscle will have minor differences if a larger CBCT FOV was acquired due to assumptions of their anatomy and density. All patients in this work were in DIBH and 9/10 were chest wall patients. Having a larger patient cohort with a wider range of situations may allow for better appreciation of the dosimetric changes under those scenarios. This work does however provides insight into the different planning techniques and flash methods for these patients and quantifies specific dose changes for both targets and OARs under different anatomical and geometrical uncertainties. The MF plans also may be of interest to those where skin flash tools are not yet available such as mentioned for VMAT in Eclipse.
In regards to Bogue J, Wan J, Lavey RS, Parsai EI. Dosimetric comparison of VMAT with integrated skin flash to 3D field-in-field tangents for left breast irradiation.
VMAT plans have superior nodal coverage to WT plans. A flash margin is needed for VMAT as demonstrated by the large drop in coverage for NF VMAT plans. RO tended to have the least variations CTVp coverage although both MF and WT were not statistically different. MF VMAT plans were time-consuming and potentially give higher max dose increases in the presence of anatomical uncertainties. CBCT changes in this sample were modest so delivered dose differences were minimal on CBCT. OAR doses were largely insensitive to geometrical and anatomical changes with little difference comparing across techniques.
Authors Contribution
All authors were involved in the preparation of the manuscript. All authors reviewed and approved the final manuscript.
Conflicts of interest
The authors declare no conflicts of interest.
Acknowledgments
This work was supported by Cancer Research UK Radnet Cambridge [C17918/A28870].
Implementation of Royal College of Radiologists Consensus Statements and National Institute for Health and Care Excellence Guidance: breast radiotherapy practice in the UK.
Hypofractionated breast radiotherapy for 1 week versus 3 weeks (FAST-Forward): 5-year efficacy and late normal tissue effects results from a multicentre, non-inferiority, randomised, phase 3 trial.
In regards to Bogue J, Wan J, Lavey RS, Parsai EI. Dosimetric comparison of VMAT with integrated skin flash to 3D field-in-field tangents for left breast irradiation.
00044-9&id=gr1.jpg){kind=link}
00044-9&id=gr2.jpg){kind=link}
00044-9&id=gr3.jpg){kind=link}