Abstract
Purpose
We address the challenges associated with applying a single isocenter treatment technique to extracranial oligometastases.
Methods
We propose a technique that uses a Single Isocenter with Distinct Optimizations (SIDO) in which all Volumetric Modulated Arc Therapy (VMAT) fields share an isocenter but each field treats only one target. When necessary, setup uncertainties from rotations and deformations are mitigated by applying a couch translation between VMAT arcs, and interplay is minimized using dynamic conformal arcs (DCA) as the starting point for inverse optimization. Using CBCTs from eleven previous SBRT cases we determined the likelihood of needing a translational shift between SIDO arcs to correct rotational uncertainties. We also compared SIDO and SIDO with DCA to single (VMAT) and dual (VMAT and DCA) isocenter plans for phantom (N=11) and patient (N=4) cases.
Results
Spatial uncertainties from inter-fractional rotations were greatest for large target separation and small margins, with ~30% of fractions requiring a correction (3mm margin and 10cm separation). SIDO with DCA greatly decreased arc modulation, with approximately 20% more modulation than DCA, and similar conformity to conventional VMAT. SIDO and SIDO with DCA had comparable conformity to single and dual isocenter plans when separation between PTVs is >3cm, while traditional single isocenter VMAT had superior conformity for <3cm. SIDO with DCA had superior GI over other planning techniques for almost all cases.
Conclusion
SIDO for extracranial oligometastases allows flexibility to mitigate spatial uncertainties from rotation and deformation, and has comparable dosimetry to traditional VMAT with low modulation when inverse optimization begins with DCA.
Keywords: Oligometastases, SBRT, single isocenter, multifocal, VMAT, SIDO
Introduction
The clinical state of oligometastases has been proposed as an intermediate state between locoregionally confined cancer and widespread metastases, in which metastases are limited in number and destination organ with a corresponding indolent natural history. Stereotactic body radiation therapy (SBRT) of extracranial oligometastases is a topic of interest in radiation oncology[1], and has been applied for numerous primary diseases including lung[2], breast[3], prostate[4], colorectal cancer[5], among others, and metastatic sites[2]. It is also the subject of ongoing multi-institutional clinical trials (including NRG BR001)[4,6]. Treatment planning for multiple target stereotactic ablative radiotherapy is challenging. It is unclear when multiple metastases are in close proximity whether treatment with dual or single isocenters is ideal.
Single isocenter treatment of multiple targets is a recent development of intracranial radiosurgery. Recent studies address the challenges associated with this technique, and show that results similar to traditional SRS are possible with added benefits including decreased treatment time[7–14]. It is possible that many of the benefits of intracranial single isocenter techniques may also apply when translated to extracranial single isocenter techniques. These benefits associated with intracranial single isocenter techniques include shorter treatment and setup times, and potentially lower monitor units due to less time on the machine and less leakage/scatter to the patient, as well as improved utility and cost effectiveness[15].
However there are also added complexities for extracranial application that make this technique more challenging at 1st glance. One major challenge is the added possibility of deformation and rotation of the multiple targets relative to each other, since extracranial rotations and deformations can be expected to be greater than intracranial, and the distance between targets may also be greater. Since SBRT utilizes highly conformal dose distributions and strict management of spatial uncertainties to minimize CTV to PTV margins, even small rotations or deformations of the patient anatomy may cause unacceptable dosimetry for targets located a distance from each other, even when an optimal patient translation is applied. A 2nd challenge is the possibility of dosimetric interplay between patient intra-fractional motion and the fluence modulation during a volumetric modulated arc therapy (VMAT) arc which has been analyzed in many studies, and while most show the effect is small, it is nonetheless a real effect and is most prevalent for treatments with few fractions such as SBRT[16–19]. In this study, we address these added challenges inherent in applying a single isocenter technique of multiple extracranial targets and we investigate treatment-planning strategies for various treatment sites and geometries (varying the distance between targets). We focus on the simple case when two extracranial targets are present, with the added complexity of a greater number of targets to be addressed in future work.
Methods and Materials
Inter-fractional Rotation and Deformation
To address the challenges of inter-fractional rotation and deformation associated with treating multiple extracranial targets simultaneously, we devised a single isocenter technique, where each VMAT arc is associated with and is optimized to deliver dose only to one specific target. By each arc delivering dose to each target separately, this technique allows for three treatment procedure options illustrated in Figure 1. If the initial CBCT indicates negligible rotational and deformational discrepancies from the planning CT and little intra-fractional uncertainties are expected, the patient may be aligned based on the initial CBCT, followed immediately by delivery of the two treatment arcs. However, if the initial CBCT indicates rotation or deformation of the two targets relative to the planning CT, then a translational shift is 1st applied to align target 1, after which the 1st arc is delivered. After delivery of the 1st arc, a 2nd translational shift may be applied to then align target 2 for treatment. This shift may be made based on the 1st CBCT for cases where intra-fractional rotations and deformations are not of concern, or if intra-fractional rotation or deformation is suspected a 2nd CBCT may be acquired for verification between arcs.
Figure 1.
A single isocenter technique with arcs treating distinct targets combines the advantage of uninterrupted treatment of targets (option 1), with the flexibility to manage inter-fractional (option 2) and intra-fractional (option 3) rotations and deformations that occur relative to the treatment plan.
We analyzed daily CBCT setup images from SBRT patients including both lung and liver, to determine the likelihood of requiring a 2nd translational correction between arcs for the single isocenter technique due to inter-fractional rotations. To do so, a 6D image registration between the CT simulation image and each pre-treatment CBCT was used to determine the daily pitch (x), yaw (y), and roll (z). For the lung and liver cases separately and combined, we calculated the fraction of daily treatments for which the CTV would fall outside the PTV, assuming a CTV to PTV margin ranging from 0.1cm to 1cm. This was calculated using the rotational axis with the largest rotational uncertainty in any direction (as a conservative estimate) and assuming normally distributed inter-fractional rotations. The threshold angle beyond which the CTV would fall outside the CTV to PTV margin was calculated using:
(1).
where m is the CTV to PTV margin and d is the separation between PTVs. The threshold angle represents the maximum rotation of the patient, beyond which the CTV for the 2nd target would fall outside the CTV to PTV margin, after the 1st CTV is aligned for treatment.
Interplay Between Intra-fractional Motion and Fluence Modulation
In an effort, to minimize dosimetric interplay, we investigated a method to combine the benefits of open fields such as those of dynamic conformal arcs (DCA) with the benefit of inverse optimization from VMAT. Static 3D beams and DCA utilize open fields, and thus have the advantage of the MLCs never blocking the target volume, therefore avoiding under-dosing the target via dosimetric interplay. Within the Eclipse treatment planning system (Varian Medical Systems, Palo Alto CA) the user can begin an inverse optimization using the current treatment plan as the optimization starting point. We investigated using a DCA treatment plan as the starting point for the VMAT optimization to determine whether the final plan would have improved dosimetry over DCA and still maintain decreased blocking of the target by the MLCs compared to the standard VMAT plan.
To evaluate the susceptibility of each plan to dosimetric interplay, we calculated a modulation factor (MF) per plan, which we defined as the ratio of the average area of the PTV in the beam’s eye view, 
, to the average of the open area of the MLC’s, 
, throughout the treatment arc. In other words:
(2).
We calculated from the open MLC area of a DCA plan with no added margin around the PTV, which may slightly overestimate . Ideally, the MF should be close to 1, with a value less than 1 corresponding to MLC coverage larger than the PTV (expected for DCA with PTV to MLC margin) and a value greater than 1 corresponding to MLC coverage smaller than the PTV. We would expect plans with a larger MF to be more susceptible to dosimetric interplay. In addition to MF, we also compared the total number of monitor units (MU) per arc and plan, after the final dose distribution was normalized. After normalizing to achieve the same PTV coverage, the total MU also serves as a surrogate for susceptibility to dosimetric interplay because lower MU can be assumed to indicate less blocking of the PTV/CTV throughout the arc. We expect total MU to be proportional to dosimetric interplay, and inversely proportional to MF.
Treatment Planning and Evaluation
All treatment plans were prepared in the Eclipse Treatment Planning Software V 13.6, using the Anisotropic Analytical Algorithm (AAA) V 13.6.23 for dose calculation. All treatment plans were made with 6 MV FFF beams. To prepare single isocenter plans with distinct optimizations, we created separate VMAT plans for each target, with a single arc, which was aligned to the multi-target isocenter. MLCs with 180 control points were added to the field and the collimator jaws were optimized to the target treated by the current VMAT arc. Jaw tracking was enabled whenever possible to minimize leakage. For cases where DCA were used as the starting point for VMAT optimization, the DCA plan was prepared and dose was calculated. VMAT inverse optimization was carried out to achieve optimal target coverage and maximum sparing of normal tissue. Inverse optimization criteria were based on NRG BR001[6]; we utilized the normal tissue objective (priority = 250) and PTV objectives of V100%≥95% (priority = 200) and Dmax≤160% (priority = 100). All plans attempted to separate the 80% isodose line between targets if possible (also a requirement of NRG BR001). Once a plan was created per PTV, the arcs were combined into a single plan, but not re-optimized. The plan was then re-normalized to account for dose contribution from each arc.
We compared SIDO, with and without DCA as the inverse optimization starting point, to three different treatment planning techniques including single isocenter VMAT (treating both targets simultaneously), dual isocenter VMAT, and single isocenter DCA. When creating the plans for the phantom and patient cases, we used consistent parameters to ensure for uniformity. These comparisons were made for phantom anatomy created virtually within the treatment planning system as well as for four patient cases. The phantoms were created with the contouring function of the Eclipse TPS. An example of one of the virtually created phantoms can be seen in Figure 2. Each consists of cylindrical geometry with two spherical targets (2cm diameter) separated by 1 to 15cm (target separation can be seen by the red distance indicator in Figure 2 and is defined as the shortest distance between the edges of each target); dose prescription was set to 3 fractions of 15 Gy each. The patient cases included one with a sacrum and iliac target (3 fractions of 10 Gy, ~2.5cm of separation), one with a left and right adrenal gland target (3 fractions of 15 Gy each, with ~3cm of separation), one with two liver targets (10 fractions of 4 Gy, ~6.25cm of separation), and one with two lung targets (5 fractions of 10 Gy each, with ~11cm of separation). With the exception of the lung case, the lesions in each patient case were in the same axial plane. For the lung case, there was approximately 2.5cm separation in the longitudinal direction from the inferior surface of the superior lesion to the superior surface of the inferior lesion. The volumes of each lesion in the patient cases can be seen in Table 1.
Figure 2.
Screen shot from the Eclipse TPS illustrating an example of one of the virtual phantoms created with the contouring feature. Targets for all virtual phantoms have a diameter of 2cm and are in the same axial plane. Separation between targets is defined as the shortest distance from the edge of the 1st target to the edge of the 2nd target (as indicated on image). Target separation is the only parameter that was changed between virtual phantoms.
Table 1.
Volumes for the two targets in each of the patient cases.
| Sacrum and Iliac | L and R Adrenal Glands | Liver | Lung | |
| Lesion 1 | 11.72cc | 70.33cc | 54.74cc | 21.34cc |
| Lesion 2 | 8.02cc | 31.56cc | 91.55cc | 9.45cc |
Plans were analyzed with respect to conformity index (CI), gradient index (GI), and modulation (MLC opening with MF and total MU). We define CI, and GI as follows:
(3).
(4).
where VPTV is the volume of the planning target volume, V100% is the volume receiving 100% of the prescription dose, and V50% is the volume receiving 50% of the prescription dose.
Results
Figure 3 shows the sacrum and iliac patient case example of the individual and combined treatment plans using SIDO with DCA as a starting point for optimization.
Figure 3.
Screen shots from the Eclipse TPS illustrating the SIDO approach. Image 1 shows a treatment plan for an iliac target (VMAT Arc 1), and image 2 shows a sacrum target (VMAT Arc 2). Image 3 is the combined SIDO plan.
Figure 4 shows the frequency in which a known shift would need to be applied as measured from eleven SBRT patients, including both lung (n=6; immobilized via Bodyfix) and liver (n=5; 2 immobilized via alpha cradle and 3 via Bodyfix) treatment sites. Spatial uncertainty caused by inter-fractional rotations is greatest when target separation is the largest, which means that targets with the largest separation would require an intra-fraction translational shift. As the CTV to PTV margin is decreased and the separation between targets increases, the probability of needing a 2nd shift when treating two targets simultaneously increases rapidly, which is a possibility using SIDO (see Figure 1) but not for traditional single isocenter treatment plans.
Figure 4.
Probability of needing to apply a 2nd shift when treating two targets simultaneously for target separations of 5, 10, and 15cm. Data from eleven lung (n=6) and liver (n=5) patient cases.
For the phantom cases, using DCA as a base plan for optimization helped maintain tight MLC conformity and decreased monitor units. The resulting plans were similar to conventional VMAT in that they maintained a high conformity index, but with lower modulation (and hence less probability of dosimetric interplay). Monitor units decreased by an average of 36% (with a maximum decrease of 62%) and conformity index remained approximately the same with an average increase of 1% for SIDO with DCA as opposed to SIDO without DCA.
Figure 5 shows conformity indices, gradient indices, and modulation factors for both phantom and patient cases. From the phantom study, it is evident that SIDO and SIDO with DCA have comparable CIs to single and dual isocenter plans when the separation between PTVs is greater than 3cm, while a traditional single isocenter plan has higher CI when the separation of PTVs decreases below 3cm. It is also evident from this figure that SIDO with DCA achieved the optimal GI over all other planning techniques for almost all distances. MF was also closest to 1 for SIDO with DCA for both phantom and patient cases. The trend for total monitor units was similar to MF: DCA plans had the lowest MU, followed by similar MU for single-isocenter and SIDO with DCA; the largest MU was for dual isocenter and SIDO plans.
Figure 5a.
Conformity indices, gradient indices, and modulation factors for single isocenter, dual (or separate) isocenter, DCA, SIDO, and SIDO with DCA for the phantom cases.
For the most part, the patient cases corroborated the trends from the phantom cases, with variations from these trends most likely being due to the more complicated geometry and heterogeneity of the patient cases. One notable exception is the low conformity index observed for the SIDO with DCA plan in the adrenal glands case. This is likely due to the complex geometry for this case combined with our attempt to break up the 80% isodose lines and the fact that the separation between the targets was 3cm, which is at the limit where SIDO achieved comparable results to other planning strategies for the phantom cases. For this case, the conformity index was better for the SIDO plan (not SIDO with DCA).
Figure 5b.
Conformity indices, gradient indices, and modulation factors for single isocenter, dual (or separate) isocenters, DCA, SIDO and SIDO with DCA for the patient cases.
We also investiaged the use of jaw tracking. Adding jaw tracking was beneficial to our plan in that it decreased the gradient index by 14.5% on average, while hardly affecting conformity. Jaw tracking was not possible, however, for DCA and SIDO with DCA plans because jaw tracking is not enabled for DCA within the optimization. The plans presented in our analysis did not use jaw tracking for planning consistency.
Discussion
Multifocal SBRT is a technique that demands long treatment times, high costs, and high accuracy. Furthermore, major challenges arise from the possibility of deformation and rotation of the multiple targets relative to each other and the possibility of dosimetric interplay between patient intra-fractional motion and the fluence modulation during a VMAT arc. We have found that the SIDO planning technique is good at combating intra and inter-fractional rotation and deformation associated with treating multiple extracranial targets, has comparable dosimetry to traditional VMAT, and maintains a low modulation factor when using DCA as a starting point for VMAT optimization.
One limitation of using SIDO and SIDO with DCA is the distance between targets. For target separations of less than approximately 3cm, SIDO performs worse than a plan that uses a single isocenter and treats both targets simultaneously. The gradient index and monitor units of SIDO and SIDO with DCA is lower at these small distances; however, due to the large decrease in conformity index these plans do not warrant benefits. Thus, if there is less than a 3cm separation between targets, treating both targets with a common isocenter simultaneously is a better approach than SIDO or SIDO with DCA. In these cases, the workflow in Figure 1 would not be possible. However, for these cases the need for the flexible workflow in Figure 1 is less essential due to the close proximity of the targets (see Figure 4).
SIDO performs the best with larger PTV separations for two reasons. The 1st reason is that the dosimetry when using SIDO is the best at larger separation distances. At larger target separations we see a higher conformity index, lower gradient index, and lower modulation factor. Also, at larger target separations we have shown that the probability of needing a translational shift greatly increases. This means that the larger the separation of PTVs, the more flexibility in treatment delivery technique that is required, which is offered by SIDO.
One limitation of the phantom study was that it was strongly controlled with respect to the size, shape, and location of PTVs (spheres of 2cm diameter evenly spaced from the center of a cylindrical chamber). With irregular shapes, various sizes and locations of PTVs, SIDO and SIDO with DCA may need to be tweaked in order to maintain high conformity. However, similar trends were observed for the patient cases, which indicate that the results obtained in the phantom study may generally apply to real cases. Another variable unaccounted for was constraints on organs at risk. Also, in this study we focused on treatment of only two targets; the planning process becomes more complex when more than two targets are present. For instance, targets sharing MLCs become non-trivial with 3 or more targets; these complexities will need to be the focus of future work as they are not addressed here.
Based on our analysis, we propose a planning technique, SIDO (Single Isocenter with Distinct Optimizations) with DCA for target separations greater than or equal to 3cm, and simultaneous treatment using VMAT for targets separated by less than 3cm. We propose using SIDO with DCA due to its advantages of lower monitor units and modulation while maintaining a high conformity and gradient index. When SIDO is used, a number of treatment workflows are possible based upon the inter and intra-fractional uncertainties (Figure 1); these options should be weighed carefully by the treatment team based upon the expected uncertainties so as to avoid dosimetric underdose from localization errors. Our findings on the SIDO treatment planning technique are encouraging, and lead to further focuses of investigation. Next steps that should be taken to progress this study include looking into different sized targets, the impact of intra-treatment respiratory motion, and different target locations.
Conclusion
This study investigated two new treatment planning methods for multiple extracranial oligometastases: SIDO and SIDO with DCA. We demonstrated that SIDO with DCA has comparable dosimetry to traditional VMAT with low modulation similar to DCA for PTV separations greater than or equal to 3cm; and the opportunity to use a known translational shift to combat intra and inter-fractional inconsistencies. For smaller separations, a single isocenter VMAT technique treating both targets simultaneously is preferable.
Acknowledgments
This research was not supported by extramural funding.
Authors’ disclosure of potential conflicts of interest
Authors’ disclosure of potential conflicts of interest
Fang-Fang Yin discloses research and licensing agreements with Varian Medical Systems, not related to this study. Justus Adamson discloses a consulting arrangement with Immunolight LLC and ownership with ClearSight Radiotherapy Products, both of which are unrelated to this study. Michael Trager and Joseph Salama have nothing to disclose.
Author contributions
Conception and design: Michael Trager, Joseph Salama, Fang-Fang Yin, Justus Adamson
Data collection: Michael Trager, Justus Adamson
Data analysis and interpretation: Michael Trager, Justus Adamson
Manuscript writing: Michael Trager, Justus Adamson, Joseph Salama
Final approval of manuscript: Michael Trager, Joseph Salama, Fang-Fang Yin, Justus Adamson
References
- Corbin KS, Hellman S, Weichselbaum RR. Extracranial oligometastases: A subset of metastases curable with stereotactic radiotherapy. J Clin Oncol 2013;31(11):1384–90. [DOI] [PubMed] [Google Scholar]
- Norihisa Y, Nagata Y, Takayama K, Matsuo Y, Sakamoto T, Sakamoto M, et al. Stereotactic Body Radiotherapy for Oligometastatic Lung Tumors. Int J Radiat Oncol Biol Phys 2008;72(2):398–403. [DOI] [PubMed] [Google Scholar]
- Milano MT, Katz AW, Zhang H, Okunieff P. Oligometastases treated with stereotactic body radiotherapy: Long-term follow-up of prospective study. Int J Radiat Oncol Biol Phys 2012;83(3):878–86. [DOI] [PubMed] [Google Scholar]
- Muldermans JL, Romak LB, Kwon ED, Park SS, Olivier KR. Stereotactic Body Radiation Therapy for Oligometastatic Prostate Cancer. Int J Radiat Oncol Biol Phys [Internet] 2016;95(2):696–702. Available from: http://www.ncbi.nlm.nih.gov/pubmed/27131082 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Takeda A, Kunieda E, Ohashi T, Aoki Y, Koike N, Takeda T. Stereotactic body radiotherapy (SBRT) for oligometastatic lung tumors from colorectal cancer and other primary cancers in comparison with primary lung cancer. Radiother Oncol 2011;101(2):255–9. [DOI] [PubMed] [Google Scholar]
- NRG Oncology. A Phase 1 Study of Stereotactic Body Radiotherapy (SBRT) for the Treatment of Multiple Metastases [Internet]. 2014. Available from: http://www.rtog.org/ClinicalTrials/ProtocolTable/StudyDetails.aspx?action=openFile&FileID=11271
- Nath S. Single-Isocenter Frameless Intensity-Modulated Stereotactic Radiosurgery for Simultaneous Treatment of Multiple Brain Metastases: Clinical Experience. Int J Radiat Oncol Biol Phys 2010;78:91–97. [DOI] [PubMed] [Google Scholar]
- Clark G. Plan quality and treatment planning technique for single isocenter cranial radiosurgery with volumetric modulated arc therapy. Pract Radiat Oncol 2012;2:306–13. [DOI] [PubMed] [Google Scholar]
- Clark G. Feasibility of single-isocenter volumetric modulated arc radiosurgery for treatment of multiple brain metastases. Int J Radiat Oncol Biol Phys 2010;76:296–302. [DOI] [PubMed] [Google Scholar]
- Hardcastle N. On a single isocenter volumetric modulated arc therapy SRS planning technique for multiple brain metastases. J Radiosurg SBRT 2012;2:1–9. [PMC free article] [PubMed] [Google Scholar]
- McDonald D. Comparison of radiation dose spillage from the Gamma Knife Perfexion with that from volumetric modulated arc radiosurgery during treatment of multiple brain metastases in a single fraction. J Neurosurg 2014;121:Suppl:51–9. [DOI] [PubMed] [Google Scholar]
- Lau S. Single-Isocenter Frameless Volumetric Modulated Arc Radiosurgery for Multiple Intracranial Metastases. Neurosurgery 2015;77:233–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Thomas E. Comparison of plan quality and delivery time between volumetric arc therapy (RapidArc) and Gamma Knife radiosurgery for multiple cranial metastases. Neurosurgery. 2014;75:409–17, discussion 417–408. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Stanhope C, Chang Z, Wang Z, et al. Physics considerations for single-isocenter, volumetric modulated arc radiosurgery for treatment of multiple intracranial targets. Pract Radiat Oncol. 2016;6:207–213. [DOI] [PubMed] [Google Scholar]
- Lester-Coll NH, Rutter CE, Bledsoe T, Goldberg SB, Decker RH, Yu JB. Cost-Effectiveness of Surgery, Stereotactic Body Radiotherapy and Systemic Therapy for Pulmonary Oligometastases. Int J Radiat Oncol [Internet] 2016;95(2):663–72. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0360301616000419 [DOI] [PubMed] [Google Scholar]
- Rossi MMG, Peulen HMU, Belderbos JSA, Sonke J-J. Intrafraction Motion in Stereotactic Body Radiation Therapy for Non-Small Cell Lung Cancer: Intensity Modulated Radiation Therapy Versus Volumetric Modulated Arc Therapy. Int J Radiat Oncol [Internet] 2016;95(2):835–43. Available from: http://linkinghub.elsevier.com/retrieve/pii/S0360301616001139 [DOI] [PubMed] [Google Scholar]
- Ong CL, Dahele M, Slotman BJ, Verbakel WFAR. Dosimetric impact of the interplay effect during stereotactic lung radiation therapy delivery using flattening filter-free beams and volumetric modulated arc therapy. Int J Radiat Oncol Biol Phys 2013;86(4):743–8. [DOI] [PubMed] [Google Scholar]
- Stambaugh C, Nelms BE, Dilling T, Stevens C, Latifi K, Zhang G, et al. Experimentally studied dynamic dose interplay does not meaningfully affect target dose in VMAT SBRT lung treatments. Med Phys [Internet] 2013;40(9). Available from: http://scitation.aip.org/content/aapm/journal/medphys/40/9/10.1118/1.4818255 [DOI] [PubMed] [Google Scholar]
- Rao M, Wu J, Cao D, Wong T, Mehta V, Shepard D, et al. Dosimetric impact of breathing motion in lung stereotactic body radiotherapy treatment using image-modulated radiotherapy and volumetric modulated arc therapy. Int J Radiat Oncol Biol Phys 2012;83(2):e251–6. [DOI] [PubMed] [Google Scholar]