- Email: [email protected]
Intensity Modulated Proton Therapy for Craniospinal Irradiation: Organ-at-Risk Exposure and a Low-Gradient Junctioning Technique
International Journal of
Radiation Oncology biology
physics
www.redjournal.org
Physics Contribution
Intensity Modulated Proton Therapy for Craniospinal Irradiation: Organ-at-Risk Exposure and a Low-Gradient Junctioning Technique Joshua B. Stoker, PhD,*,y Jonathan Grant, MD,z X. Ronald Zhu, PhD,* Rajesh Pidikiti, PhD,* Anita Mahajan, MD,z and David R. Grosshans, MD, PhDz *Department of Radiation Physics, The University of Texas MD Anderson Cancer Center, Houston, Texas; yDepartment of Radiation Oncology, Mayo Clinic Arizona, Phoenix, Arizona; and zDivision of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, Texas Received Mar 15, 2014, and in revised form Jul 2, 2014. Accepted for publication Jul 3, 2014.
Summary Intensity modulated proton therapy is a sophisticated delivery modality for which the full clinical benefit during craniospinal irradiation (CSI) is still emerging. A novel planning approach for CSI is presented which demonstrates, relative to passive scanning proton therapy, improved robustness of field junctions and superior sparing of principal organs at risk. Use of such an approach may improve clinical outcomes and enhance treatment safety and efficiency.
Purpose: To compare field junction robustness and sparing of organs at risk (OARs) during craniospinal irradiation (CSI) using intensity modulated proton therapy (IMPT) to conventional passively scattered proton therapy (PSPT). Methods and Materials: Ten patients, 5 adult and 5 pediatric patients, previously treated with PSPT-based CSI were selected for comparison. Anterior oblique cranial fields, using a superior couch rotation, and posterior spinal fields were used for IMPT planning. To facilitate low-gradient field junctioning along the spine, the inverseplanning IMPT technique was divided into 3 stages. Dose indices describing target coverage and normal tissue dose, in silico error modeling, and film dosimetry were used to assess plan quality. Results: Field junction robustness along the spine was improved using the staged IMPT planning technique, reducing the worst case impact of a 4-mm setup error from 25% in PSPT to <5% of prescription dose. This was verified by film dosimetry for clinical delivery. Exclusive of thyroid dose in adult patients, IMPT plans demonstrated sparing of organs at risk as good or better than PSPT. Coverage of the cribriform plate for pediatric (V95% [percentage of volume of the target receiving at least 95% of the prescribed dose]; 87 11 vs 92 7) and adult (V95%; 94 7 vs 100 1) patients and the clinical target in pediatric (V95%; 98 2 vs 100 1) and adult (V95%; 100 1 vs 100 1) patients for PSPT and IMPT plans, respectively, were comparable or improved. For adult patients, IMPT target dose inhomogeneity was increased, as determined by heterogeneity index (HI) and inhomogeneity coefficient (IC). IMPT lowered maximum spinal cord dose, improved spinal dose homogeneity, and reduced exposure to other OARs.
Reprint requests to: Joshua Stoker, PhD, Mayo Clinic Arizona, 5777 E. Mayo Blvd., Phoenix, AZ 85054. Tel: (801) 830-1784; E-mail: stoker. [email protected] Int J Radiation Oncol Biol Phys, Vol. 90, No. 3, pp. 637e644, 2014 0360-3016/$ - see front matter Ó 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ijrobp.2014.07.003
Conflict of interest: none. AcknowledgmentsdThe authors thank Pamela K. Allen, PhD, for running statistical analyses essential for this study.
638
International Journal of Radiation Oncology Biology Physics
Stoker et al.
Conclusions: IMPT has the potential to improve CSI plan quality and the homogeneity of intrafractional dose at match lines. The IMPT approach developed may also simplify treatments and reduce workload per patient relative to PSPT. Ó 2014 Elsevier Inc.
Introduction
Methods and Materials
Radiation therapy plays a central role in the treatment of many tumors of the central nervous system, either alone or in conjunction with surgery, chemotherapy, or both. Although cure rates for many tumor types are substantive, significant radiation-related toxicities persist. For disseminated disease or histology results with a high likelihood of subclinical dissemination, radiation to the entire craniospinal axis is commonly used. Radiation to the entire cranium and spine may impact physical, neurocognitive, musculoskeletal, hearing, and pubertal development and may be associated with an inferior quality of life for survivors (1). The physical advantages of proton therapy offer significant sparing of normal tissue for craniospinal irradiation (CSI), which may minimize the risk of some adverse effects. However, the optimal techniques for planning and delivery of proton-based CSI remain subjects of investigation. Proton-based CSI delivery may be broadly divided into passive scattering and scanning beam techniques. Relative to photon-based 3-dimensional (3D) conformal and intensity modulated radiation therapy (IMRT), passively scattered proton therapy (PSPT) has been shown to reduce dose to organs at risk (OAR) including the lens, cochlea, and temporal lobes (2-7). For delivery of protons, to date PSPT technology has been the most widely used modality clinically. However, scanning beam delivery may offer superior target coverage and further normal tissue sparing relative to that of PSPT through the use of potentially superior techniques such as intensity modulated proton therapy (IMPT). Using a superposition of pencil beams within a broad spectrum of energies, IMPT enables modulation of intermediate and proximal field doses in addition to distal target conformity. Consequently, IMPT may provide OAR sparing capabilities yet unrealized with other CSI modalities. Moreover, modulated pencil beams also are potentially capable of forming field junctions with softer dose gradients in the overlap region, improving dosimetric homogeneity and patient safety and streamlining workflow relative to PSPT. However, commercial treatment planning systems are not yet optimally configured to take advantage of potential IMPT junctioning capabilities, motivating the methods developed in this study. In the current work, we sought to develop a method for CSI by using IMPT and to compare this method to PSPT in regard to sparing of OARs and the robustness of field junctions.
Patient selection and set-up Ten patients with central nervous system tumors that had previously received CSI with PSPT were randomly identified; subjects included 5 adult and 5 pediatric cases. The IMPT planning methodology described below was used to generate comparison plans and later applied clinically for 1 additional patient not included in the comparison study, as part of a prospective clinical trial. All patients underwent computed tomography (CT)-based simulation in the supine position. The head and neck were immobilized with a thermoplastic mask and contoured headrest. The selected headrest ensured that posterior beams incident upon the cranium were not subjected to sharp changes in radiological path length. For pediatric patients requiring anesthesia, thermoplastic masks were modified to allow for placement of a nasal cannula (8). Target structures, as well as OARs were delineated by a staff radiation oncologist. The clinical target volume (CTV) consisted of the brain, including meninges, and the spinal canal, which was extended caudally to just beyond the thecal sac. For IMPT planning, the clinical target was expanded 1 cm posteriorly from the spine; nominal spot positions were expanded 0.8 cm laterally from the spine and brain, which allowed for at least 1 beam spot to be placed outside the CTV. A dose-limiting annulus surrounding the CTV facilitated shaping of the dose gradient outside the target. All plans were normalized so that 95% of the CTV received a radiobiologically equivalent (RBE) dose of 36 Gy.
Field arrangement and beam parameters PSPT plans were generated using published techniques (9). For adult IMPT, the cranium and spine were treated with a total of 5 fields, each of which used a 6.7-cm range shifter. The beam model with the range shifter has an approximately 7 to 16 mm sigma beam spot at the energies relevant to cranial targets; the beam spot sigma relevant for spine targets is approximately 6 to 12 mm. Of the 5 total fields, 2 cranial fields were mirrored anterior oblique (AO) beams, angled 75 laterally, with a superior-inferior rotation to prevent the ipsilateral eye from eclipsing the target. When it was necessary to cover the spine target, the isocenter for the cervical posterior-anterior (PA) beam, which also covered portions of the brain, was
Volume 90 Number 3 2014
IMPT craniospinal irradiation
639
Fig. 1. (A) Sagittal view demonstrating placement of PA beams along the patient’s spine. (B) Coronal view demonstrating placement and orientation of spine and cranial treatment fields. (C) Coronal view demonstrating PA beams with 45 couch rotation. LAO Z left anterior oblique; PA Z posterior-anterior; RAO Z right anterior oblique. shifted inferiorly so that its superior extent was at the level of the cochlea. Shifting the isocenter inferiorly enabled coverage to the full spine to be achieved in 3 fields for typical adult cases; the same isocenter was used for the cervical PA and AO beams if a shift was not necessary. The target defined for the cervical PA field was such that the most distal energy layer stopped more than 1 cm short of the optic chiasm. The remaining fields were PA beams covering the thoracic and lumbar spine (Fig. 1A) (10). The thoracic and lumbar PA beams were spaced equally along the craniocaudal axis. The images in Figures 1A and 1B provide representative sagittal and coronal views with marked field projections. PA beam central axes were placed as closely as possible, with a desired 10-cm overlap for junctioning as described below. The maximum field size of our system is 30 30 cm, so depending upon patient height, a nominal 45 couch rotation for PA beam set up enabled sufficient field overlap (Fig. 1C). Pediatric patients required as few as 2 PA beams.
The segmented regions ideally spanned 10 cm along the spine axis, but a minimum of 6-cm overlap was achieved in all cases. These beam-specific targets defined the beam-line settings, which determined the possible spot positions within
Field matching and optimization To achieve low-dose gradients across spine field junctions, IMPT optimization was performed using a 3-staged approach to guide the planning system towards the desired outcome. A unique target was contoured for each field prior to optimization. At the inferior end of the upper spine PA field, the target was divided along the craniocaudal axis into 4 to 10 equally sized “tapering segments,” which were used to shape the dose gradient (Fig. 2A). The superior end of the lower spine PA beam was segmented similarly. The segments varied in craniocaudal extent depending upon the height of the patient but at least exceeded the spot-spacing settings of the planning system.
Fig. 2. Sagittal view demonstrating two preparatory optimization stages for (A) anterior-oblique beams and upper and lower spine posterior-anterior fields and (B) separate thoraxlevel (mid) spine field optimization. Each optimization uses the same set of dose-tapering segments (superior-inferior locations indicated by 10-cm ruler), but the constraint values are mirrored to reverse the dose gradient.
640
Stoker et al.
the dose grid. Prior to optimization, beam-line settings were calculated to ensure they fell within machine performance limits. The treatment planning optimizer was then constrained to assign fluence for each beam only to these just-calculated positions within the dose grid. After an optimization, if any changes to a field target were made, recalculating beam-line settings was required to update possible spot positions and clear out the fluence map determined during the previous optimization. The first stage involved optimizing the AO and upper and lower spine PA beams. Each tapering segment sequentially reduced the dose from prescription to 0 dose in equally sized steps. Adjacent segments had dose constraints that overlapped by approximately 200 cGy to create a smooth gradient. This stage required multifield optimization (MFO), because each individual beam’s eye view does not see the entire target, and to enable maximal sparing of OARs located intermediately along a beam path, such as cochlea. The second optimization was a single-field optimization that contained a single PA at the thoracic level, placed midway between the central axes of the upper and lower spine PA fields. This thorax-level PA used the identical tapering segments from the initial optimization, with the segment constraints mirrored to reverse the tapered slope (Fig. 2B). The third and final stage was MFO. For this step, the stage 2 optimized field was copied and pasted into the stage 1 MFO plan. Beam-line settings were not recalculated, thereby providing the tapered fluence obtained in stages 1 and 2 as an initial condition, which guided the outcome of this final optimization to have the desired tapered field match.
Junction shifts and robustness evaluation Junction regions were shifted by altering the craniocaudal extent of the PA beam targets, including the segments used for junction tapering, and then recalculating beam-line settings. For clinical implementation, spine junctions were shifted once by approximately 2 cm over a 4-week treatment course. Delivery robustness was evaluated in part by using a previously published method (11). This method does not account for interfield setup variations. To address this, Extended Dose Range 2 (EDR2) radiographic film measurements from just beneath the spinal field junction patient were collected and compared to calculated profiles for a single patient treated as part of a prospective protocol (12). Film data were collected over multiple delivery fractions to validate dosimetric consistency (13). Then, selected films were irradiated to a known physical dose for scaling.
Dosimetric evaluation Given anatomical differences, patients were divided into adult and pediatric patients. Plan quality was evaluated by inhomogeneity coefficient derived from the equation
International Journal of Radiation Oncology Biology Physics
IC Z [D5 D95]/[Dmean], and heterogeneity index [HI Z D5/D95], where D95 is the dose-volume histogram (DVH) curve dose representing 95% of volume of the target, and D5 is the DVH curve dose representing 5% of volume of the target. Volumes of total CTV, as well as volumes of brain and cribriform plate portions of the CTV receiving at least 95% of the prescribed dose were also recorded (4). Selected dose metrics for the principal OARs were also evaluated.
Statistical evaluation Statistical analysis was performed using Stata 13 software (Stata Corp., College Station, TX) (14). Significance was determined by Wilcoxon matched-pairs signed rank test. The Wilcoxon matched-pairs signed ranks test is a nonparametric test which tests the equality of matched pairs of observations with the null hypothesis that both distributions are the same.
Results Target coverage Target volume dosimetry for IMPT and PSPT are shown in Table 1. Both techniques achieved clinically acceptable CTV coverage. Cribriform plate coverage was comparable in both techniques, within statistical error. Dose to 95% of target volume was robust relative to a 3-mm setup and 3% range error, to within 5% of planned dose (11). For adults but not children, IMPT plans did have homogeneity indices and inhomogeneity coefficients indicative of a greater range in dose within the brain target volumes. Our clinical practice progressed from using compensators for the cranial target to no compensator currently. For this group of patients, it happened that the pediatric patients had compensators and adult patients did not have compensators. Compensators create conformal dose distribution distally but increase the dose inhomogeneity slightly (10).
OAR sparing Radiation dose to OARs for pediatric and adult cases were compared between modalities (Table 2). The brain and spinal cord comprised principal components of target volume and so received high doses with minor variation across modalities. However, IMPT significantly reduced the maximum spinal cord dose. IMPT also reduced dose to the lenses in both adults and children on average by approximately half, consistent with a previous report (10). Notably, this was achieved while maintaining cribriform plate coverage, as shown in Table 1. Cochlear dose was also significantly reduced using IMPT as were doses to the parotid glands. This was more pronounced in the adult population, where mean parotid doses decreased in the range of 20 Gy to 11 to 12 Gy.
Volume 90 Number 3 2014 Table 1
IMPT craniospinal irradiation
641
Target volume coverage and dose heterogeneity indices. Pediatric patient Mean PSPT SD
Index - target V95% - whole brain V95% - cribriform plate V95% - CTV HI - whole brain IC - whole brain HI - brainstem IC - brainstem HI - spinal cord IC - spinal cord
100 87 98 1.05 0.046 1.04 0.041 1.05 0.049
Adult patient
Mean IMPT SD
1 11 2 0.01 0.007 0.03 0.031 0.01 0.011
100 92 100 1.05 0.049 1.05 0.046 1.04 0.039
1 7 1 0.01 0.001 0.01 0.005 0.01 0.008
Mean PSPT SD 100 94 100 1.03 0.027 1.03 0.024 1.05 0.047
1 7 1 0.01 0.001 0.01 0.002 0.01 0.007
Mean IMPT SD 100 100 100 1.05 0.052 1.03 0.033 1.04 0.039
1 1 1 0.01* 0.004* 0.01 0.004* 0.01 0.003
Abbreviations: V95% Z percentage of the volume of the target that receives at least 95% of the prescribed dose; CTV Z clinical target volume; D95 Z dose-volume histogram (DVH) curve dose representing 95% of volume of the target; D5 Z DVH curve dose representing 5% of volume of the target; HI Z homogeneity index Z D5/D95; IC Z inhomogeneity coefficient Z [D5 D95]/[mean dose]; IMPT Z intensity modulated proton therapy; PSPT Z passively scattered proton therapy. * Differences were significant in comparison with PSPT (P<.05), using Wilcoxon matched-pairs signed-ranks test.
In contrast, thyroid gland dose was slightly increased with the use of IMPT in both adults and children. Mean radiation doses to the lungs and kidneys were also increased in the IMPT cohort, particularly in the pediatric population. This effect is driven by the larger penumbra observed during IMPT delivery. Lateral boundaries of PSPT fields are defined by apertures, resulting in significant
Table 2
penumbral reduction relative to that of IMPT. At 15-cm depth, the 80% to 20% penumbra of a 151-MeV scanning beam (15.6-cm range) is reduced by aperture from 15.3 mm to 6.4 mm, for 10- 10-cm-field size. At 180-MeV, PSPT beam incident upon a range modulator wheel (16.9-cm range) exhibits a slightly larger 7.5-mm penumbra at 15 cm.
Selected dose metrics for principal organs at risk Pediatric patient (Gy [RBE])
Organ Brain Minimum Maximum Mean Spinal cord Minimum Maximum Mean Lenses Maximum Cochleae Minimum Maximum Mean Parotid glands Mean Thyroid Maximum Mean Lungs Maximum Mean Kidneys Maximum Mean
Adult patient (Gy [RBE])
Mean PSPT SD
Mean IMPT SD
Mean PSPT SD
Mean IMPT SD
17.6 8.6 40.4 1.0 37.1 0.3
23.7 7.6* 38.6 0.2* 36.7 0.2
27.3 7.6 39.3 0.6 36.9 0.3
30.8 2.0 38.8 0.1 37.3 0.2*
33.0 1.3 40.0 1.0 37.3 0.2
33.2 2.5 37.7 0.6* 36.4 0.4*
34.9 0.8 39.6 0.6 37.1 0.2
34.3 0.6 38.0 0.3* 36.4 0.2*
23.8 4.9
11.6 1.3*
21.1 7.5
10.5 2.0*
33.6 4.0 36.8 1.2 35.4 2.3
25.1 0.9* 30.6 0.9* 27.3 0.5*
35.9 0.5 37.2 0.7 36.6 0.5
25.4 1.0* 30.6 0.7* 27.8 0.8*
18.9 2.6
17.8 2.9
21.6 3.3
11.7 2.0*
19.6 7.4 3.8 3.2
19.3 4.5 6.9 1.7
0.4 0.7 00
9.0 5.5* 2.0 2.6*
37.6 1.0 3.4 1.5
35.2 1.3 5.8 1.8*
37.2 1.0 2.1 0.7
34.1 1.1 2.6 0.8
30.2 4.9 2.8 1.8
33.0 1.8 7.3 2.2*
21.7 12.2 0.7 0.8
18.2 9.3 1.3 1.1*
Abbreviations: IMPT Z intensity modulated proton therapy; PSPT Z passively scattered proton therapy; RBE Z radiobiological equivalent. * Differences were significant in comparison with PSPT (P<.05), using Wilcoxon matched-pairs signed-ranks test.
642
Stoker et al.
Field junctioning Classically, PSPT-based CSI is achieved with lateral or oblique cranial fields and multiple spine PA fields. PSPT collimation techniques result in steep dose falloff at the field edge. Consequently, matching field edges requires high-precision setup to achieve dose uniformity. The fluence modulation capabilities of IMPT may improve dose uniformity and setup robustness, similar to existing IMRT techniques (15, 16) with range control benefits added. Therefore, we next evaluated the benefits of IMPT versus PSPT in regard to robustness of field junctions. The graphs in Figure 3 represent the resulting dose modeled for PSPT using brass apertures (Fig. 3A) and IMPT with a 10cm field overlap (Fig. 3B). Figure 3A demonstrates a gradient (determined from the average slope between the 90% and 10% isodose) of approximately 50%/cm using PSPT, which is markedly above the 8%/cm gradient
International Journal of Radiation Oncology Biology Physics
achievable for IMPT fields (Fig. 3B). The solid lines (Fig. 3B) represent the planned delivery, and the dashed lines (Fig. 3B) represent the resulting dosimetric deviation after the modeled craniocaudal 4-mm setup error (2-mm each field). The deviation of <5% for the simulated IMPT setup error is significantly below the 25% observed for PSPT. When shifting the cervical PA field was required, the junction between the upper PA field and the AO fields was unique from the inferior spine junctions because the fields overlapped by 20 or more cm and the fields were not coplanar. Commercial optimization techniques performed reasonably well for our system under such circumstances, such that the modeled 4-mm-setup error induced a variation of 7% or less in dose. The proposed field tapering captures the robustness of similarly matched IMRT fields while maintaining the range modulation of proton delivery (16). Film measurements taken over 20 fractions for a single IMPT patient demonstrated variation of <5% between measured and calculated dose planes. A representative overlay of calculated and measured isodoses is shown in Figure 4. As facilities become more comfortable with planning and delivery methods and as robust optimization software advances, eliminating junction shifts over the treatment course may be possible but only after careful study by the treating institution, and preferably as part of clinical trials.
Discussion
Fig. 3. Dose profiles across the junction of 2 fields at the level of lumbar-thoracic spine junction. Solid lines represent relative doses from individual fields and their summed dose. Dashed lines represent resulting doses across the sampled profile after each field shifted by 2 mm. (A) Passively scanned proton therapy, which uses collimators to define field boundaries, results in deviation of 25% from planned delivery after field shifts were applied. (B) Intensity modulated proton therapy plan, which incorporates a nominal field overlap of 10 cm, resulted in a deviation of <5% from planned delivery after field shifts were applied.
This study presents a method for IMPT treatment of the craniospinal axis. We found that in comparison to PSPT, IMPT softened the field edge dose gradient for junctioned fields, and a method for achieving these soft junctions is described. The approach detailed herein is unique to date for IMPT, in that robustly matched fields reside within a single treatment plan. Because a sum plan is not generated, robust analysis is straightforward and within current treatment planning system functionality. OAR sparing during CSI, in some cases previously unachieved outside of modulated particle therapy, is also highlighted. IMPT provides superior flexibility in terms of available field arrangements and dose shaping. Junction robustness is dependent on the matched field gradients. As opposed to PSPT, manual entry of junction segments for IMPT-based CSI provides the planner a measure of control over the dose gradient. Here we highlight the capability of IMPT in achieving dose variations <5% in comparison with the 25% dose variation observed for PSPT for a 2-mm per field setup error. The clinical benefit of maintaining 5% is well established, and the potential for improved safety for patients should not be overlooked. Improvements in both setup reproducibility and robust optimization tools will hopefully facilitate routinely achieving this level of performance. Concerns driven by reduced junction shifting may potentially be addressed using established methods of field-matching verification (13).
Volume 90 Number 3 2014
IMPT craniospinal irradiation
643
Fig. 4. Isodose overlay of measured and calculated dose distributions for intensity modulated proton therapy to the spine. Solid lines represent values determined from Extended Dose Range 2 (EDR2) radiographic film measurements taken on the treatment couch just beneath supinely-positioned patient. Dashed lines were generated from the corresponding dosimetric plane calculated in the treatment planning system. We also found that IMPT further improved the OAR sparing in CSI relative to that with PSPT (10). The lens of the eye is one of the most sensitive structures in the body, with formation of cataracts and visual impairment presenting as late complications. Investigators have estimated a dose of 15 Gy to the lens causes a 50% probability of visual impairment in adults (17). The visual apparatus of children is likely more sensitive. In children <18 months of age, Hall et al (18) found a 35% to 50% increase in the risk of lens opacity for each 1-Gy increase in dose received. Also of note is the advantage of MFO-IMPT in cochlear sparing, observed here both in adults and children. PSPT fields deliver a single energy, using compensators for distal conformity to the target, but are not capable of modulating fluence at intermediate ranges, resulting in near-prescription dose to intermediately positioned OARs which eclipse the target beam’s eye view (10). Hua et al recommend cochlear doses not exceed 35 Gy in children to minimize sensorineural hearing loss risk (25). Quantitative Analysis of Normal Tissue Effects in the Clinic (QUANTEC) guidelines indicate that no threshold for sensorineural hearing loss prevention has been found in the adult population, and recommend a mean cochlear dose of at least 35 to 45 Gy (19). QUANTEC guidelines recommend a unilateral salivary gland mean dose of <20 Gy and mean bilateral salivary gland sparing of <25 Gy, each in order to avoid severe xerostomia, both of which are comfortably achieved by using IMPT (20). We did identify slight increases in doses to the thyroid, lungs, and kidneys, using IMPT. Although the observed doses would not likely result in primary organ dysfunction, further dose reductions would obviously be clinically desirable (21, 22). The current system represents a pioneering effort in IMPT. For the modeled beam, the scanning beam penumbra width is due in part to the treatment nozzle beam monitors, which increasingly scatter the pencil beam
at lower energies. Straightforward modifications, such as a retracting selected beam diagnostics or extending the vacuum portion of the nozzle, could reduce the spot size and penumbra. Penumbra measurements suggest that using apertures will reduce doses in the lungs and kidneys. The parotid dose for pediatric patients could also be reduced, and thyroid sparing could potentially be improved at least for adult anatomies to match PSPT levels. Decreasing the spot size has been demonstrated to facilitate near-target OAR sparing (23) and could be combined with aperture use. Proton therapy centers under construction are trending toward a dedicated scanning beam facility design. Aperture use will markedly diminish relative to PSPT-capable centers not justifying an in-house machine shop. If relying on third parties for infrequent machining of patient-specific apertures is not tractable, multipatient apertures of widely applicable dimensions could be used.
Conclusions An optimal beam arrangement and planning approach are critical to realizing the full benefit of IMPT. Ideally, the planner would be able to set the field overlap, and the optimizer would determine a robust solution. However, current commercially available optimization software is not capable of taking into account the multi-isocenter, interfield setup error necessary for robust field matching. Inspection of the individual field doses across a spine field junction obtained using single-stage MFO optimization reveals a nonrobust solution resulting from the optimizer’s solving for a local minima. Contrary to reported results for IMRT optimization, we found that the presence of multiple local minima traps is an impediment to finding a robust IMPT solution (24). Our 3-stage method provides a way to alter the initial conditions prior to the final optimization. Nevertheless, the method proposed herein has the potential
644
Stoker et al.
to be further simplified. Increasing field size to 40 cm would eliminate the need to shift the upper PA field, eliminating a junction for many patients. Also, we reference a previous work detailing an x-ray method for CSI. A field-in-field forward planning approach provided the IMRT optimizer low-gradient tapered upper and lower spine PA fields (15). Present IMRT optimization software incorporates this forward-planned base dose distribution, which facilitates a multistage planning process. Eventually, incorporating a base dose distribution will also be available for IMPT optimizers. Further clinical study of such techniques, including treatments forgoing junction shifts, should only be undertaken after further careful study at the treating institution and preferably as part of prospective clinical trials.
References 1. Schwartz CL. Long-term survivors of childhood cancer: The late effects of therapy. Oncologist 1999;4:45-54. 2. Barney CL, Brown AP, Grosshans DR, et al. Technique, outcomes, and acute toxicities in adults treated with proton beam craniospinal irradiation. Neuro Oncol 2014;16:303-309. 3. Cochran DM, Yock TI, Adams JA, et al. Radiation dose to the lens during craniospinal irradiation-an improvement in proton radiotherapy technique. Int J Radiat Oncol Biol Phys 2008;70:1336-1342. 4. Howell RM, Giebeler A, Koontz-Raisig W, et al. Comparison of therapeutic dosimetric data from passively scattered proton and photon craniospinal irradiations for medulloblastoma. Radiat Oncol 2012;7:116. 5. St Clair WH, Adams JA, Bues M, et al. Advantage of protons compared to conventional X-ray or IMRT in the treatment of a pediatric patient with medulloblastoma. Int J Radiat Oncol Biol Phys 2004;58:727-734. 6. Brown AP, Barney CL, Grosshans DR, et al. Proton Beam Craniospinal Irradiation Reduces Acute Toxicity for Adults With Medulloblastoma. Int J Radiat Oncol Biol Phys 2013;86:277-284. 7. Moeller BJ, Chintagumpala M, Philip JJ, et al. Low early ototoxicity rates for pediatric medulloblastoma patients treated with proton radiotherapy. Radiat Oncol 2011;6:58. 8. Owusu-Agyemang P, Grosshans D, Arunkumar R, et al. Non-invasive anesthesia for children undergoing proton radiation therapy. Radiother Oncol 2014;111:30-34.
International Journal of Radiation Oncology Biology Physics 9. Giebeler A, Newhauser WD, Amos RA, et al. Standardized treatment planning methodology for passively scattered proton craniospinal irradiation. Radiat Oncol 2013;8:32. 10. Dinh JS, Stoker J, Georges R, et al. Comparison of proton therapy techniques for treatment of the whole brain as a component of craniospinal irradiation. Radiat Oncol 2013;8:289. 11. Liu W, Zhang X, Li Y, et al. Robust optimization of intensity modulated proton therapy. Med Phys 2012;39:1079-1091. 12. Martens C, Claeys I, De Wagter C, et al. The value of radiographic film for the characterization of intensity-modulated beams. Phys Med Biol 2002;47:2221-2234. 13. Michalski JM, Klein EE, Gerber R. Method to plan, administer, and verify supine craniospinal irradiation. J Appl Clin Med Phys 2002;3: 310-316. 14. StataCorp. Stata: release 13. College Station, TX: StataCorp LP; 2013. 15. Pai Panandiker A, Ning H, Likhacheva A, et al. Craniospinal irradiation with spinal IMRT to improve target homogeneity. Int J Radiat Oncol Biol Phys 2007;68:1402-1409. 16. Yom SS, Frija EK, Mahajan A, et al. Field-in-field technique with intrafractionally modulated junction shifts for craniospinal irradiation. Int J Radiat Oncol Biol Phys 2007;69:1193-1198. 17. Henk JM, Whitelocke RA, Warrington AP, et al. Radiation dose to the lens and cataract formation. Int J Radiat Oncol Biol Phys 1993;25: 815-820. 18. Hall P, Granath F, Lundell M, et al. Lenticular opacities in individuals exposed to ionizing radiation in infancy. Radiat Res 1999;152: 190-195. 19. Bhandare N, Jackson A, Eisbruch A, et al. Radiation therapy and hearing loss. Int J Radiat Oncol Biol Phys 2010;76:S50-57. 20. Deasy JO, Moiseenko V, Marks L, et al. Radiotherapy dose-volume effects on salivary gland function. Int J Radiat Oncol Biol Phys 2010;76:S58-63. 21. Foote RL, Stafford SL, Petersen IA, et al. The clinical case for proton beam therapy. Radiat Oncol 2012;7:174. 22. Newhauser WD, Fontenot JD, Mahajan A, et al. The risk of developing a second cancer after receiving craniospinal proton irradiation. Phys Med Biol 2009;54:2277-2291. 23. van de Water TA, Lomax AJ, Bijl HP, et al. Using a reduced spot size for intensity-modulated proton therapy potentially improves salivary gland-sparing in oropharyngeal cancer. Int J Radiat Oncol Biol Phys 2012;82:e313-319. 24. Wu Q, Mohan R. Multiple local minima in IMRT optimization based on dose-volume criteria. Med Phys 2002;29:1514-1527. 25. Hua C, Bass JK, Khan R, et al. Hearing loss after radiotherapy for pediatric brain tumors: Effect of cochlear dose. Int J Radiat Oncol Biol Phys 2008;72:892-899.
Radiation Oncology biology
physics
www.redjournal.org
Physics Contribution
Intensity Modulated Proton Therapy for Craniospinal Irradiation: Organ-at-Risk Exposure and a Low-Gradient Junctioning Technique Joshua B. Stoker, PhD,*,y Jonathan Grant, MD,z X. Ronald Zhu, PhD,* Rajesh Pidikiti, PhD,* Anita Mahajan, MD,z and David R. Grosshans, MD, PhDz *Department of Radiation Physics, The University of Texas MD Anderson Cancer Center, Houston, Texas; yDepartment of Radiation Oncology, Mayo Clinic Arizona, Phoenix, Arizona; and zDivision of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, Texas Received Mar 15, 2014, and in revised form Jul 2, 2014. Accepted for publication Jul 3, 2014.
Summary Intensity modulated proton therapy is a sophisticated delivery modality for which the full clinical benefit during craniospinal irradiation (CSI) is still emerging. A novel planning approach for CSI is presented which demonstrates, relative to passive scanning proton therapy, improved robustness of field junctions and superior sparing of principal organs at risk. Use of such an approach may improve clinical outcomes and enhance treatment safety and efficiency.
Purpose: To compare field junction robustness and sparing of organs at risk (OARs) during craniospinal irradiation (CSI) using intensity modulated proton therapy (IMPT) to conventional passively scattered proton therapy (PSPT). Methods and Materials: Ten patients, 5 adult and 5 pediatric patients, previously treated with PSPT-based CSI were selected for comparison. Anterior oblique cranial fields, using a superior couch rotation, and posterior spinal fields were used for IMPT planning. To facilitate low-gradient field junctioning along the spine, the inverseplanning IMPT technique was divided into 3 stages. Dose indices describing target coverage and normal tissue dose, in silico error modeling, and film dosimetry were used to assess plan quality. Results: Field junction robustness along the spine was improved using the staged IMPT planning technique, reducing the worst case impact of a 4-mm setup error from 25% in PSPT to <5% of prescription dose. This was verified by film dosimetry for clinical delivery. Exclusive of thyroid dose in adult patients, IMPT plans demonstrated sparing of organs at risk as good or better than PSPT. Coverage of the cribriform plate for pediatric (V95% [percentage of volume of the target receiving at least 95% of the prescribed dose]; 87 11 vs 92 7) and adult (V95%; 94 7 vs 100 1) patients and the clinical target in pediatric (V95%; 98 2 vs 100 1) and adult (V95%; 100 1 vs 100 1) patients for PSPT and IMPT plans, respectively, were comparable or improved. For adult patients, IMPT target dose inhomogeneity was increased, as determined by heterogeneity index (HI) and inhomogeneity coefficient (IC). IMPT lowered maximum spinal cord dose, improved spinal dose homogeneity, and reduced exposure to other OARs.
Reprint requests to: Joshua Stoker, PhD, Mayo Clinic Arizona, 5777 E. Mayo Blvd., Phoenix, AZ 85054. Tel: (801) 830-1784; E-mail: stoker. [email protected] Int J Radiation Oncol Biol Phys, Vol. 90, No. 3, pp. 637e644, 2014 0360-3016/$ - see front matter Ó 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ijrobp.2014.07.003
Conflict of interest: none. AcknowledgmentsdThe authors thank Pamela K. Allen, PhD, for running statistical analyses essential for this study.
638
International Journal of Radiation Oncology Biology Physics
Stoker et al.
Conclusions: IMPT has the potential to improve CSI plan quality and the homogeneity of intrafractional dose at match lines. The IMPT approach developed may also simplify treatments and reduce workload per patient relative to PSPT. Ó 2014 Elsevier Inc.
Introduction
Methods and Materials
Radiation therapy plays a central role in the treatment of many tumors of the central nervous system, either alone or in conjunction with surgery, chemotherapy, or both. Although cure rates for many tumor types are substantive, significant radiation-related toxicities persist. For disseminated disease or histology results with a high likelihood of subclinical dissemination, radiation to the entire craniospinal axis is commonly used. Radiation to the entire cranium and spine may impact physical, neurocognitive, musculoskeletal, hearing, and pubertal development and may be associated with an inferior quality of life for survivors (1). The physical advantages of proton therapy offer significant sparing of normal tissue for craniospinal irradiation (CSI), which may minimize the risk of some adverse effects. However, the optimal techniques for planning and delivery of proton-based CSI remain subjects of investigation. Proton-based CSI delivery may be broadly divided into passive scattering and scanning beam techniques. Relative to photon-based 3-dimensional (3D) conformal and intensity modulated radiation therapy (IMRT), passively scattered proton therapy (PSPT) has been shown to reduce dose to organs at risk (OAR) including the lens, cochlea, and temporal lobes (2-7). For delivery of protons, to date PSPT technology has been the most widely used modality clinically. However, scanning beam delivery may offer superior target coverage and further normal tissue sparing relative to that of PSPT through the use of potentially superior techniques such as intensity modulated proton therapy (IMPT). Using a superposition of pencil beams within a broad spectrum of energies, IMPT enables modulation of intermediate and proximal field doses in addition to distal target conformity. Consequently, IMPT may provide OAR sparing capabilities yet unrealized with other CSI modalities. Moreover, modulated pencil beams also are potentially capable of forming field junctions with softer dose gradients in the overlap region, improving dosimetric homogeneity and patient safety and streamlining workflow relative to PSPT. However, commercial treatment planning systems are not yet optimally configured to take advantage of potential IMPT junctioning capabilities, motivating the methods developed in this study. In the current work, we sought to develop a method for CSI by using IMPT and to compare this method to PSPT in regard to sparing of OARs and the robustness of field junctions.
Patient selection and set-up Ten patients with central nervous system tumors that had previously received CSI with PSPT were randomly identified; subjects included 5 adult and 5 pediatric cases. The IMPT planning methodology described below was used to generate comparison plans and later applied clinically for 1 additional patient not included in the comparison study, as part of a prospective clinical trial. All patients underwent computed tomography (CT)-based simulation in the supine position. The head and neck were immobilized with a thermoplastic mask and contoured headrest. The selected headrest ensured that posterior beams incident upon the cranium were not subjected to sharp changes in radiological path length. For pediatric patients requiring anesthesia, thermoplastic masks were modified to allow for placement of a nasal cannula (8). Target structures, as well as OARs were delineated by a staff radiation oncologist. The clinical target volume (CTV) consisted of the brain, including meninges, and the spinal canal, which was extended caudally to just beyond the thecal sac. For IMPT planning, the clinical target was expanded 1 cm posteriorly from the spine; nominal spot positions were expanded 0.8 cm laterally from the spine and brain, which allowed for at least 1 beam spot to be placed outside the CTV. A dose-limiting annulus surrounding the CTV facilitated shaping of the dose gradient outside the target. All plans were normalized so that 95% of the CTV received a radiobiologically equivalent (RBE) dose of 36 Gy.
Field arrangement and beam parameters PSPT plans were generated using published techniques (9). For adult IMPT, the cranium and spine were treated with a total of 5 fields, each of which used a 6.7-cm range shifter. The beam model with the range shifter has an approximately 7 to 16 mm sigma beam spot at the energies relevant to cranial targets; the beam spot sigma relevant for spine targets is approximately 6 to 12 mm. Of the 5 total fields, 2 cranial fields were mirrored anterior oblique (AO) beams, angled 75 laterally, with a superior-inferior rotation to prevent the ipsilateral eye from eclipsing the target. When it was necessary to cover the spine target, the isocenter for the cervical posterior-anterior (PA) beam, which also covered portions of the brain, was
Volume 90 Number 3 2014
IMPT craniospinal irradiation
639
Fig. 1. (A) Sagittal view demonstrating placement of PA beams along the patient’s spine. (B) Coronal view demonstrating placement and orientation of spine and cranial treatment fields. (C) Coronal view demonstrating PA beams with 45 couch rotation. LAO Z left anterior oblique; PA Z posterior-anterior; RAO Z right anterior oblique. shifted inferiorly so that its superior extent was at the level of the cochlea. Shifting the isocenter inferiorly enabled coverage to the full spine to be achieved in 3 fields for typical adult cases; the same isocenter was used for the cervical PA and AO beams if a shift was not necessary. The target defined for the cervical PA field was such that the most distal energy layer stopped more than 1 cm short of the optic chiasm. The remaining fields were PA beams covering the thoracic and lumbar spine (Fig. 1A) (10). The thoracic and lumbar PA beams were spaced equally along the craniocaudal axis. The images in Figures 1A and 1B provide representative sagittal and coronal views with marked field projections. PA beam central axes were placed as closely as possible, with a desired 10-cm overlap for junctioning as described below. The maximum field size of our system is 30 30 cm, so depending upon patient height, a nominal 45 couch rotation for PA beam set up enabled sufficient field overlap (Fig. 1C). Pediatric patients required as few as 2 PA beams.
The segmented regions ideally spanned 10 cm along the spine axis, but a minimum of 6-cm overlap was achieved in all cases. These beam-specific targets defined the beam-line settings, which determined the possible spot positions within
Field matching and optimization To achieve low-dose gradients across spine field junctions, IMPT optimization was performed using a 3-staged approach to guide the planning system towards the desired outcome. A unique target was contoured for each field prior to optimization. At the inferior end of the upper spine PA field, the target was divided along the craniocaudal axis into 4 to 10 equally sized “tapering segments,” which were used to shape the dose gradient (Fig. 2A). The superior end of the lower spine PA beam was segmented similarly. The segments varied in craniocaudal extent depending upon the height of the patient but at least exceeded the spot-spacing settings of the planning system.
Fig. 2. Sagittal view demonstrating two preparatory optimization stages for (A) anterior-oblique beams and upper and lower spine posterior-anterior fields and (B) separate thoraxlevel (mid) spine field optimization. Each optimization uses the same set of dose-tapering segments (superior-inferior locations indicated by 10-cm ruler), but the constraint values are mirrored to reverse the dose gradient.
640
Stoker et al.
the dose grid. Prior to optimization, beam-line settings were calculated to ensure they fell within machine performance limits. The treatment planning optimizer was then constrained to assign fluence for each beam only to these just-calculated positions within the dose grid. After an optimization, if any changes to a field target were made, recalculating beam-line settings was required to update possible spot positions and clear out the fluence map determined during the previous optimization. The first stage involved optimizing the AO and upper and lower spine PA beams. Each tapering segment sequentially reduced the dose from prescription to 0 dose in equally sized steps. Adjacent segments had dose constraints that overlapped by approximately 200 cGy to create a smooth gradient. This stage required multifield optimization (MFO), because each individual beam’s eye view does not see the entire target, and to enable maximal sparing of OARs located intermediately along a beam path, such as cochlea. The second optimization was a single-field optimization that contained a single PA at the thoracic level, placed midway between the central axes of the upper and lower spine PA fields. This thorax-level PA used the identical tapering segments from the initial optimization, with the segment constraints mirrored to reverse the tapered slope (Fig. 2B). The third and final stage was MFO. For this step, the stage 2 optimized field was copied and pasted into the stage 1 MFO plan. Beam-line settings were not recalculated, thereby providing the tapered fluence obtained in stages 1 and 2 as an initial condition, which guided the outcome of this final optimization to have the desired tapered field match.
Junction shifts and robustness evaluation Junction regions were shifted by altering the craniocaudal extent of the PA beam targets, including the segments used for junction tapering, and then recalculating beam-line settings. For clinical implementation, spine junctions were shifted once by approximately 2 cm over a 4-week treatment course. Delivery robustness was evaluated in part by using a previously published method (11). This method does not account for interfield setup variations. To address this, Extended Dose Range 2 (EDR2) radiographic film measurements from just beneath the spinal field junction patient were collected and compared to calculated profiles for a single patient treated as part of a prospective protocol (12). Film data were collected over multiple delivery fractions to validate dosimetric consistency (13). Then, selected films were irradiated to a known physical dose for scaling.
Dosimetric evaluation Given anatomical differences, patients were divided into adult and pediatric patients. Plan quality was evaluated by inhomogeneity coefficient derived from the equation
International Journal of Radiation Oncology Biology Physics
IC Z [D5 D95]/[Dmean], and heterogeneity index [HI Z D5/D95], where D95 is the dose-volume histogram (DVH) curve dose representing 95% of volume of the target, and D5 is the DVH curve dose representing 5% of volume of the target. Volumes of total CTV, as well as volumes of brain and cribriform plate portions of the CTV receiving at least 95% of the prescribed dose were also recorded (4). Selected dose metrics for the principal OARs were also evaluated.
Statistical evaluation Statistical analysis was performed using Stata 13 software (Stata Corp., College Station, TX) (14). Significance was determined by Wilcoxon matched-pairs signed rank test. The Wilcoxon matched-pairs signed ranks test is a nonparametric test which tests the equality of matched pairs of observations with the null hypothesis that both distributions are the same.
Results Target coverage Target volume dosimetry for IMPT and PSPT are shown in Table 1. Both techniques achieved clinically acceptable CTV coverage. Cribriform plate coverage was comparable in both techniques, within statistical error. Dose to 95% of target volume was robust relative to a 3-mm setup and 3% range error, to within 5% of planned dose (11). For adults but not children, IMPT plans did have homogeneity indices and inhomogeneity coefficients indicative of a greater range in dose within the brain target volumes. Our clinical practice progressed from using compensators for the cranial target to no compensator currently. For this group of patients, it happened that the pediatric patients had compensators and adult patients did not have compensators. Compensators create conformal dose distribution distally but increase the dose inhomogeneity slightly (10).
OAR sparing Radiation dose to OARs for pediatric and adult cases were compared between modalities (Table 2). The brain and spinal cord comprised principal components of target volume and so received high doses with minor variation across modalities. However, IMPT significantly reduced the maximum spinal cord dose. IMPT also reduced dose to the lenses in both adults and children on average by approximately half, consistent with a previous report (10). Notably, this was achieved while maintaining cribriform plate coverage, as shown in Table 1. Cochlear dose was also significantly reduced using IMPT as were doses to the parotid glands. This was more pronounced in the adult population, where mean parotid doses decreased in the range of 20 Gy to 11 to 12 Gy.
Volume 90 Number 3 2014 Table 1
IMPT craniospinal irradiation
641
Target volume coverage and dose heterogeneity indices. Pediatric patient Mean PSPT SD
Index - target V95% - whole brain V95% - cribriform plate V95% - CTV HI - whole brain IC - whole brain HI - brainstem IC - brainstem HI - spinal cord IC - spinal cord
100 87 98 1.05 0.046 1.04 0.041 1.05 0.049
Adult patient
Mean IMPT SD
1 11 2 0.01 0.007 0.03 0.031 0.01 0.011
100 92 100 1.05 0.049 1.05 0.046 1.04 0.039
1 7 1 0.01 0.001 0.01 0.005 0.01 0.008
Mean PSPT SD 100 94 100 1.03 0.027 1.03 0.024 1.05 0.047
1 7 1 0.01 0.001 0.01 0.002 0.01 0.007
Mean IMPT SD 100 100 100 1.05 0.052 1.03 0.033 1.04 0.039
1 1 1 0.01* 0.004* 0.01 0.004* 0.01 0.003
Abbreviations: V95% Z percentage of the volume of the target that receives at least 95% of the prescribed dose; CTV Z clinical target volume; D95 Z dose-volume histogram (DVH) curve dose representing 95% of volume of the target; D5 Z DVH curve dose representing 5% of volume of the target; HI Z homogeneity index Z D5/D95; IC Z inhomogeneity coefficient Z [D5 D95]/[mean dose]; IMPT Z intensity modulated proton therapy; PSPT Z passively scattered proton therapy. * Differences were significant in comparison with PSPT (P<.05), using Wilcoxon matched-pairs signed-ranks test.
In contrast, thyroid gland dose was slightly increased with the use of IMPT in both adults and children. Mean radiation doses to the lungs and kidneys were also increased in the IMPT cohort, particularly in the pediatric population. This effect is driven by the larger penumbra observed during IMPT delivery. Lateral boundaries of PSPT fields are defined by apertures, resulting in significant
Table 2
penumbral reduction relative to that of IMPT. At 15-cm depth, the 80% to 20% penumbra of a 151-MeV scanning beam (15.6-cm range) is reduced by aperture from 15.3 mm to 6.4 mm, for 10- 10-cm-field size. At 180-MeV, PSPT beam incident upon a range modulator wheel (16.9-cm range) exhibits a slightly larger 7.5-mm penumbra at 15 cm.
Selected dose metrics for principal organs at risk Pediatric patient (Gy [RBE])
Organ Brain Minimum Maximum Mean Spinal cord Minimum Maximum Mean Lenses Maximum Cochleae Minimum Maximum Mean Parotid glands Mean Thyroid Maximum Mean Lungs Maximum Mean Kidneys Maximum Mean
Adult patient (Gy [RBE])
Mean PSPT SD
Mean IMPT SD
Mean PSPT SD
Mean IMPT SD
17.6 8.6 40.4 1.0 37.1 0.3
23.7 7.6* 38.6 0.2* 36.7 0.2
27.3 7.6 39.3 0.6 36.9 0.3
30.8 2.0 38.8 0.1 37.3 0.2*
33.0 1.3 40.0 1.0 37.3 0.2
33.2 2.5 37.7 0.6* 36.4 0.4*
34.9 0.8 39.6 0.6 37.1 0.2
34.3 0.6 38.0 0.3* 36.4 0.2*
23.8 4.9
11.6 1.3*
21.1 7.5
10.5 2.0*
33.6 4.0 36.8 1.2 35.4 2.3
25.1 0.9* 30.6 0.9* 27.3 0.5*
35.9 0.5 37.2 0.7 36.6 0.5
25.4 1.0* 30.6 0.7* 27.8 0.8*
18.9 2.6
17.8 2.9
21.6 3.3
11.7 2.0*
19.6 7.4 3.8 3.2
19.3 4.5 6.9 1.7
0.4 0.7 00
9.0 5.5* 2.0 2.6*
37.6 1.0 3.4 1.5
35.2 1.3 5.8 1.8*
37.2 1.0 2.1 0.7
34.1 1.1 2.6 0.8
30.2 4.9 2.8 1.8
33.0 1.8 7.3 2.2*
21.7 12.2 0.7 0.8
18.2 9.3 1.3 1.1*
Abbreviations: IMPT Z intensity modulated proton therapy; PSPT Z passively scattered proton therapy; RBE Z radiobiological equivalent. * Differences were significant in comparison with PSPT (P<.05), using Wilcoxon matched-pairs signed-ranks test.
642
Stoker et al.
Field junctioning Classically, PSPT-based CSI is achieved with lateral or oblique cranial fields and multiple spine PA fields. PSPT collimation techniques result in steep dose falloff at the field edge. Consequently, matching field edges requires high-precision setup to achieve dose uniformity. The fluence modulation capabilities of IMPT may improve dose uniformity and setup robustness, similar to existing IMRT techniques (15, 16) with range control benefits added. Therefore, we next evaluated the benefits of IMPT versus PSPT in regard to robustness of field junctions. The graphs in Figure 3 represent the resulting dose modeled for PSPT using brass apertures (Fig. 3A) and IMPT with a 10cm field overlap (Fig. 3B). Figure 3A demonstrates a gradient (determined from the average slope between the 90% and 10% isodose) of approximately 50%/cm using PSPT, which is markedly above the 8%/cm gradient
International Journal of Radiation Oncology Biology Physics
achievable for IMPT fields (Fig. 3B). The solid lines (Fig. 3B) represent the planned delivery, and the dashed lines (Fig. 3B) represent the resulting dosimetric deviation after the modeled craniocaudal 4-mm setup error (2-mm each field). The deviation of <5% for the simulated IMPT setup error is significantly below the 25% observed for PSPT. When shifting the cervical PA field was required, the junction between the upper PA field and the AO fields was unique from the inferior spine junctions because the fields overlapped by 20 or more cm and the fields were not coplanar. Commercial optimization techniques performed reasonably well for our system under such circumstances, such that the modeled 4-mm-setup error induced a variation of 7% or less in dose. The proposed field tapering captures the robustness of similarly matched IMRT fields while maintaining the range modulation of proton delivery (16). Film measurements taken over 20 fractions for a single IMPT patient demonstrated variation of <5% between measured and calculated dose planes. A representative overlay of calculated and measured isodoses is shown in Figure 4. As facilities become more comfortable with planning and delivery methods and as robust optimization software advances, eliminating junction shifts over the treatment course may be possible but only after careful study by the treating institution, and preferably as part of clinical trials.
Discussion
Fig. 3. Dose profiles across the junction of 2 fields at the level of lumbar-thoracic spine junction. Solid lines represent relative doses from individual fields and their summed dose. Dashed lines represent resulting doses across the sampled profile after each field shifted by 2 mm. (A) Passively scanned proton therapy, which uses collimators to define field boundaries, results in deviation of 25% from planned delivery after field shifts were applied. (B) Intensity modulated proton therapy plan, which incorporates a nominal field overlap of 10 cm, resulted in a deviation of <5% from planned delivery after field shifts were applied.
This study presents a method for IMPT treatment of the craniospinal axis. We found that in comparison to PSPT, IMPT softened the field edge dose gradient for junctioned fields, and a method for achieving these soft junctions is described. The approach detailed herein is unique to date for IMPT, in that robustly matched fields reside within a single treatment plan. Because a sum plan is not generated, robust analysis is straightforward and within current treatment planning system functionality. OAR sparing during CSI, in some cases previously unachieved outside of modulated particle therapy, is also highlighted. IMPT provides superior flexibility in terms of available field arrangements and dose shaping. Junction robustness is dependent on the matched field gradients. As opposed to PSPT, manual entry of junction segments for IMPT-based CSI provides the planner a measure of control over the dose gradient. Here we highlight the capability of IMPT in achieving dose variations <5% in comparison with the 25% dose variation observed for PSPT for a 2-mm per field setup error. The clinical benefit of maintaining 5% is well established, and the potential for improved safety for patients should not be overlooked. Improvements in both setup reproducibility and robust optimization tools will hopefully facilitate routinely achieving this level of performance. Concerns driven by reduced junction shifting may potentially be addressed using established methods of field-matching verification (13).
Volume 90 Number 3 2014
IMPT craniospinal irradiation
643
Fig. 4. Isodose overlay of measured and calculated dose distributions for intensity modulated proton therapy to the spine. Solid lines represent values determined from Extended Dose Range 2 (EDR2) radiographic film measurements taken on the treatment couch just beneath supinely-positioned patient. Dashed lines were generated from the corresponding dosimetric plane calculated in the treatment planning system. We also found that IMPT further improved the OAR sparing in CSI relative to that with PSPT (10). The lens of the eye is one of the most sensitive structures in the body, with formation of cataracts and visual impairment presenting as late complications. Investigators have estimated a dose of 15 Gy to the lens causes a 50% probability of visual impairment in adults (17). The visual apparatus of children is likely more sensitive. In children <18 months of age, Hall et al (18) found a 35% to 50% increase in the risk of lens opacity for each 1-Gy increase in dose received. Also of note is the advantage of MFO-IMPT in cochlear sparing, observed here both in adults and children. PSPT fields deliver a single energy, using compensators for distal conformity to the target, but are not capable of modulating fluence at intermediate ranges, resulting in near-prescription dose to intermediately positioned OARs which eclipse the target beam’s eye view (10). Hua et al recommend cochlear doses not exceed 35 Gy in children to minimize sensorineural hearing loss risk (25). Quantitative Analysis of Normal Tissue Effects in the Clinic (QUANTEC) guidelines indicate that no threshold for sensorineural hearing loss prevention has been found in the adult population, and recommend a mean cochlear dose of at least 35 to 45 Gy (19). QUANTEC guidelines recommend a unilateral salivary gland mean dose of <20 Gy and mean bilateral salivary gland sparing of <25 Gy, each in order to avoid severe xerostomia, both of which are comfortably achieved by using IMPT (20). We did identify slight increases in doses to the thyroid, lungs, and kidneys, using IMPT. Although the observed doses would not likely result in primary organ dysfunction, further dose reductions would obviously be clinically desirable (21, 22). The current system represents a pioneering effort in IMPT. For the modeled beam, the scanning beam penumbra width is due in part to the treatment nozzle beam monitors, which increasingly scatter the pencil beam
at lower energies. Straightforward modifications, such as a retracting selected beam diagnostics or extending the vacuum portion of the nozzle, could reduce the spot size and penumbra. Penumbra measurements suggest that using apertures will reduce doses in the lungs and kidneys. The parotid dose for pediatric patients could also be reduced, and thyroid sparing could potentially be improved at least for adult anatomies to match PSPT levels. Decreasing the spot size has been demonstrated to facilitate near-target OAR sparing (23) and could be combined with aperture use. Proton therapy centers under construction are trending toward a dedicated scanning beam facility design. Aperture use will markedly diminish relative to PSPT-capable centers not justifying an in-house machine shop. If relying on third parties for infrequent machining of patient-specific apertures is not tractable, multipatient apertures of widely applicable dimensions could be used.
Conclusions An optimal beam arrangement and planning approach are critical to realizing the full benefit of IMPT. Ideally, the planner would be able to set the field overlap, and the optimizer would determine a robust solution. However, current commercially available optimization software is not capable of taking into account the multi-isocenter, interfield setup error necessary for robust field matching. Inspection of the individual field doses across a spine field junction obtained using single-stage MFO optimization reveals a nonrobust solution resulting from the optimizer’s solving for a local minima. Contrary to reported results for IMRT optimization, we found that the presence of multiple local minima traps is an impediment to finding a robust IMPT solution (24). Our 3-stage method provides a way to alter the initial conditions prior to the final optimization. Nevertheless, the method proposed herein has the potential
644
Stoker et al.
to be further simplified. Increasing field size to 40 cm would eliminate the need to shift the upper PA field, eliminating a junction for many patients. Also, we reference a previous work detailing an x-ray method for CSI. A field-in-field forward planning approach provided the IMRT optimizer low-gradient tapered upper and lower spine PA fields (15). Present IMRT optimization software incorporates this forward-planned base dose distribution, which facilitates a multistage planning process. Eventually, incorporating a base dose distribution will also be available for IMPT optimizers. Further clinical study of such techniques, including treatments forgoing junction shifts, should only be undertaken after further careful study at the treating institution and preferably as part of prospective clinical trials.
References 1. Schwartz CL. Long-term survivors of childhood cancer: The late effects of therapy. Oncologist 1999;4:45-54. 2. Barney CL, Brown AP, Grosshans DR, et al. Technique, outcomes, and acute toxicities in adults treated with proton beam craniospinal irradiation. Neuro Oncol 2014;16:303-309. 3. Cochran DM, Yock TI, Adams JA, et al. Radiation dose to the lens during craniospinal irradiation-an improvement in proton radiotherapy technique. Int J Radiat Oncol Biol Phys 2008;70:1336-1342. 4. Howell RM, Giebeler A, Koontz-Raisig W, et al. Comparison of therapeutic dosimetric data from passively scattered proton and photon craniospinal irradiations for medulloblastoma. Radiat Oncol 2012;7:116. 5. St Clair WH, Adams JA, Bues M, et al. Advantage of protons compared to conventional X-ray or IMRT in the treatment of a pediatric patient with medulloblastoma. Int J Radiat Oncol Biol Phys 2004;58:727-734. 6. Brown AP, Barney CL, Grosshans DR, et al. Proton Beam Craniospinal Irradiation Reduces Acute Toxicity for Adults With Medulloblastoma. Int J Radiat Oncol Biol Phys 2013;86:277-284. 7. Moeller BJ, Chintagumpala M, Philip JJ, et al. Low early ototoxicity rates for pediatric medulloblastoma patients treated with proton radiotherapy. Radiat Oncol 2011;6:58. 8. Owusu-Agyemang P, Grosshans D, Arunkumar R, et al. Non-invasive anesthesia for children undergoing proton radiation therapy. Radiother Oncol 2014;111:30-34.
International Journal of Radiation Oncology Biology Physics 9. Giebeler A, Newhauser WD, Amos RA, et al. Standardized treatment planning methodology for passively scattered proton craniospinal irradiation. Radiat Oncol 2013;8:32. 10. Dinh JS, Stoker J, Georges R, et al. Comparison of proton therapy techniques for treatment of the whole brain as a component of craniospinal irradiation. Radiat Oncol 2013;8:289. 11. Liu W, Zhang X, Li Y, et al. Robust optimization of intensity modulated proton therapy. Med Phys 2012;39:1079-1091. 12. Martens C, Claeys I, De Wagter C, et al. The value of radiographic film for the characterization of intensity-modulated beams. Phys Med Biol 2002;47:2221-2234. 13. Michalski JM, Klein EE, Gerber R. Method to plan, administer, and verify supine craniospinal irradiation. J Appl Clin Med Phys 2002;3: 310-316. 14. StataCorp. Stata: release 13. College Station, TX: StataCorp LP; 2013. 15. Pai Panandiker A, Ning H, Likhacheva A, et al. Craniospinal irradiation with spinal IMRT to improve target homogeneity. Int J Radiat Oncol Biol Phys 2007;68:1402-1409. 16. Yom SS, Frija EK, Mahajan A, et al. Field-in-field technique with intrafractionally modulated junction shifts for craniospinal irradiation. Int J Radiat Oncol Biol Phys 2007;69:1193-1198. 17. Henk JM, Whitelocke RA, Warrington AP, et al. Radiation dose to the lens and cataract formation. Int J Radiat Oncol Biol Phys 1993;25: 815-820. 18. Hall P, Granath F, Lundell M, et al. Lenticular opacities in individuals exposed to ionizing radiation in infancy. Radiat Res 1999;152: 190-195. 19. Bhandare N, Jackson A, Eisbruch A, et al. Radiation therapy and hearing loss. Int J Radiat Oncol Biol Phys 2010;76:S50-57. 20. Deasy JO, Moiseenko V, Marks L, et al. Radiotherapy dose-volume effects on salivary gland function. Int J Radiat Oncol Biol Phys 2010;76:S58-63. 21. Foote RL, Stafford SL, Petersen IA, et al. The clinical case for proton beam therapy. Radiat Oncol 2012;7:174. 22. Newhauser WD, Fontenot JD, Mahajan A, et al. The risk of developing a second cancer after receiving craniospinal proton irradiation. Phys Med Biol 2009;54:2277-2291. 23. van de Water TA, Lomax AJ, Bijl HP, et al. Using a reduced spot size for intensity-modulated proton therapy potentially improves salivary gland-sparing in oropharyngeal cancer. Int J Radiat Oncol Biol Phys 2012;82:e313-319. 24. Wu Q, Mohan R. Multiple local minima in IMRT optimization based on dose-volume criteria. Med Phys 2002;29:1514-1527. 25. Hua C, Bass JK, Khan R, et al. Hearing loss after radiotherapy for pediatric brain tumors: Effect of cochlear dose. Int J Radiat Oncol Biol Phys 2008;72:892-899.