Dosimetric feasibility of single-energy proton modulated arc therapy for treatment of chordoma at the skull base

To the Editor,

Proton therapy theoretically holds two important advantages over photon therapy: 1) Proton beams have a finite penetration range, i.e. no exit dose; and 2) produce a dose distribution that increases with depth until the distal edge of the Bragg Peak [1]. The resulting dose profile has a large dose deposition at depth relative to the dose deposition proximal to the Bragg Peak. Similar to the evolution of intensity modulated radiotherapy (IMRT), proton therapy could benefit from a rotational, volume modulated arc therapy (VMAT) delivery approach. First proposed in 1997, proton arc therapy was designed to reduce the intermediate dose to healthy tissue [2]. Proton arc therapy relies on multiple gantry angles, reducing the weight of each beam angle while maintaining conformal dose to the target by escalating the dose delivered at each gantry angle. Unlike photons, protons deliver the majority of their dose at a precise depth (Bragg Peak), and therefore rotational delivery of proton pencil beams (PBS) could produce a significant dosimetric advantage over VMAT or current single field uniform dose (SFUD) proton treatment planning approaches [3,4].

This work assesses the feasibility of proton modulated arc therapy (PMAT) using a skull base chordoma treatment planning context, a preliminary step in the process of vetting this potentially advantageous delivery method for further investigation. Specifically, this work explores whether PMAT can treat chordoma at the skull base with a better conformity and homogeneity index than SFUD and IMPT plans. Motivated by the possibility of a single or reduced set of mono-energetic PMAT beams tracking a range between the distal and the proximal edges of the target while the gantry is rotated, PMAT could theoretically achieve higher conformity and homogeneity by modulating the intensity at various angles.

A cohort of six patients diagnosed with chordoma, specifically located at the base of the skull, was imaged from the top of the skull to the first thoracic vertebrae. These images were used for creation of the proton therapy plans using the Varian Eclipse v11 treatment planning system. The proton arc treatment plans were created to simulate patient treatment delivery, but were not delivered. A detailed description of the planning process with associated figures can be viewed in the supplemental material. All SFUD and IMPT plans were created under our institutional clinical protocols.

Plans were evaluated by comparing dose conformity, dose homogeneity, isodose coverage, and organs at risk (OAR) sparing for each treatment modality. Dose conformity was assessed by the conformity index (CI) and dose homogeneity assessment was conducted using the homogeneity index (HI), using the definitions from the Radiation Therapy Oncology Group (RTOG) [5–7]. The concept of HI was used to supplement CI as HI expressed information on the dose levels within structure volumes. The extent of OAR sparing was assessed by comparing the maximum dose received to the cord, optic nerve, orbits, optic chiasm, brainstem, hippocampus, temporal lobes, hypothalamus, and pituitary as specified by RTOG [5]. The mean dose to the temporal lobes and cochlea was also used as dose-volume histogram (DVH) indicators [5]. Dose constraints for these DVH indicators are summarized in the second column of Table 1.

Table 1. Average dose differences between PMAT minus SFUD and PMAT minus IMPT comparisons for each OAR and DVH indicator constraints.

OAR	DVH Indicator constraints		PMAT – SFUD (Gy)		PMAT – IMPT (Gy)
Brainstem	Dmax <67 Gy		−1.30 ± 2.78		−0.40 ± 3.56
Cord	Dmax <50 Gy		−4.32 ± 4.34		−2.16 ± 2.52
Chiasm	Dmax <60 Gy		−1.48 ± 7.45		−3.32 ± 5.56
Lt optic nerve	Dmax <60 Gy		−3.06 ± 8.32		−3.40 ± 7.04
Rt optic nerve	Dmax <60 Gy		−3.64 ± 6.40		−3.72 ± 6.91
Lt orbits	Dmax <10 Gy		−0.04 ± 1.13		0.56 ± 1.25
Rt orbits	Dmax <10 Gy		−1.40 ± 2.61		−1.2 ± 2.68
Lt hippocampus	Dmax <20 Gy		−3.45 ± 6.46		−2.93 ± 7.02
Rt hippocampus	Dmax <20 Gy		1.18 ± 5.03		1.38 ± 5.14
Hypothalamus	Dmax <10 Gy		−6.53 ± 6.22		−8.33 ± 4.26
Lt cochlea	Dmean ≤30 Gy		−10.90 ± 4.56		−2.85 ± 4.05
Rt cochlea	Dmean ≤30 Gy		−7.90 ± 3.52		−2.80 ± 5.46
Lt temporal lobe	Dmax <45 Gy	Dmean ≤25 Gy	2.88 ± 7.14	−3.25 ± 5.65	1.75 ± 7.29	−4.30 ± 5.10
Rt temporal lobe	Dmax <45 Gy	Dmean ≤25 Gy	0.95 ± 7.79	−2.93 ± 5.23	0.03 ± 8.20	−3.63 ± 4.14
Pituitary	Dmax <45 Gy		2.50 ± 4.04		0.60 ± 2.81

Statistical differences in the OAR DVH indicator dose levels were calculated using an unpaired Wilcoxon rank sum test at the 95% confidence level. A Wilcoxon rank sum test was chosen due to the number of patients in the study and two key assumptions. For one, it was a non-parametric test that assumes a specific DVH indicator’s dose levels across the patient cohort are independent of one another. In addition, this test does not assume the DVH indicator dose levels would be distributed normally in a patient population. Statistical similarity among DVH indicators would mean that the PMAT plan did not reduce the dose to the OARs significantly at the 95% confidence level.

The target dose homogeneity was similar between the plans. The average CI over six patients was 0.88 ± 0.08, 0.94 ± 0.05, and 0.97 ± 0.02 for the SFUD, IMPT, and PMAT plans, respectively. The average HI over six patients was 1.06 ± 0.03, 1.08 ± 0.04, and 1.10 ± 0.02 for the SFUD, IMPT, and PMAT plans, respectively. Numerically, both the average CI and HI were the highest, with the smallest standard deviation, for PMAT plans.

OAR DVH indicator dose levels were assessed by subtracting the dose levels from SFUD and IMPT plans from the dose levels from the PMAT plans. Negative values were interpreted as the PMAT plans having superior OAR sparing. Dose difference averages are shown in Table 1 for each DVH indicator with associated standard deviations. PMAT plans performed better than SFUD and IMPT plans except in the pituitary (SFUD and IMPT performed better), temporal lobes (SFUD and IMPT performed better), right hippocampus (SFUD and IMPT performed better), and left orbit (IMPT performed better). Table 1 contains a full summary of these dose differences. Approximately 70% of DVH indicators in both comparisons showed better sparing with PMAT, up to a 15.4 Gy difference. Of the remaining 30% of DVH indicators that displayed less dose sparing with PMAT, the maximum dose difference was 10.9 Gy. The Wilcoxon rank sum test p-values demonstrated no statistically significant differences among the plans (p > 0.05 for all comparisons), meaning that the OAR doses were statistically equivalent between the plans for this patient population despite dosimetric differences that indicate PMAT is superior.

PMAT spared 70% of OARs better than IMPT and SFUD and was statistically equivalent to IMPT and SFUD in terms of CI and HI, demonstrating the feasibility of PMAT as a chordoma treatment modality. PMAT’s numerical superiority aside, it is unsurprising that given the analysis of six patients within this feasibility study, the statistical tests for significant CI and HI difference among plans showed statistical similarity. Strictly adhering to this study’s endpoint, by virtue of its similar dosimetric performance to established SFUD and IMPT methods, PMAT appears to be feasible from a treatment planning perspective.

Undoubtedly, increasing the number of patients in the cohort would provide more powerful and meaningful statistics for the plan comparisons, potentially elucidating statistically significant CI and HI differences. However, regardless of whether a larger sample size would yield more statistically significant results for PMAT, moving beyond demonstrating feasibility towards discerning biological significance from a Wilcoxon rank sum test seems inadequate and without context. The fact that PMAT was the only modality able to deliver the additional 3–5 Gy needed to achieve the intended dose prescription underscores the existence of potential clinical significance despite statistically similar CI and HI values across all modalities.

Though CI and HI are useful metrics in demonstrating PMAT’s feasibility as a method that is at least comparable to SFUD and IMPT, they do not take into account the advantages of aggregate cognitive and functional effects as a result of lower OAR doses within neuroanatomy. Most neuroanatomical structures have been classified as radioresistant because canonical radiosensitivity assays use structural degeneration and cell death as endpoints, ignoring the legitimate possibility of functional damage despite intact structural integrity [8]. Indeed, in the neuronal synapse, one of the only neuroanatomical sites where neurological function is measured as a radiosensitivity endpoint, significant functional impairments in response to only a 2 Gy have been observed [8]. With the functional consequences of irradiating neuroanatomy largely unexplored, the potential advantages of PMAT’s ability to spare 70% of OARs cannot be assumed to be biologically insignificant.

This study’s choice to investigate the feasibility PMAT in comparison to more conventional proton techniques carries with it the caveat of dosimetric uncertainty. Though uniform dose was achieved by superimposing multiple fields from multiple directions, due to the Bragg Peak’s steep dose gradient, the potential for error in the delivered dose from range uncertainties could be large. The uncertainties in proton range are generally systematic in nature. They arise from limitations inherent in the computed tomography (CT) data (i.e. beam hardening, noise, or resolution), calibration of the CT Hounsfield unit and the transport properties of the therapeutic proton beam, CT artifacts, and uncertainty in energy dependent RBE. Random errors, such as variations in the proton beam energy, cannot be addressed within the scope of this work.

With the results of this work indicating that PMAT is a feasible chordoma treatment technique, future work will focus on incorporating more patients into this study to obtain more statistically powerful comparisons between proton modalities. As an extension of investigating the potential biological advantages associated with PMAT, future work will also endeavor to expand the concept of structural and necrosis based radiosensitivity assays for neuroanatomy to include functional characterization of radiation response.

PMAT was shown to be a viable treatment technique for chordomas located at the skull base in this feasibility study. Compared to SFUD and IMPT techniques, PMAT demonstrated comparable conformity and homogeneity indices while simultaneously reducing dose levels for several key DVH indicators. Despite statistical analysis showing no statistical differences between the treatment techniques, this study shows that PMAT offers potential advantages in the treatment of chordomas at skull base and warrants further investigation.

Janid Patricia Blanco Kiely
Department of Radiation Oncology, University of Pennsylvania, Philadelphia, Pennsylvania 19104, USA
[email protected] M. White
Department of Radiation Oncology, University of Pennsylvania, Philadelphia, Pennsylvania 19104, USA

Acknowledgments

The authors would like to thank Alejandro Carabe-Fernandez, Ph.D. for his stimulating discussions.

Disclosure statement

No potential conflict of interest was reported by the author(s).