Dosimetric comparison of distal esophageal carcinoma plans for patients treated with small‐spot intensity‐modulated proton versus volumetric‐modulated arc therapies

Background Esophageal carcinoma is the eighth most common cancer in the world. Volumetric‐modulated arc therapy (VMAT) is widely used to treat distal esophageal carcinoma due to high conformality to the target and good sparing of organs at risk (OAR). It is not clear if small‐spot intensity‐modulated proton therapy (IMPT) demonstrates a dosimetric advantage over VMAT. In this study, we compared dosimetric performance of VMAT and small‐spot IMPT for distal esophageal carcinoma in terms of plan quality, plan robustness, and interplay effects. Methods 35 distal esophageal carcinoma patients were retrospectively reviewed; 19 patients received small‐spot IMPT and the remaining 16 of them received VMAT. Both plans were generated by delivering prescription doses to clinical target volumes (CTVs) on phase‐averaged 4D‐CT's. The dose‐volume‐histogram (DVH) band method was used to quantify plan robustness. Software was developed to evaluate interplay effects with randomized starting phases for each field per fraction. DVH indices were compared using Wilcoxon rank‐sum test. For fair comparison, all the treatment plans were normalized to have the same CTVhigh D95% in the nominal scenario relative to the prescription dose. Results In the nominal scenario, small‐spot IMPT delivered statistically significantly lower liver Dmean and V30Gy[RBE], lung Dmean, heart Dmean compared with VMAT. CTVhigh dose homogeneity and protection of other OARs were comparable between the two treatments. In terms of plan robustness, the IMPT and VMAT plans were comparable for kidney V18Gy[RBE], liver V30Gy[RBE], stomach V45Gy[RBE], lung Dmean, V5Gy[RBE], and V20Gy[RBE], cord Dmax and D0.03cm3, liver Dmean, heart V20Gy[RBE], and V30Gy[RBE], but IMPT was significantly worse for CTVhigh D95%, D2cm3, and D5%‐D95%, CTVlow D95%, heart Dmean, and V40Gy[RBE], requiring careful and experienced adjustments during the planning process and robustness considerations. The small‐spot IMPT plans still met the standard clinical requirements after interplay effects were considered. Conclusions Small‐spot IMPT decreases doses to heart, liver, and total lung compared to VMAT as well as achieves clinically acceptable plan robustness. Our study supports the use of small‐spot IMPT for the treatment of distal esophageal carcinoma.


| INTRODUCTION
Esophageal carcinoma is the eighth most common cancer and the sixth most common cause of cancer deaths worldwide. There are an estimated 17 290 new cases and 15 850 deaths annually in America, and distal esophageal cancer cases are increasing rapidly in developed countries. 1,2 In recent years, trimodality therapy (neoadjuvant chemoradiation followed by esophagectomy) has improved clinical outcomes in patients with locally advanced esophageal cancers compared to surgery alone. 3,4 Concurrent chemotherapy, usually with weekly carboplatin and paclitaxel, combined with radiation doses of 41.4-50.4 Gy are considered standard treatments in the modern era. 5 The long-term results of Radiation Therapy Oncology Group (RTOG) clinical trial 85-01 confirmed that chemoradiation increased overall survival for patients with esophageal carcinoma compared with radiotherapy (RT) alone. 5 Due to proximity to surrounding organs at risk (OAR) such as heart, spinal cord, lungs, kidney, liver, and the remaining stomach, the RT planning for distal esophagus carcinoma poses special challenges. 6,7 Sufficient radiation doses must be applied to the tumor and lymph node areas, while protecting nearby critical normal structures.
Volumetric-modulated arc therapy (VMAT) is an advanced form of intensity-modulated radiation therapy (IMRT) that can deliver a highly conformal dose distribution using single or multiple arcs. 8 Compared with static-field IMRT, VMAT achieves similar OAR sparing and planning target volume (PTV) coverage with significantly shorter treatment time. 9,10 Proton beam therapy delivers highly conformal target coverage, while sparing adjacent OARs due to its unique Bragg peak dose deposition characteristics. Proton beam therapy 11 has several forms of delivery including passive scattering (PSPT), uniform scanning, and intensity-modulated proton therapy (IMPT). Unlike PSPT, IMPT uses magnetic steering of a narrow proton beam, termed a beamlet, to deliver a modulated dose to a spot of a specified size, which offers improved high-dose conformality compared with PSPT and better OAR sparing in the mid-to lowdose range compared to IMRT. 12,13 Therefore, it is hypothesized that IMPT can improve the therapeutic ratio which will result in fewer adverse effects, while achieving the same tumor control as IMRT or better. 14 However, IMPT is highly sensitive to setup and range uncertainties, as well as vulnerable to respiratory motion present in the distal esophageal regions. 15,16 The uncertainties originate from daily patient alignment, conversion of Hounsfield units to stopping power, artifacts in computed tomography, and anatomical changes in patients etc. 17 The interaction between dynamic beamlet delivery and respiratory motion, also called interplay effect, may degrade the quality of planned dose distributions, [18][19][20][21][22][23][24][25][26][27][28][29][30] compromising the safety and efficacy of the proposed treatment. In addition, large spot sizes, common in older IMPT machines, tend to lead to larger penumbras, which results in undesired dose to adjacent OARs. The majority of new proton facilities offer smaller spot sizes, which can produce smaller penumbras and better OAR sparing. However, smaller spotsize plans can exacerbate the negative consequences of patient setup uncertainty and make interplay effects more prominent. 25,31 As a result, we first need to quantify and then mitigate the impact of uncertainties and interplay effects for small-spot IMPT for the treatment of distal esophageal cancers.
Previous studies have focused on the comparison of plan quality alone among plans generated using either PSPT, large-spot IMPT (spot size σ as large as 6-15 mm depending on proton energies), and/or IMRT [32][33][34][35][36][37] with no mention of plan uncertainties or motion effect analyses. Recently, Shiraishi et al. reported that in a large cohort of esophageal cancer patients, PSPT and large-spot IMPT significantly reduced radiation exposures to the whole heart and cardiac substructures compared with IMRT. 32 Welsh et al. found that large-spot IMPT for distal esophageal carcinoma also lowered the doses to bilateral lungs and liver compared to IMRT. 37 More recently, a large, retrospective multi-institutional study also demonstrated that proton beam therapy appeared to be more clinically advantageous compared with 3-dimensionl conformal RT and IMRT in lowering the incidence of pulmonary and cardiac complications as well as the mean length of in-hospital stay. 38 To the best of our knowledge, no dosimetric study has been reported comparing plan quality and plan robustness for small-spot IMPT and VMAT in the treatment of distal esophageal cancer. The aim of this study is to evaluate the feasibility of small spot IMPT for such treatments. We compared plan quality and robustness for VMAT and small-spot IMPT. The interplay effects of small-spot IMPT were also quantified.

2.A | Patient selection
Nineteen consecutive patients with distal esophageal carcinoma treated with IMPT and 16 patients treated with VMAT between May 2014 and September 2017 at our institution were retrospectively reviewed. Table 1 shows the patient characteristics for IMPT and VMAT treatment groups. The patients included in this study were not randomly selected, but were carefully selected by an experienced physicist from the existing database to ensure that the patients from the two treatment groups did not show significant differences in age, gender, body mass, tumor volumes, motion amplitude, or prescription dose (Table 1); the cases were consecutively considered based on radiotherapy treatment start date. Patients were excluded from this study if they were less than 18 yr old and/ or not treated with curative intent. All treatment plans included in this work were approved and delivered clinically. Patients were staged using PET/CTs. All patients had an Eastern Cooperative Oncology Group performance status of 2 or better. No patients in this study had implanted cardiac devices. The commercial treatment planning system (Eclipse TM , version 13, Varian medical system, Palo Alto, CA) was used to generate treatment plans based on the simulation 4D CTs, which were used to localize the targets and OARs. Heart, cord, liver, stomach, bowel, and kidney and targets were contoured on the 4D-averaged CT. The 41.4 Gy[RBE] was also allowed for CTV low or if the patient was treated preoperatively. CTV low and CTV high were generated as follows:

2.B | Patient simulation and contouring
first, we identified the appropriate gross target volume (GTV) on the average CT or one of the respiratory phase CT scans, and the coregistered PET and CT scan. Then, a 3-4 cm expansion was added along the mucosal surface longitudinally, in addition to a 1-1.2 cm radial expansion for the CTVs which were anatomically constrained.
The lower CTV volume typically included a small expansion of elective nodal volumes in the para-esophageal region. The treating radiation oncologist also adjusted the expansion of margins based on the pathology and location of the tumor; the potential microscopic tumor extent and anatomic boundaries of heart, lungs, liver, kidneys, and bowel were also taken into consideration in the final target delineation.

2.C | Treatment planning
In VMAT treatment planning, all treatment plans were generated on the averaged CT. We used planning target volumes (PTV), formed by a 5-mm uniform expansion of CTVs, for plan optimization and evaluation in VMAT. Most commonly, 2 to 3 arcs were used. Photon optimizer (PO) model in the Eclipse TM TPS was used for VMAT optimization, and analytical anisotropic algorithm (AAA) model was used for dose calculation. The dose calculation grid size was 3 mm.
The dosimetrists created treatment plans, which satisfied institutional dose constraints (Table 2) for OAR sparing. For target coverage, V 100% of PTV high was at least 95% and D 0:03cm 3 of PTV high was no more than 110% of prescription dose.
In IMPT treatment planning, all treatment plans were also generated on the averaged CT. An optimization target volume (OTV) was formed by uniform expansion of the CTV by 5 mm to help generate a robust plan. Proton spots were placed strategically outside of the OTV as well to ensure homogenous dose distributions in the OTV.
T A B L E 1 Patient characteristics between intensity-modulated proton therapy (IMPT) and volumetric-modulated arc therapy (VMAT) treatment groups.

IMPT
VMAT P-value Usually, 2 to 3 proton beams were used. Single field optimization (SFO) was always the first option, however, multiple-field optimiza- and D mean between Monte Carlo simulation and TPS computation has been less than 3% and 2% for most of patients respectively.

2.D | Treatment delivery
The Clinac machines (Varian medical system, Palo Alto, CA) were used to deliver the VMAT plans. The related parameters including field information, energy, and estimated delivery duration are shown in Table S1. Typically, daily on-board imaging or cone-beam CT was used as the image-guided RT methods.
For IMPT, the Hitachi ProBeat-V spot-scanning proton beam machines (Hitachi, Tokyo, Japan) were used. The active scanning proton beam machine was commissioned to have an energy-dependent spot size (σ) of 2 to 6 mm, with a fixed spot spacing of 5 mm.
Our proton beam scanning machine had discrete proton energies ranging from 71.3 to 228.8 MeV. These discrete energies were carefully selected to minimize the ripples in the dose distribution along the beam direction and minimum MU effects. The energy layer switch time for all 97 energies ranged from 1.9 to 2.0 s, with an average of 1.91 s. The average spill length was 7.9 s. The average magnet preparation and verification time was 1.93 ms. 49 The related field and energy choices, estimated delivery duration, and repainting numbers are shown in the Table S2. We used orthogonal pair kV images to align to bony anatomy during the IMPT treatment.
For both IMPT and VMAT, the set up images were reviewed offline and approved by the treating radiation oncologist to make sure that the patient setup errors were within clinical tolerance.

2.F | Robustness quantification
Patient set up uncertainty is considered to be random and can be modelled as a Gaussian distribution. Range uncertainty is considered to be systematic, but range uncertainty of a large patient population can also be considered to be a Gaussian distribution. 50  For VMAT plans, seven scenarios were considered, including one nominal scenario and six perturbed scenarios. The six perturbed scenarios were created by rigidly shifting the isocenter in the same canonical directions by a distance of 3 mm, but with no range uncertainty considerations. A DVH curve was generated for each uncertainty scenario and consequently a DVH band was formed corresponding to multiple uncertainty scenarios (Fig. 1). In order to evaluate the robustness of VMAT and IMPT treatment plans, the width of the DVH band was used as a surrogate for robustness indications. The width of the DVH band was the difference between the maximum and minimum of certain DVH indices (Fig. 1)  software. 45,51 The starting phase of each field for each fraction was randomized to minimize the influence of starting phases. The interplay effects were evaluated using the fraction number in the prescription of IMPT treatment (Table S2).

2.G | Interplay effect evaluation
The proton absolute dose calibration process follows IAEA TRS-398 protocol. The absolute dose of 1MU for a specific selected beam is 1 cGy. The beam used for calibration is: range 20 cm, spread out Bragg peak (SOBP) 10 cm, and field size 10 cm × 10 cm. It is composed of 27 energy layers (120-173.6 MeV), 289 spots in each layer with 6 mm spot spacing, a total of 200 MU. In IMPT treatment planning, we also used the iso-layer repainting to mitigate interplay effects. 45,51,52 For respiratory motion less than or equal to 5 mm, the minimum and maximum MU limits in the proton machine were 0.003 and 0.04 MU, respectively. A smaller maximum MU limit of 0.01 MU was used for cases with respiratory motion greater than 5 mm. The purpose of a smaller maximum MU limit was to make the delivery system perform a larger number of iso-layer repainting, which in turn mitigated the impact of interplay effects. 52,53 During the delivery process, if the intensity of one spot was larger than the maximum MU limit, the spot was split into multiple spots, and the split spots were appended in the spot list of the same energy layer and delivered individually. For the spots with MU less than the minimum MU limit: if the intensity of one spot was larger than half of   Table 3].
Some outliers were observed. For example, heart D mean , V 20Gy [RBE] , and V 30Gy [RBE] in IMPT [Figs. 3(e)-3(f)]. These outliers were found to come from the same patient. The heart was in close proximity to the CTV high [ Fig. S1(a)], which resulted in higher dose to the heart [ Fig. S1(b)]. Similarly, close proximity of the heart to a tumor resulted in the outliers in heart V 30Gy [RBE] and V 40Gy [RBE] [Fig. 3(f)] in VMAT. To avoid vital organs including cord and lungs for this patient, we had to limit target dose coverage and homogeneity to some degree in VMAT, resulting in additional outliers in D 5% -D 95% [ Fig. 3(b)].

3.B | Plan robustness
Considering the plan robustness, the DVH index range of CTVs and OARs for both treatment groups were compared (  (Fig. 4).

3.C | Interplay effect
Interplay effects were considered for all of the IMPT plans [see

| DISCUSSION
The purpose of this comparative planning study was to evaluate the process and feasibility of small-spot machine IMPT in the treatment for distal esophagus carcinoma as well as provide a dosimetric comparison to VMAT in regard to plan quality and robustness.
Small spot-size IMPT is an attractive modality for the treatment of esophageal cancer. Our study compared IMPT with spot sizes of 2-6 mm (σ) with VMAT. We found that small-spot IMPT achieved similar plan quality as VMAT in terms of target dose coverage, homogeneity, and sparing of most OARs. More importantly, it significantly lowered heart, liver, and bilateral lung doses as compared to VMAT. As a result, IMPT will likely reduce the incidence and also severity of RT-induced cardiac and pulmonary toxicities in the longterm and perioperatively. However, the equipoise of such considerations and clinically meaningful significance over conventional treatments such as IMRT or VMAT remain undefined for distal esophageal cancer therapies.
Although small-spot IMPT is potentially capable of producing better plans due to sharper penumbra, the uncertainties due to proton range and patient setup can greatly compromise plan qualities. 56,57 Therefore, it is important to take into account the uncertainties in treatment planning when using a small-spot IMPT machine to treat patients with esophageal carcinoma. Our study demonstrates that small-spot IMPT treatment plans achieved clinically defined planning requirements in terms of plan robustness and met the clinical standards for RT. Small-spot IMPT should be considered for the routine treatment in patients with distal esophageal carcinoma. However, robustness relative to internal organ motion remains a major challenge in small-spot IMPT treatments for esophageal cance. 30,33,45,53 Compared with IMPT, VMAT has been shown to be more robust with respect to organ motions or anatomic changes. 58 Therefore, it is vital to consider and optimally mitigate the impact of respiratory motions when IMPT is clinically implemented. At our institution, we carefully consider respiratory motion The comparison of plan quality using dose-volumehistogram (DVH) indices. arrangements worked very well for most patients. However, in some scenarios, there were variants of beam arrangements for specific patients: if the patient was pretreated by radiation therapy and the dose in certain organs needed to be limited, three or four beams were used to spare certain organs (Table S2); if the patient had significant respiratory motion and two beams could not achieve the plan robustness of clinical requirements, then three or four beams were used to improve the plan robustness (Table S2).
This study has a number of limitations. First, the number of the patients included in this study was not sufficiently large nor were they matched. The results could be affected by interpatient variability and different planning skills. However, these represented consecutive samples of actual IMPT vs VMAT plans that were delivered in the clinic over a similar period of time. To address the aforementioned issues, a study with a larger patient population with both VMAT and IMPT plans generated for each patient is currently under way to further generalize our conclusions. Second, only a limited number of uncertainty scenarios were considered in this study, which might underestimate the impact of uncertainties in selected IMPT plans.
In the future, other tumor locations (cervical, proximal, and midesophageal) should be considered in further studies. Furthermore, more patient data with short/long-term clinical outcomes, perioperative complications, gradation-related toxicities, and patient-reported outcome should be reported to evaluate the potential clinical benefits of small-spot IMPT over VMAT plans.

| CONCLUSION
Compared to VMAT, small-spot IMPT significantly improves RT sparing of the heart, liver, and lungs, as well as achieves clinically acceptable plan robustness. The impact of interplay effects is small when proper treatment planning and respiratory motion measures are taken. Our results support the feasibility and acceptability for the routine clinical use of small-spot IMPT in patients with distal esophageal carcinoma.

CONF LICTS OF INTEREST
None.

ETHICAL CONSIDERATIONS
This research was approved by the Mayo Clinic Arizona institutional review board (IRB, 13-005709). The informed consent was waived by IRB protocol. Only image and dose-volume data were used in this study. All patient-related health information was removed for the study.