Beam angle comparison for distal esophageal carcinoma patients treated with intensity‐modulated proton therapy

Abstract Purpose To compare the dosimetric performances of intensity‐modulated proton therapy (IMPT) plans generated with two different beam angle configurations (the Right–Left oblique posterior beams and the Superior–Inferior oblique posterior beams) for the treatment of distal esophageal carcinoma in the presence of uncertainties and interplay effect. Methods and Materials Twenty patients’ IMPT plans were retrospectively selected, with 10 patients treated with the R‐L oblique posterior beams (Group R‐L) and the other 10 patients treated with the S‐I oblique posterior beams (Group S‐I). Patients in both groups were matched by their clinical target volumes (CTVs—high and low dose levels) and respiratory motion amplitudes. Dose‐volume‐histogram (DVH) indices were used to assess plan quality. DVH bandwidth was calculated to evaluate plan robustness. Interplay effect was quantified using four‐dimensional (4D) dynamic dose calculation with random respiratory starting phase of each fraction. Normal tissue complication probability (NTCP) for heart, liver, and lung was calculated, respectively, to estimate the clinical outcomes. Wilcoxon signed‐rank test was used for statistical comparison between the two groups. Results Compared with plans in Group R‐L, plans in Group S‐I resulted in significantly lower liver Dmean and lung V30Gy[RBE] with slightly higher but clinically acceptable spinal cord Dmax. Similar plan robustness was observed between the two groups. When interplay effect was considered, plans in Group S‐I performed statistically better for heart Dmean and V30Gy[RBE], lung Dmean and V5Gy[RBE], and liver Dmean, with slightly increased but clinically acceptable spinal cord Dmax. NTCP for liver was significantly better in Group S‐I. Conclusions IMPT plans in Group S‐I have better sparing of liver, heart, and lungs at the slight cost of spinal cord maximum dose protection, and are more interplay‐effect resilient compared to IMPT plans in Group R‐L. Our study supports the routine use of the S‐I oblique posterior beams for the treatments of distal esophageal carcinoma.

Robust optimization  has been introduced to mitigate the impact of uncertainties, and is now widely accepted in the routine proton clinical practice. However, techniques to mitigate interplay effect would still require further improvement. 48 Abdominal compression and breath holding can help limit the motion, but often result in less patient comfort during treatments. Gating 49 and repainting 13,16,50 can provide better dose distribution during motion, but will inherently prolong the treatment time. Tumor tracking 51,52 and the recently proposed 4D optimization 45,[53][54][55] are still in development, and will need further verifications prior to routine clinical use. Currently at our institution, IMPT plans are derived for distal esophageal cancer based on the averaged four-dimensional computed tomography (4D-CT) with uncertainties considered. Generated IMPT plans are then verified by plan evaluation on maximum exhalation and maximum inhalation respiratory phases, as well as a separate interplay effect calculation. 45 Previous dosimetric studies [56][57][58][59] of esophageal carcinoma treatment compared both proton and photon treatments. Zhang et al. reported that proton therapy provided significantly better sparing of the lungs than IMRT. 56 Shiraishi et al. found that in patients with mid to distal esophageal cancer, proton therapy resulted in significantly lower radiation exposure to the whole heart and cardiac structures compared to IMRT. 57 Lin et al. concluded in a large-scale multiinstitutional study that proton therapy was associated with significantly less postoperative complications and shorter length of in-hospital stay than 3D conformal radiation therapy and IMRT. 58 Liu et al. carried out a comparative study between small-spot IMPT and volumetric-modulated arc therapy (VMAT), and found that small-spot IMPT significantly improved sparing of heart, liver, and lungs with clinically acceptable plan robustness. 34 Most recently, Lin et al. reported the result of a phase IIB trial that for locally advanced esophageal cancer, proton beam therapy (PBT) reduced the risk and severity of adverse events (AEs) while maintaining similar progression-free survival (PFS) when compared with IMRT. 59 Specifically for IMPT, different configurations may lead to different outcomes, such as spot sizing and spacing. 38 The selection of beam angles is important as well. At our institution, two different sets of beam angles are typically used for IMPT planning for distal esophageal cancer: the Right-Left (R-L) oblique posterior beams (Group R-L) and the Superior-Inferior (S-I) oblique posterior beams (Group S-I) (Fig. 1). To the best of our knowledge, no comparative study has been reported comparing these two different configurations of beam angles in IMPT for the treatment of distal esophageal cancer. In this study, we compared the IMPT plan quality and robustness for distal esophageal carcinoma between R-L oblique posterior beams (Group R-L) and S-I oblique posterior beams (Group S-I). The interplay effect was also quantified for both beam orientations. To assess the impact of different beam angles, patients were chosen based on the dosimetrist's choice for beam angles at the time of IMPT planning, typically two sets, thus could be divided into two groups. In the first group, two oblique posterior beams in the S-I direction with a couch angle of 270°were employed (Group S-I). In the second group, two oblique posterior beams in the R-L direction of a couch angle of 180°were employed (Group R-L). In Group R-L, a few patients were treated with one or two additional anterior beams for better target dose distribution and adjacent organ protection (See Table S1).

2.A | Patient selection
For all patients, tumor locations were distal (ie, near the gastroesophageal junction), where the impact of the respiratory motion was important and the protection of OARs could be difficult anatomically.
In our earlier proton practice, the patients were planned and treated with mainly oblique posterior beams in the R-L direction. We then started exploring using oblique posterior beams in the S-I directions, which has become more commonly used recently. Thus, 20 patients (10 in each group) were selected for this comparison. We matched the patients by comparing their CTV volumes and also tumor motion amplitudes, so that they were most presentative clinically for statistical comparison (Table 1). Tumor motion amplitude in this paper is defined as the largest motion in one direction (S-I, A-P, or R-L, see Table S2). Additionally, no patients had any implanted electronic devices, and no range shifter was used in the treatment of any patients. All selected patients were treated with curative intent as determined by the clinical radiation oncologists.

2.B | Treatment planning
Treatment planning was carried out for all patients by using the commercial treatment planning system Eclipse TM (version 13, Varian Medical System, Palo Alto, CA). All the IMPT plans were generated on the averaged 4D-CT with CTV density override (HU = 50) for improving plan robustness related to respiratory motion. An optimization target volume (OTV) was constructed by adding a uniform 5-mm margin to the CTV to assist the robust plan generation. At least one spot was placed outside OTV to ensure the forming of a uniform dose distribution within OTV. Discrete energy layers from 71.3 to 205.3 MeV were used (Table S1).
For beam placement, two posterior beams were typically used, whereas one or two additional anterior beams could be optionally introduced as needed to achieve a clinically acceptable plan (ie, meeting the institutional dose-volume constraints and plan robustness requirements). More details regarding treatment planning for distal esophageal carcinoma can be found in Liu et al. 34 For the current study, as described in Section 2.A, all patients from Group S-I had two S-I oblique posterior beams with no additional supplemental beams required. For the 10 patients from the Group R-L, besides two R-L oblique posterior beams, two patients had one extra anterior beam each and one patient had two additional anterior beams (Table S1).
Pencil-beam convolution supposition (PCS) was used for all patients to carry out the volume-dose calculation and beamline modifiers. For optimization, single-field optimization (SFO) was usually applied as the preferred method. Alternative multiple-field optimization (MFO) would be used if SFO failed to meet the clinical requirements at first try. For the SFO approach, the PCS model was utilized; for the MFO approach, the nonlinear uniform proton optimizer (NUPO) was used. When MFO approach was used, an evaluation of dose distribution per field would be done using Eclipse TM to make sure the per field gradient was shallow, so that the plan would be robust to independent beam shifts during the delivery. In all selected patients, the plan optimizations of six patients used SFO method, with three patients in each group. After the planning on the averaged 4D-CT, two verification plans on the maximum inhalation and maximum exhalation phases without the density override were generated to evaluate the influence of respiratory motion, which in turn acted as guidance for the adjustment of the original plan generated on the averaged 4D-CT. When all three plans (the original plan on the averaged 4D-CT and two verification plans on the maximum inhalation and exhalation phases) had met the clinical requirements (Table 2), they could be considered optimal.

2.C | Treatment delivery
Hitachi ProBeat-V spot-scanning proton machines (Hitachi, Tokyo, Japan) were used to deliver the treatment at our institution. The T A B L E 1 Target characteristics between two treatment/beam angle groups.  The bandwidth was then used to evaluate the plan robustness, with the horizontal bandwidth for the dose at a reference volume and the vertical bandwidth for the volume at a reference dose. Smaller bandwidth would indicate better plan robustness.

2.F | Interplay effect evaluation
Dose calculation software was developed at our institution to assess dose distribution in the presence of interplay effect. Every spot for each field per fraction was assigned to the corresponding respiratory phase according to their temporal relationship with the spot delivery sequence and patient-specific respiratory motions.
Then, the dose of each phase was calculated with respect to the assigned spots. Finally, the calculated dose of each phase was then deformed onto the reference phase (maximum exhalation phase) through deformable image registration to get the final 4D dynamic dose. 45 Our software can handle 10 respiratory phases; however, in practice, only two equally weighted extreme phases (maximum exhalation and maximum inhalation phases) would be selected due to computation time considerations. The starting phase for each field per fraction was randomized to help mitigate its influence.
The actual fraction number of the treatment plan was used in the calculation. The multiple energy extraction (MEE) method was adopted for more efficient delivery compared to the single energy extraction (SEE) method. 61,62 Iso-layer repainting was utilized to mitigate the impact of interplay effect. During the repainting process, a minimum MU limit of 0.003 MU and a patient-specific maximum MU limit of our proton machine were employed. When the respiratory motion amplitude was within 5 mm, the maximum MU limit was set to be 0.04 MU.
Otherwise, the maximum MU limit was changed to 0.01 MU for more repainting in order to mitigate interplay effect due to increased respiratory motions. Any MU values larger than the maximum limit were split into several less intensive spots with their intensities equal to the maximum MU limit and one possible residual spot. If the intensity of a spot or a residual spot was between the minimum and maximum MU limits, it was delivered with its actual intensity. For a spot or a residual spot that had intensity less than the minimum MU limit, its delivery was adjusted according to its intensity. If the intensity was larger than half of the minimum limit, it was rounded up to be the minimum MU limit, while otherwise, the spot was dropped. Note that the maximum and minimum MU limits mentioned here were well investigated and benchmarked machine-specifically only within our institution, thus might only apply well on our machine. These DVCs are carefully generated by experienced radiation oncologists and physicists at our institution. They are generally more restrictive (thus safer) than the ones recommended by RTOG.

2.H | Statistical analysis
For fair comparison, all plans were normalized to have a CTV low D95% of 100% of the prescribed dose in the nominal scenario (without uncertainties or interplay effect considered). Wilcoxon signedrank test was applied to carry out the statistical analysis between the selected paired groups. P < 0.05 was considered to be statistically significant. Box-and-Whisker plotting was adopted to illustrate the DVH indices for all patients from each group. Any value >1.5 times of the interquartile range above or below the quartile limits was considered as an outlier.

3.D | NTCP
Pericarditis, liver failure, and pneumonitis are considered as the endpoints for heart, liver, and lung, respectively, in the NTCP calculation.
Seven of 10 patients from the Group S-I had lower NTCP for heart, 9 of 10 patients from the Group S-I had lower NTCP for liver, while the NTCP for lung from almost all patients was zero (Table 3). Significant NTCP for liver from patients of Group S-I was observed (median value: 0.0000 vs 0.0027 %; P = 0.0208). NTCP for heart and lung was comparable (Table 4).

| DISCUSSION
As the first group to do so, in this paper, we evaluated the impacts of two different beam angle configurations in the IMPT treatment  Clinically, we observed no spinal cord toxicity as a result of this approach to date. 65 Additionally, even though the slight worse hot spot control in target volume from Group S-I was observed, we believe this had little influence for the plan for two reasons. First, the difference was quite small, <2%, and evidently under clinically acceptable level. Second, during respiratory motion, the most hot spot in a plan would mostly stay within the target volume, or occasionally move to one of the adjacent organs: lung, liver, and heart.
All these organs are parallel-type organs that have relatively large volume effect, especially for lung. Thus, the dose-response may be closer to mean dose rather than the hot spot. In fact, comparable or even better mean dose results in these organs were observed in Group S-I, which could be confirmed by the comparable or even better NTCP results.
As for plan robustness, using different beam angles did not appear to produce any adverse impact with the Group S-I approach.
This is understandable because identical optimization methods were This study has certain limitations. First, the patient groups chosen were heterogeneous and small in number, and also there may exist T A B L E 3 Normal tissue complication probability (NTCP) results for heart, lung, and liver.

ETHICAL CONSI DERATIONS
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 prior to the analysis and also publication of the study.

R E F E R E N C E S SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article. Table S1. Details of Beam Angle (G for gantry, T for table), IMPT machine characteristics, and beam energy information. Table S2. Details of tumor motion amplitude in three directions.