Dosimetric comparison of scanned proton therapy and volumetric modulated arc therapy in isodose-prescribed stereotactic body radiotherapy for lung cancer

ABSTRACT

This study evaluated the feasibility of pencil beam scanning (PBS) proton therapy for stereotactic body radiotherapy using an isodose prescription in patients with isolated early-stage lung cancer, comparing dose–volume metrics with those of volumetric modulated arc therapy (VMAT). Treatment plans were generated for 10 patients with tumors located at least 2 cm from the trachea. Each plan prescribed 48 Gy(RBE) in four fractions with a 60% isodose prescription. Both multi-field optimized and single-field optimized PBS plans used four fields, while VMAT plans used two coplanar arcs. Plan robustness was assessed by simulating setup errors of ±5.0 mm and density uncertainties of ±3.5%, with worst-case target coverage also analyzed. Comparisons among the three planning strategies employed Bonferroni-adjusted multiple comparisons for target coverage, robustness, and organ-at-risk doses. PBS achieved target coverage comparable to VMAT, with similar D98% (63.2 ± 1.3 GyRBE vs. 64.0 ± 1.6 Gy; P = 0.12), mean doses (73.2 ± 0.8 GyRBE vs. 73.7 ± 1.2 Gy; P = 0.21), and robustness (D98%: 50.2 ± 0.9 GyRBE vs. 50.3 ± 1.1 Gy; P = 0.77). For the normal lung, PBS significantly reduced low-dose exposure up to V_10Gy, whereas VMAT was more effective in reducing doses from V_20Gy to the prescription level. All evaluated organ-at-risks received significantly lower mean doses with PBS than with VMAT (P < 0.05). Under a 60% isodose prescription for lung stereotactic body radiotherapy, multi-field optimized-based PBS proton therapy provides target coverage and robustness comparable to VMAT while offering superior normal tissue sparing.

INTRODUCTION

Lung cancer remains the leading cause of cancer-related deaths worldwide[1]. Non-small cell lung cancer (NSCLC) accounts for ~80–85% of cases, with treatment strategies largely dependent on the disease stage. While surgery is the standard of care for early-stage NSCLC, stereotactic body radiotherapy (SBRT) or hypofractionated proton therapy is typically employed for inoperable patients [2–6].

Conventional SBRT aims to deliver a uniform dose across the target, typically within 95–107% of the prescribed dose, as recommended by ICRU Report 83 [7]. Traditionally, the SBRT dose is prescribed relative to the isocenter [8]. Recent advances have led to the clinical adoption of isodose prescriptions that escalate the central dose, as demonstrated in the RTOG 0813 trial [9, 10]. According to ICRU Report 91, prescribing the dose to a specific isodose line remains acceptable for SBRT; however, a volumetric approach using dose–volume histogram-based metrics is also recommended [11]. Photon-based SBRT is widely implemented, with volumetric modulated arc therapy (VMAT)—a form of intensity-modulated radiotherapy delivered in arcs—representing the most advanced technique currently in clinical use [12–14].

It is well established that a BED10 (biological effective dose, α/β = 10) greater than 100 Gy is necessary to achieve local control rates of 80–97% in lung tumors treated with photon-based SBRT [15–19]. Recent evidence, including large-scale modeling studies and meta-analyses, suggests that higher maximum doses—such as target volume doses exceeding 125 Gy BED10—achieved by normalizing dose to lower isodose lines (e.g. 60–70%) are associated with improved local control, particularly for T2 and squamous cell tumors [20–22]. For example, the RTOG 0236 trial used heterogeneous prescriptions, such as the 60% isodose line, and reported excellent long-term local control [23, 24]. Although BED10 initially started around 100 Gy in photon-based SBRT, dose escalation has been used to further enhance tumor control. However, dose heterogeneity must be carefully balanced against organ-at-risk (OAR) constraints, especially in centrally located tumors. Tsujino et al. identified lung dose-volume parameters such as V_5Gy and V_20Gy as predictors of adverse events in lung cancer patients [25, 26]. These lung dose indicators help minimize toxicity by allowing tumor dose escalation while protecting surrounding normal tissue through optimized isodose prescriptions [7]. In centrally-located lung tumors, Chan et al. demonstrated that, in centrally-located lung lesions, using a lower prescription isodose line (e.g. 60–70%) enables steeper dose gradients and thereby reduces dose to critical organs at risk such as esophagus, bronchus, and major vessels, while increasing tumor dose [13].

Pencil beam scanning (PBS) is an advanced proton therapy technique for early-stage NSCLC that uses intensity-modulated proton therapy (IMPT) to provide superior flexibility in dose shaping [27–29]. Multiple studies comparing dose distributions between particle therapy and VMAT have highlighted the potential advantages of particle-based approaches [30–32]. However, although some reports have compared treatment plans using conventional dose prescriptions for IMPT and VMAT [32], no prior studies have evaluated PBS-based proton SBRT using isodose prescriptions.

Current consensus recommends a minimum prescribed dose of BED10 >109 Gy for proton therapy in lung cancer [33]. Proton beams have a relative biological effectiveness (RBE) ~1.1 times that of photons, and while they are biologically equivalent, they reduce both the mean dose to normal lung tissue and the low-dose exposure compared to photon beams. Consequently, there is growing interest in increasing BED10 in proton therapy, following the dose escalation trends seen in photon-based SBRT. Moreover, as Liu et al. noted, although proton therapy is expensive, accumulating reference data is critical to identifying patients who will derive significant clinical benefit [32]. Therefore, it is important to determine whether PBS can achieve dose distributions comparable to those of conventional VMAT.

As we consider implementing proton SBRT, we aimed to quantitatively evaluate whether proton therapy can match the quality of conventional photon SBRT delivered with standard isodose prescriptions. This study was conducted to investigate that approach.

Specifically, we assessed the clinical feasibility of proton PBS SBRT using an isodose prescription by comparing its dose distribution and robustness against the widely used VMAT technique.

MATERIALS AND METHODS

Patient information

We selected 10 consecutive early stage or metastatic NSCLC patients with isolated lung lesions treated with SBRT at Southern Tohoku General Hospital (STGH). This case was outside the +2 cm range of the trachea defined by Timmerman et al., and tumor placement was such that the target volume did not overlap with the chest wall [34]. Table 1 summarizes the patient characteristics. This study was approved by the Ethics Committees of the STGH and Osaka Proton Therapy Clinic (OPTC). All patients were placed in a supine position, with both arms raised and handles held. A vacuum cushion and a body shell were used to immobilize the body.

Table 1.

Patient characteristics and beam parameters for PBS beams

Patient ID	Gender	Age	Tumor location	TNM	GTV (cm3)	PTV (cm3)	Gantry angle (range shifter) of PBS
1	Male	86	RLL	T1aN0M0	1.5	7.6	300(3), 225 (62), 180 (5), 260 (6)
2	Male	84	RUL	T1aN0M0	2.8	11.72	150(1), 225(4), 320(60), 180(2)
3	Male	83	RUL	T1aN0M0	1.8	8.7	180(5), 300(60), 240(4), 20(60)
4	Male	75	RUL	T1bN0M0	2.8	12.7	0(60), 300(60), 35(60), 270(60)
5	Male	80	RUL	T1bN0M0	2.0	9.24	270(6), 330(2), 240(0), 210(0)
6	Male	77	LLL	T1bN0M0	1.8	9.1	135(60), 210(60), 180(60), 90(60)
7	Male	70	RLL	T1bN0M0	10	29.9	150(60), 225(62), 180(60), 290(60)
8	Male	71	RUL	T2aN0M0	12.5	33.7	0(60), 300(60), 35(60), 330(60)
9	Male	68	LUL	TxNxM1	1.0	5.9	135(3), 210(0), 90(0), 180(0)
10	Male	66	RUL	TxNxM1	0.7	9.1	330(4), 270(3), 240(1), 200(0)

RUL, right upper lobe; RLL, right lower lobe; LUL, left upper lobe; LLL, left lower lobe.

Targets and organs at risk definitions

The gross tumor volume (GTV) was delineated by an experienced radiation oncologist using respiratory-gated computed tomography with real-time position management (Varian). The clinical target volume was set equal to the GTV, following the protocol of the JCOG1408 trial [35]. No internal target volume (ITV) was defined, as the study aimed to determine whether even the smallest possible target could receive the full prescribed dose. Consequently, the treatment approach assumed breath-hold without ITV consideration. The planning target volume (PTV) was generated by expanding the GTV by 5 mm. OARs included the normal lungs, ribs, chest wall, spinal cord, heart, trachea, and esophagus.

Treatment planning

The PBS beam was optimized using a RayStation 10A treatment planning system (TPS) with OPTC's MELTHEA V (Hitachi) beam data [36]. The beams consisted of four coplanar fields and were generated for both multi-field and single-field optimization (MFO and SFO) [37]. Robust optimization accounted for setup and density uncertainties of ±5 mm (in all directions: superior–inferior, left–right, and anterior–posterior) and ± 3.5%, respectively, resulting in 21 robustness scenarios [38]. Four repaints have been used for lung cancer treatment in OPTC [39]. Range shifters with thicknesses between 60 and 66 mm were automatically selected during optimization for beams where the water-equivalent depth of the target was less than 40 mm. The treatment machine was equipped with a multi-leaf collimator (MLC) featuring a leaf width of 3.75 mm [36]; a uniform leaf margin of 5 mm was applied for all beams. An air gap of 100 mm was used for the posterior beam, while a 50 mm air gap was set for the remaining beams to avoid interference with the treatment table. The prescribed dose was 48 Gy (RBE) delivered in four fractions to D_95% of the PTV, with the GTV dose maximized while conforming to the 60% isodose line surrounding the PTV [14]. D_x% (GyRBE or Gy) indicates the dose of the contours irradiated with x% volume. A constant RBE of 1.1 was assumed, and the maximum GTV dose was adjusted to exceed 80 GyRBE, consistent with a 60% isodose prescription. We did not set an upper limit for the maximum dose to the GTV and aimed to get as close to 80 Gy as possible. Therefore, we completed treatment planning once the maximum dose to the GTV exceeded 80 Gy. Furthermore, while maintaining the maximum dose to the GTV dose under the above conditions, the D_98% of the GTV was increased to be as high as possible (Table 2). The coverage within the PTV aimed to optimize the GTV coverage while also striving to match the dose distribution around the PTV as conformity as possible. The dose calculation grid size was 1.5 mm. All PBS treatment plans were optimized using a Monte Carlo (MC) dose engine with 10 000 ions per spot and a final calculation performed with 0.5% statistical uncertainty [39]. The VMAT plans were generated using TrueBeam and Eclipse 15.6 TPS (Varian Medical Systems) [40] using 6 MV flattening-filter-free beams and two coplanar partial arc irradiations. The isocenter was placed at the geometric center of the PTV. The gantry rotations were set at two reciprocal (clockwise and counterclockwise) rotations from 0° to 180° or 180° to 360°, and the collimator angles were 30° and 330°, respectively. The MLC had a leaf width of 5 mm, and the MLC opening was optimized to be as close to the PTV as possible. AcurosXB (AXB) 15.6.06 was used as the dose calculation algorithm [41, 42]. The dose grid size was 1.5 mm, which was the same as that used in the PBS plans.

Table 2.

Dose constraints

Target/Organ	Dose/Volume	Value
GTV	Dmax	≧ 80 Gy
	D98%	As high as possible
Lungs-GTV	Mean dose	≦ 18 Gy
	V15Gy	≦ 25%
	V20Gy	≦ 20%
	V40Gy	≦ 100 cm3
Rib bones/Chest wall	V30Gy	≦ 35 cm3
Spinal cord	Dmax	≦ 25 Gy
Heart	V30Gy	≦ 15 cm3
Trachea/Main bronchus	V40Gy	≦ 10 cm3
Esophagus	V40Gy	≦ 1 cm3

Evaluations of dose statistics and target robustness

Target coverage was evaluated using D_98%, D_2%, D_0%, mean dose (D_mean) as dose statistics. Three additional evaluation indices were selected for the PTV: homogeneity index (HI), conformity index (CI), and gradient index (GI) for PTV. The HI for the GTV was defined as in Equation 1 (1) with D_2%, D_98%, and D_50%. The closer it is to 0%, the higher the homogeneity [43].

(1)

The CI and GI of the PTV were defined in Equation 2 using Paddick’s Equation [44].

(2)

The treatment volume (TV) was defined as the target volume (PTV), while the prescription isodose volume (PIV) referred to the total volume receiving either 100% of the prescribed dose (48 Gy, for conformity index [CI]) or 50% of the prescribed dose (24 Gy, for gradient index [GI]). The overlap between TV and PIV was denoted as TV∩PIV. CI and GI values closer to 100% indicate better conformity.

Dose constraints for OARs were based on the JCOG1408 trial (Table 2) [35]. Additional constraints for the chest wall followed the criteria reported by Seco et al., who compared proton and photon therapy dose distributions, with a specific constraint of V_30Gy < 35 cm³ [45]. V_xGy (%) indicates the volume irradiated with more than x (GyRBE or Gy). As additional dose statistics, we evaluated the 5 Gy interval from V_5Gy to V_45Gy of normal lungs (lungs-GTV). We also evaluated D_0% and D_mean of lungs-GTV, rib, chest wall, spinal cord, heart, and trachea.

Robustness evaluations were performed using the respective TPS features in RayStation and Eclipse. One setup error (±5.0 mm) and one range error (±3.5%) were applied independently in the superior–inferior, left–right, and anterior–posterior directions, resulting in 12 perturbed scenarios for each treatment plan. Changes in GTV D98% and D0% were compared across the three plans.

Dose statistics and robustness analyses were conducted using Bonferroni-adjusted post hoc tests. A two-tailed P-value of <0.05 was considered statistically significant. All statistical analyses were performed using the Python SciPy library.

RESULTS

Dose statistics

Figure 1 shows the dose distributions for the three irradiation patterns for a representative patient (ID: 8). The left, center, and right panels of the figure represent axial, coronal, and sagittal views, respectively, while the top, middle, and bottom panels show the MFO, SFO, and VMAT plans, respectively. Magenta and white regions indicate doses exceeding the prescription dose (48 Gy). Figure 2 and Table 3 present the results of dose statistics for the main targets and OARs. The D_98%, D_0%, and D_mean for the GTV were similar between the MFO and VMAT plans, with no significant differences. The SFO did not achieve the GTV coverage of MFO or VMAT when a 60% isodose prescription was used. The GTV HI was lower in the MFO and VMAT plans than in SFO plans (P < 0.01). The PTV CI was highest for VMAT, which was significantly greater than in the two proton plans (P < 0.05).

Fig. 1.

Dose distributions of a representative patient (ID: 8) for the MFO (upper), SFO (middle), and VMAT (lower) plans. Yellow and cyan contours indicate the PTV and GTV, respectively.

Fig. 2.

Dose–volume histograms (DVHs) (first and third rows) and boxplots (second and fourth rows) comparing MFO, SFO, and VMAT plans. In the DVH graphs, solid lines represent mean values, and shaded areas indicate the standard deviation. In the boxplots, ^** and ^* indicate statistical significance at P < 0.01 and P < 0.05, respectively.

Table 3.

Dose statistics and statistical analysis for target volumes and OARs

Target/Organ	Dose/Volume	MFO Mean ± SD	SFO Mean ± SD	VMAT Mean ± SD	MFO/SFO Differences (%)	MFO/VMAT Differences (%)	SFO/VMAT Differences (%)
GTV	D98% (Gy)	63.2 ± 1.3	60.4 ± 1.4	64.1 ± 1.6	2.9 ± 1.3 **	−0.8 ± 1.4	−3.7 ± 1.7 **
	Dmean (Gy)	73.2 ± 0.8	70.7 ± 0.8	73.7 ± 1.2	2.5 ± 1.3 **	−0.5 ± 1.1	−3.0 ± 1.6 **
	Dmax (Gy)	81.3 ± 0.7	81.1 ± 0.4	81.3 ± 0.4	0.2 ± 0.8	0.0 ± 0.5	−0.2 ± 0.4
	HI (%)	23.1 ± 2.0	27.8 ± 2.2	22.2 ± 2.0	−4.7 ± 2.3 **	0.9 ± 2.4	5.6 ± 2.3 **
PTV	Dmean (Gy)	60.4 ± 1.2	59.1 ± 0.8	60.8 ± 1.0	1.3 ± 0.9 *	−0.3 ± 0.9	−1.7 ± 1.1 **
	CI (%)	86.5 ± 2.9	82.6 ± 3.7	91.2 ± 2.2	3.9 ± 1.7 *	−4.6 ± 3.4 **	−8.6 ± 4.1 **
	GI (%)	19.1 ± 2.9	18.5 ± 2.7	20.5 ± 2.0	0.6 ± 0.7	−1.4 ± 2.1	−1.9 ± 1.8
Lungs-GTV	Dmean (Gy)	1.7 ± 0.7	1.7 ± 0.8	2.2 ± 0.9	0.0 ± 0.1	−0.5 ± 0.2	−0.5 ± 0.1
	V5Gy (%)	8.6 ± 3.5	8.8 ± 3.6	10.8 ± 4.1	−0.2 ± 0.5	−2.2 ± 1.4	−2.0 ± 1.3
	V10Gy (%)	5.6 ± 2.4	5.8 ± 2.6	6.3 ± 2.8	−0.1 ± 0.5	−0.7 ± 0.8	−0.5 ± 0.7
	V15Gy (%)	3.8 ± 1.7	3.9 ± 1.9	3.9 ± 1.9	−0.1 ± 0.3	−0.1 ± 0.2	0.0 ± 0.3
	V20Gy (%)	2.6 ± 1.4	2.7 ± 1.4	2.5 ± 1.3	−0.1 ± 0.2	0.1 ± 0.1	0.2 ± 0.2
	V25Gy (%)	1.8 ± 1.0	1.9 ± 1.1	1.6 ± 0.9	−0.1 ± 0.1	0.2 ± 0.2	0.3 ± 0.2
	V30Gy (%)	1.3 ± 0.8	1.4 ± 0.8	1.1 ± 0.7	−0.1 ± 0.1	0.2 ± 0.1	0.3 ± 0.2
	V35Gy (%)	0.9 ± 0.6	1.0 ± 0.6	0.8 ± 0.5	−0.1 ± 0.1	0.1 ± 0.0	0.2 ± 0.1
	V40Gy (%)	0.6 ± 0.4	0.7 ± 0.5	0.6 ± 0.4	−0.1 ± 0.1	0.1 ± 0.0	0.1 ± 0.1
	V45Gy (%)	0.4 ± 0.3	0.5 ± 0.3	0.4 ± 0.3	−0.0 ± 0.0	0.0 ± 0.0	0.1 ± 0.0
	V40Gy (cm3)	19.6 ± 9.6	22.0 ± 11.2	17.8 ± 9.8	−2.3 ± 1.8	1.8 ± 1.0	4.2 ± 1.7
Rib	Dmean (Gy)	8.5 ± 2.6	8.2 ± 2.4	11.4 ± 3.1	0.3 ± 0.5	−2.9 ± 1.0 *	−3.2 ± 1.2 *
	Dmax (Gy)	29.7 ± 5.7	27.7 ± 7.2	29.8 ± 7.3	1.9 ± 3.1	−0.1 ± 4.5	−2.1 ± 3.1
Chest wall	Dmean (Gy)	0.29 ± 0.1	0.3 ± 0.1	0.9 ± 0.3	0.0 ± 0.0	−0.6 ± 0.2 **	−0.6 ± 0.2 **
	Dmax (Gy)	33.7 ± 7.8	34.2 ± 8.3	35.3 ± 9.2	−0.4 ± 2.0	−1.5 ± 4.8	−1.1 ± 4.0
	V30Gy (%)	0.69 ± 1.7	0.6 ± 1.8	1.2 ± 2.45	0.1 ± 0.2	−0.5 ± 1.0	−0.6 ± 1.0
Spinal cord	Dmean (Gy)	0.0 ± 0.0	0.0 ± 0.0	0.9 ± 0.4	0.0 ± 0.0	−0.9 ± 0.4 **	−0.9 ± 0.4 **
	Dmax (Gy)	0.1 ± 0.1	0.1 ± 0.1	7.5 ± 1.9	0.0 ± 0.0	−7.5 ± 1.9 **	−7.5 ± 1.9 **
Heart	Dmean (Gy)	0.0 ± 0.0	0.0 ± 0.0	0.6 ± 0.8	0.0 ± 0.0	−0.6 ± 0.8 *	−0.6 ± 0.8 *
	Dmax (Gy)	1.8 ± 2.6	1.8 ± 2.6	3.6 ± 5.6	−0.0 ± 0.2	−1.8 ± 3.2	−1.7 ± 3.3
	V30Gy (%)	0.0 ± 0.0	0.0 ± 0.0	0.0 ± 0.0	0.0 ± 0.0	0.0 ± 0.0	0.0 ± 0.0
Trachea	Dmean (Gy)	0.5 ± 0.6	0.4 ± 0.5	3.3 ± 2.7	0.1 ± 0.2	−2.7 ± 2.3 **	−2.8 ± 2.5 **
	Dmax (Gy)	8.4 ± 8.9	6.8 ± 7.6	8.9 ± 6.1	1.6 ± 2.6	−0.5 ± 5.0	−2.2 ± 4.7
Esophagus	Dmean (Gy)	0.0 ± 0.0	0.0 ± 0.0	1.0 ± 0.4	0.0 ± 0.0	−1.0 ± 0.4 **	−1.0 ± 0.4 **
	Dmax (Gy)	0.9 ± 1.7	1.0 ± 2.0	7.4 ± 1.8	−0.1 ± 0.4	−6.5 ± 2.7 **	−6.4 ± 2.9 **
	V40Gy (%)	0.0 ± 0.0	0.0 ± 0.0	0.0 ± 0.0	0.0 ± 0.0	0.0 ± 0.0	0.0 ± 0.0

^** and ^* indicate statistical significance at P < 0.01 and P < 0.05, respectively.

As shown in the lower part of Table 3, the MFO plans performed lower than the VMAT plans for lungs-GTV V_5Gy–V_15Gy and D_mean (P > 0.05). The VMAT plans performed lower than the MFO plans at doses above V_20Gy, and MFO resulted in lower lung doses than SFO (P > 0.05). VMAT also achieved the greatest reduction in lung V_40Gy (cm³), but no significant difference was observed between the two PBS methods (P > 0.05). This was likely due to the flexibility of MFO optimization. MFO does not require a uniform dose per beam, as flexible optimization is implemented to reduce lung dose. The mean doses to the ribs, chest wall, spinal cord, heart, trachea, and esophagus were all significantly lower in the two PBS plans compared to the VMAT plan (P < 0.05). At the maximum dose, PBS significantly reduced radiation doses to the spinal cord and esophagus compared to VMAT (P < 0.05).

Target robustness

Table 4 shows the results of the GTV robustness evaluation. Figure 3 presents representative perturbed dose distributions for each plan when a 5 mm setup error (left shift) and + 3.5% density error were introduced for the patient shown in Fig. 1. Figure 4 shows boxplots of coverage (left) and dose differences (right) relative to the original plans for 12 perturbed scenarios in 10 cases, for GTV D_98% (top row) and D_0% (bottom row). The average D_98% for the 12 worst-case scenarios—based on ±3.5% density variation combined with a 5 mm setup error across 10 patients—exceeded the prescribed dose in all plans except for the SFO plans (50.2 ± 0.9 GyRBE, 47.3 ± 1.5 GyRBE, and 50.3 ± 1.1 Gy for the MFO, SFO, and VMAT plans, respectively). The mean error from the nominal plan for D_98% was not significantly different among the three groups (P > 0.05). The mean error in the maximum dose compared to the nominal plan was −1.8 ± 2.9%, −5.1 ± 3.2%, and −2.8 ± 1.4% for the MFO, SFO, and VMAT plans, respectively. The significant difference in maximum dose error between MFO and SFO indicates that the maximum GTV dose was more likely to deviate from the target of 80 GyRBE in the event of an error when using SFO.

Table 4.

Robustness and statistical analysis of GTV in the worst-case scenario

Target/Organ	Dose/Volume	MFO Mean ± SD	SFO Mean ± SD	VMAT Mean ± SD	MFO/SFO differences (%)	MFO/VMAT differences (%)	SFO/VMAT differences (%)
GTV	D98% (Gy)	50.2 ± 0.9 (48.6/51.3)	47.3 ± 1.5 (44.9/49.4)	50.3 ± 1.1 (47.9/52.0)	2.9 ± 1.2 (1.3/4.9) **	−0.1 ± 1.5 (−2.8/3.3)	−3.0 ± 1.7 (−5.3/−0.0) **
	D0% (Gy)	79.8 ± 2.4 (76.3/83.7)	77.0 ± 2.5 (73.7/81.1)	77.4 ± 1.3 (75.2/79.8)	2.8 ± 1.7 (−1.2/4.9) *	2.4 ± 1.7 (0.0/5.4) *	−0.4 ± 2.5 (−4.0/2.9)
	D98% difference of nominal plans (%)	−20.6 ± 2.4 (−24.3/−17.4)	−21.5 ± 3.5 (−28.4/−18.0)	−21.1 ± 3.1 (−25.1/−16.6)	0.9 ± 1.9 (−2.1/4.0)	0.5 ± 1.8 (−1.6/5.0)	−0.4 ± 3.3 (−4.6/7.1)
	D0% difference of nominal plans (%)	−1.8 ± 2.9 (−5.2/3.9)	−5.1 ± 3.2 (−9.2/0.3)	−2.8 ± 1.4 (−5.6/−0.7)	3.2 ± 1.9 (−0.8/6.3)*	0.9 ± 1.9 (−1.6/5.4)	−2.3 ± 2.5 (−6.1/1.7)

^** and ^* indicate statistical significance at P < 0.01 and P < 0.05, respectively.

Fig. 3.

Dose distributions in three planes for the patient in Fig. 1 when a setup error of 5 mm (left direction) and a density error of +3.5% were introduced for the three plans. dose–volume histograms show the perturbed dose–volume ranges (color-shaded area) and average perturbed dose–volume values (colored lines) for 12 error scenarios, along with the nominal plan (black line).

Fig. 4.

Boxplots showing coverage (left columns) and dose differences (right columns) relative to the original plans for 12 perturbed dose scenarios in 10 cases, for GTV D98% (upper row) and D0% (lower row). “WCS” and “All” summarize the worst-case scenario (WCS) plans for each case and all perturbed dose scenarios across patients, respectively.

DISCUSSION

In this study, we investigated whether the proton PBS SBRT method applying a 60% isodose prescription can achieve the same quality as the photon-based method for early-stage lung cancer. To the best of our knowledge, there is no research dealing with proton PBS SBRT using the 60% isodose prescription method, and these findings provide valuable data for the future expansion of proton therapy indications.

There are two clinical benefits of using isodose prescriptions in proton beam therapy. First, this approach can reduce normal lung doses, particularly in low-dose regions (V_5Gy–V_15Gy) and the average lung dose (Table 3). Proton SBRT also offers the advantage of reducing the risk of adverse events by lowering the average doses to most OARs compared to VMAT. Furthermore, isodose-prescribed proton SBRT plans, which increase the central dose, can lower lung doses and other OAR near maximum doses compared with non-isodose-prescribed plans while achieving target coverage equivalent to photon-based SBRT [7, 13, 46]. From a tumor control perspective, this method enables a dose comparable to photon-based SBRT to be concentrated and escalated near the target.

Both SFO and MFO plans were calculated using RayStation’s MC algorithm, whereas VMAT was calculated using Eclipse’s AXB algorithm. Because VMAT does not use MC calculations, differences in the underlying algorithms are expected. However, because AXB is a clinically validated algorithm that closely approximates MC results, these differences are likely to be minor [41, 42].

No statistically significant differences in GTV coverage occurred between MFO and VMAT. However, VMAT resulted in a slightly better average D_98%. This indicates that VMAT can generate plans with superior coverage compared to MFO. Furthermore, the significantly better PTV CI and GI values confirm that the dose distribution was concentrated within the PTV. MFO did not extend to this, and the reason is the expansion of the dose distribution outside the PTV due to robust optimization [47]. Furthermore, SFO could not achieve the same level of coverage as VMAT. Thus, the limited degrees of freedom of the per-beam dose distribution could be a limitation in improving the GTV coverage. Since optimization using SFO is preferable in terms of setup errors and interplay effects due to respiratory motion, it is necessary to evaluate whether the above errors are acceptable when using MFO [48]. The primary objective of this study was to verify whether PBS could achieve isodose prescriptions similar to VMAT. Therefore, a more detailed examination of coverage aspects beyond those mentioned above, including dose uniformity within the PTV, is positioned as a topic for future investigation in our subsequent report.

The current plans incorporated repainting to mitigate the interplay effect. This study did not include an ITV margin and was based on a breath-hold technique. In clinical practice, respiratory gating may be used in combination with breath-hold. However, even a narrow gating window can introduce interplay effects during actual irradiation [39], and these effects may be further amplified with isodose prescriptions. In particular, irregular breathing patterns, variable amplitudes, or baseline shifts must be considered [49–51]. Dose evaluation using 4DDD with respiratory gating is feasible [52, 53]. This simulation assumes ideal breath-hold irradiation conditions, but future research will assess the interplay effect for PBS SBRT beams designed to achieve quality comparable to VMAT.

The uniformity to the target also depends on the beam size and energy that make up the PBS beam [54]. However, these parameters vary widely from machine to machine for PBS beams. Using a beam size smaller than that of the current machine used in this study may make it possible to achieve coverage and CI beyond VMAT.

Owen et al. reported that the dose required at the isocenter to achieve 90% TCP was 176 Gy at BED10 for primary lung tumors [55], and that four fractions of 48 Gy would require at least 66.4 Gy (72.3% isodose prescription). Thus, the 60% isodose prescription used in this study is a reasonable dose as the maximum dose required for tumor control.

The patient cases were based on those described by Timmerman, involving tumors ≥2 cm from the trachea [34]. While the clinical applicability is limited, our findings indicate non-equivalence in target coverage and OAR dose between photon and proton SBRT using isodose prescriptions. As such, clinical implementation should begin gradually, starting with cases safely distant from critical OARs. This study aimed to demonstrate the non-inferiority of proton SBRT compared to photon SBRT; although differences in OAR dose were minimal, results are presented as-is to reflect the observed outcomes. Further, when considering treatment for cases with a history of irradiation, the results presented here may also be of great value. Future research will compare dose distributions for centrally located tumors within 2 cm of the trachea, where OAR dose constraints may become more critical.

A barrier to the introduction of SBRT with isodose formulation in proton beams is the reduced robustness of the proton beam due to range uncertainties. In the present study, we found that D98% ensured equivalent robustness in the worst-case GTV coverage. Previous studies have compared the robustness of proton and photon beam therapies in advanced lung cancer, head and neck, and whole pelvic regions [56–58]. The head-and-neck comparison evaluated the robustness of IMPT and intensity-modulated radiotherapy (IMRT) with robust optimization and showed comparable coverage reduction [59]. Therefore, similar to the results of this study, the PBS method with robust optimization is an important requirement for implementing SBRT.

In general, MFO plans are less robust than SFO plans, with significant dose degradation due to errors [60]. However, in this study, MFO resulted in fewer errors than SFO. Since MFO beams do not require a uniform dose to enter the target with each beam, the dose distribution tends to extend in the proximal and distal directions due to strict constraints aimed at improving uniformity around the target. This trend can be confirmed by the MFO dose distribution in Fig. 1, where a higher dose is delivered to the anterior chest wall compared with SFO. On the other hand, SFO beams optimize the dose distribution so that PTV coverage is as uniform as possible for each beam, reducing the spread of the dose distribution in the proximal and distal directions. When density errors occur, this may help reduce their effects. However, MFO is assumed to be more vulnerable to positional errors in different directions for each beam due to the non-uniform dose distribution of each beam. Since independent positional errors between beams were not considered in this study, further detailed investigation is required.

These trends may differ when the number of beams is small, as one factor contributing to robustness here was the use of four beams from widely spaced angles. In addition, as mentioned above, MFO beams increase the proximal-side dose, which may become unacceptable when beam numbers are reduced. Therefore, treatment planning should also consider proximal dose distribution when beam number or angles are limited.

No significant differences in lung volume doses up to 20 Gy were observed. However, at V_20Gy, VMAT reduced the dose more effectively than the two PBS plans. This finding was similar to the trend observed in the study by Yoshida et al. [30] comparing VMAT and carbon-ion therapy. For low and mean doses, proton beam planning with range reduces the dose, while above medium doses, VMAT can reduce the dose due to the advantages of the sharp lateral penumbra of the photon beam and the dose concentration from rotational irradiation [61]. This study was limited to particularly small target sizes, and it has been reported that PBS produces different lung dose results compared with VMAT as the target size increases [31]. The lateral advantage of VMAT may lead to different outcomes from those in this study for targets close to critical organs such as the chest wall. A comparison of dose distributions between PBS and VMAT for tumors near OARs is therefore a subject for future work.

Several limitations should be noted. First, this study was limited to isolated pulmonary lesions distant from the chest wall; results may differ for cases involving contact with the chest wall. Second, not all proton beam machines will produce similar results because of variations in machine characteristics. Third, the percentage of isodose prescription used was only 60%; several different percentages have been studied for SBRT using photons, each with different treatment planning characteristics. We plan a comparative study with various isodose prescription levels. Fourth, although this study showed that PBS with four fields is equivalent to VMAT, the effect of a smaller number of fields was not assessed. Four PBS beams per day impose a significant treatment burden, and reducing the number of fields should be investigated to determine the optimal number while maintaining equivalence with VMAT. Finally, this study addressed only the feasibility of SBRT, using a static plan that did not take ITV into account. The characteristics of the interplay effect in non-uniform target dose distributions have not yet been reported and should be investigated in the future.

In conclusion, PBS proton therapy with MFO planning, assuming SBRT with a 60% isodose prescription for isolated lung lesions, achieved GTV coverage and robustness comparable to coplanar VMAT plans while reducing mean doses to surrounding OARs.

CONFLICT OF INTEREST

The authors declare that they have no conflict of interests.

FUNDING

This study was supported by JSPS KAKENHI grant number 25K11043.

Presentation at a conference

We presented at the 37th and 38th JASTRO annual meeting.

Clinical Trial Registration number

The selected patient information was approved by the Institutional Review Boards of the Osaka Proton Therapy Clinic (No. 20241205-1) and Southern Tohoku General Hospital (No. 718).