Physics in Medicine & Biology

PAPER • OPEN ACCESS You may also like

- Thermally-driven large current-
Combined proton–photon treatment for breast perpendicular-to-plane magnetoresistance
in ultrathin flakes of vanadium diselenide
cancer Seyedamin Firouzeh, Md Wazedur
Rahman and Sandipan Pramanik

- Increased local tumor control through

To cite this article: Louise Marc et al 2021 Phys. Med. Biol. 66 235002 nanoparticle-mediated, radiation-triggered
release of nitrite, an important precursor
for reactive nitrogen species
Anna S Kim, Stavros Melemenidis, Anna-
Karin Gustavsson et al.
View the article online for updates and enhancements. - Air quality and health benefits of China’s
emission control policies on coal-fired
power plants during 2005–2020
Ruili Wu, Fei Liu, Dan Tong et al.

This content was downloaded from IP address 103.35.210.40 on 09/10/2023 at 07:52

 Phys. Med. Biol. 66 (2021) 235002 https://doi.org/10.1088/1361-6560/ac36a3

PAPER

Combined proton–photon treatment for breast cancer

OPEN ACCESS
Louise Marc1, Silvia Fabiano1, Niklas Wahl2 , Claudia Linsenmeier1, Antony J Lomax3,4 and Jan Unkelbach1
RECEIVED 1
19 August 2021
Department of Radiation Oncology, University Hospital Zurich, Switzerland
2
Department of Medical Physics in Radiation Oncology, German Cancer Research Center DKFZ, Heidelberg, Germany
REVISED 3
Center for Proton Therapy, Paul Scherrer Institute, Switzerland
29 October 2021 4
Department of Physics, ETH Zurich, Switzerland
ACCEPTED FOR PUBLICATION
4 November 2021 E-mail: [email protected]
PUBLISHED
Keywords: breast cancer, fixed proton beamline, combined proton–photon therapy
22 November 2021

Original content from this

work may be used under Abstract
the terms of the Creative
Commons Attribution 4.0
Objective. Proton therapy remains a limited resource due to gantry size and its cost. Recently, a new
licence. design without a gantry has been suggested. It may enable combined proton–photon therapy (CPPT)
Any further distribution of in conventional bunkers and allow the widespread use of protons. In this work, we explore this
this work must maintain
attribution to the concept for breast cancer. Methods. The treatment room consists of a LINAC for intensity modulated
author(s) and the title of
the work, journal citation radiation therapy (IMRT), a fixed proton beamline (FBL) with beam scanning and a motorized couch
and DOI.
for treatments in lying positions with accurate patient setup. Thereby, proton and photon beams are
delivered in the same fraction. Treatment planning is performed by simultaneously optimizing IMRT
and IMPT plans based on the cumulative dose. The concept is investigated for three breast cancers
where the goal is to minimize mean dose to the heart and lung while delivering 40.05 Gy in 15 fractions
to the PTV with a SIB of 48 Gy to the tumor bed. The probabilistic approach is applied to mitigate the
sensitivity to range uncertainties. Results. CPPT is particularly advantageous for irradiating concave
target volumes that wrap around a curved chest wall. There, protons may deliver dose to the peripheral
and medial parts of the target volume including lymph nodes. Thereby, the mean dose in normal
tissues is reduced compared to single-modality IMRT. However, tangential photon beams may treat
parts of the target volume near the interface to the lung. To ensure target coverage for range
undershoot in an IMPT plan, proton beams have to deliberately overshoot into the lung tissue—a
problem that can be mitigated via the photon component which ensures plan conformity and
robustness. Conclusion. CPPT using an FBL may represent a realistic approach to make protons
available to more patients. In addition, CPPT may generally improve treatment quality compared to
both single-modality proton and photon treatments.

Introduction

Proton beams are widely considered a superior modality of radiation therapy due to their favorable depth-dose,
which allows for a substantial reduction in dose to normal tissues compared to photons (Khan and
Gibbons 2016). However, proton therapy has two main limitations: limited availability and dose uncertainties. A
widespread use of protons is limited due to size and cost of treatment facilities, the latter being by approximately
a factor 2–3 higher than for radiotherapy with photons (Schippers et al 2018). Worldwide, around 100 proton
therapy centers with up to five treatment rooms each are in clinical operation whereas more than 12 000 linear
accelerators are used for radiotherapy with photons (DIRAC—Directory of Radiotherapy Centres 2021,
Facilities in Operation—PTCOG 2021). Consequently, a small percentage of patients is treated with protons
(Mohan and Grosshans 2017). Moreover, protons are particularly sensitive to variations in setup and anatomy as
well as range uncertainties (Lomax 2020).
The gantry is a major cost factor of the facility and is thus a hurdle for the widespread use of protons
(Schippers et al 2018). Due to its size, the installation of proton therapy in existing buildings is impractical.
Current technological developments strive to decrease the cost and size of the gantry (Schippers et al 2018,

© 2021 The Author(s). Published on behalf of Institute of Physics and Engineering in Medicine by IOP Publishing Ltd
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

Bottura et al 2020), but it still exceeds by far the current costs of conventional radiotherapy (Schippers et al 2018).
Fixed proton beamlines (FBLs), which were used in the early days of proton therapy, are now reconsidered as one
approach to reduce cost. In addition, FBLs allow implementation of protons in existing conventional treatment
rooms designed for photon therapy. However, limitations in the available beam angles with FBLs may lead to
suboptimal treatments. A suggested approach to overcome this is to position the patient on a chair that can be
rotated, translated and tilted in front of the nozzle (Devicienti et al 2010). However, accuracy and repeatability of
the patient position, as well as difficulties in image registration with pretreatment MR or PET images, remain a
concern (Devicienti et al 2010, Bortfeld and Loeffler 2017). Recently, an alternative approach to compensate for the
limited range of beam angles of an FBL was suggested and demonstrated for head and neck cancers (Fabiano et al
2020): combined proton–photon treatments (CPPT). The approach considers a treatment room composed of an
FBL with beam scanning, a standard linear accelerator (LINAC) for intensity modulated radiation therapy (IMRT)
and a robotic treatment couch. Thereby, photons and protons are delivered in the same fraction and in lying
position with accurate patient setup. The photon component may compensate for the limitations in proton beam
angles and potentially improve plan robustness compared to single-modality proton plans.
Breast cancer is a treatment site which may particularly benefit from CPPT due to both limited availability of
protons and relevant dose uncertainties. Breast cancer represents one of the three most common cancer types
worldwide (Harbeck and Gnant 2017, Sung et al 2021). Standard of care treatment includes postoperative whole
breast irradiation (Sedlmayer et al 2013). Reducing dose to the lungs and the heart can be achieved in different
treatment approaches (Xiong et al 2004) which may decrease the risk of late toxicities such as major coronary
events and radiation-induced secondary cancer (Li et al 2000). Thus, proton therapy for breast cancer (Ares et al
2010, Kammerer et al 2018) is investigated in several clinical trials (The DBCG Proton Trial: Photon Versus
Proton Radiation Therapy for Early Breast Cancer,” NCT04291378. https://clinicaltrials.gov/ct2/show/
NCT04291378?term=proton+therapy&cond=breast+cancer&draw=2&rank=6, Assessing the Rate of
Complications in X-Ray Therapy Versus Proton Beam Radiation Therapy After Breast Conserving Surgery or
Mastectomy in Treating Patients With Breast Cancer,” Identifier NCT04443413. https://clinicaltrials.gov/ct2/
show/NCT04443413). CPPT with an FBL may make proton therapy available to many more breast cancer
patients. In addition, CPPT may address the problem of uncertainty of proton dose distributions, especially at
the interface of the breast and low density lung tissue. The use of protons may allow for a relevant reduction of
normal tissue dose, while the use of photons, which are less sensitive to variability in breast-shape (Alderliesten
et al 2018), may increase the robustness of a treatment plan against range uncertainties.
In this work, we investigate CPPT with an FBL for breast cancer. The potential benefits of CPPT over the
single-modality treatments IMRT and intensity modulated proton therapy (IMPT) are analyzed for three
different breast cancer cases.

Materials and methods

Robust multi-modality treatment plan optimization

Treatment planning for CPPT is performed by simultaneously optimizing IMRT and IMPT plans based on their
cumulative dose. In order to achieve robustness against various uncertainties, a hybrid robust planning method
is employed (Tommasino et al 2020): stochastic optimization is used to mitigate range uncertainty while a 5 mm
planning target volume (PTV) margin around the clinical target volume (CTV) is applied to account for setup
uncertainty and anatomical variation. The robust treatment plan optimization problem can formally be written
as
minimize
g p
x, x
å pk f ( gd + p d k) , (1)
k

subject to gd i = å gDij gx j " i, (2)

j

p k
di = å Dilk pxl " i, k , (3)
l
g
x j , px l  0 " j , l , (4)

where the objective function f , which defines the clinical goals, is evaluated for the cumulative dose of protons
and photons for all error scenarios k and stochastic optimization minimizes the sum of objective function values
weighted by an importance factor pk . In this work we use 3 error scenarios that represent the nominal scenario
with weight pk =0.5 and two error scenarios for under- and overshoot with weight pk =0.25 (Unkelbach et al
2018). Moreover, γd and p d k represent the doses delivered by photons and protons in an error scenario k. The
elements of the dose-influence matrices for photons gDij and protons p Dilk , calculated with the open-source
radiation treatment planning toolkit matRad (Wieser et al 2017), describe the dose contribution of a beamlet j

2
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

and pencil beam l to a voxel i for unit intensities in an error scenario k. Furthermore, g X j and pX l designate the
intensities of a beamlet j and a pencil beam l. In the proton dose p d k a constant relative biological effectiveness
factor of 1.1 is included, which reflects the current clinical practice.

Uncertainty model
The modelling of range uncertainty for protons consists of two parts. Firstly, range errors due to anatomical
changes such as breast swelling (Alderliesten et al 2018) are modeled through expansion or contraction of the
patient contour with overwrite of the relative electron density (RED) for water or air, respectively. Secondly,
uncertainty in the CT imaging/calibration process is modeled by up- and downscaling of RED (Khan and
Gibbons 2016). These two components are combined to obtain two dose-influence matrices p D k for the two
most extreme errors: the undershoot scenario is modeled as expansion of the patient combined with upscaling
of RED, the overshoot scenarios by contraction and downscaling, respectively. In the following, the error
scenarios refer to an expansion/contraction of the patient’s body by ±3 mm combined with ±3% up/
downscaling of RED (denoted as ±3 mm/3%), unless stated otherwise. A 5 mm PTV margin is used to account
for setup uncertainty during planning. For evaluating the robustness of treatment plan against setup
uncertainty, setup error scenarios are modeled as a shift of the isocenter by ±5 mm in all three dimensions.

Clinical cases
Three patients were exported from the clinical planning system with approval from the Kantonale
Ethikkommission Zürich under BASEC-Nr 2017-01558 to investigate the benefit of CPPT for breast cancer.
These patients were selected to represent a spectrum of breast geometries. For all cases, the planning CT was
acquired in deep inspiration breath-hold (DIBH).

1. Patient 1 is a left-sided breast cancer with lymph node involvement (figure 1(a)). The PTV has a concave
shape, wrapping around the curved chest wall.
2. Patient 2 is also a left-sided breast cancer but without lymph node involvement (figure 4(a)). The patient has
a large breast which is lying on top of a relatively flat chest wall, leading to a mostly convex shaped PTV.
3. Patient 3 is a right-sided breast cancer with lymph node involvement with a concave shaped PTV
(figure 5(a)). The PTV borders the lungs over approximately their whole length in the sagittal plane.

Treatment planning
The treatment plans for all patients are optimized based on the same clinical goals:

• Target: A dose of 40.05 Gy in 15 fractions is prescribed to the PTV with a SIB of 48 Gy to the boost volume,
which is implemented via piece-wise quadratic penalty functions.
• Lung: minimization of mean dose.
• Heart: minimization of mean dose.
• Remaining healthy tissue (HT) including the contralateral breast: minimization of mean dose.
• Conformity: A dose falloff to half the prescription dose at 1 cm distance from the target volumes is aimed for.
Beyond 1 cm distance, dose exceeding 20.025 Gy is penalized.

The same weighting of the objective functions are used for all plans. The complete objective function and
further details of treatment plan optimization can be found in appendix A.
For left-sided breast cancer patients, the beam setting for the IMRT plan consists of 13 coplanar beams
evenly distributed between the angles 300° and 168° with an angular spacing of 19°. This approximates the best
possible partial arc plan which can be delivered with volumetric modulated arc therapy (VMAT) (Unkelbach
et al 2015). For right-sided breast cancer patients the beam arrangement is mirrored. The IMPT plan is
composed of 3–4 proton beams.

Proton beam angles

Regarding the available proton beam angles, 3 situations are investigated:

1. Horizontal FBL: This represents the proton system that may be the easiest to implement and which is used
in several centers (Facilities in Operation - PTCOG 2021). In this setup, the possible proton beam angles
that can be achieved using couch rotation lie in the coronal plane. For left-sided breast cancer patients,

3
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

Figure 1. Plan comparison for patient 1: Patient geometry (a); cumulative CPPT (b), CPPT proton (c) and CPPT photon (d) dose
distributions; IMRT (e); IMPT (f). Results are shown for an inclined FBL. Contoured structures show the PTV (blue), the boost
volume (red), the lymph nodes (yellow), the lungs (magenta), the heart (orange) and the patient’s body (green).

couch angles of 0°, 15° and 30° are used in this work. In the case of right-sided breast cancer patients, the
beam directions are mirrored at the midsagittal plane.
2. Inclined 45° FBL: To investigate if an inclined proton beamline improves plans, we investigate a 45° FBL.
Generally, inclined proton beamlines are used in several centers (Facilities in Operation—PTCOG 2021). In
that case, the possible proton beam angles fall on a 45° cone around the anterior-posterior axis. In this work,
4 proton beams with 330°, 350°, 10° and 30° couch kicks are used for left-sided breast patients. The same
beam arrangement, mirrored at the midsagittal plane, is used for right-sided breast cancer patients.
3. Proton gantry: Finally, a hypothetical concept, which assumes access to a gantry for both IMPT and IMRT,
is examined. This may provide a deeper insight into the role of photons for breast cancer if protons would be
delivered under best conditions, namely with a gantry. First, commonly used beam directions in IMPT (0°,
45°, and 90° for left-sided and mirrored for right-sided breast cancer) are investigated, then the setting of 13
proton beams (incident from the same directions as the IMRT beams) is studied.

Evaluation of optimized proton–photon combinations

Plans from CPPT are compared to single-modality IMPT and IMRT plans to quantify the benefit of the novel
treatment approach. Dose distributions are compared using the open-source radiotherapy planning research

4
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

Table 1. Mean dose in Gy for the nominal scenario in the main OARs and remaining HT for: IMRT; CPPT with a horizontal FBL (CPPT-
FBL-90°), an inclined FBL (CPPT-FBL-45°), or a gantry for protons with 3 beams; IMPT with an FBL (IMPT-FBL-90° or IMPT-FBL-45°) or
a gantry with 3 proton beams.

Mean remaining HT Mean contralateral

Mean lung dose in Gy Mean heart dose in Gy dose in Gy breast dose in Gy
Treatment modality
Patient 1 2 3 1 2 3 1 2 3 1 2 3

IMRT 4.09 1.99 6.09 0.79 0.26 1.22 3.01 1.91 3.54 3.70 3.08 6.46
CPPT-FBL-90° 2.66 1.02 4.30 0.30 0.16 0.45 2.20 1.41 2.57 1.74 2.24 3.98
IMPT-FBL-90° 2.30 1.07 3.75 0.13 0.11 0.04 1.10 0.92 1.24 0.24 0.12 0.14
CPPT-FBL-45° 2.56 1.11 4.21 0.23 0.19 0.36 2.02 1.41 2.37 2.06 2.29 4.29
IMPT-FBL-45° 3.21 1.71 4.92 0.19 0.19 0.21 0.93 1.00 1.11 0.00 0.00 0.00
CPPT-gantry 2.27 0.88 3.29 0.17 0.16 0.22 1.80 1.25 1.85 1.28 1.77 1.86
IMPT-gantry 2.69 1.44 4.27 0.15 0.18 0.13 1.04 0.89 1.12 0.19 0.05 0.09

platform CERR (Deasy et al 2003). In addition, plans are quantitatively compared in terms of mean doses in the
main OARs (heart and lung) and the remaining HT. Mean dose reductions (MDR ) achieved by CPPT are
reported as relative reductions compared to IMRT as the reference treatment modality (equation (5))
VOI
DIMRT - DCPPT
VOI
MDR = VOI
, (5)
DIMRT
where DXVOI denotes the mean dose in a volume of interest (VOI) for a specific treatment modality X. The
contralateral breast has not been considered as a main OAR. However, it has been spared by being included in
the objective functions for the remaining HT (appendix A) and the mean dose of the contralateral breast is
reported. Target coverage is compared in terms of the root mean squared error for underdosage (rmsu ) of the
PTV (equation (6)), which is the measure of PTV coverage that is included in an objective function (see
appendix A)
1
rmsu = å (40.05 - d i)+2 ,
NPTV i PTV
(6)

where NPTV is the number of voxels within the PTV and d i denotes the dose in the PTV voxel with index i. In
addition, normal tissue complication probability (NTCP) models quantifying and predicting relevant toxicities
in breast irradiation are evaluated, namely the risk of major coronary events (Darby et al 2013) and, to a lesser
degree, the risk of clinically significant radiation pneumonitis (Marks et al 2010). Since the NTCP models were
derived for standard fractionation, while we use a hypofractionated scheme, the mean doses in the heart and in
the lung are scaled up by a factor of 1.25 ≈ 50/40.05. This factor corresponds to the ratio of prescription doses
in a standard fractionated versus hypofractionated scheme. According to the NCTP model in (Darby et al 2013),
RCE increases linearly by 7.4% per gray of mean dose to the heart (MHD):
NTCP heart = RCE 0 + 7.4% / Gy · MHD · 1.25 , (7)
where RCE 0 is the patient’s baseline RCE independent of irradiation. To make the results independent of the
RCE 0, we report the reduction DNTCPheart defined as the difference between the NTCPheart value according to
equation (7) obtained for the IMRT plan and for the CPPT plan. The risk of radiation pneumonitis is described
by the Lyman–Kutcher–Burman model. Given that for lungs as a parallel organ the generalized-Equivalent
Uniform Dose (gEUD) corresponds to the mean lung dose (MLD), this model is also referred to as MLD model
(equation (8)):
1.25 · MLD - TD50 ⎞
NTCP lung = F ⎛ ⎜ , ⎟ (8)
⎝ m · TD50 ⎠
where F is the cumulative distribution function of the standard normal distribution and the parameters are the
steepness parameter m=0.45 and the position parameter TD50 =31.4 Gy.

Results

Comparison of CPPT with an FBL to single-modality treatments

Figure 1 illustrates CPPT using an inclined FBL for patient 1 in whom the target volume has a concave shape. In
the optimal combination, photons deliver most of the dose to the middle part of the PTV using beams that are
tangential to the interface of lung and breast tissue (figure 1(d)). Thereby, the volume of normal tissue outside of
the target that is traversed by photon beams is small. Protons contribute dose mainly to the most peripheral parts

5
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

Figure 2. (a)–(g): robustness evaluation for patient 1 assuming an inclined FBL comparing IMRT, CPPT, IMPT, and IMPT with
maintained target coverage (IMPT-mtc). (a)–(d): mean doses for the nominal, overshoot and undershoot scenarios in the lung (a),
heart (b), the remaining HT (c) and rmsu of the PTV (d). (e)–(g): DVHs for PTV (black), lung (blue) and heart (red) in the nominal (e),
overshoot (f) and undershoot scenario (g). (h): DVH of proton and photon contributions to the PTV of CPPT with inclined FBL for
patient 1 in the nominal case with range uncertainties of ±0 mm/3% (black), ±3 mm/3% (magenta) and ±5 mm/3% (blue).

of the PTV including the lymph node targets (figure 1(c)). Irradiating these parts with photons alone would
substantially increase the volume of normal tissue traversed by photon beams (figure 1(e)). Therefore, CPPT
decreases the dose exposure in main OARs and remaining normal tissues compared to IMRT without

6
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

Figure 3. Proton contribution of CPPT for patient 1 ((a)/(b)) and patient 3 ((c)/(d)) when using an inclined ((a)/(c)) and a horizontal
((b)/(d)) FBL. The contoured structures show the PTV (blue), the boost volume (red), the lymph nodes (yellow), the lungs (magenta)
and the patient’s body (green).

compromising target coverage. In fact, MDR of 37%, 71%, 33% and 44% are achieved in the lung, heart,
remaining HT and contralateral breast, respectively (table 1). This decrease in radiation exposure of the heart
from 0.79 Gy to 0.23 Gy mean dose corresponds to a decreased risk of major coronary events (DNTCPheart
=−5.2%). The risk of radiation pneumonitis is low for IMRT (3.1%) and is further reduced to 2.3% for
CPPT (DNTCPlung =−0.8%).
The use of photons in CPPT allows for a higher robustness against range uncertainties compared to single
modality IMPT. In order to make IMPT plans robust against range undershoot, proton beams have to
deliberately overshoot into the lung tissue and heart (figure 1(f)). This enlarges the volume of lung that is
exposed to high doses (figures 1(f), 2(e), (f)) and increases the mean doses in lung (figure 2(a)) and heart
(figure 2(b)). Figure 2 compares the robustness of CPPT to two single modality IMPT plans. When the IMPT
plan is optimized for the same weighting of objectives as the CPPT plan, IMPT underdoses the PTV in the range
undershoot scenario while CPPT maintains target coverage (figures 2(d) and (g)). Target coverage can be
restored for the IMPT plan by constraining the value of the objective for PTV coverage to be smaller or equal to
its value for the CPPT plan (IMPT-mtc plans in figure 2, see details in appendix A). However, this leads to further
deterioration of other clinical objectives in IMPT such as an increase in mean lung dose in the nominal and
overshoot scenarios (figures 2(a), (e) and (f)).

Dependence on the magnitude of range uncertainty

The photon and proton dose contributions in CPPT change when considering a different magnitude for range
uncertainties during robust treatment plan optimization. When assuming smaller range errors than ±3 mm/
3%, the proton contribution increases while accounting for larger range errors increases the photon
contribution to ensure robustness (figure 2(h)). For example, for patient 1, the photon contribution to the PTV
increases to 38%, 76% and 84% for range uncertainties of ±0 mm/3%, ±3 mm/3% and ±5 mm/3%,
respectively. Dose distributions for range uncertainties of ±0 mm/3% are provided in appendix A.

Horizontal versus inclined beamline

The benefit of CPPT over the single-modality treatments is similar for a horizontal or inclined FBL (table 1).
However, for some target geometries, such as for patient 1 with lymph node involvement, the delivery of protons
in CPPT with an inclined FBL rather than horizontal is advantageous in terms of HT sparing. In fact, the inclined
FBL allows to reduce the mean doses in the main OARs and remaining HT compared to the horizontal FBL
(table 1). The inclination angle of the FBL enables protons to irradiate the whole target volume, without
penetrating surrounding normal tissues (figure 3(a)), whereas in the case of a horizontal FBL, protons have to
pass through the ipsilateral lung to reach parts of the PTV (figure 3(b)).

7
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

Figure 4. CPPT plans for an inclined FBL for patient 2: patient geometry (a); cumulative dose (b), proton contribution (c) and photon
contribution (d). The contoured structures show the PTV (blue), the boost volume (red), the lungs (magenta), the heart (orange) and
the patient’s body (green).

Comparison to delivery with a proton gantry

Using a gantry instead of an FBL for CPPT leads to a hypothetical concept, where a gantry is accessible for both
IMPT and IMRT in the same fraction with few limitations in the available proton beam directions. When proton
gantry angles commonly used in practice are selected, the combined treatment uses both modalities even in this
situation. For patient 1 and ±3 mm/3% range uncertainty, still 68% of the dose is delivered with photons. The
proton contribution in this CPPT concept leads to a better sparing of the surrounding tissues compared to
IMRT. However, similarly to the concept of CPPT with FBL, photons ensure higher robustness against range
uncertainties, which leads to treatment plans that maintain target coverage for undershoot and limit the dose
spillage into surrounding tissues such as lung for overshoot of protons. Thus, CPPT may outperform single-
modality treatments even if a proton gantry is available. Dose distributions of the proton gantry based CPPT
plan are provided in appendix B.

CPPT for other patient geometries

For patients 2 (figure 4) and 3 (figure 5), CPPT also reveals a benefit over the single-modality treatments
independent of the angle of the FBL. Compared to IMRT, CPPT decreases the mean doses in the main OARs and
remaining tissues without compromising target coverage (see table 1). Single-modality dose distributions and
DVH comparisons are shown in appendix C.
Similar target geometries, as for patients 1 and 3, lead to similar results regarding the dose contributions
delivered by photons and protons in the CPPT plan (figures 1 and 5). For patient 3, CPPT shows MDR over
IMRT of 29%/31% in the lung, 63%/70% in the heart, 27%/33 in the remaining HT and 38%/34% in the
contralateral breast for a horizontal/inclined FBL (table 1). The higher OAR mean doses, even to the heart,
compared to patient 1 can be explained by a larger PTV that extends further in cranial-caudal direction. The
reductions in dose to the main OARs result in a reduction in the risk of radiation pneumonitis (from 4.6% to
3.3%/3.2%) for a horizontal/inclined FBL, whereas larger DNTCPheart reductions of 7.1%/8.0% are observed
for major coronary events. Also, for CPPT a benefit of using an inclined FBL over a horizontal FBL is observed
for the main OARs and the remaining HT (figures 3(c) and (d)).
For patient 2, in whom no lymph nodes are included in the target volume, the target geometry is different
(figure 4(a)). Due to the almost convex PTV shape, tangential photon beams can very effectively irradiate the
target volume while sparing the HT (figures 4(b)–(d)). In this example, with an inclined FBL in CPPT, protons
contribute dose to the boost volume, but 79% of the dose to the PTV is delivered by photons. CPPT reduces the
mean doses in the lung, heart, remaining HT and contralateral breast compared to IMRT by 49%/44%, 38%/
27%, 26%/26% and 27%/26% with a horizontal/inclined FBL. However, since the dose to lung and heart for
the IMRT plan is lower compared to patients 1 and 3 due to the target geometry, there is a smaller benefit of
CPPT in terms of NTCP reductions (from 2% to 1.7%/1.7% for radiation pneumonitis and DNTCPheart of

8
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

Figure 5. CPPT plans for an inclined FBL for patient 3: patient geometry (a); cumulative dose (b), proton contribution (c) and photon
contribution (d). The contoured structures show the PTV (blue), the boost volume (red), the lymph nodes (yellow), the lungs
(magenta), the heart (orange) and the patient’s body (green).

0.9%/0.6% for major coronary events). Independently of the inclination angle of the FBL in CPPT, protons can
irradiate the target volume without crossing additional HT. Consequently, there is no advantage of using an
inclined FBL over a horizontal FBL in CPPT for this target geometry.
For both patient 2 and 3, a benefit in sparing of surrounding HT is obtained when applying CPPT with a
gantry using common beam angles. Moreover, robustness analysis analogous to figure 2 for both patients shows
that CPPT performs better than IMPT in terms of robustness, for both concepts with FBL and gantry.

Discussion

In this paper, we investigate a novel treatment concept that potentially makes proton therapy available to more
patients who currently do not have access to it—such as breast cancer patients. The concept is based on the
combination of protons and photons within the same fraction in a treatment room equipped with an FBL for
IMPT, a standard LINAC for IMRT and a robotic couch for treatments in lying position with accurate patient
setup. CPPT for breast cancer was found to outperform single-modality treatments. Compared to IMRT, CPPT
has the potential to reduce dose to surrounding normal tissues; compared to IMPT, CPPT may improve
robustness against range uncertainties.
Higher benefits from the use of CPPT in terms of normal tissue dose reductions, and possibly reduced risk of
radiation-induced toxicity, may be expected for patients with lymph node involvement and a concave shaped
PTV. This is demonstrated in this paper for patients 1 (figures 1) and 3 (figure 5) and it can be expected that
similar CPPT plans are obtained in similar patient geometries. Patients with a mostly convex shaped PTV
instead can be treated well with tangential photon beams, and consequently little improvement is expected from
CPPT (patient 2). While major coronary events may be the main concern clinically, the decrease of lung and
integral patient dose achieved with CPPT may also reduce the risk of radiation-induced
secondary cancers (Wennstig et al 2021, Grantzau et al 2014). However, due to the uncertain dose-response
relation, we did not evaluate secondary cancer risk models.
For breast cancer, anatomical changes such as swelling may be the main potential cause of proton range
uncertainty. Up and down scaling of relative stopping power is commonly done with 3% and a slightly smaller
scaling factor, although being plausible, would not make an important difference for range uncertainties in the
breast. For the common scaling factor, however, only approximately 1 mm of range uncertainty is modeled for a
proton beam that penetrates 3–4 cm into the breast, which may underestimate range errors. In this work, we
augment the model of range uncertainty by expanding or contracting the whole patient surface by a margin
where the values of RED were overwritten by water or air, respectively. This leads to a range uncertainty model

9
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

Figure 6. Robustness evaluation of CPPT with an inclined FBL for patient 1: DVHs for CTV (a) and its boost volume (b) in the
nominal scenario and setup shifts of 5 mm in six orthogonal directions.

with one component that depends on the penetration depth and another component that is independent of the
range of the proton beam.
While range errors mostly affect the proton component of a CPPT plan, uncertainties in target position due
to variability in breath hold position and breast setup affect protons and photons alike. For this work, these
uncertainties are accounted for by a 5 mm PTV margin rather than stochastic optimization, which is common
practice in IMRT/VMAT planning in the clinic. In agreement with other publications (Tommasino et al 2020),
we observe that this hybrid approach to robust planning achieves robustness against setup errors despite the
general limitations of the PTV concept for IMPT. Indeed, when evaluating the CPPT plan shown in figure 1 for
5 mm setup errors, it is shown that setup uncertainties have little impact on the coverage of the CTV and the
boost volume (figure 6). In clinical photon treatment planning, additional postprocessing methods to improve
the robustness of photon plans are commonly used. This includes extending the fluence of tangential IMRT
fields into the air beyond the beams-eye-view projection of the breast surface, also known as skin flash (Liang
et al 2020). Although not done here, the same method could be applied to the IMRT fields in a CPPT plan.
Alternatively, future work may further develop site-specific models of anatomical variation for breast treatments
to handle these uncertainties via scenario-based robust optimization methods. For example, Dunlop et al (2019)
suggested a robust optimization approach for VMAT planning using a PTV margin and a number of simulated
CTs representing anatomical changes of the breast with motions of the breast CTV.
The primary goal of this work is to evaluate the benefit of combining an FBL and a LINAC in a treatment
room for breast cancer patients. Additionally, we include comparisons to gantry-based proton delivery to
investigate the broader question of whether there is a benefit for CPPT over single-modality IMPT even without
limitations in the available proton beam angles due to an FBL. CPPT remains beneficial when using common
proton beam directions. However, in case of handling all uncertainties and employing a very large number of
proton beams, one may expect that the need for photon beams will eventually dissolve. This is further
investigated in appendix B. In this work, IMRT is considered for the evaluation of the benefit of CPPT as it is an
appropriate treatment technique for patients with lymph nodes involved in the target volume. Other treatment
techniques such as 3D conformal radiotherapy would not achieve satisfying treatment plan quality with
tangential fields and 3D planning, thus intensity-modulated techniques are required.
The main approach to cost-effective proton therapy using an FBL is to position the patient on a rotating
chair or even treat in an upright position. Thereby, a similar degree of freedom in the available proton beam
angles as with a proton gantry can be achieved. This approach of IMPT represents an alternative approach to
achieve high quality treatment plans with an FBL. CPPT would allow for treatment in lying position with
standard immobilization devices and would therefore avoid the main concerns regarding treatment in seated or
standing position: the accuracy and reproducibility of the patient’s positioning (Bortfeld and Loeffler 2017),
which requires new image guidance solutions and difficulties in the registration of diagnostic CT, MR, or PET
images that are acquired in horizontal position (Devicienti et al 2010).
While proton treatments in seated or upright position are explored by several companies and academic
institutions (Sheng et al 2020, Leo Cancer Care 2021), CPPT is a novel and less studied concept. Further research
is needed to investigate its cost-effectiveness, its benefit across treatment sites, as well as details of the treatment
room and facility design. So far, CPPT with an FBL has been evaluated for selected head and neck and breast
cancer where it was found that both horizontal and 45° inclined FBLs lower dose to relevant OARs (Fabiano et al
2020). Future studies may further quantify the benefit of CPPT across the whole population of head and neck
and breast cancer patients. In addition, the potential benefit for other treatment sites must be evaluated,

10
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

including common cancer sites such as lung cancer. This may estimate the prospects of commercial success for a
CPPT with FBL system. In addition, it may inform decisions on treatment room design, e.g. the optimal
inclination angle of the FBL. For investigating the cost-effectiveness of the suggested concept, a cost estimation
of the treatment room containing both the FBL for IMPT and a LINAC for IMRT is needed.

Conclusion

This work evaluates the potential benefit of CPPT with an FBL for three breast cancer patients, a concept that
may increase access to proton therapy for this patient population. It was found that the combined treatment
approach may yield improvements over both single-modality treatments. Compared to IMRT, mean doses to
surrounding tissues and OARs can be lowered, potentially reducing the risk of radiation-induced toxicities,
which may be relevant for a large patient population with long-term survival. In addition, we demonstrated that
combined treatments may also excel single-modality IMPT by improving the robustness of plans.

Appendix.

For the paper

Combined proton–photon treatment for breast cancer

Appendix A. Treatment plan optimization details

For the calculation of dose influence matrices for photons and protons with matRad, a beamlet size of 5×5 mm2 is
used for 6 MV photon beamlets and a lateral/longitudinal spot spacing of 5 mm/3 mm is used for proton pencil
beams. A generic proton machine with a gaussian lateral beam profile is assumed, using energies from
31.7–236.1 MeV and an initial lateral spread (sigma at patient surface) of 5 mm for the lowest energy and 2.3 mm for
the highest energy. Simultaneous optimization of photon beamlet intensities and proton pencil beam intensities
based on their cumulative dose is performed via an own implementation of the L-BFGS quasi-Newton method. The
objectives for individual VOIs are summarized in table A1 and the full objective function is defined as follows:
1
f (d ) = å [10(48 - d i)+2 + 2(d i - 50.4)+2 ] ,
NB i B
(A1)

1
+ å [10(40.05 - d i)+2 + 2(d i - 42.0525)+2 ] ,
NrPTV i rPTV
(A2)

1
+ å (d i - dimax )+2 ,
NM1 i M1
(A3)

1
+ å (d i - dimax )+2 ,
NM 2 i M 2
(A4)

1
+ å d i,
NL i L
(A5)

Table A1. Description of VOIs and associated objective functions used for treatment plan optimization as defined in equations (A1)–(A10).

Equation VOI Volume Planning goal

A1 B Boost volume Deliver the prescribed dose of 48 Gy while penalizing dose

exceeding 105% of the prescription dose
A2 rPTV Remaining PTV: PTV without the boost volume and Deliver the prescribed dose of 40.05 Gy while penalizing
without the 0.5 cm margin around the boost dose exceeding 105% of the prescription dose
volume
A3 M1 0.5 cm margin around the boost region, excluding Achieve a dose falloff from the 48 Gy to 40 Gy within
voxels outside the PTV 0.5 cm from the boost volume
A4 M2 1 cm margin around the PTV, excluding voxels out- Achieve a dose falloff to 20.025 Gy within 1 cm from
side the patient the PTV
A5 L Lungs, without PTV Minimize mean dose
A6 H Heart, without PTV Minimize mean dose
A7 RH1 Remaining HT: Body without PTV, lung and heart Minimize mean dose
A8 RH2 Remaining HT: Body without PTV and M2 Penalize dose exceeding half the PTV prescription in vox-
els more than 1 cm away from the PTV in the patient

11
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

1
+ å d i,
NH i H
(A6)

1
+ å d i,
NRH1 i RH1
(A7)

1
+
NRH2 i RH2
å (d i - 20.025)+
2
, (A8)

whereby NVOI denotes the total number of the voxels in the respective VOI and d i the dose of the respective
voxel i. Stochastic optimization is applied to objectives (A1–A6) while (A7–A8) are evaluated only for the
nominal scenario. The two conformity objectives (equations (A3) and (A4)) are defined within a 0.5 cm margin
around the boost volume (M1) and a 1 cm margin around the PTV (M2) using maximum doses d imax given by:

d imax  48 - z iB 2 (48 - 40.05) " i Î M1, (A9)

d imax  max {40.05 - z iPTV (40.05 - 20.025) , 48 - z iB (48 - 20.025)} " i Î M2, (A10)

where ziPTV and ziB are the Euclidian distances measured in cm of a HT voxel i from the PTV and boost volume
contour, respectively. Since the contour of the boost volume is in some places close to the boundary of PTV, the
second conformity objective (equations (A4)/(A10)) is designed such that for normal tissue voxels in the margin
close to the boost volume, a higher dose is accepted.
For the IMPT-mtc included in figure 2, which maintains target coverage for range errors, two constraints are
added to the optimization problem. The expected values of the objective functions (equation (A1)) and
(equation (A2)) in the 3 range error scenarios are constrained to be at least as good as in CPPT.

Appendix B. Dependence on the magnitude of range uncertainty

Different magnitudes of range uncertainty are studied by modelling different expansions and contractions of the
patient contour. Figure B1 shows the CPPT plan for small range uncertainty (0 mm/3%), i.e. range uncertainty
is modeled only by scaling of RED in the CT without modified patient contour. Compared to figure 1,
accounting for smaller range errors yields a higher proton contribution to the target volume (figure B1(a)).
However, the qualitative features of proton and photon dose contributions remain similar.

Figure B1. CPPT plan for patient 1 assuming an inclined FBL with small range uncertainty (0 mm/3%): proton contribution (a),
photon contribution (b) and total dose distribution (c). The contoured structures show the PTV (blue), the boost volume (red), the
lungs (magenta), the heart (orange) and the patient’s body (green).

12
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

Appendix C. Comparison to delivery with a proton gantry

CPPT shows a benefit compared to single-modality treatments even in the presence of a proton gantry when
commonly used proton beam directions are employed (in figures C1 and C2 exemplary for patient 1). Photons
still contribute a substantial portion of the PTV dose as shown in figure C1(b) for patient 1. This is explained by
the fact that commonly used proton beams, here 0°, 45° and 90°, all face the problem of overshoot into the lung
tissue and underdosage of the PTV in the context of range uncertainty.
However, in case of handling all uncertainties and employing arbitrarily many proton beams, one may
expect that the need for photon beams will eventually dissolve. To investigate this, we optimize CPPT and IMPT
plans for patient 1 that use 13 proton beams, incident from the same directions as the IMRT beams. Thus, also
tangential proton beams are allowed, and dose may be delivered with both protons and photons from every
beam direction. Indeed, the photon contribution to the PTV in the CPPT plan with this beam setting is small
(around 34%) compared to the gantry setting with 3 proton beams. The DVH comparison in figure C2 shows
that the 13-beam IMPT plan can achieve robustness against the modeled range overshoot and undershoot
scenarios without excessive overshoot into the lungs or underdosage of the target volume. Moreover, CPPT
based on the beam setting of 13 proton beams showed similar robustness against setup uncertainties as for CPPT
with FBL. However, the large number of applied proton beams represents a way of irradiation which is currently
not clinically practicable and would make the treatment time inappropriately long.

Figure C1. (a)–(c)/(d)–(f) CPPT plan for patient 1 assuming a proton gantry, 3 proton beams (0°, 45°, 90°)/ 13 proton beams. Proton
contribution ((a)/(d)), photon contribution ((b)/(e)) and total dose distribution ((c)/(f)). The contoured structures show the PTV
(blue), the boost volume (red), the lungs (magenta), the heart (orange) and the patient’s body (green).

13
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

Figure C2. DVH comparison for patient 1 for PTV (black), lung (blue) and heart (red) in the nominal (a), overshoot (b) and
undershoot scenario (c), comparing IMRT, CPPT and IMPT, both with 3 proton beams, and IMPT with 13 proton beams.

14
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

Appendix D. Additional data for other patient geometries

CPPT for patient 2 and 3 decreases the dose to the OARs compared to IMRT, as can be seen in figures D1(a), (b)
and D2(a), (b), while target coverage is maintained. In order to attain robustness, IMPT plans have to
deliberately overshoot into the lungs and heart (figures D1(c) and D2(c)) for both patients.

Figure D1. (a) DVH of patient 2 for PTV (black), lung (blue) and heart (red) in the nominal scenario of IMRT, CPPT and IMPT with
an inclined FBL. ((b)–(c)) IMRT and IMPT plans assuming an inclined FBL for patient 2.

Figure D2. (a) DVH of patient 3 for PTV (black), lung (blue) and heart (red) in the nominal scenario of IMRT, CPPT and IMPT with
an inclined FBL. (b), (c) IMRT and IMPT plans assuming an inclined FBL for patient 3.

ORCID iDs

Niklas Wahl https://orcid.org/0000-0002-1451-223X

References
Assessing the Rate of Complications in X-Ray Therapy Versus Proton Beam Radiation Therapy After Breast Conserving Surgery or
Mastectomy in Treating Patients With Breast Cancer Identifier NCT04443413 (https://clinicaltrials.gov/ct2/show/NCT04443413)
Alderliesten T et al 2018 Breast-shape changes during radiation therapy after breast-conserving surgery Phys. Imaging Radiat. Oncol. 6 71–6
Ares C et al 2010 Postoperative proton radiotherapy for localized and locoregional breast cancer: potential for clinically relevant
improvements? Int. J. Radiat. Oncol. Biol. Phys. 76 685–97
Bortfeld T R and Loeffler J S 2017 Three ways to make proton therapy affordable Nature 549 451–3
Bottura L, Felcini E, De Rijk G and Dutoit B 2020 GaToroid: a novel toroidal gantry for hadron therapy Nucl. Instrum. Methods Phys. Res. A
983 164588
Darby S C et al 2013 Risk of ischemic heart disease in women after radiotherapy for breast cancer N. Engl. J. Med. 368 987–98
Deasy J O, Blanco A I and Clark V H 2003 CERR: a computational environment for radiotherapy research Med. Phys. 30 979–85
Devicienti S, Strigari L, D’Andrea M, Benassi M, Dimiccoli V and Portaluri M 2010 Patient positioning in the proton radiotherapy era
J. Express Clin. Cancer Res. 29 1–5
DIRAC - Directory of Radiotherapy Centres 2021 (https://dirac.iaea.org/Data/CountriesLight) (Accessed: 9 June)
Dunlop A, Colgan R, Kirby A, Ranger A and Blasiak-Wal I 2019 Evaluation of organ motion-based robust optimisation for VMAT planning
for breast and internal mammary chain radiotherapy Clin. Transl. Radiat. Oncol. 16 60–6
Fabiano S, Balermpas P, Guckenberger M and Unkelbach J 2020 Combined proton–photon treatments—a new approach to proton therapy
without a gantry Radiother. Oncol. 145 81–7
Facilities in Operation - PTCOG 2021 (https://ptcog.ch/index.php/facilities-in-operation) (Accessed: 09 June )
Grantzau T, Thomsen M S, Væth M and Overgaard J 2014 Risk of second primary lung cancer in women after radiotherapy for breast cancer
Radiother. Oncol. 111 366–73

15
Phys. Med. Biol. 66 (2021) 235002 L Marc et al

Harbeck N and Gnant M 2017 Breast cancer Lancet 389 1134–50

Kammerer E et al 2018 Proton therapy for locally advanced breast cancer: a systematic review of the literature Cancer Treat. Rev. 63 19–27
Khan P W S F and Gibbons J P 2016 Treatment Planning in Radiation Oncology (Philadelphia: Lippincott Williams & Wilkins (Wolters
Kluwer))
Freeman T 2020 Upright treatment could increase patient comfort, reduce proton therapy costs, Physicsworld, particle therapy, September
3, 2020. (accessed Jun. 04, 2021) (https://physicsworld.com/a/upright-treatment-could-increase-patient-comfort-reduce-proton-
therapy-costs/)
Li J G, Williams S S, Goffinet D R, Boyer A L and Xing L 2000 Breast-conserving radiation therapy using combined electron and intensity-
modulated radiotherapy technique Radiother. Oncol. 56 65–71
Liang X et al 2020 Using robust optimization for skin flashing in intensity modulated radiation therapy for breast cancer treatment: a
feasibility study Pract. Radiat. Oncol. 10 59–69
Lomax A J 2020 Proton therapy special feature : review article myths and realities of range uncertainty Br. J. Radiol. 93 1–9
Marks L B et al 2010 Radiation dose-volume effects in the lung Int. J. Radiat. Oncol. Biol. Phys. 76 70–6
Mohan R and Grosshans D 2017 Proton therapy—present and future Adv. Drug Deliv. Rev. 109 26–44
Schippers J M, Lomax A, Garonna A and Parodi K 2018 Can technological improvements reduce the cost of proton radiation therapy?
Semin. Radiat. Oncol. 28 150–9
Sedlmayer F et al 2013 DEGRO practical guidelines: Radiotherapy of breast cancer I: radiotherapy following breast conserving therapy for
invasive breast cancer Strahlentherapie Onkol. 189 825–33
Sheng Y et al 2020 Performance of a 6D treatment chair for patient positioning in an upright posture for fixed ion beam lines Front. Oncol. 10
1–9
Sung H et al 2021 Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries
CA. Cancer J. Clin. 71 209–49
The DBCG Proton Trial: Photon Versus Proton Radiation Therapy for Early Breast Cancer Identifier NCT04291378. (https://clinicaltrials.
gov/ct2/show/NCT04291378?term=proton+therapy&cond=breast+cancer&draw=2&rank=6)
Tommasino F et al 2020 Clinical implementation in proton therapy of multi-field optimization by a hybrid method combining conventional
PTV with robust optimization Phys. Med. Biol. 65 045002
Unkelbach J et al 2015 Optimization approaches to volumetric modulated arc therapy planning Med. Phys. 42 1367–77
Unkelbach J et al 2018 Robust radiotherapy planning Phys. Med. Biol. 63 22TR02
Wennstig A K et al 2021 Risk of primary lung cancer after adjuvant radiotherapy in breast cancer—a large population-based study npj Breast
Cancer 7 71
Wieser H P et al 2017 Development of the open-source dose calculation and optimization toolkit matRad Med. Phys. 44 2556–68
Xiong W et al 2004 Optimization of combined electron and photon beams for breast cancer Phys. Med. Biol. 49 1973–89

16