Radiation therapy with protons, owing to its physical properties, can be advantageous for the treatment of children. This study was conducted in order to quantify the advantages of proton therapy from a treatment planning point of view in a consecutive, realistic, and mixed pediatric/adolescent population with varying diagnoses and target locations.
Forty-five patients treated with conventional 3-dimensional conformal radiation therapy photon radiation therapy were retrospectively re-planned with scanned proton beams. Treatment sites represented were the central nervous system, head and neck, thorax, and abdomen. Median age was 8 years (range, 2-18 years). All plans were optimized with intensity-modulated proton therapy (multi-field optimization). We analyzed a number of dose-volume descriptors for planned target volumes (PTVs). Organ-specific mean doses and relevant DV -values were derived for organs at risk. In addition, homogeneity index, conformity index, treated volume, and integral dose were calculated for each treatment plan. The Wilcoxon matched-pairs signed rank test was used for studying differences between these variables for the 2 treatment modalities.
Planned target volume coverage (V95%) was similar when comparing photons and protons. Conformity and homogeneity indices were similar or better for protons for most cases. In general, doses to organs at risk were lower with protons. In cases with organs at risk in close vicinity to the PTV, the gain with protons is less.
The patient cohort benefits from reduced integral dose with protons compared with photons. Patients with tumors in the central nervous system, head and neck, upper mediastinum, and some abdominal locations will gain significantly if treated with protons compared with photons. For 7 of the 45 consecutive patients studied (whole brain, whole lung, whole abdomen, flank treatment), we found no gain with protons (difference in integral dose less than 8%).
Radiation therapy with protons has an advantage to photons because of its physical properties. The use of protons for children and adolescents is therefore appealing from an organ sparing point of view . Several comparative treatment planning studies have been made for different tumor sites in both adult and pediatric/adolescent populations [1–5]. The rationale for using protons for children is its potential for better sparing of both organs at risk (OARs) and normal tissue compared with photons. The prospect of lowering the risk for radiation-induced secondary cancers is also very attractive for pediatric/adolescent patients who will live long after their treatment [6–8]. Reduction of radiation therapy–induced side effects would help the children to live more normal lives as cancer survivors. The future use of intensity-modulated proton therapy with pencil-beam scanning will also aid in lowering the risk of secondary cancers, owing to less neutron dose compared with passive scattering , which is the most common proton technique used today.
A few studies present results for the pediatric/adolescent population for selected diagnoses [10–13]. The radiation therapy community is still divided in how favorable protons may be compared with newer highly conformal treatment techniques in photon therapy such as intensity-modulated radiation therapy and volumetric modulated arc therapy. However, for pediatric/adolescent patients, there seems to be less disagreement.
Most Swedish pediatric/adolescent patients are expected to receive their radiation therapy at the national proton center (Skandionkliniken), which is a 2-gantry facility with spot scanning only. It will start clinical operation in mid-2015.
To prepare for the start of patient treatments, we made a retrospective treatment planning study of the whole population of children/youngsters in Sweden, treated with photons during 1 year (2008), in order to not only find those patient groups that would benefit most but also to see if there might be patients who would not benefit from proton treatment . Patient data were identified from the national database (Radtox) for long-term follow-up on side effects of the treatment. It contains the same type of data as collected by the German/Austrian/Swiss RiSK Group, with whom we collaborate . In addition, we also save (when possible) full Digital Imaging and Communications in Medicine (DICOM)–radiation therapy information of the treatments, which made this study possible.
There are many practical aspects to consider when treating children with protons. The need for immobilization is crucial in proton treatment  as is the need for anesthesia, which is more frequent than in conventional radiation therapy. Imaging for treatment guidance is greatly needed not only for safe and secure treatment but also for adaptive planning in the case of changes in target volumes, body volumes, or both [17, 18]. These factors have, however, not been included in the present study.
This study was conducted in order to quantify the advantages of proton treatment from a treatment planning point of view in a consecutive, realistic, and mixed pediatric/adolescent population with varying diagnoses and target locations.
Patients and Methods
The patients in this study were retrieved from the 93 pediatric patients treated with radiation therapy in Sweden during 2008. We were able to retrospectively re-plan 45 of these patients with scanned proton beams. Those patients not re-planned received total body irradiation (13 patients) or had no accessible DICOM data (15 patients). Four patients were not registered in the Radtox database, and 1 patient denied participation. In addition, we had technical problems with 5 cases. Ten patients had no target delineated; 6 patients with acute lymphocytic leukemia (ALL) received prophylactic treatment to the brain, 2 patients received palliative treatment to vertebrae bodies, 1 patient with lymphangiomatosis received treatment toward the whole thorax, and 1 patient with Wilms disease received flank treatment. The treatment of all included patients was discussed during national telemedicine conferences  in conjunction with start of their radiation therapy. The diagnoses are tabulated in Table 1 . The ages of the children in the whole cohort varied with a median age of 8 years (range, 2-18 years).
The study was approved by the ethics committee in Umeå, Sweden (Reference No. 2014-247-32M).
Target Volumes and Organs at Risk
Target volumes were kept as originally delineated. We decided on this solution to be able to compare the resulting treatment plans. Prescribed doses are in Gy(RBE) (relative biological effectiveness) using a constant RBE of 1.10. Most other absorbed doses presented in this article are relative doses normalized to the prescribed dose.
Since we collected all consecutive cases from the 6 Swedish centers treating children, the delineation of OARs varied. To be able to compare dose distributions, an experienced radiation oncologist retrospectively added OARs so that all cases within a diagnosis group had the same set. The following OARs were defined: brain, hippocampus, brain stem, inner ears, optical nerves and chiasm, eyes and lenses, pituitary gland, inner ears , the parotid and submandibular glands, spinal cord, thyroid, lungs, heart, liver, pancreas, spleen, kidneys, bladder, and bowel where applicable.
We used Varian Eclipse (Varian Medical Systems, Palo Alto, California) for proton planning of all cases. The beam had energies from 100 to 230 MeV (range in water, 7-32 cm). Spot size (full width half maximum [FWHM]) for 120 MeV at 10-cm depth is 10.2 mm, and the distal fall off (90%-10%) is 4.4 mm. A range shifter (water equivalent thickness, 74.1 mm) was used for all beams in all treatment plans created, since the skin-to-target distance was always less than 7 cm. All proton plans were optimized with intensity-modulated proton therapy (multi-field optimization). The aim for the treatment plans was to achieve a dose distribution comparable with photons or better, with at least the same target coverage or an acceptable coverage considering nearby dose-limiting OARs. Beam angles were selected to be as robust as possible; however, no other specific robustness tests were performed. All treatment plans were optimized with the physical dose-volume objectives/constraints (for targets as well as OARs) stated in the study protocols the patients were following. All proton plans were created by the same experienced dosimetrist.
The 3-dimensional conformal radiation therapy (3DCRT) photon treatment plans used for comparison were the plans originally used for the treatment of the children/adolescents. These plans were considered optimal for treatment and in concordance with appropriate treatment protocols. There were 9 patients having 2 to 3 target volumes and dose levels to be added for the full treatment. Photon plans were then summed to the total dose prescribed. The same was done for the proton plans.
We analyzed a number of dose-volume descriptors; V95%, D98% (near minimum dose), D50%, and D2% (near maximum dose) for PTVs.
Organ-specific mean doses and relevant VD-values were derived for OARs. In addition, homogeneity index (HI = [D2% − D98%]/D50%) for PTV, conformity index (V95%/PTVvol), treated volume (V95% for the body), and integral dose (mean dose for the body) were calculated for each treatment plan [20, 21]. Wilcoxon matched-pairs signed-rank test was used to determine any significant statistical differences (P < .05) for the planned OAR doses between the 2 different modalities.
The CERR software  and RT Bench (Cureos AB, Uppsala, Sweden) were used for the data analysis and comparison of treatment plans.
Planned Target Volumes
Planned target volume coverage (V95%) was very similar when comparing photons and protons (Figure 1), which was also the aim when creating the proton plans. There were 3 exceptions, all with superficial targets. Conformity and homogeneity indices were similar or better for protons for most of the cases (Figure 2).
Organs at Risk Volumes
Results are presented for the whole cohort (Figure 3). Mean doses are reported for organs with mainly parallel architecture, while D2% (near maximum dose) is presented for serial type organs.
The whole group included 45 patients and 56 treatment plans. For those patients receiving more than 1 photon plan, the same practice was followed for proton planning. In Figure 3, summarized doses are compared with OARs. In general, mean proton doses are lower than all OARs but for the brain stem. For brain, hippocampus, inner ear, parotid glands, heart, and bowel, the mean doses are significantly lower (P < .001, P < .001, P < .001, P < .001, P < .018, and P < .009; respectively) for protons than for photons. For pituitary gland, thyroid, lungs, liver, and kidneys we did not find any significant difference (P < .408, P < .109, P < .116, P < .937, P < .176; respectively). Near maximum (D2%) doses to the optic nerves and spinal cord are significantly lower (P < .033 and P < .001, respectively), while doses to the optic chiasm and brain stem show no significance (P < .171 and P < .465, respectively). The treated volumes are most often lower for protons (Figure 4). The integral dose is on average 23% lower for protons (range, 15%-52%, P < .0001). For the majority of patients, the integral dose is lower by 25% or more; however, for a few cases, such as whole brain (2 cases where integral dose sparing is 4% and 0%, respectively), sarcoma of the chest wall (where in 1 of the 3 plans for this patient, protons do worse [15% higher integral dose with tangential fields] and for another there is only 1% sparing of integral dose). These 2 plans are boost plans. In summarizing the patient's 3 plans, the total integral dose can only be lowered by 8%; however, for this specific case, the dose to the heart is 40% lower with protons. For 3 patients treated either to the whole abdomen or a flank, the integral dose is only 1%, 4%, and 5% lower with protons.
To put our results in clinical perspective, we compared the doses received by the OARs with a recent comprehensive compilation of published organ-specific dose-volume objectives and associated endpoints made by Brodin et al  (Table 2). Note that dose-volume constraints/objectives used for OARs for treatment planning in this work are those stated in the study protocols, which may differ from those reported by Brodin et al.
Brain tissue is spared better with protons. Cochlea and parotid glands can also be spared to a better degree than with photons. The dose to the pituitary gland does not differ for the 2 modalities. This is valid for the optic chiasm, optic nerves, and brain stem as well. There is no difference in sparing of the thyroid in this material. Doses to the heart and lungs are slightly better for protons. The liver receives a lower dose with protons, but the difference in doses to the kidneys is very small. However, the objectives in Table 2 are not always met by any modality. The objectives from the literature in Table 2 are, for several organs, lower than in the treatment protocols the children were following.
We have studied and quantified the differences between proton and external beam photon therapy in a large number of consecutive pediatric patients. We did this to get improved knowledge in the area of proton planning for a pediatric/adolescent cohort and to learn in which situations these patients would gain most from being treated with protons. In the majority of the cases, the target coverage can be retained with a significant reduction in the treated volume. The integral dose (mean dose to body volume) is reduced significantly (P < .0001) with protons for the patient cohort studied.
We have used the same PTV volumes as delineated for photons. It can be argued that this is not sufficient for robust planning with protons. Likewise, the dose-volume constraints from the pediatric treatment protocol are not specifically set for protons but for photons. However, since the cohort is collected from 6 different centers in Sweden, and for ease of comparison, we decided not make any changes in the PTVs. The use of range shifter broadens the proton spot and thereby widens the penumbra. This is something to take into consideration for future protocol design. There might be a need to define target volumes specifically for the more advanced treatment technique and the technology of today, where image-guided radiation therapy and magnetic resonance in the radiation therapy departments, including proton facilities, are more frequently available.
Doses to the brain, hippocampus, inner ears, spinal cord, and parotid glands are significantly lower with protons. These findings are similar to those found by MacDonald et al  and Beltran et al . Yet, in our material, doses to the optic chiasm, optic nerves, and the thyroid are not significantly lower for protons compared with photons. We found a significantly lower dose to the heart but not to the lungs, the liver, or the kidneys. In the abdomen, there is a large healthy volume to spare, reducing acute side effects. Integral dose is on average 29% smaller with protons in those with abdominal targets, with a variation from 1% to 52%. Similar findings were made by Hillbrand et al  and Hattangadi et al . We found that the bowel received statistically significant lower proton doses, but there were no significant differences for the liver or the kidneys.
The minimum beam energy available at Skandionkliniken will be 60 MeV, which will allow for treatment of many targets without a range shifter or at least a thinner one than used in this study. This will be useful for the pediatric cohort to further decrease treated volumes and aid in keeping sharper penumbrae and higher dose gradients toward OARs and, hence, lower doses to these. Compared with the published OAR dose-volume objectives , it will be possible to keep OAR doses well below these dose-volume objectives in many instances.
In this study, we have compared 3DCRT treatments with pencil-beam scanning protons. Three-dimensional CRT is still the most common technique for pediatric radiation therapy. There is no doubt that protons can spare healthy tissue to a greater extent than 3DCRT. The results may be different if compared with the advanced photon techniques such as intensity-modulated radiation therapy, volumetric modulated arc therapy, or tomotherapy. However, we do not yet know what the “dose bath” from these techniques will result in for the young child/adolescent during the remaining lifetime. For most of the patients in our cohort, we find considerable sparing of OARs and normal tissue. To the individual patient, it will mean less acute side effects and fewer late side effects, during the rest of their lives, resulting from the radiation treatment. This finding is confirmed by Rombi et al in their overview of clinical results . A number of practical issues have to be solved, such as motion management, immobilization, and tumor changes, body changes, or both during the treatment phase. We also need to investigate skin doses and try to compare these with photons—thereby estimating skin reactions, both acute and long term. We will also have to model and measure the neutron dose contribution for spot scanning when using a range shifter, since it will be frequently used for small patients even with an energy as low as 60 MeV.
Radiation treatment with protons for children and adolescents has become more common over the years. The American survey of pediatric treatments shows this clearly. Approximately 10% of all American pediatric radiation therapy is treated with protons . The same is true in Sweden today; where at The Svedberg Laboratory in Uppsala, children are treated with passive scattering with a horizontal beam line and restricted field size. We expect to treat a larger number of the Swedish children and adolescents with protons at Skandionkliniken.
In our cohort, we had patients already treated with protons; these 5 patients were excluded since DICOM data could not be exported from the treatment planning system (TPS) used. Of the 10 patients without target, 8 would probably not gain from proton treatment, and 2 were palliative treatments who might, for that reason, not be considered for proton treatment.
The majority of the patients in this study would have gained from protons in the sense that radiation doses to their body volume and to most of their OARs would have been less with protons compared with photons. For each diagnosis group, we will need to create an optimization kit of dose descriptors in order to lower doses to OARs/body volume even more.
From this study, we conclude that most pediatric/adolescent patients will gain from protons. Less gain is seen for whole brain, whole lung, whole abdomen or flank treatment.
We have also noticed the importance of and need for updated definitions on the expansion from clinical target volume to PTV in an environment with high-precision treatment, where multimodal diagnostic imaging is available and where image-guided radiation therapy is standard in most facilities.
The present investigation serves, however, as a good base for future work, exploring these areas for subsets of patient cases in order to learn more on how to plan and treat the children with proton therapy.
ADDITIONAL INFORMATION AND DECLARATIONS
Conflicts of Interest: The authors have no conflicts to disclose.