Background
Early-stage non-small cell lung cancer (NSCLC) and lung metastases have been treated with stereotactic body radiation therapy (SBRT) with excellent clinical outcome [
1‐
4]. Locally advanced NSCLC has traditionally been treated with conventionally fractionated concurrent chemo-radiation [
5]. However, dose escalation with conventional fractionation has not demonstrated any clinical advantage in any phase III randomized controlled trials [
6]. As a result, alternative radio-therapeutic strategies are being sought. Given the clinical efficacy that has been observed with hypo-fractionated dose schedules, such as SBRT, this approach has also been increasingly considered as a treatment strategy for locally advanced and selected stage IV NSCLC in recent years [
7‐
9].
Excellent dose distribution can be achieved with image-guided and intensity-modulated radiotherapy (IG-IMRT) delivered with volumetric modulated arc therapy (VMAT) in the thorax [
10,
11]. As a result, VMAT has been quickly adopted in clinical settings to deliver thoracic radiotherapy [
12]. However, sparing thoracic organs at risk (OARs) while maintaining adequate target volume dose coverage with VMAT remains challenging in selected patients. On the contrary, particle therapy (PT), which includes proton and heavy ion therapies, may have an advantage over VMAT in OAR sparing. This improvement is due to PT’s physical properties, which allow for better normal tissue protection, while heavy ions’ increased radiobiological effectiveness (RBE) increases the tumorcidal effect of radiotherapy over photons [
13,
14]. PT’s dosimetric advantages have been shown in multiple studies [
15‐
22]. Given these advantages, PT is increasingly being considered for the delivery of hypo-fractionated thoracic radiotherapy [
23,
24]. To date, PT has been mostly delivered with passive scattering systems. With further improvements in technology, intensity modulation of particle beams with active beam scanning has been developed for better dose conformity and control of the OAR dose. However, little is known about how active scanning beams compare with VMAT in dose distribution in the treatment of lung tumors. In this study, we explored the dosimetric characteristics of intensity-modulated proton therapy (IMPT) and carbon ion therapy (IMCIT) with raster scanning beams in the delivery of hypo-fractionated thoracic radiotherapy in comparison to VMAT under limited respiratory motion.
Discussion
To the best of our knowledge, this study is the first study comparing photon VMAT with IMPT and IMCIT in the definitive treatment of loco-regionally confined lung tumors with hypo-fractionated radiotherapy. Although their clinical significance remains to be determined, our results indicated that fixed-beam IMCIT, compared with fixed-beam IMPT, may improve dose distribution for the tumor target, an effect possibly related to the sharper lateral dose gradient associated with carbon ions. The target volume dose coverage between VMAT and fixed-beam IMCIT was similar. Although the D
95 was slightly improved with IMCIT compared with VMAT, the difference may be too small to have any clinical impact. However, fixed-beam IMCIT with 3 fields is greatly limited by the number and angles of beams that can be used to generate the most conformal treatment plan. In contrast, VMAT is one of the most mature and sophisticated forms of intensity-modulated photon therapy, and it can generate excellent dose conformity to the tumor target [
11]. This difference indicates the great potential for IMCIT to further improve the dose distribution and the clinical efficacy of thoracic radiotherapy as it evolves in the future, given its known advantages over photons and protons (sharper lateral dose gradient, higher relative biological effectiveness, lower oxygen enhancement ratio, and more densely ionizing tracks that can lead to increased tumor DNA damage, less cell cycle dependence, and a stronger immunological response) [
13,
14].
Suboptimal dose conformity due to the limitations of fixed-beam PT appears to be more prominent with IMPT, because protons, compared with carbon ions, are associated with greater lateral penumbra. Improving dose conformity for IMPT is important, because it may decrease the risk of radiation pneumonitis [
31]. Although IMPT and CIRT, compared with 3D–CRT, have been shown to improve target volume coverage when multiple fields are used, IMRT has been shown to have better dose conformity and homogeneity than passively scattered PT [
17‐
22]. Our study comparing IMPT and VMAT demonstrated similar findings. IMPT arcs may lead to better dose conformity than VMAT at the cost of increased dose heterogeneity, a finding that warrants further investigation [
20].
PT has traditionally been shown to decrease the low dose volumes in the lungs [
15‐
22]. IMPT, and especially IMPT arcs, compared with passively scattered PT, may further decrease the normal lung dose [
16,
17,
19,
20]. Despite prominent low dose sparing, either passively scattered PT or IMPT may lead to higher volumes of the normal lung receiving doses >50% of the prescribed dose than photon therapy delivered with 3D–CRT or IMRT [
17,
18]. However, this unique phenomenon of proton therapy has not been previously observed in comparisons of CIRT with 3D–CRT [
21,
22]. Both IMPT and IMCIT significantly lowered the normal lung V
5 - V
20 in this study. This result was similar to those from previous studies comparing proton and photon therapies [
15‐
20]. In contrast, VMAT, compared with either IMPT or IMCIT, appears to be associated with slightly lower but clinically non-significant volumes of normal lungs receiving high doses close to the prescribed dose, probably because of the suboptimal dose conformity associated with fixed-beam particle therapy (Fig.
1). In addition, IMCIT, compared with IMPT, may further decrease the ipsilateral lung volumes receiving low doses, thus leading to reduced total lung volumes receiving low doses. These findings may be associated with both the physical properties of particle beam therapy and the better dose conformity observed with VMAT and IMCIT compared with IMPT. However, improved normal lung sparing with IMCIT compared with VMAT was not found to be as dramatic as previously observed when passively scattered CIRT was compared with 3D–CRT, possibly because of the better control of normal lung dose associated with VMAT compared with 3D–CRT [
21,
22]. Given PT’s physical properties, this sparing may be improved with arc-based methods of beam delivery [
20].
Significant sparing of the heart, esophagus, and other thoracic OARs was observed with PT compared with VMAT (Table
4, Fig.
2). This finding was consistent with those of previous studies comparing PT and photon therapy [
15‐
22]. Although no significant difference between IMCIT and IMPT was observed in the sparing of the heart, the esophagus, the spinal cord, or the major airway, the sparing of the heart and the major blood vessels in the high dose volumes may be best achieved with IMCIT. Overall, our findings suggested that although they are delivered with a limited number of beams and beams angles, fixed-beam IMPT and IMCIT may have an OAR sparing advantage over photon VMAT, which is more prominent with IMCIT than IMPT in the sparing of normal lungs and major blood vessels, owing to its improved dose conformity. Such an advantage may be further augmented as PT is more developed in the future.
Intensity modulation through active scanning may further improve dose conformity and OAR sparing in the thorax [
16,
17,
19,
20]. This advantage is significantly limited by organ motion due to changes in radiological path length as a result of organ motion adjacent to the tumor, inter-field motion, and the interplay of interference between beam and tumor motion, which may result in tumor under-dose and OAR over-dose [
32]. The interplay effect has been shown to significantly deteriorate target dose coverage and dose homogeneity when the motion amplitude is greater than 8 mm [
33]. The correlation between motion amplitude and dose heterogeneity due to interplay, which leads to increased under-dosing within the target volume, has been further demonstrated in a 4D Monte Carlo simulation [
34]. Such under-dosing has been modeled to significantly decrease the 2-year local control. However, this problem may be mitigated through fractionation, which leads to a motion-averaging effect on dose distribution, and the selection of larger spot size for the scanning beam. Together, these strategies can retain dose homogeneity with motion amplitudes of <20 mm. The mitigating effect of fractionation on interplay has also been shown in a series of 11 patients with stage III NSCLC for whom conventional fractionation with 35 fractions and hypo-fractionation with 10 fractions were modeled [
35]. These observations further support our utilization of a hypo-fractionated schedule in the current study.
Respiratory motion is mainly managed through rescanning, gating, and beam tracking [
32,
36,
37]. Rescanning mitigates the interplay effect by averaging the under-dosing and over-dosing patterns of the dose distribution. This process can be accomplished through volumetric or energy-slice by energy-slice rescanning while motion parameters are kept different between rescans. Rescanning mitigates the interplay effects without mitigating the target motion. Thus, adequate margins need to be maintained for adequate target dose coverage, possibly leading to suboptimal OAR sparing in conjunction with the dose blurring at the field borders. In addition, rescanning may be beneficial only when significant interplay exists under moderate to large respiratory motion [
38]. Gating, which is already in clinical use for photon therapy to minimize respiratory motion-related dose degradation and unnecessary OAR irradiation, has been a major approach of respiratory motion management for particle therapy to minimize interplay. Through 4D dose calculation, adequate target volume dose coverage and dose homogeneity have been demonstrated with a gating window (GW) of ≤5 mm, and lung dose can be further decreased within the clinically acceptable range with shorter GWs up to 1 mm for spot scanning proton therapy [
39]. This process supports our motion selection criteria of limiting the cranio-caudal motion to ≤5 mm through selecting the motion phases, including the maximal expiratory phase, by visual inspection of each patient’s 4D CT. Although 4D dose calculation to better account for interplay was not performed in this study, dose heterogeneities due to interplay for target motion of ≤5 mm were found to be within the clinically acceptable 5% for both carbon ions and protons at our institution (Additional file
1). How to best account for interplay in thoracic particle therapy will be more thoroughly assessed in future studies. A limitation for evaluating the effects of range uncertainties and interplay in this study is that errors may exist in the correlation between surrogate marker motion and internal tumor motion, and irregularities in the breathing pattern may also be present during real-time treatment. Such errors may be further decreased through periodical stereoscopic imaging intra-fractionally and phase-controlled rescanning (PCR) combined with gating for fast scanning particle beams [
40]. Excellent dose conformity and enhanced OAR sparing have been demonstrated for PCR combined with gating [
41]. However, rescanning during gating may not be necessary in all cases, especially when fractionated treatments are delivered, as suggested by the findings discussed above [
34,
35,
38]. To account for irregular breathing patterns throughout the entire course of fractionated particle therapy, amplitude-based gating based on tumor location observed in 4D CT datasets has been adopted and routinely used clinically [
42]. Although active tumor tracking may lead to the smallest high dose volume, this approach places a high demand on scanning speed, which must allow for rapid alteration of beam energy to adjust the Bragg peak in depth in relation to tumor motion [
43]. This longitudinal compensation may be achieved through the use of wedges [
44,
45]. However, tumor tracking remains an area of active research in thoracic particle therapy.
Owing to the scanning beam’s sensitivity to motion, 4D dose calculation for thoracic particle therapy has been advocated. The temporal density changes due to respiratory motion, which result in range uncertainties, can be incorporated into the treatment planning process. This process leads to decreased interplay and more robust treatment plans that, in contrast to 3D planning, can avoid unexpected under-dosing [
46,
47]. Robust 4D planning is preferred for thoracic particle therapy, because dose errors in 3D planning are not always dependent on motion [
47,
48]. However, this approach can be labor intensive and technically demanding. Thus, 3D planning with adequate motion management is still an acceptable approach in clinical practice [
48]. No clinically significant differences in most of the commonly evaluated dose parameters were observed in our dose recalculation. This finding may be due to the limiting of respiratory motion in this selected group of patients, because the primary goal of the study was to compare the dose distribution among photon VMAT, IMCIT, and IMPT with minimal influence from respiratory motion. Therefore, 4D planning may have a more significant impact when significant respiratory motion is encountered. Our 4D dose evaluation is only a limited approximation of the actual dose, owing to a lack of full consideration of the interplay between respiratory motion and the dynamic beam scanning, which could be achieved only with more robust approaches for 4D dose calculation that are not commercially available currently [
29,
30,
46,
47]. As a result, the limited dose variation observed after 4D dose evaluation, especially for the OAR dose parameters, may not fully capture the interplay effect and potentially be missing larger than actually observed dose uncertainties. This issue is also the major limitation of our study. Nevertheless, this study represents a step forward from 3D planning in accounting for range uncertainties that may lead to an overall decreased interplay effect [
47]. The best approaches to 4D dose calculation and how to fully account for the interplay effect are beyond the scope of the current study. Robust 4D dose calculation for scanning beam particle therapy should be further assessed, and 4D planning should be considered whenever feasible, owing to the lack of direct correlation between motion amplitude and the interplay [
29,
30,
47].
Another major limitation of this study is that fixed-angle particle beams are used. Dose distribution may be further improved with gantry-based systems through augmenting the known advantages of active scanning particle therapy with the increased number of angles for beam delivery. This question remains to be investigated in the future.