Introduction
Patients with mediastinal lymphoma commonly are young with curable histologies. While radiotherapy is an effective treatment for mediastinal lymphoma, there has been reluctance to irradiate these patients given increased risk of radiation-associated late cardiac toxicity [
1‐
3] and secondary cancers of the breast [
4,
5] and lung tissue [
6]. Strategies to reduce toxicity in these patients have included reducing treatment volumes (involved site/node radiation), lowering radiation dose, and improving radiation delivery. Several techniques have been used in the latter setting, including butterfly intensity modulated radiation therapy (IMRT) [
7], deep inspiration breath hold (DIBH) [
8], and proton therapy [
9].
Proton therapy’s physical dose distribution, which is associated with a steep dose fall-off beyond the Bragg peak, makes it an attractive radiation technique in patients with mediastinal lymphoma. However, there are unique treatment planning considerations that arise with proton therapy. Mostly, pencil beam analytical (PBA) dose algorithms have been used to calculate dose distributions, although PBAs may not calculate proton dose in lung tumors accurately given the heterogeneous tissue interfaces that proton beams must traverse [
10,
11]. In addition, there is compelling theoretical and laboratory evidence indicating that within the Bragg peak, proton linear energy transfer (LET) and, therefore, relative biological effectiveness (RBE) varies [
12,
13], although a constant clinical RBE of 1.1 is currently used for proton therapy planning and outcome assessment. In sites such as the mediastinum, uncertainties in the proton biological dose distribution (RBE × physical dose) arise from both the dose calculation algorithm and uncertainties associated with proton RBE. In our clinic, anterior or anterior oblique proton beams are typically used for mediastinum lymphoma treatment planning. There is a possibility that cardiac structures that lie at the distal end of these beams may be exposed to high linear energy transfer protons with an RBE greater than 1.1.
Numerous comparative dosimetric studies have compared proton and photon-based techniques for mediastinal lymphoma [
14], although most, if not all, of these were performed using PBA and a constant clinical RBE of 1.1. Commercial Monte Carlo-based (MC) dose calculation algorithms have recently become available for clinical use. MC, which is regarded as the gold standard for physical dose calculations, also has the potential to incorporate variable RBE (vRBE) models that account for spatial variations in proton kinetic energy and LET within the Bragg peak. We evaluated the role of MC dose algorithms for proton treatment planning in the mediastinum and compared dosimetry between photons and proton plans that had been optimized and calculated with MC. A secondary goal was to explore the potential impact of vRBE on cardiac doses relative to
60Co γ-rays and MV x-rays. We strived to accomplish these goals by performing (i) dosimetric comparisons between proton PBA and MC based plans, (ii) dosimetric comparisons between photon and proton MC plans on DIBH and free breathing (FB) scans, and (iii) dosimetric comparison of cardiac structure doses for biological dose distributions [(physical dose × vRBE) vs (physical dose × 1.1)] for DIBH and FB proton plans and the photon DIBH plans (RBE = 1.0). Note, photon plans were only constructed on DIBH scans given that prior studies [
15] have suggested that IMRT in DIBH, proton therapy in FB, and proton therapy in DIBH each significantly reduced estimated late effects and life-years lost compared to IMRT in FB. Therefore, we wanted to use to “best” photon plan as comparison to proton plans. Both protons in DIBH and FB were used as comparisons given that at many proton centers, including ours, DIBH may not be in routine use given lack of volumetric image guidance (e.g. cone-beam CT) and patients may instead be treated using FB.
Discussion
To our knowledge, this is the first study evaluating the impact of MC dose algorithm and vRBE on proton dosimetry among mediastinal lymphoma patients. Although our study is limited by small patient numbers, MC dosimetry revealed reduced target coverage under PBA-based planning and increased dose heterogeneity, consistent with findings in lung cancer patients [
10,
11]. Our study also revealed that dosimetric endpoints could be maintained if MC planning is used for both initial plan optimization and final dose calculation. These findings highlight the limitations of the PBA dose calculation algorithm to calculate dose across heterogeneous tissue interfaces (e.g. soft tissue, bone, lung) and suggest that MC dose calculation algorithm should routinely be used for proton treatment planning for patients with mediastinal lymphoma. Despite under coverage of the CTV/ITV, which in the worst-case scenario, 90% of the prescription dose covered 100% of the volume, none of the 6 patients treated with the nominal proton plans calculated with PBA developed recurrence at a median follow-up of 28.4 months from proton RT start, though our numbers are small.
Although this study is performed for pencil beam scanning proton therapy, the findings may still apply to passive scatter or uniform scanning proton techniques. The underlying analytical pencil beam dose calculation algorithm [
37] used for proton pencil beam planning is the same for uniform scanning and passive scattering. For plans with similar anatomy and beam configuration, PBA plans may show similar limitations for dose calculations for these modalities.
The dosimetric superiority of proton over photon plans has been established [
14], although prior comparative dosimetry studies used PBA, which may inaccurately estimate target volume coverage. Because of under-coverage of the target volume with PBA, re-optimization of plans using MC to improve target coverage was associated with slightly higher dose to nearby organs at risk. There has been recent interest in evaluating the impact of DIBH to photon or proton techniques [
38‐
40], with the hypothesis that in certain subsets of patients, photons with DIBH may provide similar cardiac and/or lung sparing compared to protons with free breathing. Within our cohort, mean lung dose, lung V5Gy, spinal cord, and mean breast dose were lower with proton therapy free breathing compared with photon DIBH. Proton therapy with DIBH was also associated with lower mean heart dose and V20Gy compared with photon DIBH. Whether differences in mean dose to the breast, heart, and lung, which ranged in magnitude from 2 to 5 Gy, is clinically meaningful depends in part on the patient’s age [
41], sex [
41], prior treatments [
3,
42], baseline co-morbidities [
43], and family history [
44], which can each modulate the risk of late toxicity from radiotherapy. For example, while there is a linear, no threshold relationship between dose to the breast tissue and risk of secondary cancer in Hodgkin lymphoma survivors [
45], the relative risk is higher for a young (< 35 year old) female patient with intact ovarian function compared with a female in her fourth decade of life, whose fertility lifespan is more limited [
41]. Therefore, any sparing of breast tissue from radiation will be more clinically meaningful in a young female patient. Age is only one dimension of risk; on top of this, other risk factors, as mentioned earlier, increase a patient’s baseline risk, and with it, the absolute excess risk of radiation-associated secondary cancer. For this reason, providing an objective cut off of “acceptable” dose differences (e.g. between photons and protons) is challenging given various other clinical and treatment factors that also modulate the risk of late radiation toxicity.
Addition of DIBH to proton therapy did not improve mean heart or breast dose, but as expected, DIBH improved lung metrics. Our findings are consistent with a larger cohort of 21 mediastinal lymphoma patients with lower mediastinal involvement, in which addition of DIBH did not impact mean heart dose with IMRT or proton therapy. Proton therapy had similar or lower dose to the heart, lung, and breast tissue [
46]. In contrast, in the largest study to date from the University of Copenhagen, life years lost (LYL) attributable to late effects after radiotherapy for mediastinal Hodgkin lymphoma was calculated based on normal tissue dose generated with IMRT and proton plans, with and without DIBH [
15]. Compared to IMRT-FB, proton therapy and IMRT-DIBH was associated with significantly lower LYL, but no difference was seen between proton therapy-FB and IMRT-DIBH. However, extent of mediastinal disease was not reported, which may be an important factor in which mediastinal lymphoma patients benefit from DIBH [
8]. Recent consensus recommendations from ILROG highlight which mediastinal patients may benefit from proton therapy, including those with mediastinal disease that extends below the origin of the left main coronary artery [
47].
Given that primarily anterior beams are used for our proton plans and concerns that the RBE at the proton beam’s end of range may exceed the currently used clinical RBE = 1.1, we also explored the biological dose to cardiac substructures using a published vRBE model for the endpoint of DSB induction. Our analysis of representative mediastinal patient plans indicates that, despite the potential for very large end-of-track RBE effects biological dose in cardiac substructures is not substantially increased. Biological dose was estimated using a single model; currently, no consensus exists on the most appropriate vRBE model to calculate biological dose.
Of note, although the RBE for DSB induction is one of the most biological significant forms of initial molecular damage and is closely related to cell survival (see Appendix), clinical endpoints such as local tumor control and normal tissue complications may not exhibit the same general trends in vRBE as molecular or cellular surrogates. On the other hand, vRBE models developed for the endpoint of reproductive cell survival have been in routine clinical use for high LET carbon ion therapy for some time with little or no evidence of unexpected normal-tissue damage or compromised local tumor control [
48‐
51]. On the relevant spatial scales (few mm), corrections for vRBE modeling in carbon ion therapy are much larger than vRBE corrections in proton therapy (i.e., on the order of 3–5 compared to on the order of 1.0 to 1.4). Also, the same general molecular and cellular RBE mechanisms of action are largely believed the same for protons and carbon ions in vitro and in vivo [
36,
52]. Although distal to the Bragg peak, RBE for cell survival is larger than the RBE for DSB and has the potential to become as large as 2–3.7 (blue shaded region in Additional file
1: Figure S1), the impact of differences in RBE estimates (DSB versus cellular survival) on biological dose is minimal as the proton physical dose is rapidly decreasing over these few millimeters beyond the Bragg peak.
Proton therapy was associated with lower dose to the left ventricle, aortic valve, mitral valve, and tricuspid valve, although the clinical significance of these differences is not clear. Risk of valvular heart disease (VHD) after cardiac irradiation is non-linearly related to dose to the affected valve, in which risk of VHD increases by only 2.5% per Gy with valve doses <=30 Gy [
53]. As most patients within our cohort were treated with doses <=30 Gy (i.e. within the shallowest slope of the dose-response relationship), decrease in valvular dose may be associated with minimal changes in VHD risk. Mean left ventricle dose was relatively low among our cohort (range, 4.2–6.2 Gy across photon and proton plans), and while protons was associated with significantly lower dose, the clinical impact may be very small, if any, as the risk of heart failure is non-linearly related to mean left ventricular dose [
54].
In conclusion, MC-based dosimetry revealed reduced target coverage under PBA-based planning. Proton plans optimized with MC dose algorithm confirmed modest sparing of normal tissue over photon techniques with DIBH, although the relative benefit varies between patients. MC-based dose algorithms should be used in proton treatment planning for patients with mediastinal lymphoma. Our preliminary study suggests that end of range RBE effects do not significantly impact biological dose in cardiac substructures for mediastinal targets, although these dosimetric estimates will require validation with late toxicity data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.