2.2. Volumes Definition, Dose Prescription, and Planning Objectives
For all the patients, an expert radiation oncologist delineated two clinical target volumes (CTVs) according to previously reported definitions [
19,
20]. A high-risk CTV (HR-CTV) including the gross tumor volume (GTV) of the primary tumor was determined by clinical information, endoscopic procedure, and magnetic resonance imaging (MRI), before any CHT, with a margin in agreement with the compartment-related definition by Claus et al. [
21]. A low-risk CTV (LR-CTV) was defined, including bilateral nodal levels Ib–III and retropharyngeal nodes irradiated electively, in compliance with the international guidelines [
22]. Planning target volumes (HR-PTVs and LR-PTVs, respectively) were generated by adding a 3 mm margin to the corresponding CTVs. Although in clinical practice different dose prescriptions and fractionations schemes could be used, based on the histology, disease stage, extension, and response to induction chemotherapy, we used the same dose prescription scheme in order to better compare VMAT and IMPT for the study purpose. Thus, all plans were re-optimized with a simultaneous integrated boost (SIB) modality, with a prescribed dose of 70 Gy(RBE) and 56 Gy(RBE) in 35 fractions, to HR-CTV and LR-CTV, respectively. For IMPT plans, we used a fixed RBE value of 1.1. The contoured OARs included optic chiasm, optic nerves, retinae, anterior chambers, eyeballs, lacrimal glands, brainstem, spinal cord, temporal lobes, cochleae, and lenses. Parotid glands, mandible, and glottic larynx were also considered for patients needing elective or curative neck irradiation, but were not included in the present analysis. All optic structures were classified into ipsi-lateral and contra-lateral, according to primary tumor proximity. When tumor localization was central and symmetric respect to paired OARs, the structure receiving the higher mean dose has been considered as ipsi-lateral. In absence of clearly established MRI images, visual field tests analysis helped to determine the more impaired, potentially expendable side of the ocular structures. Structures located within the CTVs were not contoured because either they were absent due to resection or it was assumed that symptoms would be tumor-related and not therapy-related. Thus, we did not expect any outcome improvement from PT. In order to reduce inter-observer bias, an additional expert radiation oncologist checked for contoured OARs plausibility. For the investigation purpose, we have included the updated evidence in terms of suggested radiation dose-volume constraints for a variety of normal tissue complications related to head and neck cancer treatments, published after the QUANTEC reports (early 2010) and including Patient Reported Outcome measures (PROs), where available. The plan optimization process aimed to increase as much as possible the CTVs coverage without exceeding the constraints to selected OARs. In particular, we opted for a priority order for planning objectives and constraints. We gave the highest planning priority to the following structures: brainstem, spinal cord, optic chiasm, and contra-lateral optic nerve at least, to preserve mono-lateral vision. CTVs coverage represented a second priority and the lowest priority was given to the remaining OARs sparing. We considered as clinically acceptable and potentially deliverable to the patient all the plans passing the criteria summarized in
Table 1, column 2.
2.4. Plan Analysis and Comparison
For each patient case, we performed a DVH analysis similarly to the quoted references [
19,
26]. For target coverage evaluation, we focused only on the CTVs, since the PTVs were not included in the IMPT optimization. Regarding OARs sparing, we considered a combination of several dose parameters possibly associated to ocular and neurological toxicities, based on the literature as shown in
Table 1, column 3. In particular, the volume of retina receiving doses higher than 50 Gy(RBE) (V
) and 55 Gy(RBE) (V
) was evaluated as potentially related to radiation retinopathy, according to [
27,
28].The volume of optic nerves and chiasm receiving doses higher than 55 Gy(RBE) was also estimated, being related to the risk of radiation-induced optic neuropathy (RION), as reported in QUANTEC series [
14,
18,
29,
30]. The dose to the lacrimal glands was also investigated and we considered both the volume receiving 30 Gy(RBE) (V
) and the mean dose, being possibly related to dry eye syndrome (DES) [
31] as described in [
32,
33]. Finally, the volume of the brain receiving doses higher than 25 Gy(RBE) (V
) and 35 Gy(RBE) (V
) was also assessed, as potentially leading to fatigue or memory impairment [
34]. For the dosimetric comparison we calculated the relative percentage difference for all the DVH parameters reported in
Table 1 (column 3) between photon and proton plans. (DVH
and DVH
respectively) as
DVH
= 100*[(DVH
− DVH
)/DVH
]. Finally, we also computed the Homogeneity Index (HI) = (D
− D
)/D
, where D
is the prescription dose, and the Conformity Index (CI) = TVD
/TV, where TVD
and TV represented the total volume encompassed by the D
and the target volume respectively. Since PTVs were not used in the IMPT plans, the CTVs were used as TV. We applied the Wilcoxon signed-rank test with a p-value of 0.05 for testing the null hypothesis that a certain DVH parameter is equal for both the two sets of 22 VMAT and IMPT plans. Therefore, in the plan comparison analysis only the DVH indices with a statistically significant difference between the two samples were included. Afterwards, we converted
DVH
to an arbitrary variable (DVH*) which can assume three discrete values (−1, 0, 1), according to the following criteria: if
DVH
was higher than +20
revealing an evident advantage from IMPT, then DVH* = +1; if
DVH
was lower than −20
indicating a distinct benefit from VMAT, then DVH* = −1. In all other cases, DVH* = 0, meaning that the two radiation techniques are comparable for that specific OAR under investigation.
After a comprehensive review of the literature, eight NTCP models were used for plan comparison in this study (
Table 2). We based the model selection on a focus on SNC-specific toxicities relying on clinical experience and considering studies developing, validating, or applying NTCP models with available parameters for evaluation. In this work the clinical toxicities endpoints were divided into two categories, intermediate and severe, depending on their impact on patients Quality of Life (QoL), as detailed in
Table 2, in brackets. Late neurological toxicity with devastating clinical consequences or potentially life-threatening, such as blindness [
12], brain, brainstem and spinal cord necrosis [
15], temporal lobe injury [
35], were defined as severe. Otherwise, other relevant rare adverse effects, which still have a significant but less tremendous impact on patients QoL, were referred as intermediate. We established the acute overall ocular toxicity ≥ Grade 2 as intermediate, according to the radiation toxicity criteria of Radiation Therapy Oncology Group (RTOG) and European Organization for Research and Treatment of Cancer (EORTC) as reported by Batth et al. [
11], since the authors contemplated a wide spectrum of toxicity with variable impact on patients’ Qol, including conjunctivitis, keratitis and corneal ulceration. In contrast, DES was scored as a severe toxicity [
36] since it is related to acute radiation reactions that ultimately resulted in compromised vision according to RTOG Grade 3 and 4 toxicities and National Cancer Institute’s Common Terminology for Adverse Advents (NCI CTCAE) for DES Grade 2 and 3. Subsequently, we defined late brain necrosis as intermediate [
16] sinche the authors chose brain necrosis CTCAE v4.0 ≥ Grade 2 endpoint derived from MRI and clinical symptoms for their study. The net difference in NTCP for specific endpoint (
Table 2) between photon and proton plans (phNTCP and pNTCP respectively) was calculated as
NTCP =
NTCP –
NTCP, for severe and intermediate toxicities (
NTCP
,
NTCP
, respectively) and used for further analysis. We thus introduced a supplementary selection criterion for plan comparison as a mixed
NTCP-
DVH parameter called total score (TS), determined as reported in Equation (
1). TS consisted in a weighted sum that considers the 8 NTCP models described in
Table 2, four models for severe and four models for intermediate toxicities (
NTCP
and
NTCP
, respectively) were adopted, together with
DVH for the m DVH parameters that, according to Wilcoxon test, were statistically significant in terms of the percentage difference between IMPT and VMAT plans.
A weighting factor multiplies each term in Equation (
1). We applied a relative weight of 20 to the
NTCP
(w
), while we assigned a unitary weight factor for
NTCP
(w
), given the impact on the patient QoL. Finally we assigned to w
a value of 10, being both severe and intermediate toxicities considered in the third term of the equation. If at least one of the following two conditions was met, we expected the selected patient case to benefit from IMPT in terms of reduced risk of radiation-induced side effects:
- 1
(a) ΔNTCP
exceeded a threshold of 20% (similar to [
10]) for at least three of all the investigated intermediate toxicities side effects.
(b) NTCP exceeded a threshold of for a single severe toxicity.
- 2
TS was higher than a certain arbitrary threshold of 250.