Introduction
Radiation dermatitis (RD) or skin toxicity is one of the most common acute side effects of radiation in head and neck cancer patients. If higher grade, it can cause unwanted delay or untimely stoppage of treatment, and considerable physical as well as psychological suffering of the patients. Radiobiologically, the development of RD has long been known to be dependent on dose-volume and overall-treatment-time [1]. Concurrent chemotherapy [2] increases the risk of RD significantly, as well as other acute side effects such as oral mucositis and dysphagia. Modern conformal techniques, like intensity-modulated radiotherapy (IMRT) [3] and volumetric-modulated arc therapy (VMAT), are now standards of care in head and neck squamous cell carcinoma. Though highly conformal as per target volume coverage and Organs-at-risk sparing [4, 5], these modern techniques have been observed to result in a relative overdosage of patient’s skin surface [6]. Apart from only a handful of previous studies, neither recommendations nor any guidelines are available [7] to use skin dose constraints to prevent the development of RD. Our study aims to correlate the pattern of VMAT dose distribution to the skin with the different grades of RD, and to assess whether specific skin dose-volume threshold values may be used as possible predictive factors for higher grades of toxicity in future studies. We hope that our study will help researchers in future to prospectively validate our data, and create a specific skin dose-constraint guideline so that treatment can be planned abiding those constraints to prevent acute skin toxicity of higher grade.
Materials and methods
Sample and treatment features
80 treatment plans of histopathologically proven inoperable non-metastatic head and neck squamous cell carcinoma patients of different subsites already treated with cisplatin-based definitive concurrent chemoradiation in VMAT technique at our institution between November 2022 and November 2023 were randomly selected, excluding the bolus-using plans (superficial gross tumour or skin infiltration). During treatment, concurrent cisplatin [8] was administered at a dose of 100 mg/m2 3 weekly for patients with an Eastern Cooperative Oncology Group (ECOG) Performance Status of 0–1 with dose modification, if needed, according to a kidney function test and general condition. For the different subsites, radiation dose was 66–70 Gy at conventional 2 Gy per fraction (33–35 fractions) in a phasic plan or 30 fractions in the simultaneous integrated boost (SIB) technique as per physician’s choice. Initial pre-treatment nutritional status was assessed using Simplified Nutritional Appetite Questionnaire (SNAQ) and noted down [9]. Standard supportive care measures, such as pain management, maintenance of oral hygiene and nutritional counselling, were implemented on a routine basis.
For each of the patients, a personalized thermoplastic head-neck-shoulder mask was created. An institutional Philips Brilliance 16-slice CT Scan machine was used to acquire simulation scans of 3mm slice width. Target volume contouring was done according to international consensus guidelines [10] at Varian Somavision workstation. 7mm margin was given for planning target volume (PTV) around the clinical target volume (CTV) as per institutional protocol. For PTV, a negative 3mm margin was delineated from patient’s body surface contour to avoid excessive accumulation of skin dose. Double arc-VMAT Treatment plannings were done using Eclipse v15.5 (VARIAN medical systems) software (RapidArc) and Anisotropic Analytical Algorithm (AAA) for dose calculation and optimization. The patients were treated in VARIAN TrueBeam linear accelerator (serial number- 3279). Informed consent was obtained before starting treatment from all individual patients included in this study.
Analysis of skin dose distribution
The selected 80 VMAT plans were retrieved from our digital archive. For each plan, 1 new ring structure of 3 mm thickness was created below the external skin surface up to 3mm depth as all the PTV margins were cropped 3mm from the skin surface as per institution protocol, further supported by published data [11, 12], and thus this 3mm-thick ring structure was a perfect model to study the skin dose parameters. These volumes of interest extended between the upper and lower limits of PTV plus a fixed 1cm margin in both cranial and caudal directions. For this ‘3mm ring’ structure of each patient, the following parameters were collected — V40, V50, V60 and Dmax, where V40, V50 and V60 equal to the volume of the ring structure receiving a minimum of 40Gy, 50Gy and 60Gy dose respectively, and Dmax represents the maximum absolute point dose anywhere within the structure. These parameters were correlated with different grades of RD as per Common Terminology Criteria for Adverse Events (CTCAE) v5.0 [13], as assessed by at least two of the authors at a time, once weekly during treatment and documented in the individual patient’s treatment record files.
Outcome measures and statistical analysis
Descriptive statistics were used to report patient ECOG Performance Status, age, sex, smoking history, disease [primary site, American Joint Committee on Cancer (AJCC) 8th edition tumor–node–metastasis (TNM) staging], treatment (RT compliance, RT schedule) and skin dose distribution (explorative dosimetric parameters) — related characteristics as median and range for continuous variables. These continuous variables were tested with Mann-Whitney test as it is a common nonparametric test for comparing two groups of independent samples [14], while categorical variables were tested by Fisher’s exact test due to its validity in analysing small sample of categorical variables and providing an exact p-value [15]. A p-value of < 0.05 was considered significant.
By plotting the receiver operating characteristic (ROC) curves, the optimal cutoff value was determined using Youden’s approach. The Youden Index is a commonly used summary measure of the ROC curve, and it measures the effectiveness of a diagnostic marker or a test, and enables the selection of an optimal threshold value (cutoff point) for the marker or test [16]. Sensitivity, specificity, and area under the curve (AUC) at the cutoff value were calculated in order to study the ability of the variables predicting the risk of developing higher grades of RD. Risk ratio (RR) for the association between each variable and the risk of developing higher grades of RD were obtained by univariate analysis.
All the statistical analysis was done using MedCalc software version 22.021.
Results
The selected 80 patients’ characteristic features are summarized in Table 1. In brief, median age was 65 years; 85% of the patients were male, and 67.5% of the sample population were heavy smokers (> 20 pack–years). 77.5% of the patients initially had moderate malnutrition as per SNAQ criteria, whereas 22.5% were well-nourished. The proportion of patients in subsets of oropharynx, hypopharynx and larynx were almost similar (28.75%, 25% and 27.5%, respectively) whereas 15% patients had disease of nasopharyngeal origin. Human papilloma virus (HPV) testing was not done in the oropharyngeal primary patients due to unavailability of the facility in our institution. A large proportion (63.75%) of the patients had stage IVA disease according to TNM staging 8th edition. TPF (docetaxel, carboplatin and fluorouracil)-based Induction chemotherapy was prescribed for only 5 of these 80 patients (6.25%). Overall, treatment compliance was good as per the patients’ attendance documented in the treatment room register, with only 6.25% having prolonged (> 4 days) RT treatment discontinuation due to toxicities. Treatment features and toxicity rates are shown in Table 2. 47.5% patients had grade 1 RD, 42.5% suffered from grade 2, and only 10% had grade 3 RD as per CTCAE criteria. No patients in our study suffered from grade 4 acute skin toxicity.
|
Characteristics |
No. of patients (%) (n = 80) |
|
Median age Years [range] |
65 (16–78) |
|
Sex Male Female |
68 (85%) 12 (15%) |
|
ECOG performance status 0 1 2 |
59 (73.75%) 17 (21.25%) 4 (5%) |
|
Initial nutritional status |
|
|
Well nourished |
18 (22.5%) |
|
Moderately malnourished |
62 (77.5%) |
|
Severely malnourished |
0 |
|
Smoking history (pack–years) 0 0–10 10–20 > 20 |
7 (8.75%) 11 (13.75%) 8 (10%) 54 (67.5%) |
|
Primary tumour subsite Nasopharynx Oropharynx Hypopharynx Larynx Others |
12 (15%) 23 (28.75%) 20 (25%) 22 (27.5%) 3 (3.75%) |
|
AJCC Stage (8th edition) III IVA IVB |
24 (30%) 51 (63.75%) 5 (6.25%) |
|
Characteristics |
No. of patients |
|
VMAT schedule Sequential SIB |
26 (32.5%) 54 (67.5%) |
|
RT compliance No interruptions Temporary interruptions Median of interruptions (days, range) < 3 days ≥ 4 days |
62 (77.5%) 18 (22.5%) 4 (1–16) 13 (16.25%) 5 (6.25%) |
|
Radiation dermatitis (RD) Grade 1 Grade 2 Grade 3 Grade 4 |
38 (47.5%) 34 (42.5%) 8 (10%) 0 |
Patient, disease, treatment-related features and skin dose-volume parameters are summarized in Table 3 as per development of grade 1 and grade 2/3 RD. Statistically significant association of higher grade of RD was found with higher skin volumes receiving 40 Gy, 50 Gy or 60 Gy dose, and with the higher maximum point dose in the skin ring structure. From plotting ROC curves, statistically significant optimal cutoff values were found using Youden’s approach — 71.6632 cc for V40 (AUC = 0.972, p-value < 0.001), 29.98 cc for V50 (AUC = 0.982, p-value < 0.001), 7.624 cc for V60 (AUC = 0.859, p-value < 0.001), and 66.56Gy for Dmax (AUC = 0.680, p-value = 0.04). RR with 95% confidence interval (CI) were calculated by univariate analysis, and all of the 4 variables (V40, V50, V60 and Dmax) were significantly correlated with grade 2/3 skin toxicity as depicted in Table 4. V50 was found to have the best predictive accuracy among the four parameters (AUC 0.982) as the higher AUC values in ROC curve analysis indicate better test performance. Figures 1–4 show the ROC curves plotted in MedCalc software. Figure 5 shows the dose-volume histogram (DVH) of the 3mm ring structure in a sample VMAT treatment plan.
|
Characteristics |
Grade 1 RD (n = 38) |
Grade 2/3 RD (n = 42) |
p-value |
|
Age < 65 years > 65 years |
20 (51.28%) 18 (43.90%) |
19 (48.71%) 23 (56.09%) |
0.508822 |
|
Sex Male Female |
33 (48.52%) 5 (41.67%) |
35 (51.47%) 7 (58.33%) |
0.66073 |
|
ECOG PS 0 1 2 |
28 (47.45%) 8 (47.06%) 2 (50%) |
31 (52.54%) 9 (52.94%) 2 (50%) |
0.994319 |
|
Initial nutritional status Well nourished Moderately malnourished Severely malnourished |
11 (61.11%) 27 (43.54%) 0 |
7 (38.89%) 35 (56.45%)
|
0.188992 |
|
Smoking history (pack–years) 0 0–10 10–20 > 20 |
5 (71.43%) 6 (54.54%) 4 (50%) 23 (42.59%) |
2 (28.57%) 5 (45.45%) 4 (50%) 31 (57.41%) |
0.49967 |
|
Primary subsite Nasopharynx Oropharynx Hypopharynx Larynx Others |
5 (41.67%) 11 (47.83%) 9 (45%) 12 (54.54%) 1 (33.33%) |
7 (58.33%) 12 (52.17%) 11 (55%) 10 (45.45%) 2 (66.67%) |
0.9254 |
|
Stage (AJCC 8th edition TNM) III IVA IVB |
10 (41.67%) 26 (50.98%) 2 (40%) |
14 (58.33%) 25 (49.02%) 3 (60%) |
0.7089 |
|
VMAT schedule Sequential SIB |
13 (50%) 25 (46.296%) |
13 (50%) 29 (53.7%) |
0.7560 |
|
PTV (size, cc; median, range) |
161.4 (96.8----247) |
202.6 (128.9----274.6) |
0.1247 |
|
Skin ring 3 mm V40 (size, cc; median, range) |
51.4182 (18.05---80.095) |
83.4892 (70.645----119.542) |
< 0.00001 |
|
Skin ring 3 mm V50 (size, cc; median, range) |
21.868 (4.37----37.812) |
47.93 (30.1146----85.06) |
< 0.00001 |
|
Skin ring 3 mm V60 (size, cc; median, range) |
2.647 (0.0012----7.264) |
10.4082 (0.000333----40.6) |
< 0.00001 |
|
Skin ring 3 mm Dmax (point value, Gy; median, range) |
66.56 (56.412----74.243) |
67.964 (60.697----74.26) |
0.030084 |
|
Variable |
Grade 1 RD (n = 38) |
Grade 2/3 RD (n = 42) |
Relative risk (95% confidence interval) |
p-value |
|
V40 ≥ 71.66 cc < 71.66 cc |
6 32 |
40 2 |
14.7826 (3.8353–56.9772) |
0.0001 |
|
V50 ≥ 29.98 cc < 29.98 cc |
6 32 |
42 0 |
57.2449 (3.6487–898.1137) |
0.004 |
|
V60 ≥ 7.624 cc < 7.624 cc |
2 36 |
18 24 |
2.2500 (1.5973–3.1694) |
< 0.0001 |
|
Dmax ≥ 66.56 Gy < 66.56 Gy |
18 20 |
38 4 |
4.0714 (1.6346–10.1410) |
0.0026 |
Discussion
IMRT is a highly precise conformal radiotherapy technique in which multiple small beamlets, each with a non-uniform intensity profile, create dose distributions that can accurately conform to convex and concave structures alike [17]. The introduction of IMRT into the treatment of head and neck cancer patients made the reduction of treatment-related morbidity possible along with exquisite dose distribution [18], although with a substantial risk of marginal geographic miss of target and increased integral dose. Lee et al. [6] in 2002 first considered the skin of the neck as a separate structure for IMRT optimization, and showed that the volume of the skin exposed to > 45 Gy could be brought down by 20% compared with regular IMRT planning, as well as decreasing its mean dose by 6%. Price et al. [19] applied a 4-mm negative margin to the PTV from the body surface contour, and found the occurrence of superficial hotspots above 110% to be minimal, along with better rotational IMRT plan conformity with this increase of PTV to skin distance. In another study by Penoncello et al. [20] published in 2016, VMAT, a special form of IMRT delivering single or double arcs of precise radiation beams while continuously rotating around the patient’s body [21] showed 5.6% reduction of mean dose to the skin compared with static IMRT. However, the more modern refinements of IMRT techniques introduced neither any specific constraints to use for the skin nor any indications how to contour it in the irradiated area. Different OAR guidelines [22] recommend 3–6 mm skin thickness to be contoured, the variation depending on different skin thicknesses in different regions of interest. Studer et al. [11] in 2011 found a positive association between the incidence of Grade 3/4 RD and the radiation dose delivered to the larger skin volume in patients receiving cetuximab as concurrent chemotherapy. It was assessed by measuring skin doses > 50 Gy and > 60 Gy in the subdermal area 3mm below the skin. Severe RD was found in V50 91cc and V60 50 cc where grade 0–2 reactions were seen in V50 61cc and V60 27cc. In 2019, Mori et al. [23] found that a 2-mm thick superficial body layer DVH was associated with the risk of developing acute RD in patients treated with tomotherapy. They found V56/V64 to be the most predictive parameters for grade 2/grade 3 RD. Optimal cutoff values were found to be 7.7 cc and 2.7 cc for V56 and V64, respectively. V64 < 3 cc constraints should keep the risk of Grade III toxicity lower than 10%. Also in 2019, Bonomo et al. [12] created 3 ring structures, 2 mm, 3 mm and 5 mm below the skin to assess the correlation between skin dose-volume and Grade III/IV RD in 90 head and neck cancer patients treated with tomotherapy. They compared a concurrent cisplatin cohort with a cetuximab cohort. In multivariate analysis, they found statistically significant correlation of > 10 kg weight loss and performance status > 1 with the development of severe RD. The best predictive accuracy they found was at 2mm: an AUC 0.61 with V50 of 19.9cc and V60 of 5.8 cc. Our retrospective work adds to the available literature providing hypothesis-generating results, though the limitations must be acknowledged. Firstly, the relatively smaller sample size reduces the generalizability of our results. Secondly, the data collected retrospectively from our institutional datasheets may have interobserver differences. Thirdly, all the patients underwent the VMAT treatment according to the original plan. No weight loss was noted down, and no adaptive RT was considered due to heavy workload and logistical reasons. For that reason, we could not correlate the occurrence of moderate to severe RD with significant anatomic change (weight loss). But we think that proper prospective validation with a larger sample size and multivariate analysis are warranted based on our results, and specific skin dose-constraints can be recommended in future international RT guidelines to decrease moderate to severe acute skin toxicity.
Conclusions
The dose distribution pattern to a 3 mm thick skin layer below the surface may help predict the development of moderate to severe radiation dermatitis in head and neck cancer patients receiving definitive concurrent chemoradiation. An optimum cut-off value of 29.98 cc for the volume of skin receiving at least 50 Gy radiation dose was found to be the strongest predictor of grade 2 and 3 radiation dermatitis in our study. This data may help future researchers to design large prospective studies incorporating these constraints. This will further validate the applicability of the present study in clinical setting to prevent radiation-induced skin toxicity.
Conflict of interest
The authors declare no potential conflict of interest.
Funding
None declared.