Dermoscopy of acute radiation-induced dermatitis in patients with head and neck cancers treated with radiotherapy

Head and neck cancer (HNC) was the seventh most common cancer in the world in 2018. Treatment of a patient may include surgery, radiotherapy (RT), chemotherapy, targeted therapy, immunotherapy, or a combination of these methods. Ionizing radiation used during RT covers relatively large volumes of healthy tissue surrounding the tumor. The acute form of radiation-induced dermatitis (ARD) are skin lesions that appear usually within 90 days of the start of RT. This is a prospective study which compares 2244 dermoscopy images and 374 clinical photographs of irradiated skin and healthy skin of 26 patients at on average 15 time points. Dermoscopy pictures were evaluated independently by 2 blinded physicians. Vessels in reticular distribution, white, yellow or brown scale in a patchy distribution, perifollicular pigmentation and follicular plugs arranged in rosettes were most often observed. For these dermoscopic features, agreement with macroscopic features was observed. Two independent predictors of severe acute toxicity were identified: gender and concurrent chemotherapy. Knowledge of dermoscopic features could help in the early assessment of acute toxicity and the immediate implementation of appropriate therapeutic strategies. This may increase the tolerance of RT in these groups of patients.


Patients
The study group consisted of 26 patients who underwent RT due to HNC (24 squamous cell carcinomas, one lymphoepithelial carcinoma, and one undifferentiated nasopharyngeal carcinoma) at the Maria Skłodowska-Curie National Research Institute of Oncology, Gliwice Branch, between September 2020 and March 2021.The inclusion criteria were age > 18 years, radical treatment signed consent.Patients treated with biological drugs (bio radiodermatitis), and with active dermatoses that could affect the clinical and dermoscopic picture of the examined skin area under observation were excluded from the study.Details of the patients' clinical and histopathological characteristics, the location of the tumor, are shown in Table 1.The control group consisted of skin regions not exposed to ionizing irradiation from the same patients (748 images).

Treatment
In seven cases, induction chemotherapy (indCHT) prior to radiochemotherapy (CHRT) was given; CHRT and RT alone was applied in 12 and four patients, respectively.The median total RT dose was 70 Gy (50-72 Gy) given in 25-40 daily fractions.Radiotherapy was delivered for over 7 weeks by incorporating five fractions per week combined with chemotherapy (CHT) (cisplatin, 100 mg/m 2 days (d) 1, 22, 43) or as a concomitant boost (CB) with seven fractions per week without CHT.Clinical target volume 1 (CTV1) included a primary tumor and involved lymph node groups with a margin.Clinical target volume 2 (CTV2) included CTV1 and areas at risk of harboring microscopic spread of primary tumor and elective lymph node groups.All patients were treated with Lower (3), middle (7), and upper (2) pharynx; epiglottis (1), glottis (1); larynx (8); palatine tonsil (2), an alveolar triangle of palatine tonsil (1); metastasis to the lymphatic system of the neck from an unknown primary site (1) Carcinoma planoephiteliale (24)  Lymphoepitelial carcinoma (1) Undifferentiated nasopharyngeal carcinoma (1)

Statistical analysis
A photographic database of 2244 dermoscopic photographs and 374 clinical photographs was analyzed in the final statistical assessment.Concordance based on Cohen's κ coefficient in the assessment of dermoscopic and macroscopic photographs between two independent observers in 89% of the results was greater than or equal to 0.9.In particular, the value of κ ranges between − 1 and + 1 (κ equal to + 1 implies a perfect agreement between the two ratings, while that of − 1 implies perfect disagreement; if κ assumes the value 0, then this implies that there is no relationship between the two ratings, and any agreement or disagreement is random).Univariate and multivariate binary logistic regression was applied to evaluate the impact of the RT fractions on binary skin diagnostic outcomes.In turn, to estimate the influence of the collected risk factors on the observed dermoscopic features, a multivariate ordinal logistic model was used.The statistical outcomes were expressed by a classical odds ratio (OR) with a 95% confidence interval (CI); a p value of < 0.05 was considered statistically significant.Due to repeated measures with consecutive RT fractions for each patient, the regressions were extended for random effects.The statistical outcomes were expressed by a classical odds ratio (OR) together with a 95% confidence interval (CI 95%) and a p value.

Ethical approval
The authors have received approval from the local ethics committee of the National Research Institute of Oncology (reference number KB/430-44/19).The study was conducted in accordance with the Helsinki Declaration of 1964, and its later amendments.All subjects provided informed consent to participate in the study as well as for publication.

Results
There were oral cavity carcinoma, oropharyngeal carcinoma, hypopharyngeal carcinoma, laryngeal carcinoma nasopharyngeal carcinoma and neck lymph nodes tumor as a metastatic cancer from unknown primary in 1, 8, 3, 10, 3 and 1 patients, respectively.There were five women and 21 men with the mean age of 60.5 years (range 34-74) in this group.All patients (26) observed during the course of RT developed ARD.The highest noted grade according to RTOG/EORTC, at the end of the RT treatment, was grade II in 14 patients, grade III in 10 patients, and the remaining two developed grade IV ARD.Grade I was observed in the first week (on average on Day 4.69) (Fig. 1A), grade II in the third week of the follow-up (Day 20.69) (Fig. 1C), grade III in the 6th week of the followup (Day 37.81) (Fig. 1E), and grade IV in the 5th week of the follow-up (Day 34.66) (Fig. 1G).The percentage occurrence of dermoscopic features depending on the grade of radiation-induced dermatitis per RTOG/EORTC 13 is presented in Table 2 and Fig. 1B,D,F,H (Table 2).
Summary of dermoscopic findings: vessels in each grade of ARD were polymorphic.The arrangement of the vessels was also heterogeneous, and there was no typical arrangement for a particular grade of ARD.In healthy skin, we did not observe vessels in reticular distribution, but their presence was detected in every degree of ARD.Unspecific distribution was more common in healthy skin than in ARD.In each grade of ARD, a patchy scale was observed and the frequency of scale occurrence increased with the grade of development according to RTOG without characteristic color was observed.However, the incidence of scale increases with the degree of development in RTOG (Fig. 1H).Moreover, a feature present in all grades but not observed in healthy skin was follicular plugs arranged in rosettes.
Statistically significant results are underlined in bold in Table 3.A relationship between the observed dermoscopic and clinical features was checked using κ coefficient (Table 3).
The agreement between dermoscopic and clinical features was 0.03-0.54and bright erythema, epilation, dry and moist desquamation, moderate edema, and dermoscopic features such as vessels in reticular distribution, white, yellow, brown scale and patchy scale distribution, follicular plugs arranged in rosettes and perifollicular pigmentation.Negative results mean incompatibility: when a given macroscopic feature is present, the dermoscopic feature is not present.In the next step, dermoscopic and clinical features were analyzed in terms of the influence of time, age, gender, induction chemotherapy, concurrent chemotherapy, total radiation dose, fractional dose, tumor location, as well as the histopathological diagnosis during the whole RT treatment on the skin diagnostic outcomes using logistic regression.The statistically significant relationships between clinical features and possible ARD risk factors-time, age, gender, indCHT, concurrent CHT, and fractional dose-are expressed by odds ratios reported in Table 4 whereas OR is a measure of association between radiation exposure and a clinical outcome; OR > 1 indicates the increased occurrence of any event, while OR < 1 a protective exposure) (Table 3).
Based on the results in Table 4, we observed the relationship between the presence of vessels in reticular distribution and time, age, gender, induction chemotherapy, and concurrent CHT (Table 4).The statistical interpretation of the OR (univariate regression) may be as follows: 1 day of observation generates an increased risk of vessels in reticular distribution by 8%, and 5 days of observation (1.08 5 = 1.47), so by almost one and a half.A 10-year difference in the age of patients generates a (1-0.97 10 ) × 100% = 26% reduction in the occurrence of vessels in reticular distribution.The risk of vessels in reticular distribution is 64% lower in men than in women.Induction chemotherapy reduces the risk of vessels in reticular distribution almost three times (OR = 2.94).Concurrent CHT reduces the risk of vessels in reticular distribution by 1.83 (OR = 1.83).The results regarding the effect of collected risk factors on skin reaction in a multivariate model showed that the effect of gender and induction chemotherapy increased.Moreover, in the multivariate model, the lack of concurrent CHT reduces the risk of vessels in reticular distribution by 53% (see the right panel of Table 4).Other results in the table should be interpreted analogously.Considering individual factors affecting clinical response, each day of observation during RT treatment statistically generates a higher chance of occurrence of vessels in reticular distribution (Figs.1D,F,H, 2A,E), white scale and yellow scale (Fig. 2B), and brown scale (Fig. 2C) with patchy distribution (Figs.1B,H, 2C), perifollicular pigmentation (Fig. 2D), follicular plugs arranged in rosettes (Fig. 2E), while the chance of unspecific distribution of vessels decreases (Fig. 2D).In the context of a macroscopic response, each day of observation during RT treatment statistically generates a higher chance of occurrence of bright erythema (Fig. 1C,E), epilation (Fig. 1C,E), dry and moist desquamation (Fig. 1C,E), moderate (Fig. 1C) and pitting edema (Fig. 1E), and ulceration (Fig. 1G) while the chance of follicular and faint erythema decreases (Fig. 1A).The results regarding the effect of collected risk factors on skin reaction in a multivariate model were comparable.In a univariate analysis, age was a significant factor for vessels in reticular distribution, vessels in unspecific distribution, follicular plugs arranged in rosettes, and perifollicular pigmentation as well as in the group of macroscopic features for follicular erythema.However, multivariate analysis did not show this relationship for the unspecific distribution of vessels, follicular plugs arranged in rosettes, or perifollicular pigmentation (the association is on the border of statistical significance, i.e., p < 0.1).Gender is important for the occurrence of vessels in reticular distribution, vessels in unspecific distribution, white, yellow, patchy scale, follicular plugs arranged in rosettes, www.nature.com/scientificreports/and for macroscopic features for faint and bright erythema and dry desquamation.Multivariate analysis showed that the gender effect was stronger in each case.The risk of vessels in reticular distribution, white scale, yellow scale, patchy scale, and follicular plugs arranged in rosettes is 85%, 89%, 80%, 95%, and 69% lower in men than in women, respectively.The risk of faint erythema is 118% higher for men than women, while the chance of bright erythema and dry desquamation is 88% and 74% lower in men than in women, respectively.Induction chemotherapy increases the risk of yellow scale and reduces the risk of vessels in reticular distribution.In the multivariate model, these dependencies increase, and we observe that induction chemotherapy increases the risk of yellow scale occurrence by two-thirds (OR = 0.34) and reduces the risk of vessels in reticular distribution almost six times (OR = 5.90).Concurrent chemotherapy is important for the occurrence of vessels in reticular distribution, follicular plugs arranged in rosettes, perifollicular pigmentation and macroscopic features such as follicular erythema, tender erythema, and moist desquamation.In turn, multivariate analysis did not show this relationship for follicular plugs arranged in rosettes, follicular erythema and moist desquamation (the association is on the border of the statistical significance, i.e., p < 0.1).Non-concurrent chemotherapy reduces the risk of vessels in reticular distribution, perifollicular pigmentation and tender erythema by 53%, 51% and 94%, respectively.

Discussion
Graham et al. emphasized the importance of archiving photographs, which are a useful source of documents for auditing and monitoring radiotherapy-induced skin toxicity 18 .In turn, the study by Ni et al. used deep learningbased method for the automatic assessment of radiation-induced dermatitis in patients with nasopharyngeal carcinoma 19 .In our study, 2244 dermoscopic photographs and 374 clinical photographs were archived, creating www.nature.com/scientificreports/ a database that in the future could be used as a database to automate clinical assessment.In the current literature, only one study used dermoscopy, but only for the presence of erythema in ARD 20 .So far, only clinical features have been described, and there are no data on the analysis of dermoscopic features in ARD.One of the previous studies reported dermoscopic changes in the surrounding tissue of basal cell carcinoma in patients who underwent brachytherapy 21 .Radiation-induced dermatitis occurs in about 90-95% of patients exposed to ionizing radiation [22][23][24] .Published reports on the share of individual grades per RTOG are ambiguous.This is probably due to many variables affecting the development of this type of skin toxicity.Elliot et al. showed in their observation that 1% of patients did not develop any grade of ARD, 20% developed grade I, 57% grade II, and 23% grade III or IV 25 .Kang et al. observed radiation-induced dermatitis of the maximum grade I-IV in 46.6%, 18.0%, 5.5%, and 0.9% of the patients, respectively 26 .In turn, in the report from Franco et al., the toxicity profile at the end of RT was Grade 0 in 3.5% of patients, Grade I in 32%, Grade II in 61%, Grade III in 3.5% 27 .Mild erythema may appear as early as a few hours after exposure to ionizing radiation 28 , but usually develops about 7-10 days after starting therapy 29 .Dry desquamation (RTOG/EORTC grade I) usually occurs after 3-4 weeks from the start of treatment.More intense erythema, hair loss, and hyperpigmentation are usually observed between 2 and 4 weeks of therapy 30 .Moist desquamation (RTOG/EORTC grade II) usually occurs after 4 weeks when the total RT dose to the skin is 40 Gy or higher 31,32 .In the study of Franco et al., grade II appeared between treatment weeks 4-5; for those having grade III acute skin toxicity, this event mainly began during weeks 5 and 6 27 .Data variability is also likely to be influenced by treatment and clinical risk factors.ARD can lead to pain, discomfort, reduced quality of life, and premature discontinuation of treatment.Therefore, it is important to make a rapid diagnosis when the first symptoms appear and to implement appropriate prevention and treatment.Dermoscopy can be a complementary tool to support macroscopic ARD evaluation.Our study is the first in the published papers to attempt to identify the correlations between the clinical and dermoscopic features of ARD with its dermoscopic follow-up.The importance of the total dose during RT is well known 33,34 .Moreover, in our study, we selected patients scheduled for RT at comparable total doses to minimize the risk of a dose effect.A statistical dependence of the influence of days of observation during RT was observed for the features correlating in the test of compatibility of clinical and dermoscopic features.Predicting the risk of radiation-induced dermatitis is essential for proper prevention and treatment.Kawamura et al. in their study created a scoring system taking into account V60Gy, concurrent chemotherapy status, age, and body mass index 35 .Age ≥ 67 years was significant in their study for the development of ARD.Meyer et al. showed that gender is important in the context of the development of radiation-induced dermatitis 33 .Kawamura et al. showed that concurrent chemotherapy with platinum and cetuximab (cetuximab > platinum) had significant importance in the development of radiation-induced dermatitis.Gold standards of management have not yet been established, and treatment, as well as prevention, are common and empirical, based on personal experience supported by weak scientific evidence 9,36 .Based on the study by Robijn et al., there could be a strong recommendation to use photobiomodulation therapy (PBMT) in the prevention and management of ARD in cancer patients 37 .The identified dermoscopic features may facilitate the selection of topical Table 3. Level of agreement between the presence of selected dermoscopic features 17 and clinical features 13

Figure 1 .
Figure 1.Macroscopic images (A,C,E,G) of ARD in grades (G) from G1 to G4, clinically assessed in line with RTOG criteria 13 and dermoscopic findings (B,D,F,H) described in line with the consensus of experts in nonneoplastic dermatoses on behalf of the International Dermoscopy Society in one of the patients observed during the course of the RT treatment.(A) Faint erythema (G1); (B) dermoscopic image (G1) of ARD reveals linear branched and linear curved vessels in clustered distribution and white structureless areas; (C) bright erythema, epilation, moist desquamation and moderate edema (G2); (D) dermoscopic image of ARD (G2) shows linear branched and linear curved vessels in reticular distribution of vessels, and follicular plugs arranged in rosettes; (E) bright erythema, epilation, confluent moist desquamation and pitting edema (G3); (F) dermoscopic image (G3) with linear branched vessels in reticular distribution, perifollicular pigmentation and follicular plugs arranged in rosettes; (G) ulceration in ARD (G4); (H) dermoscopic image (G4) reveals linear branched vessels in reticular distribution, white, yellow, patchy scale.

Figure 2 .
Figure 2. Dermoscopic images of ARD described in line with the consensus of experts in non-neoplastic dermatoses on behalf of the International Dermoscopy Society by Errichetti et al.17.(A) Dermoscopic image of ARD reveals linear branched, linear curved vessels with reticular distribution and perifollicular white color; (B) dermoscopic image of acute radiodermatitis reveals white and yellow, patchy scale; (C) dermoscopic image reveals linear branched and linear curved vessels with clustered distribution, brown, patchy scale; (D) dermoscopic image reveals linear branched and linear curved vessels with unspecific distribution, and perifollicular pigmentation; (E) linear branched and linear curved vessels with reticular distribution and follicular plugs arranged in rosettes.

Table 1 .
Clinical characteristics, location of the tumor, and histopathological type of the group of observed patients.

features Clinical features Faint erythema Bright erythema Epilation Dry desquamation Moist desquamation Moderate edema Pitting edema Ulceration
in ARD assessed with values of κ statistics.

Table 4 .
Statistically significant ORs (p < 0.05) of the influence of clinical data on the occurrence of dermoscopic and macroscopic features in ARD (univariate and multivariate logistic regression).Significant values are in [bold].