Acute Radiation Dermatitis Evaluation with Reflectance Confocal Microscopy: A Prospective Study

Background: During radiotherapy (RT), most breast cancer patients experience ionizing radiation (IR)-induced skin injury—acute radiation dermatitis (ARD). The severity of ARD is determined by a physician according to CTCAE or RTOG scales, which are subjective. Reflectance confocal microscopy (RCM) is a noninvasive skin imaging technique offering cellular resolution. Digital dermoscopy (DD) performed in conjugation with RCM can provide more information regarding skin toxicity. The purpose of this study is to create an RCM and DD features-based ARD assessment scale, to assess the association with CTCAE scale and possible predictive value. Methods: One hundred and three breast cancer patients during RT were recruited; every week, clinical symptoms of ARD (CTCAE scale) were evaluated and RCM, together with digital dermoscopy (DD), was performed. Results: According to RCM; after 2 RT weeks, exocytosis and/or spongiosis were present in 94% of patients; after 3 weeks, mild contrast cells (MMCs) were detected in 45%; disarrayed epidermis (DE) was present in 66% of patients after 4 weeks and in 93% after 5 weeks; abnormal dermal papillae (ADP) were present in 68% of patients after 5 weeks. The coefficients of RCM features (RCMcoef) alone and together with dermoscopically determined erythema (RCM-ERYcoef) were significantly associated with ARD severity grade. RCMcoef is a significant predictive factor for the clinical manifestation of ARD. Conclusions: RCM features of irradiated skin appear earlier than clinical symptoms, have a characteristic course, and allow the severity of ARD to be predicted.


Introduction
Whole breast radiotherapy (WBRT) after breast-conserving surgery (BCS) is the standard treatment for patients with breast cancer [1]. During RT, most breast cancer patients experience radiation-induced skin injury-acute radiation dermatitis (ARD)-and up to 30% of them suffer from a moderate or severe form of this side effect [2].
One of the most important risk factors for the severity of ARD is the IR dose to the skin. Definite erythema appears when a dose of more than 12 grays (Gy) of IR is delivered to the skin (2-3 weeks of RT). Dry desquamation can be diagnosed after 20 Gy (3-4 weeks of RT), and moist desquamation after a 30 Gy dose (>4 weeks of RT) [3].
In everyday practice, the severity of ARD is determined by a physician according to Common Terminology Criteria for Adverse Events (CTCAE) or Radiation Therapy Oncology Group (RTOG) scales. These scales are used to assess symptoms such as skin erythema, dry and moist desquamation, edema, bleeding, necrosis, and ulceration. The same scales are used in clinical studies investigating the efficiency of topical agents for ARD

Digital Dermoscopy
Dermoscopic images were taken before every RCM imaging with a VivaCam ® digital dermoscope (MAVIG GmbH, Munich, Germany) that uses a 2 MP CCD camera. These full-HD resolution images were used to assess the erythema (ERY) of the skin. A total of three radiation oncologists separately evaluated skin ERY according to a scale from 0 to 2, where 0 means "no erythema", 1 means "faint erythema", and 2 means "severe erythema". The evaluations were combined, and the final grade was determined when at least two evaluators agreed.

Assessment of Skin Lesions with RCM
RCM images were taken with the commercially available VivaScope ® 1500 (MAVIG GmbH, Munich, Germany) medical device, which uses 830 nm wavelength, a 20 mW (max power) laser diode for tissue illumination, and a 30× magnification 0.9 NA water immersion objective for laser beam focusing onto the tissue and reflected light collection. The VivaScope ® 1500 provides a lateral resolution of 1.25 µm and an axial resolution of 5.0 µm. The size of individual images is 500 × 500 µm, with a frame rate of 9 images per second. The maximum mapped field is 8 × 8 mm 2 , the image resolution is 1024 × 1024 pixels, and the resultant magnification is 520×.
In the beginning of the imaging session, firstly, mosaic confocal reflectance images of 4 mm by 4 mm in the spinose-granular skin layer were taken. After that, 500 µm by 500 µm images every 5 µm up to 100 µm of depth were taken at three selected points. All of the images were evaluated by two observers. A lesion was graded as 1 if it was present in less than 50% of images and 2 if in more than 50%.

Statistical Analysis
Statistical analysis was used to describe characteristics of patients to compare different patient groups. For the descriptive statistics, scale variables were described by median and standard deviation and categorical variables by frequency of distribution. Ordinal logistic regression with sorted cases by subject and within-subject variables, and a Chisquare (X2) test, were used to determine the association between variables with similar radiotherapy fractions. Binary logistic regression was used to evaluate the association between variables with different radiotherapy fractions. The differences were considered statistically significant if a p-value was less than 0.05. Statistical analysis was performed using IBM-SPSS Statistics 21 (SPSS, Inc., Chicago, IL, USA).

Results
After the exclusion of a patient with missing RCM imaging, a total of 103 subjects were included in the further analysis, except for a few cases when moist desquamation appeared at the end of RT and RCM was discontinued.
All of the patients were females from 28 to 76 years old, with an average age of 57.1 years ± 8.97 (CI: 55. 35-58.85). Study patient characteristics are shown in Table 2.

Clinical ARD Assessment
Clinical ARD assessment according to the CTCAE grading scale is presented in Table 3. The patients in this study experienced the same clinical manifestation of ARD (Table 3) as other patients during the whole breast 3D-RT [32].

Dermoscopic Skin Imaging
Before RT, the control measurement of DD was performed on all patients. During treatment, dermoscopic images were taken after every 5 fx. Consecutive DD images revealed the progression of skin ERY during RT, which is depicted in Figure 1.
Clinical ARD assessment according to the CTCAE grading scale is presented in Table 3. The patients in this study experienced the same clinical manifestation of ARD (Table  3) as other patients during the whole breast 3D-RT [32].

Dermoscopic Skin Imaging
Before RT, the control measurement of DD was performed on all patients. During treatment, dermoscopic images were taken after every 5 fx. Consecutive DD images revealed the progression of skin ERY during RT, which is depicted in Figure 1. The ERY in the DD images was evaluated by three radiation oncologists according to a scale from 0 to 2, where 0 means "no ERY", 1 means "faint ERY", and 2 means "severe ERY". In the left upper corner of Figure 1, the dermoscopic view of the patient's skin before RT (0 fx) is demonstrated with no visual signs of ERY. All subsequent measurements were compared to the control. The example of the dermoscopic view of the faint ERY in a patient's skin after 5 fx and 10 fx is demonstrated in the middle and right upper corner of Figure 1, and severe ERY after 15 fx, 20 fx, and 25 fx in the bottom of Figure 1. The results of skin ERY evaluations from all patients' DD images are presented in Table 4. The ERY in the DD images was evaluated by three radiation oncologists according to a scale from 0 to 2, where 0 means "no ERY", 1 means "faint ERY", and 2 means "severe ERY". In the left upper corner of Figure 1, the dermoscopic view of the patient's skin before RT (0 fx) is demonstrated with no visual signs of ERY. All subsequent measurements were compared to the control. The example of the dermoscopic view of the faint ERY in a patient's skin after 5 fx and 10 fx is demonstrated in the middle and right upper corner of Figure 1, and severe ERY after 15 fx, 20 fx, and 25 fx in the bottom of Figure 1. The results of skin ERY evaluations from all patients' DD images are presented in Table 4. DD-digital dermoscopy; ERY-erythema; deg-degree of erythema in DD imaged, where 0 means "no erythema", 1 means "faint erythema", and 2 means "severe erythema"; fx-radiotherapy fractions; NPnot performed.
Ordinal logistic regression showed that a higher degree of skin ERY, determined with DD, was associated (p < 0.0001) with a higher degree of ARD according to CTCAE scale.
In early fx (5 and 10), ERY using DD were diagnosed more frequently than according to the clinical evaluation (Table 3).

Skin Lesion Assessment with RCM
Before RT, the control measurement of RCM was performed on all patients. The representative view of the normal honeycomb pattern of epidermis before RT is shown in Figure 2a and the normal view of dermal papillae at the level of the dermo-epidermal junction (DEJ) in Figure 2b. All subsequent measurements were compared to the control.    The analysis of the first ten patients [33] revealed that spongiosis, exocytosis, and mild contrast cells (MCCs) are the early signs of radiation-induced skin injury, while disarrayed epidermis (DE) and abnormal dermal papillae (ADP) appear at the end of RT.
After 5 fx, the aggregates of round-to-polygonal cells corresponding to exocytosis ( Figure 2c) were detected in 55.3% of all patients, but there were no alterations in the DEJ (Figure 2d). After 10 fx, an additional to exocytosis, the darker area relative to the surrounding epithelium of the stratum spinosum with intercellular spaces between larger than normal keratinocytes corresponding to spongiosis is demonstrated in Figure 2e, but still no alterations in the DEJ (Figure 2f). After 15 fx, the first signs of disarrayed epidermis (DE) were present in some patients' RCM images (Figure 2g). Simultaneously, at the level of DEJ a single, aggregates or diffuse round-to-polygonal, mildly refractive cells (MCCs) (Figure 2h) corresponding to inflammatory cells were detected. Exocytosis and spongiosis are the earliest and nonspecific symptoms characteristic for other inflammatory conditions as well [17], while MMCs are considered as the "hallmark" of IR-induced skin injury [3]. These three RCM symptoms do not cause structural damage of the skin. The harm to skin structure can be diagnosed when the normal honeycomb-like architecture at the level of the stratum spinosum is lost (disarrayed epidermis (DE)) or when bright papillary rims at the DEJ are absent (abnormal dermal papillae (ADP)). After 20 fx, broad DE (Figure 2i) and MMCs in the level of the DEJ (Figure 2j) were present. After 25 fx, extensive DE ( Figure 2k) and ADP at the level of the DEJ (Figure 2l) could be detected.
The manifestations of RCM features of all study patients are presented in Table 5. RCM-reflectance confocal microscopy; Int.-intensity of RCM features, where 1 means that the symptom is present in less than 50% of an image, and 2 in more than 50%; fx-fractions of radiotherapy; MMC-mild contrast cell; DE-disarrayed epidermis; ADP-abnormal dermal papillae; NP-not performed.
By comparing these results with the degree of ARD (Table 3), we can clearly see that RCM features appear earlier.

Clinical ARD Association with RCM and DD
In order to compare the microscopic signs of radiation-induced skin injuries (spongiosis, exocytosis, MMCs, DE, and ADP) with clinical ARD evaluation (CTCAE scale), the coefficient of RCM features was calculated (RCM coeff ). Firstly, the sum of the RCM (RCM sum ) features was calculated according to Formula (1) as follows: Here, spongiosis and exocytosis were counted as one symptom when only the maximum value of one of these symptoms was included in further calculations (S,Emax). The values of other symptoms (MMCs, DE, and ADP) were taken from Table 5.
RCM sum was combined with RCM coeff , according to symptom acuity. Nonspecific inflammatory changes such as low-grade spongiosis or exocytosis were assigned an RCM coeff value of 0. RCM symptoms that were reversible and corresponded to non-structural changes (such as higher grade spongiosis or exocytosis and MMCs) were assigned an RCM coeff value of 1. High-grade symptoms causing damage to the skin structure were assigned an RCM coeff value of 2. RCM coeff values depending on RCM sum are presented in Table 6. RCM coef values of all study patients are presented in Table 7. Ordinal logistic regression revealed that a higher RCM coeff is associated (p < 0.0001) with a higher degree of ARD, according to the CTCAE scale.
In order to compare the combination of RCM (spongiosis, exocytosis, MMCs, DE and ADP) and DD (ERY) with clinical ARD manifestation (CTCAE scale), the coefficient of RCM and DD features was calculated (RCM-ERY coeff ). Firstly, the sum of the RCM and DD (RCM-ERY sum ) features was calculated according to Formula (2) as follows: RCM-ERY sum = RCM sum + ERY (2) Here, RCM sum is calculated as mentioned above (Formula (1)), and ERY is the degree of skin erythema according to DD from Table 5.
RCM-ERY coef values of all of the study patients are presented in Table 9. Ordinal logistic regression showed that a higher RCM-ERY coeff is associated (p < 0.0001) with higher degree of ARD, according to the CTCAE scale.
A X 2 test was used in order to compare clinical ARD assessment (CTCAE scale) with ERY, RCM coeff , and RCM-DD coeff when measurements were performed after the same number of fx (Table 10).  After 20 RT fractions, the levels of ARD clinical grade (CTCAE scale) were significantly associated with the levels of both RCM coeff (p = 0.038) and RCM-ERY coeff (p = 0.008), as the frequency of the higher subgroup of CTCAE increased with the higher subgroup of RCM coeff or RCM-ERY coeff , while after 25 fx, the ARD clinical grade (CTCAE scale) was significantly associated with the values of both RCM coeff (p = 0.01) and ERY (p = 0.038).
The next question was whether RCM and DD features, determined in the beginning of RT (from 10 fx to 15 fx), could significantly distinguish CTCAE grade 1 from grade 2 determined at the end of RT (20 fx and 25 fx). For the same patient, an RCM coeff value of 1 after 10 fx and 15 fx had a significantly higher odds ratio of grade 2 in the CTCAE scale after 20 fx compared to grade 1 (Table 11). Similarly, for the same patient, an RCM coeff value of 1, determined after 15 fx, had a significantly higher odds ratio of grade 2 according to the CTCAE scale after 25 fx compared to grade 1 (Table 12). In addition, the AUC estimates for RCM coeff after 10 fx and 15 fx were also significant regarding CTCAE after 20 fx and 25 fx (Tables 9 and 10). Therefore, the results of this analysis showed that RCM coeff , determined at the beginning of RT, significantly distinguished between grades 1 and 2 of the CTCAE scale at the end of RT and could be considered as a promising predictive factor.

Discussion
ARD is a relevant clinical problem, but until now the efficacy of topical treatment has not been proven in large, randomized studies [2]. There is a lack of new methods that could improve the diagnostics of ARD and accelerate the development of new topical agents for the prevention and treatment of this side effect.
The aim of this study was to create an RCM and DD features-based ARD assessment scale, to assess the association with CTCAE scale and possible predictive value. For this purpose, we recruited 103 breast cancer patients during RT and every week we evaluated the clinical symptoms of ARD (CTCAE scale) and performed RCM. Before every RCM, the full-HD image of the same skin region was taken with DD. Given that RCM can be difficult to apply in everyday practice, we also analyzed DD images ( Figure 1) and compared them with clinical ARD manifestation (CTCAE scale).
To our knowledge, this is the first prospective clinical study with more than 100 patients analyzing radiation-induced skin injuries with RCM and DD.
The first and, so far, the only study analyzing dynamic skin changes during RT using RCM was performed by Vano-Galvan and colleagues [27]. In our study, we investigated 103 patients and performed RCM once a week, while the Vano-Galvan study was conducted with 6 patients and RCM imaging every three weeks.
In our investigation, we performed more than 600 imaging sessions with RCM and DD. We found that by using RCM, the first symptoms of radiation-induced skin injury could be detected very early. After 5 fx, exocytosis or spongiosis was detected for 65% of all patients, and after 10 fx, 94% of patients experienced IR-induced skin injury (Table 5). During clinical evaluation, however, an ARD degree of 1 was diagnosed for 13% of patients after 5 fx, and after 10 fx only 40% of all patients were diagnosed with an ARD degree of 1 or 2 ( Table 3).
According to RCM (Table 5 and Figure 2), radiation-induced skin changes have a typical course determined with RCM. After 2 weeks of RT (10 fx), exocytosis and/or spongiosis was detected for 94% of patients; after 3 weeks (15 fx), inflammatory cells (MMCs) were detected in 45% of patients, and this number increased until the end of RT; after 4 weeks (20 fx), DE was detected in 66% of patients and increased to 93% at the end of RT; after 5 weeks (25 fx), ADP was detected in 68% of patients.
Radiation-associated epidermal atypia was the main finding in RCM for basal cell carcinoma patients when the coetaneous response to RT was imaged in vivo [41]. The authors of this study suggested that epidermal atypia is likely a sign of ARD, which corresponds well with the findings of our study showing that at the end of RT, the majority of patients had pronounced DE.
For the more detailed comparison of RCM and DD symptoms with the clinical assessment of ARD, we created coefficients based on RCM and DD imaging results.
Ordinal logistic regression showed that the higher degree of ARD, according to the CTCAE scale, is associated with higher values of RCM coeff , ERY, and RCM-ERY coeff (Table 10). Comparing measurements after the same number of fx, the strongest association between RCM or DD features and the CTCAE scale was found at the end of RT. After 4 RT weeks, RCM coeff and RCM-ERY coeff were significantly associated with the CTCAE scale grade, while after 5 weeks, similar results showed RCM coeff and ERY (Table 10).
The final step was to analyze RCM and DD features as possible predictive factors for the severity of ARD determined by the CTCAE scale. According to binary logistic regression, RCM coeff determined in the beginning of RT (from 10 fx to 15 fx) showed significant predictive value to distinguish the CTCAE grade 1 from grade 2 determined in the end (20 fx and 25 fx) of RT (Tables 11 and 12).
These results are promising for integrating RCM into clinical studies or in everyday practice for the early detection of IR-induced skin changes and severity prediction. However, performing RCM requires time, effort, and a qualified specialist, which could be major limitations for applying it in clinical practice. On the other hand, DD is much easier to perform and interpret. Therefore, DD images could be analyzed by a physician and RCM by employing telediagnostics [42].
One of the possible limitations of this study is that all of the patients were Caucasian women. Some studies show that people with white skin have more severe clinical reactions to irritants compared to those with black skin [43]. A similar study with black skin patients should be conducted.
As mentioned above, there is a lack of clinical studies focusing on RCM application for skin changes during RT, but there are similar investigations with laboratory animals. For example, MPM used for early-stage radiation skin injury in a mouse model showed that after IR exposure, epidermal cells and intercellular spaces became large and irregular [44]. These symptoms correspond well with exocytosis and spongiosis in our study.
It was demonstrated that OCT angiography could non-invasively detect IR-induced changes in mouse skin, such as skin thickening, the dilation of large blood vessels, and irregularity in vessel boundaries [45].
Based on the results of this study, we propose RCM and DD as a sensitive tool for evaluation of IR-induced skin changes. The next step should be to analyze the benefits of RCM and DD in randomized clinical trials investigating the efficiency of topical agents for ARD prevention or treatment.

Conclusions
RCM reveals radiation-induced skin changes at the cellular level with no harm for the patient and could improve the assessment of skin condition during RT. RCM features of irradiated skin appear earlier than clinical symptoms, have a characteristic course, and allow for a prediction of the severity of ARD. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: All data will be made available upon reasonable request to the corresponding author.