Dendritic cells in reflectance confocal microscopy are a clue for early melanoma diagnosis in extrafacial flat pigmented melanocytic lesions

Abstract Differential diagnosis of extrafacial flat pigmented lesions with dermoscopic reticular and/or homogeneous pattern is challenging. Dendritic cells upon reflectance confocal microscopy (RCM) still represent a pitfall. This study aims to determine the role of dendritic cells upon RCM in the epidermis and dermo‐epidermal junction (DEJ), together with common RCM features for melanoma and nevi, in dermoscopically equivocal extrafacial flat pigmented lesions. A retrospective evaluation of RCM images of melanocytic extrafacial flat pigmented lesions with reticular and/or homogeneous dermoscopic pattern and with histopathological diagnosis, was performed. A multivariate model of RCM features was used to obtain a score of independent risk factors. A total of 698 lesions were included. Increasing patient age, epidermal dendritic cells, many dendritic cells in the DEJ (>30%) and many (>5/mm2) round atypical cells were independent risk factors for melanoma. Edged papillae and melanophages were indicative of nevus. A score based on these features was developed to assist in melanoma differential diagnosis. The RCM observation of abundant (>30%) dendritic cells in the DEJ is highly suggestive of malignity. This independent risk factor should also be considered for improved differential diagnosis of extrafacial melanoma.


| INTRODUC TI ON
Dermoscopy improves skin cancer detection sensitivity and reduces benign/malignant ratio of excised lesions. 1,2 Reflectance confocal microscopy (RCM) can further improve skin lesion diagnostic accuracy with in vivo visualization of the epidermis and superficial dermis in real time correlating well with dermoscopic and histopathologic findings. 3 However, RCM is limited by the lack of nuclear staining, limited imaging depth, difficulty in visualizing nodular lesions and distinguishing dendritic melanocytes in pagetoid pattern from Langerhans cells, which can occasionally simulate pagetoid spread. 4 Therefore, RCM evaluators need to consider all cellular and architectural lesion characteristics to improve agreement with final histopathologic diagnosis. 5 Common differential diagnostic RCM features have been previously published. Melanoma is typically observed with a disarranged honeycomb pattern and bright-nucleated cells in a pagetoid spread in the suprabasal layers and non-edged dermal papillae, with atypical melanocytes and/or foci with loss of the dermal papillae at the DEJ. 4,6,7 The presence of atypical dendritic cells, in particular infiltrating the hair follicle (folliculotropism) at the DEJ, has been proven to be a highly specific RCM pattern for facial lentigo maligna (LM)/lentigo maligna melanoma (LMM) diagnosis. [8][9][10] For extrafacial lesions, melanoma was described as characterized by pagetoid infiltration of round cells and/or dendritic cells and the focal proliferation of dendritic pagetoid cells in the epidermis in selected lesions on chronically sun-damaged skin, 11 and the presence of junctional cytological atypia (where roundish and dendritic cells were considered together) was the strongest RCM predictive factor for in situ melanoma. 12 Differential diagnosis between atypical melanocytes and Langerhans cells with cell morphology visualized at RCM alone cannot be achieved. However, Segura et al. identified dendritic cells in pigmented basal cell carcinomas and, with the aid of immunohistochemistry, concluded that dendritic cells in the tumoral basaloid nests correspond to melanocytes, whereas dendritic cells in the epidermis correspond to Langerhans cells. 13 Therefore, we hypothesize that cell distribution, depth and architecture may assist in RCM differential diagnosis.
The aim of the current study was to correlate common RCM patterns of the epidermis and DEJ, (dendritic and roundish cells considered separately) and their distribution, depth and architecture in extrafacial melanocytic lesions with common dermoscopic presentation (i.e. reticular and/or homogeneous pattern) and without dermoscopic melanoma specific clues, diagnosed nevi or melanoma.
The combination of features indicative of melanoma and the subsequent development of a score to assist in differential diagnosis is the secondary aim of the study.

| Study data set
We performed a retrospective analysis of consecutive extrafa-

| Image acquisition and analysis
Standardized polarized dermoscopic clinical images were obtained with DermLite Photo (3Gen) mounted on a Canon G16 camera. In vivo RCM images (Vivascope 1500; Mavig GmbH) were captured according to a standardized procedure previously described. 19 RCM mosaic images were obtained at the suprabasal epidermis (spinous and granular layers), DEJ and papillary dermis. All three RCM mosaic images/lesion were evaluated by a single clinician for the presence of RCM parameters, published previously, see Table S1. 6,[19][20][21][22][23] Percentage of presence of selected features were calculated by counting the squares of the mosaic block when observed. The reader was blinded to final histopathological diagnosis.
Biopsy specimens were analysed by a dermatopathologist following fixation in formaldehyde, embedding in paraffin, sectioning and staining with haematoxylin-eosin.

| Confocal dendritic cells-index: A predictive score for melanoma diagnosis
Based on prognostic factors identified with logistic regression, a score for melanoma diagnosis probability was devised. Briefly, each prognostic variable is assigned a score (0-10). For simplicity, age ranges were created and dendritic cells at the DEJ combined both absent (0%) and <10%. Total confocal dendritic cells-index (CDC-I) scores (0-52) correspond to melanoma diagnosis probabilities. Logistic regression model (stepwise forward selection) was used for association between parameters and to identify prognostic factors. Intercept-only model was fitted and individual score statistics were evaluated (p < 0.05), removing insignificant variables before adding variables. Data were expressed as odds ratio (OR), 95% confidence interval (CI). p < 0.05 was considered statistically significant.

| Statistical analysis
A nomogram for predicting melanoma probability (including univariate and multivariate logistic regression analyses) screened for fit predictors. Nomogram predictability was assessed with area under the curve (AUC), by receiver operating characteristic (ROC) analysis.   Retrospective RCM image analysis revealed dendritic cells in the epidermis in over two thirds of the lesions. According to histopathological diagnosis, dendritic cells were observed in over 80% of the in situ and invasive melanomas, whilst they were less frequently observed in junctional and compound nevi. Almost all in situ and invasive melanomas had atypia in the epidermis, and almost all lesions with a regular epidermis were associated with compound and junctional nevi diagnoses (p < 0.001), see Table 1. Overall, most lesions did not have atypical round and /or oval cells (72.2%), and any eventual presence and higher density was mostly associated with in situ or invasive melanomas (p < 0.001).

| Confocal dendritic cells-index
Confocal dendritic cells-index (CDC-I), including predictive variables identified with regression analysis, was developed from the Nomogram analysis. A worksheet ( Figure 1) enables clinicians to calculate a personalized CDC-I score, corresponding to a probability of a melanoma diagnosis demonstrates the application of CDC-I in two lesion examples (Figures 2,3). The CDC-I predictive accuracy was high (AUC = 0.84).

| DISCUSS ION
This large study confirms common RCM features for differential diagnosis between melanomas and nevi for extrafacial lesions 4,6 and more recent findings associating dendritic cells in the epidermis of chronic sun-damaged skin with melanoma. 11 The presence of abundant dendritic cells in the DEJ is a differential diagnostic feature.
Pellacani et al., hypothesized that dendritic cells could be the RCM hallmark of slow-growing melanomas. 7 In facial LM studies, tangled lines are reported as a specific feature, 8 The current study is limited by a retrospective design, image evaluation by a single clinician only, the lack of inter-personal evaluator agreement and lesion selection bias, excluding unequivocal nevi.
Analysis did not include assessment of the features according to melanoma invasiveness, but could be considered in future studies. Further, the lack of immunohistochemistry analysis does not enable a clear distinction in this study between melanocytes and Langerhans' cells. 27 The authors recommend lesion excision if: • Dendritic cells/ tangled lines are present in >30% of lesion surface, or • Edged papillae are absent • CDC-I score is >15 (50% probability of a melanoma diagnosis).
Despite lacking correlation between RCM dendritic morphology and benign or malignant melanocytes or Langerhans cells upon histology, the extent (>30%) and density of dendritic cells at the DEJ seem to be indicative of melanocytic proliferation, and therefore indicative of melanoma. Our study confirms that the abundant presence of round cells in the DEJ assist in identifying invasive melanoma. The reproducibility of the CDC-I score requires validation in other samples prior to being applied to clinical practice.

ACK N OWLED G EM ENTS
The patients in this manuscript gave written informed consent to publication of their case details. Open access funding enabled and organized by CRUI.

CO N FLI C T O F I NTE R E S T
None to declare.

AUTH O R CO NTR I B UTI O N S
All authors have read and approved the final manuscript.
LG performed the research, designed the research study, contributed essential reagents or tools and analysed the data. SK designed the research study, contributed essential reagents or tools, analysed the data and wrote the paper. JC designed the research study, analysed the data and wrote the paper. CL contributed essential reagents or tools. SC performed the research and contributed essential reagents or tools. FF performed the research and analysed the data. GP designed the research study, analysed the data and wrote the paper.

E TH I C S S TATEM ENT
This retrospective study was approved by the local ethics committee (Prot. AOU 0008852/20 of 25/03/2020.).

PATI E NT CO N S E NT S TATE M E NT
All patients consented to the storage and further analysis of clinical and lesion data for research purposes.

PE R M I SS I O N TO R E PRO D U CE M ATE R I A L FRO M OTH E R S O U RCE S
Not applicable.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.