Reflectance confocal microscopy for noninvasive examination of nonmelanocytic tumors and virus‐associated skin lesions in organ transplant recipients

Drug‐induced immunosuppression is necessary to prevent rejection of the foreign organ in transplanted patients, but neoplastic and virus‐associated skin diseases are frequent complications. Reflectance confocal microscopy (RCM) recently emerged as a promising tool for the early diagnosis of skin lesions.


| INTRODUC TI ON
An organ transplantation is considered to be a life-saving procedure for patients with severest organ damage. The continuous intake of various immunosuppressants can prevent a rejection of the foreign organ by the patient's own immune system. The number of organ transplanted patients has increased to 1 000 000 worldwide. 1 Verrucae and skin tumors are the two most frequent skin diseases in organ transplanted patients, with skin tumors also accounting for 40%-50% of the malignant tumors. The malignant tumor incidence increases with the duration of the immunosuppressive therapy. In Western Europe and in the United States, 5% of all organ transplant recipients are suffering from malignant skin tumors in the first year post-transplantation. 10 years after the transplantation 10%-27% and after 20 years 40%-60% of the patients are affected.
In Australia, the tumor incidence after 20 years even increases to 70%-82%, 2 with the risk of nonmelanocytic skin tumors being especially high. These tumors account for 95% of all skin tumors in long-term transplant recipients. 3 In addition, multiple lesions in the sense of a field cancerization are often detectable in nonmelanocytic skin tumors in organ transplant recipients. The mortality rate of immunosuppressed patients is also increased, exceeding 5%. 4 Verrucae and skin tumors in organ transplanted patients occur at different times in the corresponding skin areas, with over 90% of the skin tumors manifesting in UV-exposed skin areas. 5,6 Moreover, the type, dose, and length of immunosuppressive treatment, an increased age of the patient at transplantation, the kind of the transplanted organ, and dialysis prior to kidney transplantation play an important role. 4,7,8 The prognosis for neoplastic and virus-associated skin diseases in immunocompetent patients can be excellent, provided these diseases are diagnosed and treated in an early stage.
In organ transplant recipients, however, it is to be considered that even inconspicuous skin lesions tend to quickly progress and infiltrate deeply. These skin diseases often occur in multiple number at unusual locations, and in many cases, their morphology is atypical. Therefore, they may present a considerable diagnostic challenge for the investigator. Diagnostic problems may occur especially when it comes to micronodular basal cell carcinoma (BCC) in the face which is frequently associated with multiple benign hyperplasia of the sebaceous glands that are also generated under immunosuppression, to sclerodermiform BCCs or to superficial BCCs without any dermatoscopic criteria. 9 The current diagnostic standard consists of a histological examination after invasive skin biopsy. In general, an in situ squamous cell carcinoma (SCC), which is divided into actinic keratosis (AK) and its specific form of Bowen's disease, is clinically diagnosed. However, since it is often impossible to distinguish between an in situ and an early invasive SCCs, the histological evaluation remains to be the gold standard. 10,11 Furthermore, verrucae are also clinically diagnosed. However, if the mostly hyperkeratotic lesions cannot be distinguished from AKs and early invasive SCCs by differential diagnostics, a sample should be excised in these specific cases, too. 12 Compared with the non-transplanted test population, the incidence of BCC in organ transplanted patients is 10 times higher. 13 In addition, their risk of developing AK is 250-fold higher and their risk of an invasive SCC genesis is 100 times higher. 14 Within the first 5 years of immunosuppression, 40% of all organ transplanted patients are developing AKs. 4 Overall, it can be said that it is necessary to repeatedly excise multiple skin samples from organ transplanted patients. For this patient cohort, it also has to be considered that they often receive a continuous co-medication with an anticoagulant, which may result in an increased risk of bleeding. Due to the chronic immunosuppression, the risk of wound infections in the surgery area is increased. Thus, the development of reliable noninvasive and time-saving investigation methods is indispensable as an alternative to routine histology. Reflectance confocal laser scanning microscopy (RCM) is an optical high-resolution noninvasive diagnostic procedure permitting the visualization of dermatological processes at a cellular level. Like an in vivo biopsy, it enables in real time to image epidermal and superficial dermal structures down to the upper stratum reticulare. 15,16 Since organ transplanted patients are often showing widespread extensions of the skin changes and subclinical lesions, RCM can be used to determining the most suitable biopsy site. In this way, scars caused by unnecessary invasive biopsies are avoided. In recent years, the applicability of RCM for in vivo diagnostics of neoplastic and virus-associated skin changes in immunocompetent patients was investigated. In all studies, this method was described as a promising tool for noninvasive diagnostics. [17][18][19][20][21][22][23][24] The confocal microscopic evaluation criteria of BCC are well known. 18,[25][26][27][28][29] Most of them become visible at the level of the superficial dermis (D) and the dermo-epidermal junction (DEJ), respectively. 30 A large multicenter study showed high sensitivity and specificity values for selected parameters, which correlated well with those of the conventional histology. 18  The purpose of the present study was to assess the accuracy of RCM for the diagnosis of neoplastic and virus-associated skin changes in organ transplanted patients by using selected RCM evaluation criteria in a clinical setting. The gold standard was the conventional routine histology. In the initial experiments, the hitherto published RCM criteria of the individual skin changes were reproduced. Subsequently, the sensitivity and specificity data were determined for both the RCM diagnosis and the individual parameters.
Finally, it was assessed whether the investigators had reached concordant results. The method is supposed to be highly reliable if all RCM criteria are detected by the independent investigators (INVs) and concordant RCM diagnoses are made, respectively.

| Patients
Once the approval of the competent ethics committee had been obtained, the study participants were recruited at the Skin Tumor Center of the Charité-Universitätsmedizin Berlin. The patients were examined during their risk profile-adapted prophylactic checkups and follow-ups, respectively. A total number of 35 patients, nine of them female and 26 of them male patients, with skin types I-III according to the Fitzpatrick scale, aged between 46 and 83 years (mean of 64.5 years) were included in the study. Thirty one of these patients had a kidney transplant, three a heart transplant, and one patient had simultaneously received a kidney and a heart transplant. Overall, 61 skin changes were clinically, dermatoscopically, confocal microscopically, and histologically evaluated. 20 lesions were considered suspicious for BCC, six for Bowen's disease. AK was clinically diagnosed 23 times, and 12 skin lesions appeared as verrucae.

| Evaluation by reflectance confocal microscopy
At first, a dermatoscopic image (VivaCam®) of the skin change was made, which was then evaluated using RCM (VivaScope®). Thereby, the dermatoscopic images and RCM images were superimposed so that a concordance between the images was obtained. With the VivaBlock® function, horizontal maps were captured at the levels of the stratum spinosum (SS) and the superficial dermis. In addition, one VivaStack®, each, was generated at 5-μm intervals down to the maximum depth of 250 μm, that is, from the stratum corneum (SC) through the entire epidermis, down to the level of the upper papillary dermis, with extra single images being taken in the areas of specific interest. The video capture module was used to produce short videos at the DEJ/D level, where vessels were observed, in order to visualize the blood flow.

| Histology
To confirm the suspected clinical diagnosis and the correlation of the RCM findings with the criteria of the conventional histology, a sample was biopsied. The punch biopsy diameter was 4-6 mm, on average. Subsequently, the specimen was evaluated at the division of dermatohistology of the Charité-Universitätsmedizin Berlin. In the clinically suspected BCC cohort, the diagnosis could be confirmed by the histological result 10 times, whereas 10 further lesions were found to be other benign and malignant skin changes (non-BCC lesions). In the Bowen cohort, the clinical diagnosis for 4 out of 6 lesions was confirmed by histology. In the cohort with clinically suspected AKs, the diagnosis could be histologically confirmed for 18 out of 23 lesions. The verruca cohort included 12 lesions, seven of which were verrucae and five non-verrucae.

| Morphological analyses
The RCM images were retrospectively subjected to a morphological analysis by two independent dermatologists (INV I and INV II).
One investigator was already well versed in RCM, whereas the skills of the second investigators were considered to be moderate. Thereby, it was possible to create study conditions reflecting the overall dermatological community that consists of investigators with different RCM experience. The INVs were aware of the suspected clinical diagnosis. The lesions were morphologically described and systematically evaluated for the presence and absence, respectively, of specific RCM criteria for each diagnosis, which are known in the current literature. If ≥50% of the generally described RCM criteria were found, the clinical diagnosis was confirmed by RCM.

| Technical principles of reflectance confocal microscopy
The reflectance confocal microscope (RCM) consists of a light source, a focusing and an objective lens, and a detector. 16 A diode laser beam serves as source for monochromatic, coherent light which penetrates into the skin through a beam splitter and the objective lens and focuses a small spot in the tissue. 15 Contrary to histology with its established vertical incision, the laser of the RCM scans in horizontal direction. Thus, the image interpretation is based on virtual horizontal tissue incisions. 38 The light is reflected, scattered, or absorbed by various refractive indices on boundary layers, cell organelles, and cellular microstructures. 39 The light signals reflected from the skin return to the detector in front of which a tiny pinhole aperture is installed. This ensures that only light from the focused level is processed. 15 Subsequently, the light signals are transferred to a computer that generates twodimensional sectional images of various gray shades from the differentiated reflexion patterns. 40 Thereby, the contrast of the RCM images is based on the different refractive indices of "endogenous contrast media". 39 Structures with a high refractive index result in a bright confocal image. 15 With a refractive index of n = 1.72, melanin is the strongest contract provider. Further natural contrast media are keratin (n = 1.51) and hydrogenated collagen (n = 1.43).
They are surrounded by structures with a lower refractive index, such as epidermis (n = 1.34) and dermis (n = 1.41. 16,38 In this study, a commercial RCM suitable for in vivo application (VivaScope® 1500; Lucid-Tech Inc, MAVIG GmbH) was used. The laser energy at tissue level is below 30 mW, thus not inducing any tissue or eye damage (laser category Ia). 38 The lateral resolution is <1.25 μm, thus being at cellular level. The lateral solution is around <5 μm, thus corresponding approximately to the layer thickness of conventional histology. 15,41 Skin structures can be represented down to a total penetration depth of 250-350 μm. 39 By positioning the VivaCam® onto a stainless steel ring, which is fixed to the patient's skin area, a digital image of dermatoscopic quality can be generated. The macrocamera generates single images of 5 megapixel size on a 10 x 10 mm image area and ensures a correlation of the macroscopic images with the confocal images of the VivaScope®. By displacing the x-, y-, and z-axes of the laser, imaging at three levels is possible. Single images are 500 x 500 μm in size. By moving the x-and y-axes of the laser, a skin layer of maximally 8 x 8 mm in size can be scanned. In this way, a map of 16 x 16 single images is generated which corresponds to a VivaBlock®.
The resolution of VivaBlocks® is 2 pixel/μm. The VivaStack® function generates images in the z-axis, thus providing an optical punch biopsy. VivaCube® combines VivaBlock® with VivaStack®.
This function permits to completely scan a skin area in the x, y, and z directions (length x width x depth). 38,42

| Statistical analysis
Statistical data evaluation was conducted using the statistic programs GraphPad Prism 7 and GraphPad QuickCalcs (online). The sensitivity and specificity were determined for both the individual RCM parameters and the diagnosis. The gold standard was routine histology. Since only a very small number of lesions could be evaluated for Bowen's disease, a statistical analysis was considered unreasonable in this case. The zero hypothesis (H0) formulated for this study is that RCM and conventional histology result in different diagnoses. Accordingly, the alternative hypothesis (H1) assumes that RCM and histology yield consistent diagnoses. At a P-value of ≤.05 (error probability 5%), the result was considered statistically significant and the zero hypothesis (H0) was discarded.
At a P-value of ≤.01, the result was rated very significant, and at a P-value of ≤.001 highly significant. 43,44 Finally, it was checked whether the two investigators had made concordant diagnoses and found the same RCM parameters, respectively. The degree of concordance was expressed by Cohen's Kappa coefficient (k).

| Morphological RCM analyses of basal cell carcinoma
Loss of honeycomb structure with slightly atypical keratinocytes: At the stratum granulosum (SG)/stratum spinosum level, a disordered keratinocyte pattern is visible ( Figure 1A Figure 1D). Gap formation between tumor cells and tumor stroma: A dark region separates the neoplastic cell aggregates from the fibrous stroma ( Figure 1D).
It contains amyloid and muzine plaques. 25,29 Inflammatory infiltrate: It consists of small, bright, clearly distinguishable cells without nucleus which frequently appears in clusters ( Figure 1E).
Multiplication, strong winding, and dilatation of the blood vessels: Vessels become entangled between the tumor nests due to neoplastic angiogenesis ( Figure 1F). 31

| Statistical analyses of basal cell carcinoma
The first cohort exhibited 20 lesions with a clinical BCC diagnosis.  (Table 1A). Consequently, if this criterion is found, the lesion can be diagnosed as BCC at highest probability. Highest sensitivity was also demonstrated for the loss of honeycomb structure, with the specificity being high as well (90%). However, the inflammatory infiltrate with only moderate values was of minor importance for both the specificity (60%) and the sensitivity (70%). The analyses of INV II also showed highly sensitive (100%) and very specific (80%) values for the loss of honeycomb structure (Table 1B). The multiplication, strong winding, and dilatation of the blood vessels showed highest sensitivity (100%) at a poor specificity of only 50%. Streaming yielded high sensitivity and specificity values of 90% and 80%, respectively. Except for the inflammatory infiltrate (P = .1775), INV I reached the significance level for all RCM criteria (Table 1C). The highest significance values were reached for the loss of honeycomb structure (P < .0001), the tumor cell islands (P < .0001), and the streaming (P = .0010). Also, INVII reached the highest P-value of .0003 for the loss of honeycomb structure. Except for the inflammatory infiltrate (P = .3613) and the dendritic cells (P = 1), INV II was able to reach the significance level for the remaining criteria.
Except for the inflammatory infiltrate, good to medium concordance data could be reached for all RCM criteria (Table 1D). A good concordance according to Altman was achieved for the loss of honeycomb structure and for streaming.

| Morphological RCM analyses of Bowen's disease
The RCM criteria of Bowen's disease are similar to those of AK. Therefore, only the parameters deviating from AK are described hereinafter. Strong loss of honeycomb structure with atypical keratinocytes: The loss of honeycomb structure is more pronounced, the cellular atypia can affect the entire epidermis ( Figure 2A Figure 3F). Solar elastosis: A moderate solar elastosis exhibits an unorganized, hyporefractile collagen in the reticular dermis ( Figure 3G); if the damage is more severe, highly refractile curled fibers are visible ( Figure 3H).

| Statistical analyses of actinic keratosis
The third cohort included 23 lesions with clinical AK diagnosis. shown for the dendritic cells (Table 2D), whereas the concordance regarding the loss of honeycomb structure was only low. is the rete ridge lengthening and papillomatosis ( Figure 4D). 46,47 Increased vascularization/vascular dilatation: An increased amount of blood components can leak through the vessels into the SC.

| Morphological RCM analyses of verruca
Therefore, the verrucae often appear darkish ( Figure 4E). to be of highest specificity and very sensitive, respectively (Table 3A).

| Statistical analyses of verruca
Consequently, these RCM parameters are the most reliable ones for the diagnosis of verruca. Of minor importance is papillomatosis with moderate specificity (60%) and poor sensitivity (50%) results, respectively. INV II, however, was able to reach the highest sensitivity for koilocytosis at good specificity of 80% (Table 3B)

| D ISCUSS I ON
According to our knowledge, data exclusively addressing the evalu-

| Basal cell carcinoma
In the literature, sensitivity values ranging between 48.8% and 100% and specificity values ranging between 53.6% and 100%, respectively, have been reported for BCC diagnosis by means of RCM. 18 analyses could show a good interobserver concordance for this criterion. 28,29 In concordance with Nori et al, the two investigators obtained high sensitivity values of 90% and 100%, respectively, for the multiplication, strong winding, and dilatation of the blood vessels. As this is partially observed also in benign lesions such as verrucae or in inflammatory dermatoses such as psoriasis, the specificity strongly declined to 80% and 50%, respectively. 51 These values also correlated very well with Nori et al, 18 who published a specificity of 53.6%. Referring to the gap formation between tumor cells and tumor stroma, Peppelman et al 29 stated an 80.5% prevalence in the lesions evaluated by them.
Likewise, in that study high prevalences of 80% and 90%, respectively, were found for the peritumoral gap. On the other hand, the analyses of Risphon et al reported these changes exclusively for every second BCC lesion. 30 Also the prevalences described by Longo et al, 28 with 77.3% in nodular and 13.6% in infiltrating BCCs are enlightening. According to our results, the most important RCM criteria for BCC diagnosis are the loss of honeycomb structure, the tumor cell nests, and streaming. We could show that the occurrence of these parameters in a lesion makes the presence of a BCC extremely possible. Moreover, it was demonstrated that RCM renders invasive biopsies unnecessary for most skin changes with the clinical diagnosis of a suspected BCC.

| Limitations of the study
The evaluation of reflectance confocal microscopic images requires special education and experience. The level of experience and expertise can vary to a high extent depending on the investigator. Both investigators within this study were moderately to highly skilled at the evaluation of confocal images. Therefore, in clinical practice evaluation quality, sensitivity, and specificity might differ from the results found in our study.
Furthermore, during in vivo measurements the investigator will always be able to clinically evaluate the lesion, which may easily bias the microscopic evaluation.

| CON CLUS ION
The present study involving larger cohorts of patients has been the first to use RCM for analyzing neoplastic and virus-associated skin changes in organ transplant recipients. It proved to be a well reproducible and reliable method for the diagnosis of BCCs and AKs. The study has already resulted in a very good definition of RCM parameters permitting a specific diagnosis of these skin diseases. RCM could provide a huge potential as a genuine alternative to histological diagnostics by invasive skin biopsy. Due to the few numbers of patients with Bowen's disease investigated by us, a conclusion is difficult. It must be taken into account, however, that this disease is a rare one. Although verrucae occur relatively often and are easily detectable, it has to be considered that multiple new hyperkeratotic lesions often occur under immunosuppression, where verrucae, but also epithelial tumors, such as in situ or invasive SCC, have to be considered as differential diagnosis. In this context, RCM represents a valuable diagnostic tool that may provide important information complementing clinical or dermatoscopic findings. There is no doubt that RCM plays and will continue to play a central role in the diagnostics of neoplastic and virus-associated skin diseases in organ transplant recipients.