COMPARATIVE EVALUATION OF THE CYTOLOGICAL, HISTOPATHOLOGICAL AND IMMUNOHISTOCHEMICAL FINDINGS OF CANINE CUTANEOUS AND SUBCUTANEOUS MASSES

In this study, we compared the cytological, histopathological, and immunohistochemical diagnoses of 71 canine cutaneous and subcutaneous masses. Cytological diagnoses included 56 tumors (21 mesenchymal, 15 epithelial, 16 round cell, four melanocytic), 13 inflammatory reactions, and two cysts. Of the 21 cytologically diagnosed mesenchymal tumors, three were later confirmed non-tumoral (hematoma, granulation tissue, fibroepithelial polyp). Thirteen out of 15 epithelial tumors were correctly diagnosed cytologically, whereas two cases were confirmed to be non-tumoral (fibroepithelial polyp, granulation tissue) after histopathological examination. One mast cell tumor was later confirmed as fibrous hyperplasia; diagnoses were correct in other round cell tumors. Cytological diagnoses were correct for all melanocytic tumors and cystic lesions. Five cases which had been cytologically diagnosed as inflammatory reactions were diagnosed as tumors (lymphoma, papilloma, sebaceous adenoma, and squamous cell carcinoma) after histopathological examination. Immunohistochemistry confirmed the histopathological diagnoses of all epithelial and round cell tumors, while the diagnoses of six mesenchymal tumors were changed after the immunohistochemical examination. The total accuracy of cytology in the diagnosis of tumoral/non-tumoral masses was 84.5%, and the accuracy in the determination of benign/malignant behavior was 83%. Diagnostic accordance between histopathology and immunohistochemistry was 86.6%. High success rates obtained with cytological diagnoses prove that cytology is a reliable diagnostic tool. The main diagnostic challenge remains with mesenchymal tumors and tumors accompanied by inflammatory reactions. The results suggest that immunohistochemistry is fundamental for diagnoses of most mesenchymal tumors.


INTRODUCTION
Cytology is a labor-and cost-effective method that yields results in a very brief time compared to histopathology in the diagnosis of cutaneous and subcutaneous masses [1][2][3]. However, a proper sample cannot be obtained in many cases, and the knowledge and experience of the examiner directly affect the rate of accurate diagnosis. Since the tissue structure is not observed in cytology, histopathology remains an integral part of the diagnosis [3].
Numerous studies have been carried out to compare the cytological and histopathological diagnoses of both healthy and abnormal structures in humans [4][5][6][7] and in animals [8][9][10][11][12]. The accuracy rate has ranged between 70.5-91% in studies evaluating the reliability of cytology in the diagnosis of cutaneous and subcutaneous masses [8][9][10]12,13]. In a retrospective study conducted by Simeonov [14], the cytological slides of 300 cutaneous and subcutaneous palpable lesions from dogs were compared with histopathological slides. The total agreement rate between cytology and histopathology was 88.7%, sensitivity was 90.47%, specificity was 97.22%, positive predictive value was 98.44%, and the negative predictive value was 63.63%.
In this study, we compared the cytological diagnoses of 71 cutaneous and subcutaneous masses with histopathological and (in some cases) immunohistochemical examination results. The predictive value of cytology for the differentiation of tumoural/nontumoural masses and benign/malignant nature was evaluated.

MATERIAL AND METHODS
A total of 71 masses were excised from the skin and subcutis of dogs, and the cytological, histopathological, and (in some cases) immunohistochemical staining results were compared. Approval for the study was obtained from the Animal Experiments Local Ethics Committee (decree no: B.30.2.ULU.0.8Z.00.00-13).

Preparation and suitability of cytological slides
Cytological slides were prepared by using fine-needle biopsy, impression smear, or scraping techniques [2] and stained with Hemacolor™ (Merck Millipore, Darmstadt, Germany). In cases in which a sample could not be collected with fine-needle biopsy, an impression smear was made on the cut surface of the masses after surgical removal. Scraping technique was used only in masses that yielded very low cellularity with fineneedle aspiration or impression. The suitability of the cytological slides was assessed according to a classification from a previous study [15] and all slides except those of poor quality (insufficient cell number, thick preparation, etc.) were included.

Tissue processing for histopathology and immunohistochemistry
Tissue samples from the masses were fixed in 10% neutral buffered formalin and were processed routinely; 3-5 μm thick sections were stained with hematoxylin-eosin (Merck Millipore). Information about the primary antibodies is summarized in Table 1. All antibodies were known to cross-react with canine tissues, as confirmed with previous studies or indicated by the manufacturer. Cases from our archives with confirmed diagnoses were used as the positive control, while negative controls were generated by applying PBS instead of the primary antibody.

Examination of the slides
The cytological slides were examined by one pathologist (VI), histopathological and immunohistochemical slides were additionally evaluated by two different pathologists with consultation (ITC, AA), who were uninformed about the cytological diagnosis.
Three main criteria were sought while determining the accuracy of cytology: accuracy in diagnosing whether the mass was a tumor or non-tumor, and if tumoral whether it had a benign or malignant character. In cases in which the diagnosis of a tumor was changed to another tumor after the histopathological examination (e.g., diagnosis of a trichoblastoma was changed as sebaceous epithelioma), statistical analysis was also performed. Finally, the accuracy rate of histopathological diagnosis was calculated after the final diagnoses were determined with immunohistochemical staining.

Statistics
Sensitivity, specificity, positive predictive value, negative predictive value, and total accuracy rate for cytology were determined as reported previously [22,23] and according to the following equations (Table 2)

Vim-Cystic lesions (2)
Epidermal inclusion cyst (2) Epidermoid cyst (2) ---*Immunohistochemical staining was not performed in "---" marked cases. Incompatible diagnoses are written with bold characters. ** E-Cad: E-Cadherin, Tryp: Tryptase, Vim: Vimentin, CK: Cytokeratin Histopathological examination revealed a papilloma which was cytologically diagnosed as inflammatory reaction. Two trichoblastoma diagnoses in cytology was changed to sebaceous epithelioma and perivascular wall tumor, and in one case, malignant fibrous histiocytoma diagnosis in cytology was changed as trichoepithelioma after histopathological examination. In two cases, in which inflammatory reaction diagnosis was made cytologically, squamous cell carcinoma diagnoses were made after histopathological examination. A diagnosis of perianal gland adenoma was changed to perianal gland epithelioma histopathologically. In one case, cytologically, an inflammatory reaction was diagnosed due to the existence of neutrophils, but this diagnosis was changed as sebaceous hyperplasia/adenoma after histopathological examination. One case (Fig. 1a) that was cytologically diagnosed as carcinoma was histopathologically diagnosed as sebaceous adenocarcinoma (Fig. 1b).
A melanocytoma diagnosis in cytology was changed as malignant melanoma after histopathological examination. Another malignant melanoma diagnosis by cytology was compatible with histopathological diagnosis.
Diagnoses of benign mesenchymal tumor, malignant mesenchymal tumor, and trichoblastoma in cytology were changed as perivascular wall tumor after histopathology. In two cases, benign mesenchymal tumor diagnoses were hemangioma and perineuroma after histopathology. Hemangiopericytoma diagnosis in cytology was changed as fibroma after histopathology. Malignant mesenchymal tumor diagnosis in cytology was malignant fibrous histiocytoma after histopathology.
In a case which was cytologically diagnosed as mastocytoma, histopathological diagnosis was made as fibrous hyperplasia due to intense fibrous tissue hyperplasia with scattered mast cells in the dermis. Cytological and histopathological diagnoses were compatible in four histiocytomas (Figs. 4a and 4b) and three histiocytic sarcomas (Figs. 5a and 5b). In a case diagnosed as lymphoma, cytological and histopathological diagnoses were compatible. In another case, cytology revealed numerous yeasts resulting in a diagnosis of Malassezia dermatitis, but the final diagnosis after histopathology was epitheliotrophic lymphoma associated with Malassezia. Malignant fibrous histiocytoma diagnosis in cytology (Fig. 6a) was changed as malignant histiocytosis in histopathology (Fig. 6b).
Seven inflammatory reaction diagnoses cytologically were compatible with histopathological diagnoses. In a case, cytologically numerous atypical large epithelial cells with abundant cytoplasmic vacuoles were observed, and adenocarcinoma was diagnosed, but after histopathological examination, diagnosis was changed as chronic granulomatous reaction.
Cytologically, a malignant mesenchymal tumor was diagnosed in a case, but the diagnosis was changed as fibroepithelial polyp by histopathology. In another case, cytologically a fibropapilloma was diagnosed, but histopathologically fibroepithelial polyp was diagnosed. In one case cytological diagnosis of hemangiosarcoma was changed as hematoma in histopathology. In the cytological examination of a case, benign mesenchymal tumour was diagnosed, however histopathologically, granulation tissue formation was diagnosed. In a case, cytological inflammatory reaction diagnosis was changed as sebaceous hyperplasia in histopathology.
In two cases, epidermal inclusion cysts were cytologically diagnosed and these diagnoses were changed as epidermoid cyst by histopathology.

Comparison of histopathological and immunohistochemical findings
All tumoral cases except melanocytic tumors and lipomas were stained immunohistochemically. The histopathological diagnoses were confirmed in 40 tumors, but changed in six cases, all of which were of mesenchymal descent.
Cytokeratin and vimentin immunohistochemical stainings were used for all epithelial tumors. All cases were positive for cytokeratin and negative for vimentin except one sebaceous adenocarcinoma case, where both cytokeratin and vimentin were positively stained (Figs. 1c and 1d). Histopathological and immunohistochemical diagnoses were compatible for all epithelial tumors. CD31, CD204, cytokeratin, desmin, Melan-A, neurofilament, S100, sarcomeric actin, smooth muscle actin, and vimentin antibodies were used for the diagnosis of mesenchymal tumors, except two hemangioma cases where only CD31 positivity was considered sufficient due to the typical histopathological appearance. All histiocytic sarcomas were positive for CD18, CD204 (Fig. 5c), E-cadherin (Fig. 5d) and vimentin, but negative for CD3, CD20, and tryptase. Final diagnosis of these cases was made as histiocytic sarcoma. All mesenchymal tumors were negative for cytokeratin and positive for vimentin. In one case with a histopathological diagnosis of fibroma, positivity was observed for S100, sarcomeric actin, and vimentin, and the final diagnosis was made as rhabdomyoma.
One of the three histopathologically diagnosed perivascular wall tumors was positive for S100 and vimentin, and the diagnosis was changed as hemangiopericytoma. Two other perivascular wall tumors were positive only for vimentin, therefore the diagnosis was redefined as undifferentiated sarcoma. In one case of undifferentiated sarcoma, tumor cells stained with anti-sarcomeric actin (Fig. 2c), SMA (Fig. 2d), and vimentin, so the final diagnosis was hemangiopericytoma. In another case which was histopathologically diagnosed as haemangioma, tumor cells were negative for CD31 and all other mesenchymal markers except vimentin, and the final diagnosis was defined as benign mesenchymal tumor.
The antibodies used for the immunohistochemical diagnosis of round cell tumors were as follows: tryptase for mast cell tumours; CD3, CD18, CD20, E-cadherin, and tryptase for lymphomas; and CD3, CD18, CD20, CD204, E-cadherin, and vimentin for histiocytic tumors. All seven mastocytoma cases were confirmed after positive immunostaining with tryptase. CD18 positivity (Fig. 4c) together with CD3, CD20, CD204, and tryptase negativity were observed in all histiocytoma cases. E-cadherin positivity (Fig. 4d) was seen only in one histiocytoma case and, therefore, a Langerhans cell origin was suspected. In a malignant histiocytosis case, CD18, CD204 (Fig. 6c), E-cadherin and vimentin (Fig. 6d) positivity were observed and the diagnosis was confirmed.
One histopathologically diagnosed lymphoma case was positive for CD3 and CD18, but negative for CD20, E-cadherin, and tryptase, and a final diagnosis was made as T-cell lymphoma. The other lymphoma case was positive for CD18, but negative for CD3, CD20, E-cadherin and tryptase, and the final diagnosis in this case was considered as null-lymphoma.
Compatibility rate of histopathological and immunohistochemical diagnoses were found as 86.6 %. All control tissues were positive in terms of the relative cell components. Additionally, some tumoral cases were used as self-control tissue for normal components on the slides. All negative control slides were negative for used antibodies.

Accuracy of cytology in characterizing benign/malignant nature
In order to determine the accuracy of cytological diagnosis in defining benignancy/ malignancy, cytological and immunohistochemical staining results were compared because benign/malignant differentiation of some cases could be possible after immunohistochemistry. Non-tumoral lesions were considered as benign formations.

Accuracy of cytology in the specific-general diagnosis of tumors
For the evaluation of the accuracy rate for specific and general diagnosis between cytology and histopathology, statistical analysis was performed. Specific diagnosis such as hemangioma, lipoma, trichoblastoma, mastocytoma etc. and general diagnosis such as benign mesenchymal tumor, malignant mesenchymal tumor, carcinoma etc. were compared. All changing diagnosis were evaluated in this statistical comparison without considering diagnostic accordance for the character of the tumor. The values were calculated as follows: sensitivity (true positive rate) = 73.4 %, specificity (true negative rate) = 40.9 %, positive predictive value = 70.5 %, negative predictive value = 40.9 %, total accuracy rate = 63.3 % (Table 8).

DISCUSSION
Cytology is a quick and reliable tool in most pathology laboratories. In dogs, the accuracy of cytological diagnoses was tested in various studies yielding different accuracy rates [1,10,[12][13][14]. Cohen et al. [10] evaluated 269 masses cytologically and histopathologically and 63.2% accuracy rate was calculated when insufficient samples were excluded. Ghisleni et al. [12] have compared the cytological and histopathological findings of 292 palpable cutaneous and subcutaneous masses in 242 dogs and 50 cats between 1999-2003. As a result, 90.9% of the cases showed agreement between cytological diagnosis and histopathological diagnosis. In another retrospective study conducted by Simeonov [14], the cytological preparations of 300 cutaneous and subcutaneous palpable lesions of dogs were compared with histopathological slides and total agreement rate between cytology and histopathology was 88.7%.
It should be noted that changes in the nomenclature of tumors in time and failure to use standardized criteria may lead to controversies in histological diagnosis, which naturally affects compliance with cytological diagnosis [24][25][26]. In our study, histopathological diagnoses were made according to the previous tumor classifications and additional immunohistochemical stainings were performed in order to eliminate the above mentioned disputes.
In our study, a mesenchymal malignant tumor, a haemangiosarcoma, and an adenocarcinoma were misdiagnosed in cytology and diagnoses were changed as fibroepithelial polyp, organizing haematoma and chronic granulomatous inflammation, respectively. It is difficult to distinguish well-differentiated haemangiosarcomas from benign haemangiomas and/or haematomas (which form reactive fibroblasts in the wall), and histological examination is often necessary for the definitive diagnosis [27]. It has been reported that inflammation can induce dysplastic changes and such cases can be misinterpreted as a tumor [2]. Large epithelioid macrophages and proliferating fibroblasts are commonly observed in inflammatory lesions and these cells can be similar to malignant tumor cells. Also squamous cells, fibroblasts, and transitional epithelial cells can exhibit marked anisonucleosis, distinct nucleoli, mitotic figures, and variations in nucleus:cytoplasm ratio, features all of which can be confused with tumor cell features [2,28].
In the study of Simeonov [14], 19 cases with a diagnosis of inflammation in cytology were diagnosed as tumor in histopathology. It should be remembered that inflammation can coexist in tumors [29]. In our study, cytological diagnoses of five inflammatory reactions were changed as a papilloma, a sebaceous adenoma, a sebaceous hyperplasia, and two squamous cell carcinomas after histopathology. This demonstrates the importance of careful examination of tumors with concurrent inflammatory reaction.
Origin of melanocytic tumors was determined accurately in cytology by observing dense melanin pigmentation. In our study, one case of melanocytoma diagnosis in cytology was changed as malignant melanoma in histopathology. The reason of this incompatibility was difficulty of nuclear malignancy evaluation due to dense pigmentation. In the study of Simeonov [14], all melanocytic tumor diagnoses in cytology were confirmed with histopathology. In that study, tumor character was not evaluated in cytology, but all cases were diagnosed as benign in histopathology. In our study, malignancy could be evaluated in other cases due to the low pigmentation of cells in cytology.
We did not subclassify mesenchymal tumors cytologically in most cases in our study. Instead, they were classified as benign or malignant mesenchymal tumors in nine cases, and these diagnoses were consistent with the final diagnoses. Similarly, Ghisleni et al. [12] reported that cytomorphology was inadequate for determining the cell type in 45 of 69 (67.2%) mesenchymal tumors and these cases were defined as sarcoma. In Simeonov's study [14], 33 of the mesenchymal tumors could not be subclassified, of which 25 were later identified as fibrosarcoma, five as liposarcoma, two as osteosarcoma, and one as myxosarcoma.
Mesenchymal tumors comprise 8-15% of canine cutaneous and subcutaneous tumors and are formed mostly by spindle-shaped cells [30]. Differential diagnosis of these tumors is difficult due to similar morphological patterns [17,20]. Immunohistochemically vimentin expression is suggestive of a mesenchymal origin [31]. In accordance with this, all mesenchymal tumors were vimentin positive in our study. Even with additional immunohistochemical diagnostic markers, a mesenchymal tumor origin may not always be determined [19,32]. An immunohistochemical panel for definitive differentiation of mesenchymal tumors has not been defined yet [33]. Due to this ambiguity, sometimes these tumors are named as spindle cell tumors of the soft tissue [31,34,35]. In our study, an exact diagnosis could not be reached in two mesenchymal tumors and these tumors were diagnosed as undifferentiated sarcoma. Histopathologic diagnosis of two perivascular wall tumors were undifferentiated sarcoma after immunohistochemistry due to vimentin positivity alone. Additionally, a case of haemangioma was negative for all markers except vimentin and the final diagnosis was changed as benign mesenchymal tumor depending on the histopathological pattern.
In our study, vimentin and S100 positivity were observed in cases histopathologically diagnosed as myxoid PNST (Figs. 3c and 3d) and perineuroma. GFAP and S100 are suggested as useful markers for the diagnosis of PNSTs [20]. Peripheral nerve sheath tumors originate from Schwann cells, modified Schwann cells, fibroblasts or perineural cells [36]. Haemangiopericytoma should be primarily considered in the differential diagnosis of PNSTs. Especially, the perivascular spiral pattern is characteristic for haemangiopericytomas. These patterns have also been observed in PNSTs especially around collagen fibres instead of capillaries [31]. Similarly, we observed these patterns around nerve fascicles which included fine collagen fibres in the case of perineuroma and these areas were positive for S100. Although high positivity rates have been reported with neurofilament [37], in our study neurofilament immunostaining was negative in all cases. Chijiwa et al. [32] reported that α-SMA may be useful for differentiation of PNSTs and hemangiopericytomas. They observed that 17 PNSTs were negative and five hemangiopericytomas were positive for α-SMA. In another study which involved cases of hemangiopericytoma, leiomyosarcoma, and hemangiosarcoma, α-SMA was positive in six of 18 hemangiopericytomas [38]. We observed α-SMA positivity in only one case which was histopathologically diagnosed as undifferentiated sarcoma. In this case, vimentin and sarcomeric actin were also positive, thus the diagnosis was changed as hemangiopericytoma. In another case diagnosed as perivascular wall tumor in histopathology, vimentin and S100 positivities were seen. Due to some reports mentioning S100 positivity in peripheral wall tumors [17,32] and the observation of a whirling pattern in our case, the final diagnosis was made as hemangiopericytoma.
Malignant fibrous histiocytoma is in the "fibrohistiocytic" category according to the 2002 WHO Classification [39]. In humans, malignant fibrous histiocytoma is classified as storiform-pleomorphic, giant cell, inflammatory, myxoid, and angiomatoid type according to the common cell type [40]. However, only three types have been reported in domestic animals [41] and these types were reclassified as undifferentiated pleomorphic sarcoma in the new classification [39]. In our study, CD3, CD18, CD20, CD204, E-cadherin, tryptase, and vimentin immunostaining was performed for the differentiation of malignant fibrous histiocytoma from histiocytic sarcoma. Negativity of CD3, CD18, CD20, E-cadherin, but positivity with CD204 and vimentin were observed. While canine histiocytic sarcomas are regularly CD18 positive and mostly vimentin negative, malignant fibrous histiocytomas have a vimentin positive and CD18 negative phenotype [41][42][43]. Ko et al. [44] performed α-SMA, CD68, desmin, S100 and vimentin in three canine malignant fibrous histiocytomas and they observed positivity for vimentin, positivity in only giant cells for CD68, and negativity for α-SMA, desmin, and S100. These researchers suggest that even only vimentin positivity without any other special markers is useful for the diagnosis of malignant fibrous histiocytoma. Similarly, in our mentioned case, due to the negativity of CD18 and positivity of vimentin and CD204, a diagnosis of malignant fibrous histiocytoma was made.
In one study, the accuracy of cytology in round cell tumors has been found to be less than 50% [10]. Ghisleni et al. [12] also reported that there were certain limitations in determining the origin and benignancy/malignancy of 10 round cell tumors. In our study, cytological diagnoses of mastocytomas, and histiocytomas were confirmed in all cases. Mastocytoma diagnosis is generally considered simple due to distinct cytoplasmic granules [8,45,46], and high accuracy rates between cytology and histopathology have been reported [12,14]. Although certain studies report high accuracy rates for cytological diagnosis of histiocytomas [12,47], some histiocytic tumors cannot be differentiated from other round cell tumors (lymphoma, plasmacytoma) and only a round cell tumor diagnosis can be made with cytology [12,14,45]. Distinguishing histiocytic sarcoma from lymphoma and plasmacytoma can also be difficult [47].
In canine cutaneous round cell tumors, histopathology is generally not sufficient for the exact diagnosis and immunohistochemical stainings should also be utilized [48]. CD18, E-cadherin, and MHC-II have been suggested as useful markers for the immunohistochemical diagnosis of histiocytic tumors. Furthermore, CD3 and CD79 have been suggested as markers for the differentiation of B and T cell lymphomas [42,47,49]. Fernandez et al. [48] examined 72 canine cutaneous round cell tumors and used tryptase, chymase, and serotonin for mast cells; CD1a, CD18, and MHC-II for histiocytes; CD3 for T lymphocytes; and CD79a for B lymphocytes and plasma cells. We used only tryptase for mastocytomas; CD3, CD18, and CD20 for lymphomas; CD18, CD204, E-cadherin, and vimentin for histiocytic tumors. CD18 was used for differentiation of malignant fibrous histiocytoma from histiocytic sarcoma. Tryptase positivity was deemed sufficient in diagnosis of mastocytomas, and no further immunohistochemical stainings were performed to keep costs low.
For differentiation of canine lymphomas from other round cell tumors, immunohistochemical staining should be performed in many cases [49]. In our study, one lymphoma case revealed CD3 and CD18 positivity and the final diagnosis was made as T cell lymphoma. In other case which was diagnosed as lymphoma, positivity of CD18 and negativity of CD3 and CD20 were observed. CD18 is a formalinresistant integrin molecule on the surface of leukocytes and dendritic cells originating from bone marrow. For this reason, CD18 (or MHC-II) positive and CD3 or CD79a negative tumoral cells are considered as of a histiocytic origin [42,43,47,50]. However, atypical immunophenotype of tumor cells have also been reported rarely [51,52]. As an example null lymphomas do not express CD3 and CD79a or light and heavy chains of immunoglobulins [53][54][55].
Null lymphoma has been rarely reported in dogs. This tumor has been suggested to have a "natural killer" cell origin [53]. CD3 and CD79a negativity of these tumors does not necessarily mean that these are not lymphoma. Granularity feature of tumoral cells may be useful for diagnosis, but additional immunohistochemical staining may also be needed [56]. Therefore, we performed E-cadherin and tryptase in the case of nulllymphoma and these two markers were found negative.
In our study CD18 was used for the determination of leukocytic origin and all round cell tumors showed positivity. Unlike a previous report [21], we observed E-cadherin positivity in only one histiocytoma case, but not in other three histiocytomas. In dogs, E-cadherin is a marker used for histiocytomas [57,58], but not a specific marker for Langerhans cells [21]. Pazdzior-Czapula et al. [47] showed limited positivity for CD18, E-cadherin, and MHC-II expressions in canine cutaneous histiocytomas. They suggested that E-cadherin has limited value for the diagnosis of cutaneous histiocytomas. According to Valli et al. [51], Langerhans cells are unique cells which express E-cadherin. However, according to some other studies, due to its expression in plasmacytomas and epitheliotropic lymphomas, E-cadherin is not considered a unique marker for Langerhans cells and it cannot be used for the exact differentiation of canine cutaneous round cell tumors [47,59].
It has been reported that canine histiocytic sarcomas express CD1, CD11c, CD18, CD45 and MHC-II but not express E-cadherin [18] and these tumors originated from dendritic cells [21,42]. In another study, because of E-cadherin positivity in two histiocytic sarcomas, it was thought that Langerhans cell origin of these tumors [47]. Hirako et al. [60] reported that primary cutaneous histiocytic sarcoma originated from Langerhans cells in a male Pembroke Welsh Corgi dog. They showed immunopositivity for vimentin, HLA-DR antigen, Iba1, CD18 and E-cadherin, thus they suggested Langerhans cell origin due to the E-cadherin positivity. In our study, E-cadherin positivity was observed in three histiocytic sarcomas. Furthermore, CD18, CD204 and vimentin positivity were observed in same cases.

CONCLUSION
As a result of our study, we suggest that cytopathology is a reliable and quick technique with high accuracy rate (84.5% for tumoral/non-tumoral comparison and 83% for comparison of benignancy/malignancy) as compared to histopathology. The main diagnostic challenge in cytology is the mesenchymal tumor. Even histopathology is not sufficient for the accurate diagnosis of mesenchymal and round cell tumors, and specific immunohistochemical panels should be performed for accurate diagnosis. Furthermore, histopathology is the method that confirms or refutes the cytopathological diagnosis, and also guides the immunohistochemistry in some cases.