Article Text
Abstract
Aims: To describe the cytological appearances of melanoma in fluid specimens and potential diagnostic pitfalls in interpreting such specimens, which have been infrequently reported in the literature.
Methods: Cases of melanoma diagnosed at a single institution between 1993 and 2008 in cytology specimens of fluids (pleural, ascitic, cerebrospinal and other fluids), but excluding fine needle biopsy specimens, were identified and reviewed.
Results: 32 fluid specimens containing metastatic melanoma (from 26 patients) were identified. Most of the specimens were moderately cellular and showed moderate to marked nuclear pleomorphism. Mitotic figures and intranuclear cytoplasmic invaginations were identified in 11 (34.4%) and seven (21.9%) cases, respectively. Melanin pigment was seen in eight (25.0%) cases. Variable numbers of histiocytes were present, and mesothelial cells were present in body cavity fluid specimens.
Conclusions: In fluid specimens, reactive mesothelial cells and histiocytes may mimic epithelioid melanoma cells. Awareness of the morphological features and diagnostic pitfalls of melanoma in fluids is necessary to avoid the potentially serious consequences of misdiagnosis.
Statistics from Altmetric.com
Metastatic melanoma has a wide range of morphological appearances in cytological preparations.1–9 Cytological material may be obtained directly from a clinically suspicious lesion using fine needle biopsy (FNB), which is a rapid, safe and minimally invasive technique1 10 and an accurate means of diagnosing metastatic melanoma.11 12 Alternatively, fluid in body cavities (such as pleural fluid, peritoneal fluid, cerebrospinal fluid), fluid collections in other body sites or fluid within cystic tumours may be aspirated, or washings of various body sites (eg, bronchial washings) may be obtained and subjected to cytological examination. The cytological features of melanoma in FNB specimens have been well described in the literature, along with the differential diagnoses and difficulties in accurate diagnosis in this setting.1–9 11 12 Less often described13 are the cytological appearances of melanoma in fluid specimens and potential diagnostic pitfalls in such specimens, and these are the subjects of this study.
METHODS
Cytological fluid specimens (pleural fluid, ascitic fluid, cerebrospinal fluid and other fluids, but excluding FNB specimens) examined between January 1993 and April 2008 and diagnosed as “melanoma” or “consistent with melanoma” were extracted from the laboratory information system of the Department of Anatomical Pathology, Royal Prince Alfred Hospital, Sydney, Australia. Slides, which included air-dried Romanowsky-stained smears, wet-fixed Papanicolau-stained smears and (when sufficient cellular material was present) cell block preparations stained immunohistochemically for S-100 protein, HMB-45 and/or Melan-A, were retrieved from the archives.
The slides were reviewed and the following cytopathological features were assessed: cellularity (graded as low, moderate and high); arrangement of cells (dispersed or aggregates); cell shape (epithelioid or spindle); binucleation and multinucleation; nuclear pleomorphism (graded as mild, moderate or marked); nucleoli (graded as inconspicuous, small, intermediate or prominent); intranuclear cytoplasmic invaginations (INCIs, pseudoinclusions); mitotic figures; necrotic debris; pigment and the localisation of pigment (in melanocyte cytoplasm, in macrophages or in the background); immunochemistry findings; and the presence, frequency and type of other cells.
RESULTS
During the study period, fluid specimens from 26 patients diagnosed as “melanoma” or “consistent with melanoma” were identified. The patients ranged in age from 39 to 91 years (median 65 years, mean 65.3 years). There were 17 men (median 64 years, mean 64.9 years, range 49–83 years) and nine women (median 70 years, mean 66.0 years, range 39–91 years). Table 1 lists the clinical features.
Table 2 summarises the cytological features. There were 32 fluid specimens including pleural fluid (14 specimens), ascitic fluid (six specimens), cerebrospinal fluid (four specimens), specimens from brain tumours (three brain tumour fluid specimens and one brain tumour washing specimen), bronchial washing (one specimen), ocular aqueous fluid (one specimen), left iliac fossa fluid (one specimen) and axillary seroma fluid (one specimen).
The specificity of the diagnosis of melanoma was 100%, as the diagnosis was not altered in any of the specimens on review. However, the diagnostic sensitivity could not be determined owing to the methodology of case selection, as discussed below.
The smears ranged in cellularity from low to high. The cells were singly dispersed in all cases, and tumour cell aggregates were identified in four (12.5%) cases (two brain tumour fluid specimens and two pleural fluid specimens). The cell shape was epithelioid in all (96.9%) but one case (3.1%), which was composed of spindle cells. Binucleation and multinucleation were identified in 19 (59.4%) and five (15.6%) cases, respectively. Nuclear pleomorphism was moderate to marked, and nucleoli were moderately to markedly prominent in 23 (71.9%) cases (fig 1A). A single case showed scattered bizarre giant tumour cells. Mitotic figures and INCIs (pseudoinclusions) were identified in 11 (34.4%) and seven (21.9%) cases, respectively (fig 1B).
Necrotic material was present in two (6.3%) cases, both of which were tumour fluid samples from metastatic melanoma deposits in brain. Melanin pigment was seen in eight (25.0%) cases and was located in the cytoplasm of tumour cells in all eight cases, as well as in macrophages (four cases, 12.5%) and extracellularly (one case, 3.1%) (fig 1A). The presence of melanoma was confirmed by immunochemistry for S-100, HMB-45 and/or Melan-A in cell block preparations in 25 (78.1%) cases, and in histological preparations from biopsy specimens in three additional cases (9.4%). S-100, HMB-45 and Melan-A were positive in 25/25 (100%), 18/21 (85.7%) and 6/8 (75.0%) specimens in which they were assessed.
In some body cavity fluid specimens, reactive mesothelial cells and histiocytes were present. They were rounded or polygonal in shape, contained rounded or oval nuclei and moderate to abundant amounts of cytoplasm, and exhibited various degrees of cytological atypia (but less so than melanoma cells) (fig 1C,D).
DISCUSSION
Melanoma can have a wide range of histological appearances, and its cytological features are also protean. Knowledge of the cytological features of melanoma is important to prevent misdiagnosis and its potentially deleterious consequences.
Melanoma in FNB smears
Virtually all of the literature describing the cytological features of melanoma is based on FNB specimens.1–9 11 FNB smears composed of frankly malignant cells with cytoplasmic pigment and INCIs as well as giant cells, with abundant cellular material in the cell block for ancillary studies, can be readily diagnosed as melanomas.7 The cells are usually dispersed singly, but variable numbers of loose (or less often tight) cell groups may be seen in some cases.1 6 The cell groups may show tubules and pseudorosettes or cell-in-cell patterns, raising the differential diagnosis of carcinoma.1 2 The cells are variable in size and may be epithelioid (50–60%), spindled (up to 14%) or mixed (up to 18%); less often, clear cells, small cells, round cells or granular cells may also be present.1 2 5 7 The nuclei are often eccentrically placed within the cells (particularly in epithelioid cells, producing a plasmacytoid appearance) and show variable degrees of cytological atypia. Binucleation and multinucleation (including giant cells) may be seen. The nuclei usually contain finely granular chromatin. Prominent nucleoli (macronucleoli) and INCIs are identifiable in approximately half of the cases.1 2 5–7 Macronucleoli are common in epithelioid tumours, but are decreasingly common in mixed and spindle cell tumours.1 Mitotic activity is variably identified (up to 17% of FNB cases).1 2 The cytoplasm is often granular, with well-defined cytoplasmic borders. Other cytological variations include cytoplasmic vacuolation, clear cells, signet-ring cells or rhabdoid cells.1 2 14 Cytoplasmic vacuolation is common (up to 25% of cases).2 4 6 The vacuoles may be fine and diffusely distributed in the cytoplasm or large, filling the cytoplasm, separated by thin cytoplasmic strands and compressing the nucleus to one edge of the cell.2 15 The background may be clean or may contain blood,6 haemosiderin,6 melanin, lipid, chronic inflammatory cells or necrotic debris.1 9 11 14
Melanin pigment is seen in 20–80% of melanomas in FNBs and may be found in the cytoplasm of tumour cells and/or macrophages, or extracellularly.1 2 5–7 9 11 16 Cytoplasmic melanin is usually fine and non-refractile, and appears blue–black with Romanowsky stains and brown–black with Papanicolau stains; however, it may condense into coarse granules resembling haemosiderin (which appears dark green with Romanowsky stains and golden brown with Papanicolau stains).2 14 16 The distinction between melanin and haemosiderin can be difficult with conventional stains, and melanin pigment can be confirmed histochemically with Schmorl or Fontana Masson stains.5 16 Other cytoplasmic pigments such as phagocytosed carbon or lipofuchsin are rarely encountered in fluid specimens. However, the presence of melanin pigment is not diagnostic of melanoma, as it can be seen in a variety of tumours, including medullary thyroid carcinomas, schwannomas, ganglioneuroblastomas, bronchial carcinoid tumours, pigmented dermatofibrosarcoma protuberans and paragangliomas, as well as benign conditions such as dermatopathic lymphadenitis.14 16 These are more likely to occur as solid tumours than in an effusion and are therefore more likely to present diagnostic difficulty at FNB than in a fluid specimen.
Owing to the wide variation in appearances of melanoma in FNB smears, ancillary studies (cytochemistry for melanin pigment, immunocytochemistry and/or electron microscopy) have been shown to be very useful in making a conclusive diagnosis,14 17 and this would also apply to fluid specimens.13 Melanoma is usually positive for S-100, HMB-45 and Melan-A, as well as a host of other melanocytic markers.4 7 Melan-A and MART-1 are highly sensitive and specific immunochemical markers for melanoma when applied to cytological preparations.18
Melanoma in exfoliative fluid specimens and differential diagnostic considerations
When evaluating fluid specimens from patients with a known history of melanoma, the possibility of other malignancies must always be considered, as second malignancies develop either synchronously or metachronously in up to 20% of patients with melanoma.1 7 11 Epithelial and non-epithelial tumours, such as poorly differentiated carcinoma, lymphoma and pleomorphic sarcoma may resemble epithelioid melanoma, especially when melanin pigment is not identifiable.1 4 15 Correlation with the clinical history and ancillary studies are important in achieving an accurate diagnosis.4 13 19
Adenocarcinomas are composed of epithelioid cells arranged in various combinations of cell groups and dissociated cells, with malignant nuclear features and prominent nucleoli.14 Although cell aggregates are unusual in melanoma, they were found in 12.5% of cases in this study. A predominance of cell aggregates over cell dissociation raises the possibility of carcinoma. Cell aggregates in melanoma are generally syncytial with indistinct cellular borders, while those in carcinomas are often three-dimensional, sometimes with central acinar formation, and show distinct cell borders.1 Metastatic adenocarcinoma of the breast (particularly lobular carcinoma) and poorly differentiated gastric carcinoma may show prominent cellular dissociation and plasmacytoid morphology, closely resembling melanoma (fig 2A,B).14 20
Smears of melanoma composed of a dispersed population of atypical cells may resemble atypical lymphoid cells21–23 (such as those seen in Epstein–Barr virus infection),22 23 small or large cell lymphoma,1 6 7 21 24 25 acute myeloid leukaemia, plasmablastic lymphoma26/plasma cell myeloma27 if the cells exhibit plasmacytoid features.13 Smears of plasmablastic lymphoma are composed of dispersed cells with scant to moderate amounts of cytoplasm, eccentrically placed rounded nuclei with vesicular chromatin, one or more prominent nucleoli, along with multinucleate cells and mitotic figures. The atypical plasma cells are positive for CD138, CD79a and either kappa or lambda, and negative for CD20 and melanocytic markers.26 Large binucleate tumour cells with eosinophilic nucleoli may be mistaken for Reed–Sternberg cells, and a conclusion of Hodgkin lymphoma may be reached,5 especially in the presence of a concomitant inflammatory cell component in the background. In addition to immunochemistry in cell block preparations, which may be impossible to perform because of low cellularity (especially in cerebrospinal fluid (CSF) specimens), flow cytometric evaluation of fluid samples may be helpful in confirming a diagnosis of lymphoma, particularly B-cell non-Hodgkin lymphoma.28
Smears of melanoma may be composed of a predominance of cells with rhabdoid morphology (large, eccentrically placed nuclei, prominent nucleoli and abundant densely staining cytoplasm containing prominent globular inclusions), appearances seen in ∼5% of FNB specimens of metastatic melanoma.19 These cases may raise the possibility of a rhabdomyosarcoma2 or other tumours with rhabdoid morphology, such as renal and extrarenal rhabdoid tumour,19 the rare atypical teratoid/rhabdoid tumour29 and rhabdoid meningioma30 in CSF specimens, and various carcinomas and sarcomas.19
In body cavity fluid specimens (peritoneal and pleural fluid) and in body cavity washing specimens (peritoneal and pelvic washings), mesothelial cells and histiocytic cells may resemble epithelioid melanoma cells (fig 3A,B),13 31 32 particularly in the setting of reactive conditions, such as collagen vascular disease, cardiac failure, infection, cirrhosis, renal failure, peritoneal dialysis, pancreatitis, neoplasia and exposure to radiation or chemotherapeutic agents.33 34 Occasionally, reactive mesothelial cells show moderate to marked anisokaryosis and increased nuclear/cytoplasmic ratios, prominent nucleoli and occasional cell groups associated with psammoma bodies (fig 3A,B).31 32 Multinucleated cells and mitotic figures may be identified.33 Reactive mesothelial cells may also exhibit spindle shapes (usually with admixed epithelioid cells), causing difficulties in distinction from spindle cell melanoma.35 When identifiable, the presence of a predominant monolayer arrangement, intercellular “windows”, fine even nuclear chromatin and a gradation of cellular changes from typical benign mesothelial cells to those with reactive features are useful diagnostic features,31 but, in rare cases, the reactive mesothelial cells may contain large nuclei with irregular nuclear membranes, granular hyperchromatic chromatin and prominent nucleoli, mimicking malignancy.33–35 Confirmatory immunochemistry (positivity for mesothelial cell markers and negativity for melanocytic markers) is often helpful in arriving at the correct diagnosis.34–36
Take-home messages
Melanoma can show a broad range of cytological appearances in exfoliative fluid specimens.
Reactive mesothelial cells and histiocytes/macrophages may also mimic melanoma cells, particularly in body cavity (pleural, peritoneal) fluid specimens.
Malignancies that may mimic melanoma in fluid specimens include adenocarcinomas and haematological neoplasms, especially those exhibiting plasmacytoid morphology.
Careful evaluation of cytomorphological features, along with clinicopathological correlation, will usually allow accurate diagnosis.
Epithelioid histiocytes and macrophages are very common in fluid specimens. Often, the macrophages contain cytoplasmic haemosiderin pigment, which can resemble melanin. Some cases of melanoma may be composed of tumour cells with eccentrically placed nuclei with prominent nucleoli (but not macronucleoli) and finely dispersed chromatin,1 14 which are features exhibited by reactive macrophages (fig 3B).37 In CSF specimens, reactive cells of the monocyte/macrophage lineage and reactive lymphoid cells may be present in various settings (both reactive conditions and secondary to the presence of malignancy elsewhere in the brain). These reactive cells may be large and may exhibit many morphological features in common with melanoma cells.38
Because no diagnoses were altered on review, the specificity for a diagnosis of melanoma was 100%. However, it was not possible to determine the sensitivity for melanoma diagnosis in this series, as our laboratory information system does not permit the kind of case search that would accurately identify all false negative diagnoses of melanoma in fluid specimens and hence determine the sensitivity of melanoma diagnosis in this situation. In order to determine the sensitivity rate, we would have needed to review all fluid samples reported at our institution during the study period and to correlate them with clinical and histological findings. It was due to the enormous requirements of such an undertaking that we designed the study focusing simply on the findings in specimens containing melanoma.
In conclusion, melanoma may metastasise to a variety of body sites within which melanoma cells may be exfoliated, as shown by the results of this study. It is important to consider the possibility of metastatic melanoma when evaluating malignant cells in fluid specimens, particularly in patients with a previous history of melanoma. On the other hand, in patients with known melanoma, fluid specimens containing atypical cells should be critically evaluated for features of metastatic melanoma, but also for the various neoplastic and non-neoplastic conditions that can mimic the cytological features of melanoma. Careful assessment of the cytomorphology of the cells, use of ancillary cytochemistry and immunocytochemistry, and correlation with the clinical findings will enable accurate diagnosis.
Acknowledgments
We acknowledge the support of the Cancer Institute New South Wales, the Australian National Health and Medical Research Council, and colleagues at the Sydney Melanoma Unit and the Department of Anatomical Pathology, Royal Prince Alfred Hospital.
REFERENCES
Footnotes
Competing interests: None.
Funding: RM and RAS are Cancer Institute New South Wales Clinical Research Fellows.