RESUMO
Although the clinical and pathologic diagnosis of some melanomas is clear-cut, there are many histopathologic simulators of melanoma that pose problems. Over-diagnosis of melanoma can lead to inappropriate therapy and psychologic burdens, whereas under-diagnosis can lead to inadequate treatment of a deadly cancer. We used existing data on DNA copy number alterations in melanoma to assemble panels of fluorescence in situ hybridization (FISH) probes suitable for the analysis of paraffin-embedded tissue. Using FISH data from a training set of 301 tumors, we established a discriminatory algorithm and validated it on an independent set of 169 unequivocal nevi and melanomas as well as 27 cases with ambiguous pathology, for which we had long-term follow-up data. An algorithm-using signal counts from a combination of 4 probes targeting chromosome 6p25, 6 centromere, 6q23, and 11q13 provided the highest diagnostic discrimination. This algorithm correctly classified melanoma with 86.7% sensitivity and 95.4% specificity in the validation cohort. The test also correctly identified as melanoma all 6 of 6 cases with ambiguous pathology that later metastasized. There was a significant difference in the metastasis free survival between test-positive and negative cases with ambiguous pathology (P=0.003). Sufficient chromosomal alterations are present in melanoma that a limited panel of FISH probes can distinguish most melanomas from most nevi, providing useful diagnostic information in cases that cannot be classified reliably by current methods. As a diagnostic aid to traditional histologic evaluation, this assay can have significant clinical impact and improve classification of melanocytic neoplasms with conflicting morphologic criteria.