Molecular testing of borderline cutaneous melanocytic lesions: SNP array is more sensitive and specific than FISH

Abstract

Melanocytic lesions with borderline features are diagnostically challenging. Single-nucleotide polymorphism (SNP) arrays, which detect genomic copy number alterations (CNAs), can be helpful in distinguishing between nevi and melanoma. Fluorescence in situ hybridization (FISH) has been used as a more rapid, less expensive alternative to SNP array, using a panel of probes that are often gained or lost in melanoma. We used SNP array data from 63 borderline cutaneous melanocytic lesions and 44 definitive melanomas to predict the performance of FISH testing. Lesions were considered positive by "virtual FISH" if 1 or more of the 5 FISH-probed loci demonstrated appropriate CNAs by SNP array. Cases were classified as positive by SNP array if ≥3 CNAs were present, based on internal validation studies, or if FISH criteria were met. Conventional FISH was performed in 33 cases (17 borderline lesions, 16 melanomas). Of the 63 borderline cases, 44 (70%) were positive by SNP array and 30 (48%) were positive by virtual FISH. A higher proportion of melanomas were positive by SNP array (41/44, 93% sensitivity) and virtual FISH (36/44, 82% sensitivity). Virtual FISH had 61% sensitivity in the borderline group using SNP array as the gold standard, whereas specificity was 84%. There was good correlation between conventional and virtual FISH, with agreement in 30 of 33 (91%) cases. Although FISH is highly effective in distinguishing between nevi and melanoma in cases where the histological diagnosis is straightforward, it is not nearly as sensitive or specific as SNP array when applied to borderline lesions.