Characterization of germline pathogenic variants in melanoma patients.

Abstract

9590 Background: Approximately 10% of melanomas arise from hereditary predisposition. NCCN guidelines recommend genetic counseling for patients with ≥3 invasive melanomas, or a mix of invasive melanoma, pancreatic cancer, and/or astrocytoma in an individual or family. However, predictors of germline pathogenic/likely pathogenic variants (P/LPV) in patients undergoing germline testing remain poorly characterized. Methods: A retrospective study of all melanoma patients referred for any indication to the University of Michigan Cancer Genetics Clinic from 2015-2024 was performed. Patients who declined testing or had known hereditary cancer syndromes (personal or familial) were excluded. Baseline demographics, referral reasons, melanoma characteristics, and genetic testing results were collected. Chi-square and Fisher’s exact tests were used to evaluate associations with P/LPV (p<0.05). Results: Among 624 patients, median age at melanoma diagnosis was 52 (IQR 38-63); 60.1% were female. The most common referral reasons were new melanoma diagnosis (64.4%), family history of non-melanoma cancers (57.1%), or family history of melanoma (39.4%). Overall, 88 (14.1%) harbored P/LPV (92 variants across 26 genes). P/LPV in established melanoma-risk genes were identified in 42 patients (6.7%): CDKN2A (n=16), MITF (n=8), POT1 (n=5), BRCA1/2 (n=6), BAP1 (n=3), TP53 (n=3), and PTEN (n=1). The strongest associations with any P/LPV were personal history of ≥2 additional non-skin primary cancers (OR 2.35; p=0.014) and ≥2 primary melanomas (OR 1.80; p=0.017). Diagnostic category (in situ, invasive, metastatic) and subtype (cutaneous, uveal, mucosal, acral, unknown primary) showed statistical significance with P/LPV (p<0.001), though effect sizes were modest. Age at diagnosis (<45, <55, <65), sex, and family history of melanoma or pancreatic cancer in 1˚ or 2˚ relatives were not associated with P/LPV. In a secondary analysis restricted to established melanoma-risk genes (n=42), ≥2 primary melanomas remained strongly predictive (OR 4.11; p<0.001) and 1˚ family history of melanoma became significant (OR 2.01; p=0.03), while ≥2 additional non-skin primary cancers was not associated (p=0.40). Conclusions: In this melanoma cohort excluding patients with known hereditary syndromes, 14.1% had a P/LPV. Personal tumor burden, not family history, was the strongest predictor of a positive test, suggesting current NCCN criteria may miss high-risk individuals by emphasizing family history. For established melanoma-risk genes specifically, both personal history of multiple melanomas and 1˚ family history of melanoma were predictive, supporting a gene-specific approach to risk assessment. These findings support expanding referral criteria to include patients with ≥2 primary melanomas or ≥2 additional non-skin primary cancers, agnostic of family history, to better identify individuals most likely to benefit from germline testing.