The Melanoma Risk Loci as Determinants of Melanoma Prognosis

ABSTRACT Background Genetic risk factors of human cancer emerge as promising markers of clinical outcome. Recent melanoma genome-wide scans (GWAS) have identified loci associated with disease risk, nevi or UV/pigmentation, but prognostic potential of these variants is yet to be determined. In this study, we performed the first-to-date systematic evaluation of the association between established melanoma risk loci and progression. Methods 891 melanoma patients prospectively accrued and followed up at NYU Medical Center were studied. We examined the association of 108 melanoma susceptibility single nucleotide polymorphisms (SNPs), selected or imputed from recent GWAS on melanoma, nevi or pigmentation, with recurrence-free survival (RFS) and overall survival (OS). The genotyping was performed using Sequenom I-plex. Univariate and multivariate Cox PH model were used to test association between each SNP and RFS and OS. Multivariate adjustments included age at diagnosis, gender, tumor stage, histological subtype, thickness, ulceration status, and mitotic index. ROC curves were used to measure utility of SNPs in predicting 3-year recurrence. Results Strong associations were observed from the multivariate analysis for rs7538876 (RCC2) with RFS (HR = 1.48, 95% CI 1.20 – 1.83, p = 0.0002) and rs9960018 (DLGAP1) with both RFS and OS (HR = 1.43, 95% CI = 1.07-1.91, p = 0.0159, HR = 1.52, 95% CI = 1.09-2.22, p = 0.0138, respectively). Subgroup analyses for ulceration and tumor thickness have identified several SNPs which show associations for RFS and OS among these different groups. In addition, we identified the classifiers rs7538876 and rs9960018, that when combined with stage and histological type, achieve a higher area under the ROC curve (AUC = 84%), compared to the baseline (AUC = 78%), in predicting 3-year recurrence. Conclusions Our data revealed associations between specific melanoma susceptibility variants and worse disease outcome. The strength of associations observed for rs7538876 and rs9960018 suggest biological implication of these loci in melanoma recurrence. The observed predictive patterns of associated variants with clinical outcome provide the evidence for the potential utilization of genetic markers in melanoma prognostication. Disclosure All authors have declared no conflicts of interest.