Abstract 2861: Independent validation of a prognostic gene-signature based risk score in formalin-fixed paraffin-embedded melanomas

Current staging of melanoma, as defined in 2009 by the American Joint Committee on Cancer (AJCC), is based mainly on histopathological criteria but is limited in predicting outcome. Complementary molecular markers are not available for routine prognostic assessment. We have previously identified and validated a prognostic nine-gene signature expressed in fresh-frozen (FF) primary cutaneous melanomas (training cohort: n=91; validation cohort: n=44). A signature-based risk score predicts patient overall survival (OS) independently of AJCC staging (multivariate regression analysis: p = 0.0004; hazard ratio: 3.8). However, clinical application requires adaptation to formalin-fixed, paraffin-embedded (FFPE) melanomas. Therefore, we have transfered signature expression analysis onto FFPE melanomas. From FFPE melanomas matching the training and validation cohorts of the above FF melanoma study (n=125), RNA was prepared and transcribed into cDNA. Following cDNA pre-amplification, expression of the 9 signature genes, 2 additional candidate genes, and 3 housekeeping genes was quantified by real-time PCR. Correlation of gene expression with OS was evaluated using Cox regression analysis. Expression of a signature of 8 out of 11 genes (risk gene: KBTBD10; protective genes: DCD, GBP4, COL6A6, PIP, SCGB1D2, SCGB2A2, KRT9) was associated with OS in univariate regression and Kaplan Meier analysis. A signature-based risk score predicted OS independently of AJCC staging (multivariate analysis: p=0.0059, hazard ratio 3.09). The misclassification rates were 20% overall, 13.8% for low risk, and 5.7% for double low-risk (combined with AJCC staging). The risk score complemented and refined conventional AJCC staging. Thus, the FF melanoma risk score was successfully transfered onto FFPE melanomas. In order to validate the FFPE melanoma risk score, we analyzed signature expression in an independent cohort of 130 selected FFPE melanomas, which were particularly difficult to classify by AJCC staging (misclassification rate 40.8%), in order to stringently test the performance of the risk score. The misclassification rate of the FFPE melanoma risk score was comparable, even slightly better (39.2%) than that of AJCC staging, confirming its prognostic performance. The FFPE melanoma risk score was also externally validated in a Molecular Diagnostics Lab (Dermatologikum Hamburg). The concordance of melanoma classification exceeded 85%, demonstrating technical robustness of the risk score.