The genomic signature of the DecisionDx-Melanoma™ test was discovered, developed and validated in a series of studies at Castle Biosciences that involved numerous collaborative centers (see flow chart below). Work began in 2009 to identify a putative genomic signature and initiate clinical studies.
Work from an initial development study including three collaborators was presented at ASCO 2012 (see Figure 2 below). This study had strong indicators that the 31 gene expression profile test, or a smaller number of genes, had the potential to more accurately predict individual metastatic risk in patients with lower risk Stage I or II disease than the current AJCC staging system. This finding was viewed as a potentially significant advance in the staging of patients with cutaneous melanoma.
The collaborative group was rapidly expanded to include seven centers, and the initial training and independent clinical validation studies were completed in early 2013. These data were presented at ASCO 2013.
Several studies are ongoing, including a second validation study of the use of the DecisionDx-Melanoma test in predicting metastatic disease as well as other potential uses.
Development study data reported at ASCO 2012
Results of the development studies, presented at ASCO 2012, showed a highly accurate assay (ROC=0.92; accuracy=89%) able to significantly stratify Class 1 cases with 5-year survival of 100%, from Class 2 cases, with 5-year survival of 38% (p<0.0001); Figure 2.
Comparison of DecisionDx-Melanoma gene expression profile to other prognostic factors used to predict recurrence in cutaneous melanoma patients indicated that the DecisionDx-Melanoma assay is an independent predictor of metastatic risk. Cox univariate and multivariate regression analysis were performed for those patients in the 107 sample development cohort, following prediction of metastatic risk, who had clinical Stage I or II cutaneous melanoma. GEP was compared to the factors comprising AJCC pathologic stage: Breslow’s thickness, mitotic rate, and ulceration status. As shown in Table 2, under univariate analysis, Breslow’s thickness, ulceration status and GEP Class 2 are significant predictors of metastasis (p<0.001). However, multivariate analysis indicates that only GEP Class 2 is a significant predictor (Hazard ratio = 148; p = 0.0001).