The DecisionDx-Melanoma™ gene expression profile test is a robust, high-accuracy, high-complexity, multivariate assay that stages patients with a low risk of developing metastatic disease from those with a high risk. The test, along with specimen collection techniques, has been validated on archival tissue from formalin-fixed, paraffin-embedded (FFPE) samples.
Castle Biosciences Inc., which discovered and developed the assay, analyzed over 300 cutaneous melanoma samples obtained from seven centers in the U.S. RNA was isolated from formalin-fixed, paraffin-embedded (FFPE) biopsies or wide local excisions of primary cutaneous melanoma (CM) from patients with Stage I-IV (82% were Stage I or II) converted to cDNA and analyzed using RT-PCR. The training set was comprised of 164 samples. Independent validation was performed on an additional 104 CM samples.
Using a pre-specified predictive algorithm, the gene expression profile (GEP) signature of a cutaneous melanoma tumor analyzed with the DecisionDx-Melanoma assay was compared to the expression pattern of the training set of patients with known outcomes.
Analysis of the training set showed a sensitivity of predicting metastatic disease of 85% with an area under the Receiver Operating Characteristic (ROC) curve of 91%. Prediction of metastatic risk for the independent validation set resulted in a sensitivity of predicting metastatic disease of 89% and an ROC of 91% (Table 2).
Independence of the DecisionDx-Melanoma test was evaluated using Cox regression analysis for patients with Stage I or II melanoma. Cox regression analysis was performed using both the individual AJCC staging criteria of Breslow’s thickness, ulceration, and mitosis as well as the combined factors of AJCC stage. The analysis showed that Breslow’s thickness is an important predictor of metastatic disease as is AJCC stage. The DecisionDx-Melanoma test was also shown to be an important predictor of metastasis.
Importantly, multivariate analyses found the DecisionDx-Melanoma test to be superior to and independent of Breslow’s thickness, ulceration status and mitotic index in predicting metastatic disease.
Similarly, multivariate analyses found the DecisionDx-Melanoma test to be superior to and independent of AJCC stage (Stage I, IIA vs IIB, IIC) in predicting a patient’s individual risk of metastatic disease.