A 19-gene prognostic GEP signature (DecisionDx-Thymoma) to determine metastatic risk associated with thymomas.

Abstract

7106 Background: The treatment of thymomas is predominantly based on the stage of disease. Histologic classification is of limited value as all types of thymomas can give rise to metastases. In order to better predict the metastatic behavior of these tumors, we performed genome-wide gene expression analysis and identified a set of genes associated with presence or absence of metastases. In the current study, we sought to further develop and validate the gene signature using quantitative RT-PCR analysis. Methods: Thymomas with archived blocks and long-term follow-up data were reviewed. This training set study consisted of 50 cases, including 34 cases on which the discovery microarray analysis had been performed. RNA was extracted from 5 x 10-micron thick sections in a CAP-accredited CLIA certified laboratory and analyzed by RT-PCR using custom TLDA cards on an ABI7900HT instrument. Expression data and biostatistical analysis were performed using GeNorm and JMP Genomics (SAS). Predictive modeling using Partition Tree Analysis (PTA) and Logistic Regression Analysis (LRA) was performed. Metastasis-free survival (MFS) was assessed using Kaplan-Meier analysis. Results: A 19-gene expression profile (GEP) signature was developed using a cohort of 50 thymomas for predicting metastasis. PTA yielded ROC of 0.97 (met. accuracy = 96%, non-met. accuracy = 81%), while LRA yielded ROC of 0.895 (met. accuracy = 87%, non-met. accuracy = 85%). PTA classification showed 5-year MFS rates of 100% and 31% for predicted low risk (Class 1) and high risk (Class 2) of metastasis (median MFS = NR and 4.1 yrs, resp., P<0.0001 Log-Rank), respectively. LRA showed 5-year MFS rates of 100% and 17% for predicted Class 1 and high risk Class 2 of metastasis (median MFS = NR and 2.9 yrs, resp., P<0.0001 Log-Rank), respectively. Analysis of additional cohorts is ongoing. Conclusions: We have successfully completed development of a 19-gene signature (DecisionDx-Thymoma) that appears to predict metastatic behavior of thymomas more accurately than traditional staging. If validated in larger cohort, this signature will provide insight for the future management of patients with this rare malignancy.