A 19-gene prognostic GEP signature to determine metastatic risk associated with thymomas.

Abstract

68 Background: Therapeutic decisions for patients with thymomas are made on tumor stage as histologic classification is of limited prognostic value. A molecular prognostic tool may better predict the likelihood of metastasis in thymomas and enable better management of patients. Methods: Reverse-transcriptase-polymerase-chain-reaction (RT-PCR) assay for 23 genes (19 test and 4 reference genes) was performed on archival paraffin-embedded primary thymomas (n=36). The gene expression levels were used to compute gene expression profile (GEP) signature and classify tumors into low or high risk. The GEP was validated in an independent test set (n=75) of patients. Results: We developed and validated a 19-gene prognostic GEP signature for predicting metastatic behavior in thymomas. Using radial basis machine (RBM) modeling in the training set, both 5-year and 10-year MFS rates were 85% and 22% for predicted low risk and high risk of metastasis (p<0.0005, log-rank), respectively. For the validation set, 5-year MFS rates were 93% and 35% for predicted low and high risk patients (p<0.0018 log-rank), respectively. The 5-year MFS rates for Masaoka stage in the validation set were 53% and 41% for stage I/II and stage III/IV(p<0.0433 log-rank), respectively. In a multivariate Cox model, the GEP was the only independent predictor of metastasis free survival compared to age at diagnosis and stage (p=0.048). Conclusions: Using archival tissues, we identified and validated a 19-gene signature that predicts likelihood of metastases of thymomas more accurately than traditional staging. The signature provides insights for both patient stratification and targeted therapy.