A multi-gene assay to predict long-term mortality in early stage lung adenocarcinoma

Abstract

7569 Background: The likelihood of long-term mortality for patients with early stage lung adenocarcinoma is poorly defined by clinical stage and histopathological findings. Our hypothesis was that a multigene quantitative polymerase chain reaction (PCR) assay can predict risk of mortality among patients with early stage lung adenocarcinoma. Methods: We identified 65 genes that were previously identified as prognostic for long-term mortality in early stage lung cancer in 3 published microarray studies and 2 PCR-based studies. RNA was extracted from 124 fresh-frozen tumor samples from consecutive patients with completely resected lung adenocarcinoma with at least 3 years of clinical follow-up. 80 samples were randomly assigned to a test group and the remainder assigned to a validation group. Real-time PCR of the 65 identified genes were run on the test set using Taq-man assays. A prediction model was created using a proportional hazards model of normalized gene expression levels using backwards model selection. A model score was calculated for each patient using model coefficients and individual gene expression levels. Patients were defined as high-risk if the model score was greater than the median score. Results: Adequate real-time PCR profiles were identified in all 80 patients. Eighteen genes were included in the final model. The proportion of patients identified as high-risk and low-risk was 52% and 48% percent, respectively. The Kaplan-Meier estimated five-year survival in the low-risk group was 82% and 5% in the high-risk group (P<0.001, log-rank test). Median survival was 22 months in the high-risk group and was not reached in the low-risk group. In multivariate survival analysis, the prognostic score predicted survival independent of tumor stage and size (P<0.001). Prognostic score predicted mortality better than clinical stage, based on model log-likelihood values (P<0.001). Conclusions: This multi- gene assay aids in predicting long-term mortality among patients with surgically resected early stage lung adenocarcinoma. We are currently validating this model in our test set of samples. No significant financial relationships to disclose.