A Gene Expression Signature to Predict Metastasis and Survival in Non-small Cell Lung Cancer

Background: Current histopathological methods are inadequate for predicting outcome and recurrence in patients with non-small cell lung carcinoma (NSCLC) after surgery. In this study, we investigated the use of gene expression signatures to predict outcome and metastasis in lung cancer patients. Methods: Gene expression was studied by microarray and the real-time reverse transcriptase polymerase chain reaction (RT-PCR) in normal and lung tumor tissue of 188 NSCLC patients who underwent surgical resection. The 5 cancer-related genes and 1 reference gene expression levels measureed by real-time RT-PCR were used in a prospectively defined algorithm to determine the risk for each patient. Finally, we used an independent cohort to verify the 5 gene-based predictive model derived from decision tree analysis. Results: The 5 gene-based decision tree model was able to predict the prognosis. The recurrence rate at 36 months was 53% in the low-risk group versus 83% in the high-risk group (P=0.002). The 5 gene-based model could also predict overall survival (P<0.001). In multivariate analysis, the decision tree model predicted that high-low dichotomy and stage were both significant for recurrence. In addition, it could also predict metastasis and survival of NSCLC patients within the stage I-II subgroups. A similar result was found using an independent cohort of NSCLC patients. The high-risk patients had a significantly poorer overall survival than the low-risk patients (P=0.005). We also found distinct gene signatures which could distinguish between NSCLC, and normal tissue and histology subtypes. Conclusions: A gene expression signature can predict metastasis and survival of NSCLC patients. Keywords: Survival and metastasis analysis