Association of RASSF1A and p63 with Poor Recurrence-Free Survival in Node-Negative Stage I–II Non–Small Cell Lung Cancer

Purpose: This study was aimed at analyzing the recurrence-related prognostic significance of 12 candidate molecular biomarkers in node-negative stage I–II non–small cell lung cancer (NSCLC). Experimental Design: We retrospectively analyzed promoter methylation of eight genes using methylation-specific PCR in formalin-fixed and paraffin-embedded tissues from 328 node-negative stage I–II NSCLCs. The expression of Bcl-2, E-cadherin, p53, and p63 proteins was also assessed by immunohistochemistry. Results: Recurrence was found in 145 (44%) of 328 node-negative stage I–II NSCLCs with a median follow-up period of 6.2 years. No association was found between recurrence and alteration of individual biomarker in univariate analysis. We defined recurrently divergent groups on the basis of recursive partitioning analyses for 12 biomarkers and found a significant association of co-alteration of RASSF1A and p63 with poor recurrence-free survival (RFS). Cox proportional hazards analysis showed that hypermethylation of RASSF1A and negative expression of p63 was associated with poor RFS [HR, 1.93; 95% confidence interval (CI), 1.13–5.47; P = 0.009] compared with those without co-alteration of RASSF1A and p63, after adjusting for age, adjuvant therapy, histology, and tumor size. Random forest classifier including RASSF1A and p63 showed best performance in the prediction of recurrence in node-negative stage I–II NSCLCs: area under receiver operator characteristic curve for random forest was 0.91 and error rate for the model was 17%. Conclusion: The present study suggests that RASSF1A and p63 may be independent prognostic indicators for RFS in node-negative stage I–II NSCLCs. Clin Cancer Res; 19(5); 1204–12. ©2012 AACR.