Purpose: Many patients die of recurrent NSCLC. Current clinical and molecular tests fail to predict patients at high risk for recurrence. We investigated the relationship between gene promoter methylation and recurrence of NSCLC and its prediction accuracy.

Materials and Methods: We prospectively enrolled patients who underwent surgery for T1-2N0 NSCLC. We collected primary tumors, regional lymph nodes (RLN) and mediastinal lymph nodes (MLN). A total of 112 patients were included; cases (n = 39) and controls (n = 73) were defined based on cancer recurrence within 40 months of surgery. Promoter methylation of eight cancer-related genes (CDO1, TAC1, SOX17, p16, CDH13, RASSF1A, APC and MGMT) was obtained by using nanoparticle-based DNA extraction followed by qMSP. Logistic regression analysis adjusted by age, gender, race, tumor size and pack year was used to study the relationship between the gene promoter methylation and the risk of recurrence of NSCLC; machine learning assessed its prediction accuracy.

Results: Cases had significantly higher p16 methylation levels than controls in all three tissue types by univariate and multivariate analysis after adjusting for confounders (age, gender, race, tumor size, smoking history, and other gene methylation levels). SOX17 was significantly methylated among cases in both RLN and MLN samples, and remained significant in MLN after adjusting for confounders. Table 1 shows the prediction accuracy of p16 and SOX17 using tumor tissue alone and combined with RLN and/or MLN. Prediction accuracies were increased after adding biomarkers from lymph nodes.

Conclusion: Increased methylation status in the promoter region of p16 and SOX17 is associated with an increased risk of cancer recurrence among patients with T1-2N0 NSCLC treated with curative resection. Adding molecular analyses of lymph nodes to primary tumor samples significantly increased the diagnostic accuracy of predicting recurrence in these patients.

Table 1.

Prediction accuracy using tumor and combined with RLN and/or MLN.

Tissue SamplesSensitivitySpecificityPPVNPV
Tumor 0.41 0.84 0.52 0.76 
Tumor+RLN 0.68 0.96 0.90 0.84 
Tumor+MLN 0.56 0.88 0.71 0.79 
Tumor+RLN+MLN 0.69 0.90 0.79 0.85 
Tissue SamplesSensitivitySpecificityPPVNPV
Tumor 0.41 0.84 0.52 0.76 
Tumor+RLN 0.68 0.96 0.90 0.84 
Tumor+MLN 0.56 0.88 0.71 0.79 
Tumor+RLN+MLN 0.69 0.90 0.79 0.85 

Citation Format: Chen Chen, Andrew Yang, Devlin Danielle, Kristen Rodgers, Peng Huang, Zhihao Lu, Candace Griffin, Beverly Lee, Richard Battafarano, Fenglei Yu, Tza-Huei Wang, Stephen Baylin, James Herman, Alicia Hulbert, Malcolm Brock. DNA methylation as biomarkers to predict early recurrence of T1-2N0 lung cancer. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 4455.