Multivariate Analysis Of A Panel Of Protein Biomarkers For Early Detection Of Non-Small Cell Lung Cancer

Proceedings: AACR 101st Annual Meeting 2010‐‐ Apr 17‐21, 2010; Washington, DC Introduction: Lung cancer is the leading cause of cancer mortality in the US; it has been estimated that 219,440 individuals will be diagnosed with lung cancer and 159,390 will die from the disease in 2009 alone. Low dose CT, while very sensitive at visualizing early stage tumors, also identifies non malignant solitary pulmonary nodules (false positives) in a significant fraction of smokers. Biomarker panels may have considerable value when combined with imaging protocols in helping to discriminate benign from malignant nodules. We have employed a novel mass spectrometry-based approach to identify serum biomarkers which we have previously shown to detect non-small cell lung cancer (NSCLC) representing all 4 stages of disease. In this study we extend these findings to a cohort highly representative of early stage lung cancer. Methodology: Validation of non small cell lung cancer markers was performed by ELISA analysis on a cohort of sera collected at the NYU Langone Medical Center. In this study lung cancer cases represent predominantly early stage disease (65 stage I/ 91 all cancer). Controls comprise current and former healthy smokers (n=90) as well as subjects with non-malignant lung disease (COPD, n=46). Samples were randomly divided into a training set (cancer n=39; smokers n= 38; COPD n=20) and a test set (cancer n=52; smokers n=52; COPD n=26). Logistic regression analysis revealed several multi-marker panels capable of distinguishing malignant samples from matched controls in the training set. A classifier developed for each panel was then applied to the test set. Multivariate approaches were employed to analyze this data together with other key clinical parameters including: tumor stage, size, histology and lung-function data. Results: Markers for the panel were selected based on both individual marker performance and complementation of the performance of the relevant biomarker panel. Logistic regression analysis on the training set revealed a 9-marker panel which resolved malignant samples with 90% sensitivity at 96% specificity (AUC=0.977). The panel resolved malignant lesions under 1 cm and demonstrated good discrimination of COPD patients. A 6-marker panel was identified with similar performance characteristics (AUC=0.974). Applying the classifier generated from the training set to the test set also showed strong performance in this study, with an AUC of 0.92 for the 9 marker set in preliminary analyses. Conclusion: Panels of lung biomarkers identified initially through proteomic analysis have shown robust performance in a study cohort strongly biased toward early-stage disease. The findings from these studies suggest that these biomarker panels provide the performance and flexibility to design tests suitable for a variety of diagnostic applications such as screening high risk subjects prior to CT scanning and improving the discrimination of nodules identified by radiologic imaging. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 4568.