Abstract P3-10-18: Early Detection of Recurrent Breast Cancer Using Metabolite Profiling

Abstract

Abstract Background: The growing field of metabolite profiling (or metabolomics) focuses on the quantitative detection of multiple small molecules in biological systems that result from pathophysiological stimuli. As is the case for early detection of primary breast cancer, it is anticipated that early diagnosis of breast cancer recurrence will not only improve survival but also help clinicians determine the best therapeutic strategies for patients by avoiding under or over treatment. We report on the development of an early test for recurrent breast cancer using metabolite profiling methods. Methods: We applied a combination of nuclear magnetic resonance (NMR) and gas chromatography-mass spectrometry (2D GC-MS) to analyze the metabolite profiles of 257 serial serum samples from breast cancer patients consisting 116 samples from breast cancer recurrence and 141 samples from breast cancer patients with no evidence of disease (NED). NMR and GC-MS data were analyzed by combining advanced univariate and multivariate statistical methods and comparison of individual spectral features between patients with and without recurrent breast cancer. Results: From multivariate analysis of hundreds of spectral features, ten metabolite markers (7 from NMR and 3 from GC-MS) were targeted to build logistic regression model, which yielded a prediction model with a high accuracy (AUROC >0.89 using 5 fold cross validation) with a sensitivity of 86% and specificity of 89%. When the model was tested using leave one patient cross validation it yielded a sensitivity of 75% and a specificity of 83% (AUROC >0.88). In addition, strikingly, over 55% of the patients could be correctly predicted to have recurrence as early as 13 months before the recurrence was actually diagnosed clinically, representing a large improvement over the current diagnostic assay CA 27.29. A second statistical approach using independent training and testing sets yielded very similar results. To the best of our knowledge, this is the first study to develop and validate a prediction model for early detection of recurrent breast cancer based on the metabolic profiling. The combination of complementary NMR and MS analytical methodologies is particularly useful for building accurate profiles. Conclusion: Metabolite profiling methods provide a powerful approach for the development of diagnostic tests for monitoring breast cancer patients. Citation Information: Cancer Res 2010;70(24 Suppl):Abstract nr P3-10-18.