 |
|
| MicroRNA expression profiles predict recurrence after surgery for stage 1 non-small cell lung cancer | |
Sai Yendamuri1, Steen Knudsen2, Todd L. Demmy1, Santosh Patnaik1; 1Roswell Park Cancer Institute, Buffalo, NY; 2Medical Prognosis Institute, Horsholm, Denmark
Methods: Formalin fixed paraffin embedded (FFPE) tissue specimens from 79 patients with pathologic stage 1 NSCLC were used for analysis. Tissue was deparaffinized and miRNA extracted. After quality control assessments of the extracted RNA, hybridization was performed to a locked nucleic acid based array platform containing probes for all miRs in miRBase version 11. Data from the arrays were background corrected and Loess normalized. In a leave-one-out cross validation, miRNAs differentially expressed between patients with recurrence and patients without, were selected with a t-test, using a multiple testing correction leaving a false discovery rate of 1%. The resulting miRNAs were subjected to Principal Component Analysis. The five most important components trained a multivariate classifier using the classification algorithms: K nearest neighbor, nearest centroid, neural network and support vector machine. The left out sample was predicted by majority vote among the classification algorithms into “Good Prognosis” or “Poor Prognosis”. A Kaplan-Meier plot was prepared of the time to recurrence for the “Good Prognosis” and “Poor Prognosis” groups. A log-rank test for statistical significance of the difference between the two groups was performed. As a leave one out cross validation was performed, separate internal training and test sets were not created. Results: Of 79 samples, 78 samples passed the quality control conditions for hybridization. Data analysis performed as detailed above led to a model containing over 100 miRNA included in the five principal components. This model predicted outcome in a statistically significant fashion (Figure 1). Median time to recurrence in “Good Prognosis” tumors had not been reached, whereas the median time to recurrence in “Poor Prognosis” tumors was 22 months (p<0.01). Conclusion: This miRNA microarray profile predicts recurrence after surgery for stage 1 NSCLC and deserves validation by datasets from other institutions. Furthermore, ease in the handling of input material (avoiding frozen tissue) and stability of miRNA to degradation makes this platform more practical than mRNA-based technologies in all clinical environments. 
Back to 2009 Annual Meeting Back to Program Outline Back to Main Program |
RSS Feeds
We Model Excellence |
|
Copyright © American Association for Thoracic Surgery. All rights reserved.
Read the Privacy Policy.
IMPORTANT REMINDER: The preceding information is intended only to provide
general guidance and not as a definitive basis for diagnosis or treatment in any particular case.
It is very important that you consult a doctor about any specific medical problem or question.
Read the Privacy Policy.
IMPORTANT REMINDER: The preceding information is intended only to provide
general guidance and not as a definitive basis for diagnosis or treatment in any particular case.
It is very important that you consult a doctor about any specific medical problem or question.