Comparison of microarray analysis of fine needle aspirates and tissue specimen in thyroid nodule diagnosis

Microarray technology provides a new opportunity to improve the diagnostic accuracy of fine needle aspiration (FNA) in evaluating thyroid nodules. Here, we evaluate whether ex vivo FNA and tissue samples can be used interchangeably in microarray and whether the method of acquisition affects the precision of the gene list that is generated. To assess whether FNA samples provide adequate material for reliable gene expression analysis, paired tissue and FNA samples were collected from 13 thyroid nodules; 7 malignant, 6 benign. RNA was extracted from each specimen, converted to complimentary DNA and hybridized to AffymetrixU-133 GeneChips. Cluster analysis was then performed using 61 genes predetermined to differentiate benign from malignant nodules. Clustering patterns were evaluated using 2-group K-means and hierarchical analysis. Twelve concordant pairs were used to generate differentially expressed genes between the sampling methods. Twenty-five of 26 samples clustered concordantly with the pathologic diagnosis. The sensitivity, specificity, and accuracy were 100%, 100%, and 100% for FNA and 85.7%, 100%, and 92.3% for tissue, respectively. Two-group K-means revealed an adjacent grouping for 12 of 13 pairs. Hierarchical analysis clustered 8 of 13 pairs together. Sixty-seven genes were differentially expressed between FNA and the tissue sampling methods. These genes predominantly represented stromal components and were upregulated in the tissue compared with FNA samples. We conclude that FNA is a reliable alternative to tissue samples in predicting malignancy with microarray.