To detect changes in gene expression data from DNA microarrays, a fixed threshold value is used in various studies. However, it is not always guaranteed that a threshold value which is appropriate for highly expressed genes is suitable for genes with low expression. To address this issue, we have proposed adaptive threshold, which has different values for different expression levels. In this study, the performance of the adaptive threshold method was investigated through simulations. The sensitivity in various noise conditions was in a range between 72.7 and 100% while the specificity was better than 99% for all noise conditions. These results demonstrated the good performance of the proposed method.