UICC World Cancer Congress 2006

Objective: Liver metastasis and lymph node involvement in patients with colorectal cancer (CRC) are major factors for predicting their clinical outcome and choosing appropriate adjuvant therapy after surgery, respectively. However, conventional diagnosis for above factors based on clinicopathological parameter is sometimes impertinent for accurate individual prediction of clinical outcome. It is important to develop the ability to precisely predict the metastatic potential in CRC. Methods: We investigated whole human gene expression profile in 155 patients with CRC (50 localized, 63 regional with node-positive and 42 with liver metastasis) using 30K DNA chip. Unsupervised analysis revealed that each expression profile was quite different. Four supervised prediction approaches were used to identify the metastasis-associated molecular signatures consisted of key role genes and discriminates that can distinguish between patients with and without metastasis. Then we analyzed their survival based on molecular signatures. Results: Mean of the overall accuracy for predictions of liver metastasis and nodal status were 87.5% and 82.9%, respectively in each external validation samples. Five-year survival rates between the subgroups in the patients on metastasis-associated signatures are significantly different (P<0.001), and the signature was indicated an independent prognostic parameter(P = 0.026). Conclusion: The clinical value is that the patients with metastasis-positive signature could receive more aggressive and appropriate treatments and closer follow up. Our findings suggest that microarray-based technique may provide better decision-making for the treatment strategy in CRC and new potential targets against carcinogenesis or new diagnostic markers. These benefits on molecular analysis may improve the classification of CRC.