[703] Can We Predict Lymph Node Count in Colorectal Cancer Resection Specimen Using Clinical, Pathologic and Molecular Variables?

S Ogino, K Nosho, N Tanaka, JL Hornick, CS Fuchs. Brigham and Women's Hospital/Harvard Medical School, Boston, MA; Dana-Farber Cancer Institute, Boston, MA; Harvard School of Public Health, Boston, MA

Background: The number of recovered lymph nodes is associated with improved survival of colorectal cancer patients. However, little is known on how the node count is influenced by clinical, pathologic and molecular variables including host immune response, the length of resected colorectum, microsatellite instability (MSI) and CpG island methylator phenotype (CIMP).
Design: We constructed a multivariate linear or logistic regression model to predict the negative or total node count in 732 colorectal cancer cases, using clinical and tumoral features, including tumor location, stage, the length of resected colorectum, lymphoid reaction, mucin, signet ring cells, KRAS, BRAF mutations, p53, MSI and CIMP.
Results: The two most significant predictors for the raw negative node count were the length of colorectum and stage, followed by tumor location, MSI, KRAS mutation and peritumoral lymphocytic reaction. R-square of the multivariate model was only 0.19, indicating that 81% of variability in the negative node count remained unexplained. In multivariate logistic regression to predict ≥12 total nodes, the length of resected colorectum, tumor location and MSI were significantly associated with ≥12 total nodes.

ROC (receiver operator characteristics) curve based on the multivariate model showed an area under curve of 0.70, indicating a modest ability to predict ≥12 total node count.

Conclusions: The length of colorectum, marked peritumoral lymphoid reaction, stage II and MSI-high are associated with a high negative node count. The negative or total lymph node counts vary greatly even after accounting for clinical, pathologic and molecular variables in a multivariate linear or logistic regression model.
Category: Gastrointestinal

Tuesday, March 23, 2010 11:30 AM

Platform Session: Section E, Tuesday Morning


Close Window