Abstract P4-03-03: Determination of lymph node status using the primary tumor’s gene expression signature

Abstract

Abstract Introduction:Lymph node status is one of the most important clinical parameters of breast cancer. Axillary lymph node dissection and sentinel lymph node biopsy have considerable morbidity associated with them and a method to accurately predict lymph node positivity would be clinically useful. To date, no gene expression signature has been established that is capable ofestimatin lymph node positivity. Our aim was to develop a new predictor using a large set of patient samples and compare its performance to known clinical variables. Methods: An integrated database was constructed using publicly available Affymetrix microarrays with known lymph node status. The patients were divided into the four molecular subtypes of basal, luminal A, luminal B and HER2 positive using the St. Gallen criteria and the gene chip based gene expression data of estrogen receptor, HER2 receptor and MKI67. ROC analysis was performed for each gene within each subtype. Then, the top 10 genes of each list were combined, and using the expression data of these 40 genes across all patients, Manhattan or Euclidean distanceswere computed to calculate the distance between each sample. All the patients were ranked, and lymph node status was defined by using the proportional lymph node positivity of “x” nearest ranked patients. Multiple regression was performed to compare the classification using the gene expression data to known clinical variables. Statistical significance was set at p<0.01. Results: The database includes 2756 patients, of whom 983 are lymph node positive. To optimize the classification, different number of genes (all/top100/top40) and different number of closest patients (x = 1/2/5/10/25/25/100/250/500) and two different distance metrics (Euclidean and Manhattan distance) were assessed. The best performance was achieved using the top40 genes with Manhattan distance to the 100 nearest patients. This setting reached a sensitivity of 0.70, specificity of 0.73 and accuracy of 0.72 across all patients. When compared to ESR1 status, HER2 status, MKI67 expression, grade, size, and agein the multiple regression, only the gene-expression based classification (1.56E-31) and size (1.7E-24) were significant. Discussion:We have established a pipeline capable of determining lymph node positivity using the gene expression data of 40 genes. Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P4-03-03.