A twenty eight-gene signature and survival in completely-resected non-small cell lung cancer (NSCLC)

10583 Background: Current staging methods are imprecise for predicting the outcome of treatment of non-small-cell lung cancer (NSCLC). We have developed a 28-gene signature that is closely associated with recurrence-free and overall survival. Methods: We used whole-genome gene expression microarrays to analyze frozen-tumor samples from 174 patients (pT1&2, N0&1, MO), who had undergone complete surgical resection in 5 European institutions. Randomly generated numbers were used to assign 2/3 of the samples to an algorithm training group with the remaining 1/3 set aside for independent validation. Cox proportional hazards models were used to evaluate the association between the level of expression and patient survival. We used risk scores and nearest centroid analysis to develop a gene-expression model for the prediction of treatment outcome. Leave-one-out cross validation was used to prevent model over-training. Results: 28 genes that correlated with survival were identified by analyzing microarray data and risk scores. Based on the expression of these genes, patients in training and validation groups were classified as either high (48%) or low (52%) risk. Analysis of predicted risk groups revealed significantly different survival distributions for patients in both the training set (p<0.001) and independent validation set (p=0.006). Genes in our prognostic signature encode for several membrane-bound proteins with previously demonstrated involvement in cell cycle regulation and cell proliferation processes. Conclusions: Our 28-gene signature is closely associated with time to recurrence and overall survival of completely-resected NSCLC patients. [Table: see text]