The Use of Gene-Expression Profiling to Predict Outcome of Follicular Lymphoma Patients Treated with Rituximab and Chemotherapy.

The anti-CD20 monoclonal antibody rituximab has been extensively evaluated for treatment of follicular lymphoma (FL) and is now an integral component of many treatment strategies. However, it is not known when and how to use this combination modality best in the overall treatment course of FL patients. Here we have analyzed the pattern of gene expression in lymph nodes from 28 FL patients who received rituximab in combination with chemotherapy. We demonstrate a novel feature selection method, in which genes whose expression correlates with treatment outcome are devided into groups based on similar expression patterns. By selecting representatives from these groups we built a three-gene predictor (for example RRAD, MGC5254, MARCO) which turned out to be successful in classifying all 28 samples correctly with respect to the responders versus nonresponders distinction. A similar four-gene predictor (for example USP9Y, CUL-4B, NSFS/REST, MADH) correctly classified the responding patients into two subgroups with very different event-free survival rates at the median follow up of 21 months (median not reached vs. 15 months, p<0.0001). The latter predictor also efficiently delineated patients within specific FLIPI risk groups. Our data demonstrate the ability of these microarray-based predictors to classify FL patients treated with rituximab and chemotherapy to subgroups with significantly distinct outcomes.