Several gene expression signatures are predictive of prognosis in diffuse large B cell lymphoma (DLBCL), but the lack of practical methods for a genome scale analysis has restricted their clinical applicability. Towards construction of a molecular predictor amenable to rapid testing on routinely obtained diagnostic clinical specimens, we studied genes previously reported to be associated with survival in DLBCL, testing and validating risk scoring models with robust survival associations in the current therapeutic era.

We identified LMO2 expression as a robust univariate predictor of survival and cell of origin classification of DLBCL, with independent prognostic value. We examined bivariate models combining expression of LMO2 with other genes, and identified TNFRSF9 as a tumor microenvironment gene with independent prognostic influence. A combined model integrating both LMO2 and TNFRSF9 expression was independent of “cell of origin” classification, “stromal signatures,” International Prognostic Index (IPI), and added to the predictive power of IPI. A composite model was validated in multiple independent patient cohorts using public microarray data. Using routinely obtained formalin fixed, paraffin embedded diagnostic specimens from an independent cohort, we developed a simple assay validating the clinical utility of this 2-gene model, as well as a composite model integrating the IPI. In conclusion, measurement of a single gene expressed by tumor cells and a single gene expressed by the immune microenvironment is sufficient to predict overall survival in patients with DLBCL treated with R-CHOP. A combined model serves as a robust clinical risk assessment tool.

This talk is also presented as Poster A27.

Citation Information: Clin Cancer Res 2010;16(14 Suppl):PR5.