Introduction: Diffuse large B-cell lymphoma (DLBCL) is a biologically heterogeneous diagnostic category with important prognostic subgroups, including activated B-cell (ABC) and germinal center B-cell (GCB) cell-of-origin (COO) subgroups. The GCB COO subgroup shares more molecular features with follicular lymphoma (FL) than other DLBCL subpopulations, and is itself heterogeneous, with a subgroup dependent on activation of the PI3K/AKT pathway (via loss of PTEN). When combined with chemotherapy, obinutuzumab (GA101; G) was clinically superior to rituximab (R) in first-line (1L) FL in the GALLIUM trial. However, the combination of G and CHOP chemotherapy (G-CHOP) was not more effective than R plus CHOP (R-CHOP) in 1L DLBCL in the GOYA trial (NCT01287741) (Vitolo et al. J Clin Oncol 2017). Here, we performed exploratory biomarker analyses to evaluate if a subgroup of patients (pts) derived benefit from G-CHOP in the GOYA 1L DLBCL population. Treatment effect was assessed using the Linear Predictor Score (LPS) from the Nanostring Lymphoma Subtyping Test (LST), as this continuous variable informs on the likelihood of a DLBCL pt having a germinal center-derived tumor (which has a similar biological origin to FL) based on the gene expression profile [Wright et al. Proc Natl Acad Sci USA 2003].Methods: Biomarker testing was performed on formalin-fixed, paraffin-embedded tumor tissue collected prior to treatment and tested retrospectively in central laboratories. COO was assessed using the NanoString Research Use Only LST (NanoString Technologies Inc., Seattle, WA, USA) in 933 pts (representing 66% of the intent-to-treat population in GOYA). Additional biomarker analyses used for molecular characterization included DNA-targeted sequencing of 467 genes using the FoundationOne® Heme (FOH) panel (n=499) and identification of BCL2 translocations using Vysis LSI Dual Color Break Apart FISH Probes (n=644; FISH cut-off, 50%). Multivariate Cox regression and elastic net penalized regression were used to evaluate biomarker treatment effects. In addition, bootstrap simulations were conducted to identify the optimal LPS to reflect the robustness of the observed treatment effect in GOYA and the generalizability of treatment effect to independent study populations. Multiple testing adjustment was done by estimating false-discovery rates (FDRs) using the Benjamini-Hochberg procedure (significance, <5% FDR). Pathway enrichment analysis was performed using a hypergeometric test. All pts consented to the biomarker analyses.Results: Strong expression of a germinal-center gene expression profile (by LPS) was linked with a benefit in outcome from treatment with G-CHOP vs R-CHOP among pts in GOYA. Bootstrap simulations identified an optimal LPS cut-off for predicting G-CHOP benefit as the 25% (233/933) of GOYA pts with the lowest LPS scores. These pts are referred to as ‘strong-GCB’ pts, and comprised 43% (233/540) of evaluable GCB pts in GOYA. Strong-GCB pts treated with G-CHOP achieved significantly better clinical outcomes in terms of investigator-assessed progression-free (HR=0.33, p=0.0007), event-free (HR=0.47, p=0.003), and overall survival (HR=0.41, p=0.019; but not end-of-treatment PET-complete response) than those treated with R-CHOP (Table 1; Figure 1). In multivariate analyses, the observed benefit was independent of baseline demographics and disease characteristics. Safety was similar with either regimen. In gene-set analyses on the FOH data, compared with other GCB pts, strong-GCB pts were significantly enriched for mutations that are characteristic for FL pts. In particular, BCL2 translocations and mutations in several m7-FLIPI genes were highly enriched in strong-GCB pts vs other DLBCL subgroups (Figure 2). There was no evidence for transformed indolent non-Hodgkin lymphoma in the strong-GCB subset on central pathology review.Conclusions: Analyzing the data from the GOYA trial, we have identified a new distinct molecular subgroup of GCB DLBCL that comprises around 25% of all DLBCL pts, referred to as “strong-GCB”. This distinct subgroup is identifiable by gene expression profiling and characterized by mutations that are also commonly identified in FL pts [Morin et al. Nature 2011]. Strikingly, treatment with G-CHOP appears to confer substantial clinical benefit over R-CHOP in this new subgroup of DLBCL. [Display omitted] DisclosuresOestergaard:F. Hoffmann-La Roche Ltd: Employment. Bolen:Genentech: Employment, Equity Ownership. Mattiello:Roche: Employment. Sehn:Abbvie: Consultancy, Honoraria; Seattle Genetics: Consultancy, Honoraria; Celgene: Consultancy, Honoraria; Roche/Genentech: Consultancy, Honoraria; Amgen: Consultancy, Honoraria; Janssen: Consultancy, Honoraria. Trněný:Janssen: Consultancy, Honoraria; Abbvie: Consultancy, Honoraria; BMS: Consultancy, Honoraria; Celgene: Consultancy, Honoraria, Research Funding; Roche: Consultancy, Honoraria, Research Funding; Takeda: Consultancy, Honoraria; Gilead: Consultancy, Honoraria. Vitolo:Roche: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding; Gilead: Honoraria; Celgene: Honoraria, Membership on an entity's Board of Directors or advisory committees; Takeda: Honoraria; Janssen: Honoraria, Membership on an entity's Board of Directors or advisory committees; Mundipharma: Honoraria. Nielsen:F. Hoffmann-La Roche Ltd: Employment, Equity Ownership. Morariu-Zamfir:Roche: Employment. Fingerle-Rowson:F. Hoffmann-La Roche Ltd: Employment, Equity Ownership. Lenz:Celgene: Consultancy, Honoraria, Research Funding; Bayer: Consultancy, Honoraria, Research Funding; Janssen: Consultancy, Honoraria, Research Funding; Roche: Consultancy, Honoraria.