Collective Analysis of Genomic Data Sets to Assess the Impact of Eligibility Criteria on Molecular Subsets in First-Line Clinical Trials for Patients with Diffuse Large B-Cell Lymphoma

Introduction

Rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP) has been the standard first-line (1L) therapy for diffuse large B-cell lymphoma (DLBCL). DLBCL is a molecularly heterogeneous disease with variable clinical outcomes by genetic subtypes (Alizadeh Nature 2000; Chapuy Nat Med 2018; Ennishi JCO 2019; Wright Cancer Cell 2020). While the incorporation of polatuzumab vedotin in 1L led to superior progression-free survival (PFS) in clinically defined high-risk patients by International Prognostic Index (IPI), data indicate difference in benefit of polatuzumab vedotin across genetic subtypes (Palmer NEJM 2023; Morschhauser ASH 2023). Intriguingly, the GUIDANCE-1 trial assigning treatment by LymphGen classification showed that adding biologically rational targeted treatment to R-CHOP may improve outcomes (Zhang Cancer Cell 2023). Although genetic subtypes should be considered in future trial designs for treatment of 1L DLBCL, a significant barrier is anticipated when allocating treatment by genetic subtypes due to the molecular heterogeneity of DLBCL and associated rarity of each subtype. Approaches to anticipate the impact of eligibility criteria on molecular subgroups of patients for future 1L trials for DLBCL are needed.

Methods

We identified DLBCL genomic data sets that had individual patient-level clinical data including IPI variables and survival outcomes following 1L treatment with R-CHOP that are linked to molecular data including whole exome or whole genome sequencing, copy number alterations, gene expression profiling (GEP) groups including double-hit signature or dark zone signature, and genetic subtypes including Consensus Cluster classification and LymphGen subtypes. We applied IPI-based eligibility criteria to define a patient cohort for a hypothetical 1L DLBCL clinical trial, with chosen eligibility criteria derived from expert consensus recommendations for 1L studies incorporating R-CHOP plus targeted therapy (Harkins Blood Adv 2022) and including age at diagnosis ≥18-80 years, Eastern Cooperative Oncology Group performance status 0­­­­­­-2, and Ann Arbor stage ≥2. We compared frequencies of molecular alterations, stratification by GEP and molecular subgroups, and survival outcomes following treatment with R-CHOP (PFS and overall survival [OS]) by trial eligible and ineligible groups. Within the eligible cohort, we also compared copy number alterations and mutational frequencies within each genetic subtype, and we evaluated survival outcomes by GEP group and genetic subtype.

Results

Four data sets met criteria for analysis from Chapuy et al. (n = 259), Ennishi et al. (n = 297), Urata et al. (Blood Adv 2023; n = 1,034), and Shen et al. (Transduct Target Ther 2023; n = 1,001). Baseline clinical characteristics were similar across data sets. Following application of selected eligibility criteria, the most commonly observed molecular alterations in the eligible group included 18q and 3q copy number gain and mutations in PIM1 and SOCS1, whereas the most commonly observed alterations in the ineligible group included 7q amplification, 6q deletion, and mutations in PIM1 and KMT2D. Stratification by GEP subgroup showed enrichment in the activated B-cell (ABC) subtype in the eligible (Ennishi: 32.1%; Urata: 45.4%) vs. ineligible (Ennishi: 24.1%; Urata: 39.6%) groups. Two-year PFS and 2-year OS in eligible patients with ABC subtype were 58% and 71% (Ennishi) and 58% and 76% (Urata), respectively. Genomic subtypes represented in the eligible (vs. ineligible) group most commonly included Consensus Clusters 5 (24.2% vs. 17.2%), C2 (20.3% vs. 24.1%), and C1 (18.1% vs. 10.3%; Chapuy) and LymphGen subgroups TP53 (15.4% vs. 11.2%), BN2 (11.2% vs. 10.6%), and MCD (9.8% vs. 14.1%; Shen). Two-year PFS and 2-year OS in eligible patients were 58% and 75% for C5 and 60% and 82% for MCD, respectively.

Conclusions

By applying selected eligibility criteria to genomic data sets, we identified mutational frequencies, GEP subgroups, and genomic subtypes for eligible and ineligible groups. This work illustrates the feasibility and value of leveraging published data sets incorporating patient-level clinical and molecular data to predict the genomic subtype prevalence and outcome benchmarks for ensuing trial cohorts and thereby facilitate trial design for future 1L clinical trials incorporating treatment assignment by genomic subtype.