EZH2 and CREBBP mutation status identify a subset of patients who benefit from bendamustine- over CHOP/CVP-based immunochemotherapies in follicular lymphoma: Findings from the GALLIUM trial and validation in a large international cohort

Abstract Background: Follicular lymphoma (FL) remains a clinical challenge due to its relapsing course and difficulty in tailoring therapies to individual risk. Frontline treatment commonly includes an anti-CD20 antibody with either bendamustine or CHOP/CVP. Bendamustine has been associated with longer progression-free survival (PFS) compared to CHOP/CVP in the GALLIUM trial. However, bendamustine was also associated with more grade 3-5 infections, prolonged T-cell reduction, fatal events, and second neoplasms(Hiddemann, 2017). Additional concerns include its immunosuppressive effects and potential to impair subsequent T-cell based therapies including CAR-T cells (Iacoboni, 2024). This highlights the need to better identify patients who benefit most and least from bendamustine. Here, we used our gene mutation data to identify predictive biomarkers of bendamustine benefit and validated our findings in a large international real-life cohort. Methods: The GALLIUM trial (NCT01332968) enrolled 1,202 patients with previously untreated, advanced-stage FL (stage III/IV or bulky stage II) and ECOG 0–2, all requiring treatment per GELF criteria. Patients were randomized to receive rituximab or obinutuzumab with CHOP, CVP, or bendamustine allocated by the treating center (Marcus, 2017). Targeted sequencing of recurrently mutated genes was available for diagnostic biopsies from 418 evaluable cases. Kaplan-Meier and Cox regression analyses were used to correlate gene mutations with clinical outcomes. The median follow-up in the GALLIUM cohort was 6.9 years; 283 patients (68%) received bendamustine, 135 (33%) received CHOP or CVP. The validation cohort consisted of 473 patients from Australia (N=208), USA (N=138), and Canada (N=127). The median follow-up in the validation cohort was 5.5 years; 181 patients (38%) received bendamustine- and 292 (62%) received CHOP- or CVP-based immunochemotherapies. Results: In this updated analysis with longer follow-up, we confirmed that the previously described clinicogenetic risk model m7-FLIPI (Pastore, 2015) predicts PFS in CHOP/CVP–treated patients, but not in those receiving bendamustine. The mutation status of two of the seven m7-FLIPI genes was associated with differing PFS outcomes in patients treated with CHOP/CVP vs bendamustine: Patients with wild-type EZH2 had significantly shorter PFS with CHOP/CVP-based therapies compared to EZH2-mutated cases (HR=0.43, p=0.020); notably, no difference in PFS was observed in the bendamustine cohort. In addition, CREBBP mutations were associated with shorter PFS after CHOP/CVP (HR=2.05, p=0.037), but not after bendamustine. None of the other m7-FLIPI gene mutations predicted PFS in bendamustine-treated patients. Based on these findings, we combined the EZH2 and CREBBP mutation status to stratify patients. Only patients with both a CREBBP mutation and wild-type EZH2 showed a clinically relevant benefit from bendamustine, with significantly longer PFS compared to CHOP/CVP-treated patients (HR=0.44, p=0.00011) and similar OS (HR=0.62, p=0.27). All other patients had similar PFS with either regimen (HR=1.05, p=0.86), but we observed a trend to longer OS with CHOP/CVP (HR=2.19, p=0.11). These results were confirmed in the validation cohort: EZH2 mutations were associated with longer PFS in patients receiving CHOP/CVP (HR=0.62, p=0.025), while CREBBP mutations were associated with shorter PFS (HR 1.42, p=0.033). Again, none of the m7-FLIPI gene mutations were predictive of PFS in bendamustine-treated patients. Stratification based on EZH2 and CREBBP mutation status validated the previous results: only patients with both a CREBBP mutation and wild-type EZH2 had a PFS benefit from bendamustine (HR 0.61, p=0.027), again with similar OS (HR=0.75, p=0.48). Likewise, all other patients experienced similar PFS with either regimen (HR=1.07, p=0.73), but again had a trend towards longer OS when treated with CHOP/CVP (HR=1.75, p=0.068). Conclusion: Our findings suggest that EZH2 and CREBBP mutation status can inform treatment selection in FL. Patients with CREBBP mutations and wild-type EZH2 benefit most from bendamustine with significantly longer PFS and similar OS; all other patients had similar PFS with either regimen but a trend towards longer OS with CHOP/CVP. These data may indicate a distinct biology in FL with CREBBP mutations and wild-type EZH2, and support the use of simple mutation testing to guide personalized therapy in newly diagnosed advanced-stage patients.