Risk Profiling of De Novo DLBCL Patients By Measuring Circulating Tumor DNA

Sarit Assouline and Won Seog Kim contributed equally

Background: Approximately 30–45% of patients with de novo diffuse large B cell lymphoma (DLBCL) relapse or are refractory to standard-of-care (SOC) treatment, rituximab (R)-CHOP [cyclophosphamide, doxorubicin, vincristine, prednisone]; these patients have a poor prognosis and may benefit from improved therapies (Chaganti, et al. Br J Haematol 2016). Available prognostic factors are not sensitive/specific enough to identify patients that may fail first-line (1L) SOC therapy. Advances in circulating tumor DNA (ctDNA) assessment have shown promise for identifying patients with de novo DLBCL at a higher risk of progression, with potential to inform future trials (Kurtz, et al. J Clin Oncol 2018); however, additional data are required. The aim of this study was to assess the utility of ctDNA in identifying high-risk patients with DLBCL and evaluate measures of ctDNA at baseline and during therapy.

Methods: Samples (n=40) from a Phase Ib/II study in patients with de novo DLBCL treated with R-CHOP plus atezolizumab following an initial cycle of R-CHOP (NCT02596971; Younes, et al. Blood 2018) were used for targeted sequencing. The clonoSEQ assay (Monter, et al. Expert Rev Mol Diagn 2019) was used to assess baseline (n=40) and end-of-induction (n=32) peripheral blood mononuclear cells (PBMCs). Next generation sequencing analyses were performed with a modified Roche AVENIO ctDNA workflow based on CAPP-Seq technology (Kurtz, et al. 2018), and using baseline (n=26), Cycle (C) 2 Day (D) 1 (n=26), C3D1 (n=26) and end-of-consolidation (EoC; n=13) samples. CAPP-Seq uses mean mutant molecules per mL (MMPM), a measure of the average allelic mutational frequency, to assess ctDNA levels. Pre-treatment MMPM levels and on-treatment dynamics (fold-change from baseline/clearance) were used to assess if ctDNA was associated with progression-free survival (PFS) and could risk-stratify DLBCL, although add-on-therapy at C2D1 may dilute potential risk stratification post-baseline. Pre-defined fold-changes of 2 and 2.5 (log10) for C2D1 and C3D1 samples, respectively, were used to define high-risk patients. Clearance of ctDNA (defined by cross-monitoring analysis of samples against a resampled distribution of tumor-specific gene variants) was assessed separately for patient stratification.

Results: Clone identification in tissue (n=32/40) and clone recovery in baseline PBMCs (n=14/32) was identified in 35% of patients by clonoSEQ while ctDNA prevalence was observed in all 26 patients at baseline by CAPP-Seq. ctDNA was detected at higher rates in on-treatment plasma samples by CAPP-Seq (C2D1: 85%; C3D1: 65%; EoC: 46%). Downstream analyses were performed with CAPP-Seq due to the higher sensitivity. Pre-treatment MMPM levels trended prognostic for PFS (univariate hazard ratio [HR], 0.51 [95% CI: 0.09–2.76]; adjusted for age, sex, Eastern Cooperative Oncology Group performance status and international prognostic index (adj) HR, 0.25 [95% CI: 0.03–2.13]). On-treatment dynamics at the pre-defined cut-offs C2D1 and C3D1 showed that 42% of patients were high-risk (95% CI: 23–61; Figure A). On-treatment ctDNA also trended prognostic for PFS for C2D1 (univariate HR, 0.34 [95% CI: 0.06–1.88; Figure B]; adj HR, 0.36 [95% CI: 0.06–2.02]) and C3D1 (univariate HR, 0.56 [95% CI: 0.11–2.76]; adj HR, 0.62 [95% CI: 0.09–4.07]). Clearance of ctDNA identified a small number of patients that did not have any PFS event after C2D1 (n=5/26; Figure C) and C3D1 (n=9/26). We identified individual tumor-specific mutations that increase post treatment, which may enable personalized treatment decision making (Figure D). Associations between PD-L1 expression and ctDNA levels will be presented.

Conclusions: Within the confinements of this study, our preliminary data suggest that ctDNA is a potentially promising tool for identifying high-risk patients, and could enable personalized risk-adapted therapy for the 1L treatment of DLBCL. Further studies are needed to assess the full utility of this assay.