Prognostic Utility of Minimal Residual Disease (MRD) after Curative Intent Induction Therapy for DLBCL: A Prospective Real-World Ctdna Study

Background: The prognostic value of MRD using ultra-sensitive ctDNA assays after 1L DLBCL immunochemotherapy was recently reported (Roschewski, ASH 2022). However, real-world evidence (RWE) for the utility of ctDNA MRD assays to predict survival outcomes after standard-of-care (SOC) 1L treatment for DLBCL remains less clear. Here, we report longitudinally measured ctDNA levels, without use of tumor biopsies, in a real-world population of DLBCL patients treated with curative intent using current anthracycline-based induction regimens.

Methods: We prospectively enrolled patients with DLBCLs from Samsung Medical Center undergoing 1L therapy for ctDNA profiling at 3 pre-defined milestones (Baseline, Interim after 3 cycles (C4D1), and end of treatment [EOT]). A total of 99 patients had baseline pretreatment plasma available with sufficient tumor burden for inclusion in this study. 364 samples were profiled in a blinded manner by Phased Variant Enrichment & Detection Sequencing (PhasED-Seq, Foresight Diagnostics). Tumor-derived Phased Variants (PVs) were identified directly from baseline plasma (median 4 mL) without the use of tumor tissue, with matched constitutional DNA from paired PBMCs used to censor germline variants and CHIP. PVs were used to longitudinally assess ctDNA-MRD after therapy, including at the C4D1 and EOT timepoints. MRD levels were compared to standardized responses by PET/CT (Lugano 2014), PFS, and OS. Patients/samples were reported as MRD positive when ctDNA levels exceeded an analytical detection threshold (~1:10⁶ cfDNA molecules) corresponding to 98% specificity.

Results: Among the 99 evaluable patients, the median age was 58 years and 38% were women. 67% had stage III-IV disease, 53% had IPI scores ≥3, and 6% had MYC and BCL2 translocations. All patients were treated with anthracycline based therapy, with 94% receiving RCHOP and 6% receiving DA-EPOCH-R. The median follow-up was 44 months.

We tested ctDNA performance at each landmark for predicting outcomes. ctDNA-MRD levels by PhasED-Seq at each landmark were prognostic for both PFS and OS. Specifically, when dividing pretreatment ctDNA levels at the median, ctDNA significantly predicted outcomes (PFS: P=0.004, HR=2.7; OS: P=0.0004, HR=5.1). At the C4D1 and EOT landmarks, we divided patients into those with detectable versus undetectable ctDNA-MRD. Both timepoints significantly stratified patients for outcomes including at C4D1 (PFS: P=0.0002, HR=5.0; OS: P=0.0046, HR=4.1) and at EOT (PFS: P<0.0001, HR=7.8; OS: P<0.0001, HR=10.8).

As the current standard for assessing response in DLBCL is imaging, we also examined the prognostic value of PET/CT scans at the same time-point as ctDNA. While interim PET/CT scans in our cohort were prognostic for PFS, they were not significantly prognostic for OS (PFS: P=0.02, HR=2.5; OS: P=0.06, HR=2.4). At EOT, PET/CT scans were prognostic for both PFS and OS (PFS: P=0.011, HR=3.4; OS: P=0.011, HR=4.0). At both milestones, PET/CT appeared less discriminatory than ctDNA-MRD. Exemplifying this, the overall survival of patients at 24 months was better separated at EOT by ctDNA-MRD than it was by PET/CT scan (OS24 ctDNA-MRD+: 37%, ctDNA-MRD-: 94%, compared with PET/CT PR/SD/PD: 50%, PET/CT CR: 86%).

Finally, to further compare the prognostic value of ctDNA-MRD in the context of the current standard PET/CT scan, we performed a Cox proportional hazards model to predict PFS and OS, using both PET/CT and ctDNA-MRD as univariate predictors, as well as a multivariate model considering both PET/CT and ctDNA. As expected, interim and EOT ctDNA-MRD and PET/CT scans were predictive of outcomes when considered in univariate analysis. However, when considered in combination, while the prognostic value of ctDNA-MRD for OS remained at both interim and EOT timepoints, PET/CT scan was no longer significantly prognostic of outcomes (Fig 1).

Conclusions: These data both demonstrate the feasibility and the prognostic utility of ctDNA-MRD during and after SOC induction therapy for DLBCL in a real-world population using an ultrasensitive ctDNA-MRD assay. The higher predictive value and accuracy of detectable ctDNA-MRD as compared with PET/CT suggest opportunities for integration of such assays in lymphoma response criteria, to potentially inform future clinical decision making.