Dynamics of Circulating Tumor DNA As a Predictor of Outcomes with Lisocabtagene Maraleucel in Relapsed or Refractory Large B-Cell Lymphoma from the Transcend NHL-001 Study

Background

Ultrasensitive circulating tumor DNA (ctDNA) based minimal residual disease (MRD) methods have been established to be prognostic for outcomes following 1L therapy for diffuse large B-cell lymphoma (LBCL) (Roschewski M, et al. Blood 2022). Additionally, the feasibility and utility of such methods as a biomarker for disease surveillance in patients receiving CD19-targeted CAR T-cell therapy has also been explored (Sworder BJ, et al. Cancer Cell 2023; Stepan L, et al. Blood 2023). However, its potential in patients receiving lisocabtagene maraleucel (liso-cel) as 3L+ therapy for relapsed or refractory (R/R) LBCL remains unclear. The seamless design of the TRANSCEND NHL-001 trial (NCT02631044) demonstrated statistically significant clinical response and durability of the response with a median follow up of 19.9 months (Abramson JS, et al. Blood 2024). Here, we report longitudinally measured ctDNA levels from plasma samples in the TRANSCEND NHL-001 trial and evaluated its feasibility and utility as an outcome predictor in these patients.

Method

Out of 269 patients from the TRANSCEND NHL-001 trial, we included a subset of patients with LBCL (n=47) in our analysis, utilizing 216 plasma samples and 46 PBMC samples. Plasma samples were longitudinally collected prior to liso-cel infusion and at day 15, day 30, day 90, day 180, and day 360 after liso-cel infusion. Baseline plasma and matched PBMC samples were used to censor germline variants and genotype single nucleotide variant (SNV) using Cancer Personalized Profiling by Deep Sequencing (CAPP-Seq). ctDNA levels were quantified at each timepoint and expressed as haploid genome equivalents/mL (hGE/mL). At each longitudinal timepoint after liso-cel infusion, we used bootstrap resampling optimized for log rank statistics to identify the optimal ctDNA fold change thresholds to stratify progression-free survival (PFS). The prognostic utility of absolute and dynamic ctDNA levels and standardized PET/CT criteria (Lugano 2014) were evaluated in relation to PFS and overall survival (OS) using Kaplan-Meier survival curve and Cox proportional hazards model.

Results

Among these 47 patients, 46 had an available germline sample to allow for SNV determination. SNVs (median: 164, IQR: 78-287) were identified in the pretreatment sample from 42 (89%) patients. The median pretreatment ctDNA level was 689 hGE/mL (IQR: 311-2306) as compared to 136 hGE/mL (IQR: 35-696) in a previously cohort of the TRANSFORM trial (Stepan L, et al. Blood 2023), suggesting that our subset of LBCL patients was enriched for patients with high burden disease. The median follow-up was 19.9 months.

We assessed ctDNA performance at each timepoint for predicting outcomes. When stratifying ctDNA levels based on bootstrap resampling thresholds, patients with ctDNA level above the ctDNA fold change threshold had inferior PFS and OS as early as 1 month (PFS: P<0.05; OS: P<0.01), but most significantly at 3 months (PFS: P <0.001; OS <0.01). At both timepoints, all patients with ctDNA level above the respective threshold developed progressive disease.

To further assess how ctDNA compares in prognostic value in context of current standard of care PET/CT scans, we conducted a Cox proportional hazards model to predict PFS and OS. We first evaluated absolute ctDNA level at 2 weeks, 1 month, and 3 months after liso-cel infusion as independent predictors in univariate analyses. ctDNA was prognostic as early as 1 month (PFS: P<0.05, HR=1.6; OS: P<0.01, HR=2.0) and at 3 months (PFS: P<0.01, HR= 2.4; OS: P<0.05, HR=2.0). In contrast, PET/CT scan did not predict clinical outcomes at 1 month. The prognostic value of PET/CT scan at 3 months was not obtainable as the response at 3 months was either complete remission or progressive disease.

Mutations in individual genes at baseline were also genotyped using CAPP-Seq and will be described at the time of presentation.

Conclusion

Absolute and dynamic ctDNA measurements are predictive for clinical outcomes in R/R LBCL patients receiving liso-cel as 3L+ therapy in the TRANSCEND NHL-001 trial. The prognostic value of ctDNA was seen as early as one month following liso-cel infusion and was most significant at 3 months. The high prognostic value of ctDNA as compared with PET/CT demonstrates the feasibility and utility of ctDNA as an outcome predictor in this setting, indicating significant potential for its integration into future clinical trials and response criteria.