Circulating Tumor DNA Kinetics at Day 14 after CAR T-Cell Therapy Identifies Patients with B-Cell Lymphoma at Risk for Early Treatment Failure

Introduction

CAR-T cell therapy has been established in relapse treatment of aggressive B-cell lymphomas and demonstrated superior outcome in patients with early and late relapsed/refractory B-cell lymphoma. However, only 50% of patients are cured, making biomarkers to predict response and to guide therapeutic intervention very important. Circulating tumor DNA (ctDNA) is an emerging tool for characterization, monitoring and stratification of aggressive lymphomas while ctDNA kinetics have been shown to identify high-risk patients early during the course of treatment.

Here, we describe the usefulness of ctDNA to assess early response to CAR-T cell therapy to identify patients in need of interventional treatment to improve response to CAR-T cell treatment.

Methods

Fifty-three patients (pts) with relapsed disease (42 DLBCL, 2 HGBCL, 5 MCL, 2 FL, 2 PMBCL) and serial plasma samples taken before and after lymphodepletion, d0 and d14 after CAR-T cell infusion were obtained from CAR-T pts treated in our institution.

For initial tumor genotyping, FFPE samples for 36/53 pts and plasma samples obtained prior to lymphodepletion (range -101 to -12 days prior to CAR-T therapy) for the remaining 17 pts were sequenced using a targeted capture assay (Euroclonality-NGS DNA Capture assay (EC-NDC), Univ8® Genomics, UK) for the identification of single nucleotide variants (SNVs), structural variants and immunoglobulin (IG) receptor rearrangements.

At follow-up, 30 ng cfDNA (cell-free DNA) from d0 and d14 was sequenced with a median sequencing depth of >2700x. Three patients with <10 ng cfDNA at d14 were excluded from further analysis. ctDNA levels were calculated by multiplying the maximal variant allele frequency (VAF) by cfDNA levels represented by haploid genome equivalents per ml (hGE/ml) and converted to a logarithmic value. Pts were grouped according to ctDNA levels at d14 into high and low ctDNA groups. ctDNA levels were correlated with PET-CT-Response at month 3 (M3), progression-free (PFS) and overall survival (OS).

Data were analyzed by ARResT/Interrogate (Bystry, Bioinformatics, 2017) with an adapted pipeline for targeted capture in cfDNA. A minimum of 4% VAF in FFPE and 0.5% VAF in plasma samples was used for variant calling. Detectable MRD was called with ≥1 unique read per marker. ctDNA fraction was estimated based on the highest VAF detected.

Results

Bridging prior to CAR-T treatment was applied in 47/53 pts. PET response at M3 was achieved in 32/50 pts (31 metabolic CR, one PR, three NA). Tumor genotypes could be obtained in 50/53 (94%) patients with a mean of seven markers per patient (range 1 – 26). Clonal IGH rearrangement was detected in 38/53 pts (72%). Disease-specific translocations were detected in 7/53 pts (5/5 t(11;14) for MCL and 2/2 t(14;18) in FL).

Prior to lymphodepletion, ctDNA levels were comparable among responders and non-responders, while in the 50 cases with plasma samples from day 0, ctDNA level were significantly lower in pts achieving PET response (2.10 log₁₀hGE/ml vs. 3.81, p=0.021). This finding reflects the impact of bridging therapy thus resulting in lower tumor burden on outcome.

ctDNA levels on d14 ranged from 0.0 to 6.1 log₁₀hGE/ml (median 3.2). Patients were stratified to high/low ctDNA group at d14 with a cut-off of >3.125 Log₁₀hGE/ml as determined as optimal cut-off by Youden's J statistic. According to this threshold, 28 of 47 evaluable pts were grouped into the high and 19 into the low ctDNA group.

ctDNA levels on d14 in 47 evaluable pts were significantly different among response groups (p=0.023) and low ctDNA correlated with PET-response at M3 (p=0.001). PET-CT prediction by ctDNA had an overall accuracy of 0.71, a sensitivity of 89% and a specificity of 58%. The median PFS for the high ctDNA group was 3.75 months and was not reached in the low ctDNA group at a median observation time of 14.8 months (p=0.009). Hazard ratio (HR) for PFS in the ctDNA high group was 3.03 (95% CI: 1.27 - 7.22), p=0.012. Median OS in the high ctDNA group was 14.3 months and not reached in the low ctDNA group (p=0.009). HR for OS in the high ctDNA group was 4.62 (95% CI: 1.32 – 16.2), p=0.017.

Conclusion

ctDNA kinetics on d14 identifies patients at very high risk of CAR-T cell failure and early progressive or refractory disease. This enables targeted early intervention with immune modulatory drugs or bispecific antibodies to prevent CAR-T treatment failure and potentially improve outcomes.