IntroductionCirculating tumor DNA (ctDNA) has become a well-established dynamic biomarker in Large B Cell Lymphoma (LBCL). Previous studies have shown that decrease in ctDNA quantities after standard R-CHOP treatment is significantly correlated with prognosis (Kurtz et al, Blood 2015 and Kurtz et al, J Clin Oncol 2018). However, no studies have described the very early kinetics of ctDNA following immunochemotherapy. Moreover, low-pass whole-genome sequencing (lpWGS) of cfDNA can identify DNA copy number alterations (CNAs) and be used to define focal CNA score (FCS). High FCS, denoting genomic instability, has been shown to be linked to inferior response rates to CAR-T therapy for LBCL patients (Cherng et al, Blood 2022). We developed a prospective clinical trial, LYMPHOCLEAR, which included 24 LBCL patients with planned first-line (L1) R-CHOP treatment, for whom ctDNA samples were taken at multiple timepoints within the first 48 hours post R-CHOP and at C2D1. We sought to describe early ctDNA kinetics and propose a clinically relevant model in order to potentially identify a prognostic early ctDNA analysis timepoint. We also sought to describe variants at each timepoint to see whether distinct natural evolutions could be observed very early post R-CHOP. We sought to evaluate whether the impact of FCS on response could be observed at very early timepoints.

Patients and Methods We included 24 LBCL patients (17 DLBCL, 4 PMBL, 2 HGBCL, 1 B Cell lymphoma NOS) who required hospitalization for C1 R-CHOP, according to their physician's decision. Median age was 64.5 years (range 24-85). IPI was 1-2 for 30.4% of patients and 3-5 for 69.6%. Ann Arbor stage was 4 in 70.8% of patients, with bulky mass in 29%. We performed capture-based NGS and lpWGS using DNA extracted from patient plasma samples at H0, H4, H8, H12, H16, H24 and H48 post R-CHOP initiation and at C2D1. All patients had a FDG PET-CT performed at baseline, post C2 and at end of treatment (EOT). Total Metabolic Tumor Values (TMTV) were measured using a fixed SUVmax > 4 threshold. A latent class mixed model (LCMM) was used to modelize ctDNA kinetics and delineate groups of patients with similar trajectories (Proust-Lima et al, J. Stat. Softw. 2017).

Results The application of a LCMM segregated four classes of ctDNA kinetics with robust statistical characteristics: Bayesian Information Criterion (BIC) was 330 and the post-hoc probability of a patient belonging to a specific class was close to 0.9. Classes 1 (n=3) and 2 (n=2) had lower baseline ctDNA values (median log 10 1.71 and 0.72 respectively) but class 1 had decreasing values from H0 to H48 while class 2 had increasing values from H0 to H48. Classes 3 (n=5) and 4 (n=14) had higher baseline ctDNA values (median log 10 3.42 and 3.21 respectively) but ctDNA quantities increased from H0 to H48 in class 3 whereas they decreased in class 4. Classes 1 and 2 had lower baseline TMTV and LDH than classes 3 and 4 (TMTV: 370.2cm3 and 61.1cm3 respectively versus 1347.3cm3 and 703.1cm3 and LDH: 212 and 220 respectively versus 775 and 340), also highlighting their lower tumor burden.

We confirmed that patients with Deauville score (DS) 4-5 at PET2 had significantly higher ctDNA quantities at C2D1 than patients with DS 1-3 (p=0.0048). Interestingly, we also showed that ctDNA values at H4 tend to be higher in patients with positive PET2 and that median ctDNA variant allele frequency at H4 was higher in patients with DS 4-5 at EOT PET.As expected, FCS was positively and strongly correlated to ctDNA quantities and tended to be higher in patients with DS 4-5 at EOT PET. Additional data regarding cfDNA fragmentome evolution at these early timepoints will be reported. As compared to baseline profiles, we observed that additional mutations over time mainly targeted intronic and µ Switch regions, suggesting very early enrichment of ctDNA by additional subclones following immuno-chemotherapy.

ConclusionsWe have shown that it is possible to describe the kinetics of ctDNA quantities and variant types in the first 48 hours post R-CHOP treatment initiation, leading to the identification of 4 patient classes, as well as the observation of very early ctDNA enrichment by additional subclones. We confirmed known impact of C2D1 ctDNA quantity but also highlighted potential impact of H4 ctDNA quantity and VAF on patient response. If confirmed in a larger cohort, this would allow for exceedingly early detection of poor responders to L1 R-CHOP.

This content is only available as a PDF.
Sign in via your Institution