Baseline Prognostic Factors Do Not Predict End of Treatment Ctdna MRD Status and Have Limited Impact on MRD Prognostic Performance in DLBCL

Background: Baseline clinical factors used for DLBCL risk stratification, such as the International Prognostic Index (IPI), have had limited use for guiding therapy selection. Quantification of circulating tumor DNA (ctDNA) before and during therapy predicts treatment failure for DLBCL and informs disease monitoring and risk stratification. Detectable measurable residual disease (MRD) by ultrasensitive ctDNA assays at end of first-line chemoimmunotherapy (EOT) has strong prognostic value (Roschewski et al., ASH 2022). The impact of clinical features on EOT MRD detection and the clinical performance of MRD status is unknown but has implications for clinical adaptation and trial design.

Methods: We used a previously reported dataset assessing plasma ctDNA during and after curative-intent 1L anthracycline-based therapy combining multiple prospective cohorts (Goldstein et al., ASH 2023). The regimens included RCHOP or EP[O]CH alone or with acalabrutinib, lenalidomide, obinutuzumab or polatuzumab. CtDNA was evaluated using phased variant enrichment and detection sequencing (PhasED-Seq, Foresight Diagnostics), an ultrasensitive ctDNA assay. We assessed the ability of baseline clinical features to predict MRD status at EOT, including IPI [stratified as Low (0-1), Intermediate (2-3) and High (4-5)], Stage, Age, Cell-of-origin subtype (COO), along with interim findings on PET. We assessed ctDNA-specific features, such as pre-treatment ctDNA level, early molecular response (EMR; > 2 log-fold ctDNA reduction after 1 cycle) and major molecular response (MMR; > 2.5 log-fold ctDNA reduction after 2 cycles). We used time-dependent ROC curves to assess the impact of IPI and pre-treatment ctDNA on EOT MRD performance for predicting PFS.

Results: 145 patients were evaluable for MRD at EOT, with 26% Low-risk, 54% Intermediate-risk, and 17% High-risk IPI, while 58% were GCB subtype. At a median follow-up of 18.1 mo., 35 (24%) progressed. MRD was detectable at EOT in 16% of cases with Low-risk IPI, 26% of Intermediate risk, and 36% of High risk. The likelihood of detectable MRD at EOT was not significantly higher for Intermediate-risk IPI (HR 1.8, 95% CI [0.7-5.4]) and High-risk IPI (3 [0.9-10]) relative to Low-risk. Advanced stage disease was associated with higher rate of detectable MRD at EOT at 31% vs 7% early stage (5.9 [2-26]). COO, age, and sex were not associated with MRD detection by PhasED-seq at EOT.

Higher baseline ctDNA levels were associated with higher rates of detectable MRD at EOT at 37% compared to 18% for patients with lower baseline ctDNA (2.8 [1.2-6.9]). Lack of a MMR was also associated with detectable MRD at EOT, and all 6 patients with <MMR had detectable MRD, while only 18% of patients who achieved MMR had detectable MRD at EOT (p=0.0002). EMR and interim PET were not associated with significant differences in EOT MRD status.

Detectable EOT MRD was associated with inferior PFS (HR=30, p=9.9e-14). High-risk IPI was prognostic for worse PFS relative to Low-risk IPI (HR=3.65, p=0.02), but Intermediate-risk was not (HR = 1.8, p = 0.22). In a multivariable Cox proportional hazards model, EOT MRD status remained associated with worse outcomes (HR=30, p=1.8e-13), but IPI was not. Stratifying IPI Low-, Intermediate- and High-risk, EOT MRD remained highly prognostic in each subgroup. The sensitivity and specificity of the PhasED-seq assay at EOT between Low-, Intermediate- and High-risk IPI were 74%, 86% and 77% and 92%, 96%, and 100% respectively. The PPV and NPV were 73%, 90%, and 100%, and 92%, 94% and 87% for Low-, Intermediate-, and High-risk strata respectively. Performance of EOT PhasED-Seq for predicting PFS was high regardless of IPI, with AUCs of 0.83, 0.91, and 0.89 for Low-, Intermediate-, and High-risk. PhasED-Seq had high predictive ability for PFS for both low and high pre-treatment ctDNA burden with AUCs of 0.92 and 0.85 respectively.

Conclusions: MRD status maintains high performance at EOT for predicting PFS regardless of baseline IPI risk classification and baseline ctDNA. Specificity and PPV of MRD status at EOT improved with higher IPI. Although higher rates of MRD-positivity are seen as IPI increases, baseline variables do not precisely predict MRD status at EOT. These data suggest the prognostic utility of ctDNA-MRD at EOT is independent of clinical features, and ctDNA-based risk-adapted trial designs should consider factors beyond IPI to identify high-risk patients.