Prediction of Clinical Response By Phased Variants in Circulating Tumor DNA (ctDNA) in the VALENTINE-PTCL01 Trial of Patients with Relapsed or Refractory (R/R) Peripheral T-Cell Lymphoma (PTCL)

Introduction: PTCL is an aggressive non-Hodgkin lymphoma with limited treatment options for patients with R/R disease. Phased variant (PV) enrichment and detection sequencing (PhasED-Seq) has been successfully utilized in diffuse large B-cell lymphoma. Detection of genetic aberrances implicated in PTCL could inform disease monitoring and clinical decision making in PTCL. Valemetostat tosylate (valemetostat) is a novel and potent dual inhibitor of enhancer of zeste homolog (EZH)2 and EZH1 approved in Japan for the treatment of R/R PTCL and adult T-cell leukemia/lymphoma. VALENTINE-PTCL01 (DS3201-A-U202; NCT04703192) was an open-label, single-arm, global phase 2 trial of valemetostat in patients with R/R PTCL. To explore patient-specific PhasED-Seq as a potential biomarker for disease status in R/R PTCL, we assessed changes in ctDNA prior to and following valemetostat treatment and correlations with clinical response in a subset of patients from VALENTINE-PTCL01. Methods: Biomarker-eligible patients were ≥ 18 years of age with confirmed PTCL, R/R disease after ≥ 1 prior line of systemic therapy, and a sufficient baseline (BL) sample. Patients received oral valemetostat 200 mg/day in continuous 28-day cycles until disease progression or unacceptable toxicity. PVs for ctDNA detection (ie, ctDNA-measurable residual disease [MRD]) were identified by comparing whole genome sequences of BL tumors to paired normal tissues. Custom capture panels were used to monitor ctDNA-MRD in plasma on cycle 1 day 1 (C1D1), C2D1, and C5D1. Response was assessed via a computed tomography scan according to Lugano 2014 criteria. Results: VALENTINE-PTCL01 included 133 patients with PTCL, and 41 had samples for PV assessment; of these, 9 were excluded due to low tumor purity, precluding PV detection. The 32 remaining cases had ≥ 2 PVs, allowing for PhasED-Seq ctDNA-MRD disease monitoring. PTCL subtypes included PTCL, not otherwise specified (n = 15), angioimmunoblastic T-cell lymphoma (n = 7), T-follicular helper phenotype PTCL (n = 1), anaplastic lymphoma kinase-negative anaplastic large cell lymphoma (n = 1), and other PTCLs (n = 8). Of the 32 patients with ≥ 2 PVs, ctDNA-MRD was detected in samples from at least 1 timepoint (C1D1, C2D1, or C5D1) in 31 (97%). Compared with sequencing data from a panel focused on T-cell lymphoma driver variants, mean variant allele frequency (VAF) for identified PVs correlated with panel-based VAF (Pearson R > 0.8), confirming that identified PVs were specific for PTCL tumor cells. ctDNA-MRD concentration at C1D1 closely correlated with tumor burden (ie, tumor size and VAF levels) (R > 0.9, P = 7.9×10-14), thus demonstrating its applicability as a marker of disease dynamics. Overall, 19/31 patients achieved a response, including 8 with a complete response (CR) and 11 with a partial response (PR); 4 achieved stable disease (SD) as best overall response (BOR), 6 had progressive disease (PD), and 2 had no BOR data available. All patients that achieved CRs had at least a 5-fold decline in ctDNA-MRD between BL and C2D1; this threshold was thus used to predict clinical response. Among patients with samples at C1D1 and C2D1, those with > 5-fold decreases in ctDNA-MRD over that period (7/7 [100%] CR, 7/8 [87.5%] PR, 0/8 [0%] SD, 1/4 [25%] relapse/PD) achieved longer median PFS than patients that had < 5-fold ctDNA-MRD decreases (> 5-fold decrease, 11.1 months [median follow up (mFU) 8.3 months]; < 5-fold decrease, 3.5 months [mFU 3.2 months]; p < 0.0001). At C5D1, patients who achieved a CR had the lowest ctDNA-MRD, followed by those with PR, and then nonresponders (SD/PD) (median ctDNA for CR, PR, and nonresponders was 0.26, 1.49, and 172.0 hGE/mL, respectively; CR/PR vs SD/PD Wilcoxon p = 0.017). Of the 3 patients with undetectable ctDNA-MRD at C5D1, 1 ultimately relapsed (mFU 8.3 months) and 2 maintained their responses (mFU 8.3 and 13.8 months), indicating that ctDNA-MRD concentrations at C5D1 could predict disease status. Conclusions: Analysis of ctDNA-MRD by PhasED-Seq is applicable in PTCL utilizing a customized approach. Declines in ctDNA-MRD at C2D1 and C5D1 correlated with response to therapy and PFS. These data suggest that ctDNA-MRD monitoring could serve as a noninvasive biomarker assessment to help predict response and survival in patients with R/R PTCL and during valemetostat treatment.