Prospective Validation of Circulating Tumor DNA Measurable Residual Disease After First-Line Therapy in Large B-Cell Lymphoma

Abstract

PURPOSE End-of-treatment (EOT) response evaluation by positron emission tomography (PET) remains suboptimal in patients with large B-cell lymphoma (LBCL), because of its limited positive predictive value (PPV). Circulating tumor DNA (ctDNA)—based measurable residual disease (MRD) detection offers a minimally invasive approach and may improve prognostication. We prospectively evaluated EOT MRD using phased variant enrichment and detection sequencing (PhasED-Seq) in patients with first-line LBCL. METHODS Patients were enrolled in the HOVON-902 prospective cohort and received curative-intent first-line treatment. Phased variants (PVs) were identified and tracked using tumor biopsies or pretreatment plasma. The prognostic significance of EOT ctDNA-MRD status in progression-free survival (PFS) and overall survival (OS) was compared with that of the International Prognostic Index (IPI) and EOT PET-computed tomography (CT). RESULTS PV identification was successful in 134 of 136 (99%) using either tissue or plasma. At EOT, 83% of patients were MRD-negative and 17% of patients were MRD-positive. MRD positivity was strongly associated with inferior outcomes: the 3-year PFS was 17% in MRD-positive versus 85% in MRD-negative patients (hazard ratio [HR], 9.8 [95% CI, 5.1 to 19]; P = 9.63 × 10 −12 ), and the OS was 43% versus 92%, respectively (HR, 7.7 [95% CI, 3.4 to 17.4]; P = 1.27 × 10 −6 ). In multivariate analysis, MRD was an independent prognostic factor when controlling for IPI and EOT PET-CT. MRD positivity had a higher PPV for 2-year PFS than positive PET (68% v 56%, P ≤ .001), whereas negative predictive value was similar between negative MRD and PET (89% v 88%, P = .71). MRD positivity was associated with a significantly higher relapse risk within both complete metabolic response (CMR) and non-CMR subgroups. CONCLUSION This study validates ultrasensitive ctDNA-MRD detection using PhasED-Seq in a uniformly treated, prospective real-world LBCL cohort. These findings support further evaluation of MRD integration into clinical response assessment.