Abstract Background: Diffuse large B-cell lymphoma (DLBCL) is the most common subtype of non-Hodgkin lymphoma, with relapse rates of 30–40% despite standard frontline chemoimmunotherapy. Early identification of high-risk patients remains challenging. Circulating tumor DNA (ctDNA) is a promising biomarker for disease monitoring and prognosis, particularly when minimal residual disease (MRD) is assessed via immunoglobulin (IG) gene rearrangements. However, the prognostic value of dynamic ctDNA changes combined with PET/CT imaging has yet to be fully established. To address this, we conducted a large-scale prospective study assessing the prognostic utility of serial ctDNA-based MRD tracking, both alone and integrated with PET/CT imaging. Methods: In this prospective multicenter study (ChiCTR2400087307, “MeDIG Study”), consecutive patients with newly diagnosed DLBCL were enrolled from November 2022 to March 2025. Eligible patients were ≥18 years old, had detectable tumor clonotypes identified from diagnostic tissue biopsies, and planned to receive standard first-line chemoimmunotherapy. Clinical responses were assessed according to the Lugano 2014 criteria. Plasma samples were collected at baseline, during treatment, and throughout a 2-year post-treatment follow-up period. MRD was assessed by IG VDJ rearrangements using the NEOMRD® assay. Results: A total of 724 plasma samples were collected longitudinally from 117 treated patients at baseline, during treatment, and throughout follow-up, with a maximum follow-up duration of 32 months. The median age was 56 years (range, 18–83), with 47.0% males. Over half of patients (59.0%) had stage III–IV disease, and 58.1% had an IPI score ≥2. At the end of treatment, the objective response rate (ORR) was 86.3% and complete response (CR) rate was 83.8%. After a median follow-up of 17 months, the estimated 2-year progression-free survival (PFS) and overall survival (OS) rates were 75.2% and 86.5%, respectively. Baseline ctDNA concentration ≥ 1 LG/mL was observed in 37.1% of patients and was strongly associated with higher IPI (p 0.0001) and advanced stage (p 0.0001). Those patients with elevated baseline ctDNA had inferior 2-year PFS compared to those with lower level (70.0% versus 83.7%; HR = 2.95; 95% CI, 1.07–8.14; p = 0.028). MRD negative rates increased over time and MRD status at each timepoint was highly prognostic. After the first treatment cycle, 52.9% of patients achieved MRD negativity; MRD-positive patients had a significantly worse outcomes compared with MRD-negative patients (2-year PFS: 48.6% vs. 89.5%; HR = 5.93; 95% CI, 1.69–20.82; p = 0.0016). After cycle 2, MRD negativity rose to 69.6%, with MRD-positive patients showing a higher risk (2-year PFS: 43.5% vs. 82.3%; HR = 4.36; 95% CI, 1.65–11.49; p = 0.0013). At end-of-treatment, MRD negativity reached 79.7%, and MRD positivity was associated with a dramatically inferior prognosis (2-year PFS: 13.3% vs. 83.9%; HR = 12.32; 95% CI, 5.41–43.16; p 0.0001). In multivariate models adjusting for age, stage, LDH, and IPI, MRD positivity at cycle 1 (adjusted HR = 4.87; p = 0.022), cycle 2 (adjusted HR = 4.25; p = 0.014), and end-of-treatment (adjusted HR = 11.69; p 0.001) remained independent predictors of inferior PFS. Furthermore, Integration MRD and PET/CT results after cycle 2 defined four distinct risk groups (p 0.0001). Patients with double-negativity (MRD-/PET-) showed the best 2–year PFS of 90.0%, while double-positive patients (MRD+/PET+) had the worst (14.3%). Discordant groups demonstrated intermediate outcomes: MRD-/PET+, 87.5% and MRD+/PET, 51.8%, suggested that MRD negativity was a stronger predictor of favorable PFS than PET-CT negative alone. Conclusions: In this prospective study of newly diagnosed DLBCL patients, ctDNA-based MRD monitoring provided early and independent prognostic information that outperformed early/interim PET-CT. Integrating MRD and PET assessment significantly improves risk stratification. These findings support the combined use of ctDNA and imaging to enhance risk stratification, guide treatment decisions, and ultimately improve patient outcomes.
Zhang et al. (Mon,) studied this question.