Analysis of cell-free DNA in lymphomas: from sample collection to genotyping and minimal residual disease monitoring

Visual Abstract

Abstract

Circulating tumor DNA (ctDNA) is increasingly investigated in lymphomas because it enables noninvasive molecular profiling, longitudinal assessment of clonal evolution, and quantification of minimal residual disease (MRD), which reflects residual tumor burden and treatment response and serves as a clinically validated prognostic biomarker. The clinical utilities of ctDNA include supporting diagnosis, enabling early detection of relapses, and resolving ambiguous imaging findings. Current approaches for ctDNA assessment in lymphomas include droplet digital polymerase chain reaction, immunoglobulin clonotype sequencing, hybrid-capture next-generation sequencing with unique molecular identifiers or duplex barcoding, and phased sequencing. Establishing ctDNA as a clinical-grade assay requires rigorous quality control and standardization across all technical steps, from blood collection and plasma processing to cell-free DNA extraction, quantification, and analytically validated genotyping and MRD measurement. Large prospective trials and international standardization efforts are underway to define ctDNA-based MRD assessment as a reproducible and clinically actionable tool in lymphoma care. In this review, we outline key preanalytical and analytical workflows for ctDNA assessment in lymphomas and discuss unresolved challenges and future directions in the field.

Introduction

Circulating tumor DNA (ctDNA) has achieved clinical maturity in solid organ malignancies, with US Food and Drug Administration–approved companion diagnostics enabling detection of actionable mutations to guide targeted therapy in lung, breast, and prostate cancers.¹ Beyond treatment selection, ctDNA demonstrates high clinical validity for minimal residual disease (MRD) detection following curative-intent therapy, with postoperative ctDNA positivity strongly predicting recurrence risk across multiple tumor types.^2,3 Additionally, ctDNA enables longitudinal monitoring of treatment response and early detection of resistance mutations.^4,5

Cell-free DNA (cfDNA) has been investigated across a range of hematologic malignancies.6, 7, 8, 9, 10 Studies in myeloid tumors and leukemias have shown limited contribution of cfDNA beyond the typing of circulating malignant cells. In contrast, lymphomas typically lack circulating cancer cells, making cfDNA a particularly valuable tool for noninvasive molecular profiling, disease monitoring, and detection of MRD.11, 12, 13, 14, 15 Consequently, this review focus on lymphomas, as they represent the hematologic malignancy in which cfDNA has been most extensively studied and applied in translational and clinical research.11, 12, 13^,15,16

Ensuring optimal sample quality for ctDNA analysis

The biological origins and elimination mechanisms of cfDNA in healthy and diseased states directly dictate the necessary requirements for preanalytical sample handling (Table 1).

Table 1.

Sources and clearance of cfDNA in health and disease: implications for preanalytical handling

Domain	Key points	Implications for cfDNA/ctDNA analysis
Physiological origin of cfDNA	cfDNA derives from normal cellular turnover, mainly apoptotic hematopoietic cells.6,17,18 Fragments are predominantly double-stranded, nucleosome-associated, histone-bound, and ∼165 bp in length.6,17,19 In pregnancy, fetal/placental DNA contributes to plasma cfDNA.17	It establishes baseline background cfDNA present in healthy individuals.
Physiological kinetics	cfDNA has a short half-life (minutes to hours) and low steady-state plasma concentration (1-10 ng/mL). Rapid clearance maintains low background levels.6,17,20, 21, 22, 23	It supports use of cfDNA as a dynamic biomarker reflecting real-time biological changes.
Definition of ctDNA	In patients with cancer, a fraction of cfDNA originates from tumor cells and is termed “circulating tumor DNA” (ctDNA).19	ctDNA represents the tumor-derived component measurable in plasma.
Determinants of ctDNA detectability	ctDNA levels depend on tumor biology, tumor burden, microenvironment, and host clearance mechanisms.17,19,22, 23, 24	Detection sensitivity varies substantially between patients and disease settings.
Tumor burden and stage	Larger tumors and metastatic/disseminated disease release more ctDNA due to higher cellular turnover. Early-stage or very small tumors may produce ctDNA below assay detection limits.17,19	Advanced cancers generally yield higher assay sensitivity than localized disease.
Tumor location and vascularization	Highly vascularized organs (eg, liver, lung) shed more ctDNA; sanctuary sites (eg, brain) shed less.25	Anatomical site influences plasma ctDNA abundance.
Tumor biology	High proliferation, apoptosis, necrosis, hypoxia, genomic instability, and aggressive biology increase ctDNA shedding. Dominant tumor clones contribute more ctDNA than minor subclones.19,26	ctDNA levels may reflect tumor aggressiveness and clonal architecture.
Therapy-related ctDNA changes	Treatment-induced tumor cell death can transiently increase ctDNA levels.27	Timing of blood collection relative to therapy is important for interpretation.
Typical ctDNA fractions	Early-stage cancers: often <1% of total cfDNA. Advanced/metastatic cancers: typically 1%-10%. Highly aggressive tumors: may exceed 50%.25,28, 29, 30	Assay sensitivity requirements depend on disease context.
Renal clearance	Small protein-free cfDNA fragments (∼150-200 bp) can be filtered and excreted in urine.10,30,31	Renal dysfunction may increase background cfDNA levels.
Hepatic/macrophage clearance	Kupffer cells and hepatic macrophages take up and degrade cfDNA.10,29, 30, 31, 32	Liver dysfunction may impair cfDNA clearance.
Enzymatic degradation	DNases (eg, DNASE1L3) degrade cfDNA in plasma/extracellular fluids. Shorter fragments are cleared faster; nucleosome-bound fragments are more resistant.32, 33, 34	Fragment size and chromatin association influence cfDNA persistence.
Physiological modifiers	Circadian variation and fasting have minimal effects. Intense exercise can transiently elevate cfDNA several-fold, usually normalizing within 30-60 minutes.10,35	Exercise timing should be standardized when feasible.
Inflammation and tissue injury	Trauma and inflammatory conditions increase background cfDNA production.10,30,31	Elevated normal cfDNA may dilute ctDNA and reduce detectability.
Pregnancy-related confounding	Placental cfDNA introduces paternally inherited variants that may mimic tumor mutations.30,36,37	Risk of false-positive variant interpretation.
Transplant-related confounding	Organ transplant recipients may have donor-derived cfDNA. In allogeneic stem cell transplant recipients, circulating hematopoietic DNA is donor-derived, whereas tumor mutations remain recipient-derived.29, 30, 31, 32,38	Germline and CH interpretation become complex in transplant settings.
CH	CH mutations (eg, DNMT3A, TET2, ASXL1, TP53, JAK2, SF3B1, PPM1D) arise from non-malignant hematopoietic clones and may appear in plasma cfDNA.6,39	It represents a major source of false-positive plasma variants.
Approaches to identify CH	Compare plasma cfDNA with matched leukocyte DNA; similar VAF strongly suggests CH origin. Evaluate gene context, tumor relevance, VAF stability, patient age, and longitudinal behavior. Bioinformatic CH classifiers may assist.29,39	Parallel leukocyte sequencing is considered the gold standard.
Preanalytical recommendations	Standardize blood collection conditions when feasible. Avoid sampling immediately after intense exercise or during acute inflammation. Maintain consistency across serial samples.10,30,31,35	This reduces biological variability and improves longitudinal interpretation.
Importance of serial monitoring	Relative ctDNA changes over time are often more informative than absolute values.27,40, 41, 42	Longitudinal sampling is clinically valuable for MRD and response assessment.
Source of germline DNA	Tumor-free DNA is required to distinguish somatic mutations from inherited variants and CH. In allogeneic stem cell transplantation, nonhematopoietic tissues (eg, skin fibroblasts, hair follicles) are required for germline controls.28,32	Choice of germline comparator depends on transplant status.
CH testing in transplantation	Donor-derived blood cells should be evaluated for CH after organ or stem cell transplantation.13,16,43,44	Necessary for accurate plasma variant interpretation.
Tumor tissue vs plasma for baseline genotyping	Tumor tissue provides superior initial variant recovery and higher MRD sensitivity. Plasma captures spatial heterogeneity and confirms ctDNA detectability.45	Tissue-informed MRD assays generally outperform plasma-only approaches.
Limitations of tissue biopsy	Core biopsies may yield limited DNA; FFPE processing may degrade DNA and impair sequencing quality.20,28,46	Tissue may be insufficient or technically suboptimal.
Advantages of plasma genotyping	Plasma may identify mutations absent from biopsy (∼20% of patients) due to tumor heterogeneity. Some DLBCL patients (1%-5%) may have undetectable ctDNA despite disease presence.12,47,48	Plasma complements tissue analysis.
Current evidence gaps	Direct prospective comparisons between tissue-informed and plasma-informed MRD strategies in lymphoma are lacking.	Optimal baseline strategy remains uncertain.
Recommended practical strategy	Biobanking both pretreatment plasma and tumor biopsy is currently the most comprehensive approach.	Minimizes risk of MRD assay failure and maximizes future analytical flexibility.

bp, base pair; CH, clonal hematopoiesis; DLBCL, diffuse LBCL; FFPE, formalin-fixed paraffin-embedded.

To minimize preanalytical variability, maximize ctDNA yield, and ensure accurate detection of tumor-specific genomic alterations, consensus expert reviews and best practice recommend standardized preanalytical procedures, including tube type, timely processing, and proper plasma handling, which are essential to ensure analytical validity and reproducibility of ctDNA assays.^46,49, 50, 51

Proper blood collection and processing procedures prevent contamination with genomic DNA from lysed blood cells. The American Society of Clinical Oncology (ASCO) and the College of American Pathologists (CAP) jointly recommend collecting blood for ctDNA testing in cell stabilization tubes, which allow for delayed processing without significant leukocyte lysis for up to several days at room temperature.²⁸ When processing is performed promptly, standard EDTA tubes may also be used, but plasma separation should occur within 6 hours after collection. Plasma is the preferred matrix for ctDNA testing rather than serum, because clotting during serum preparation releases genomic DNA from leukocytes, diluting the tumor-derived DNA fraction and reducing analytical sensitivity.⁵²

Blood should be drawn from peripheral veins, and the phlebotomist should follow the tube manufacturer’s instructions. A 10 mL blood draw typically yields ∼4.7 mL plasma. This represents the minimum recommended plasma volume required to obtain sufficient cfDNA input for highly sensitive ctDNA detection assays. Immediately after collection, the tube should be gently inverted 8 to 10 times to ensure proper mixing of the stabilizing reagent with the blood, avoiding vigorous shaking, because this may result in hemolysis. Tubes should be stored and transported at ambient temperature (18-25°C), avoiding refrigeration (<10°C) or exposure to high heat (>37°C), because extreme temperatures can affect cell stability and DNA preservation. Samples can be shipped at room temperature using protective packaging to avoid physical stress (vibration, impact, freezing, or overheating).53, 54, 55, 56, 57

After collection, plasma must be separated from cellular components before freezing, as freezing unspun whole blood can degrade ctDNA. Upon arrival at the laboratory, a 2-step centrifugation (eg, 1600g for 10 minutes, then 16 000g for 10 minutes) is usually performed to remove cellular debris before freezing plasma at −80°C. Processed plasma should be aliquoted into single-use fractions (eg, 1 mL vials) to avoid multiple freeze-thaw cycles, which can degrade nucleic acids and reduce ctDNA detectability.⁵⁸

Paired genomic DNA from leukocytes can be extracted from the buffy coat of EDTA or cell stabilization tubes. Although cell stabilization tubes are designed to stabilize blood cells and prevent lysis, the leukocytes remain intact in the tube.59, 60, 61, 62, 63

Extraction of cfDNA from plasma

Spin column–based and magnetic bead–based kits are the most commonly used and validated approaches for cfDNA extraction, with spin column methods being used in most published studies. The Association for Molecular Pathology (AMP) and the CAP recommend selecting methods that are validated for cfDNA recovery, especially for clinical assays.⁵⁸ cfDNA fragment size reflects biological origin: normal cfDNA consists predominantly of nucleosome-associated fragments of ∼166 base pairs (bp) from apoptotic hematopoietic cells, whereas tumor-derived cfDNA is enriched in shorter fragments (100-150 bp) due to altered chromatin structure and aberrant nuclease activity in cancer cells.^33,64,65 This size difference facilitates tumor detection, because extraction methods that preferentially recover shorter DNA fragments lead to relative enrichment of ctDNA. This enrichment enhances the sensitivity of downstream mutational analyses by increasing the number of detectable tumor-specific variants and improving the likelihood of identifying low-frequency mutations.⁶⁵ Different extraction kits vary in their efficiency for recovering short vs long cfDNA fragments. For applications requiring high yield of short fragments (eg, mutational analysis), kits that favor short fragment recovery should be selected. For quantitative analysis, methods with high reproducibility and minimal size bias should be prioritized. Automated extraction platforms can reduce hands-on time and improve reproducibility.^66,67 Always quantify cfDNA yield and assess integrity after extraction using fluorometric analysis and polymerase chain reaction (PCR)–based methods to ensure suitability for downstream applications.⁶⁸

Increasing the amount of cfDNA input in the assay generally improves the sensitivity of ctDNA detection, because a larger number of DNA molecules increases the probability of capturing rare tumor-derived fragments. However, this benefit reaches a plateau once the optimal input level is achieved, typically around ∼30 to 50 ng of cfDNA. Beyond this range, the assay approaches saturation, and additional cfDNA provides only marginal gains in analytical sensitivity.69, 70, 71 Consequently, further increasing cfDNA input, which would require collecting blood volumes beyond the standard 10 to 20 mL, results in only limited incremental improvement in detection performance. This reflects the practical balance between maximizing molecular input and maintaining feasible and minimally invasive blood collection protocols.^45,72

For digital droplet PCR (ddPCR) mutation analysis, the minimum recommended input is typically 10 to 20 ng of cfDNA per reaction, with optimal input being 20 to 50 ng. This range allows for reliable detection of low-frequency variants, balancing sensitivity and assay precision, as supported by technical validation studies and clinical chemistry assay standards. For ddPCR, the lower limit of detection (LOD) is constrained by the number of mutant DNA copies present. Validation studies have shown that reliable quantification is achieved with at least 35 to 64 mutant copies per reaction.^58,73

For targeted next-generation sequencing (NGS) methods such as cancer personalized profiling by deep sequencing (CAPP-seq) or phased sequencing (PhasED-seq), the minimum cfDNA input is generally 10 ng, with recommended input of 30 to 50 ng per reaction based on assay validation protocols.^58,69 For high-sensitivity applications (eg, MRD), higher input (≥30 ng) is preferred to maximize molecular complexity and detection of rare events^47,58,69; sensitivity improves with greater input, although practical cfDNA yields often limit achievable amounts.⁴⁷ AMP and the CAP recommend reporting the input cfDNA for NGS assays and discourage using the extract without quantitation.

Approaches for ctDNA qualification and quantification

The principal methodologies used for ctDNA qualification and/or quantification in lymphoma are PCR-based and NGS-based assays including ddPCR, immunoglobulin sequencing, hybrid-capture NGS with chemical error suppression, hybrid-capture NGS with digital error suppression, PhasED-seq, and whole-exome sequencing (WES).

ddPCR

ddPCR is robust for single-nucleotide variants, copy number alterations, and gene rearrangements, and is less affected by DNA quality than conventional quantitative PCR, allowing analysis of fragmented or low-input samples. Limitations include the need for prior knowledge of the target mutation and lower multiplexing capacity compared with NGS-based approaches.74, 75, 76, 77

Immunoglobulin sequencing

Immunoglobulin gene rearrangements, primarily the variable (diversity) joining (V(D)J) junctions of the immunoglobulin heavy and light chains, are tracked using immunoglobulin sequencing, which relies on universal primers to amplify and sequence immunoglobulin heavy and light chain genes to identify lymphoma clonotypes.^78,79 High mass of input cfDNA and deep sequencing are necessary to detect rare clonotypes in plasma, with 50 to 100 ng of ctDNA and unique coverage exceeding 10 000× per target being required to achieve sensitivity down to 10⁻⁴. Advantages include high tumor specificity and broad applicability across B-cell lymphomas. False-positive risk arises from background B-cell clonotypes unrelated to the tumor, and from lower specificity of light chain clonotypes, which may overlap with normal B-cell populations. Additionally, technical artifacts and contamination from lysed peripheral blood mononuclear cells can confound results.80, 81, 82, 83

Issues in detection of tumor clonotype in baseline cfDNA include low ctDNA abundance before therapy, which may limit sensitivity; the possibility that the dominant tumor clonotype in tissue is not represented in plasma due to spatial heterogeneity or sampling bias; and problems in tumor clonotype detection in immunoglobulin hypermutated lymphomas, including somatic hypermutation in the immunoglobulin gene, which can disrupt primer binding and lead to false negatives or incomplete clonotype recovery. This is a particular challenge in germinal center–derived lymphomas such as follicular lymphoma, in which hypermutation is extensive. Formalin-fixed paraffin-embedded (FFPE) biopsy samples of the diagnostic lymphoma biopsy are often used to inform on the tumor clonotype that has to be tracked in cfDNA. FFPE samples often yield fragmented or chemically modified DNA, reducing the efficiency and accuracy of clonotype identification and calibration for subsequent MRD tracking. This can result in lower calibration rates and missed detection of the tumor clonotype.^80,81,84

Hybrid-capture NGS with chemical error suppression

Hybrid-capture NGS (eg, CAPP-seq) targets panels of lymphoma-relevant genetic aberrations (single-nucleotide variants, translocations, copy number changes). They typically include between 30 and 475 genes. Larger panels aim at capturing the full spectrum of recurrent and driver genetic alterations in various lymphoma subtypes.⁴⁰ Hybrid-capture NGS integrates both chemical and computational error-suppression strategies to enhance analytical sensitivity and specificity.⁸⁵ Although error filtering can reliably detect and filter a large proportion of false positives, such as those arising from sequencing artifacts (especially at fragment ends), cytosine deamination (C:G>T:A transitions), polymerase errors, technical artifacts such as those arising from homopolymers, pseudogenes, and high GC content regions leading to misaligned reads, or platform-specific noise, additional refinement is almost always necessary. Cytosine deamination is particularly increased in cfDNA compared with cellular genomic DNA, likely due to prolonged exposure to plasma nucleases and the extracellular environment, contributing to elevated substitution error rates, especially C:G>T:A transitions, which are not restricted to fragment ends and can persist through error correction.⁸⁶

Chemical error suppression uses unique molecular identifiers (UMIs) or duplex sequencing. UMIs are short, random nucleotide tags ligated to each DNA fragment before PCR amplification. During analysis, reads sharing the same UMI are grouped into “families,” and a consensus sequence is generated for each family. This consensus approach eliminates random PCR and sequencing errors, as true variants are expected to be present in multiple reads within a family, whereas errors are typically unique to individual reads. Increasing the family size (number of reads per UMI) further reduces error rates.⁸⁷ Duplex sequencing is a molecular error-suppression technique in which both strands of each original double-stranded DNA molecule are independently tagged with UMI, sequenced, and then computationally compared to generate consensus sequences that represent the original DNA template. During library preparation, distinct UMIs are ligated to both the forward and reverse strands of each DNA fragment. After sequencing, reads are grouped by UMI and strand orientation. Only variants confirmed in consensus sequences from both strands are considered true positives, whereas discordant or single-strand variants are filtered out as error.⁸⁸ The recommended sequencing depth of coverage when using UMIs for accurate ctDNA detection in lymphoma is at least 10 000× unique, deduplicated coverage per target region, which typically translates to a raw sequencing depth of 30 000× to ≥100 000×, depending on the assay design and cfDNA input. This depth ensures sufficient read families (ie, consensus reads derived from multiple raw reads sharing the same UMI) to confidently detect low-frequency variants and suppress technical errors. However, increasing sequencing depth reaches a point of saturation in sensitivity. This saturation occurs because the number of unique original DNA fragments is limited by cfDNA input (∼10 000 cell equivalents per 60 ng), imposing a ceiling on deduplicated depth that cannot be exceeded regardless of sequencing reads. Beyond this point, further increases in coverage do not improve the LOD, which becomes constrained by background error rates inherent to sequencing chemistry and sample preparation.^85,89,90

Hybrid-capture NGS with digital error suppression

Digital error suppression uses computational approaches to reduce sequencing errors without physical molecular barcodes. This includes grouping reads with identical mapped start and end positions into consensus families, under the assumption that reads sharing exact genomic coordinates likely originate from the same original molecule, analogous to how UMI-tagged reads are grouped, but using position rather than barcodes. Additionally, platform-specific error suppression involves subtracting a predefined “blacklist” of stereotyped sequencing artifacts known to recur in particular genomic contexts or instrument workflows.⁹¹ By filtering out these recurrent technical errors, the pipeline markedly reduces false-positive variant calls, especially at very low variant allele frequencies (VAF). Together, UMI-based chemical barcoding and digital error filtering creates a highly robust error-suppression system, enabling reliable detection of rare ctDNA mutations that would otherwise be obscured by background noise.^88,92, 93, 94

ctDNA assays can be broadly categorized as tumor-agnostic (library-based, fixed panels detecting recurrent mutations without previous tumor sequencing) and tumor-informed (personalized assays tracking patient-specific variants identified from tumor tissue). Tumor-agnostic approaches enable rapid deployment and do not require tumor material, making them broadly applicable in clinical settings; however, they are generally limited by lower sensitivity (typically, VAF of ≥0.1%) and may miss rare or patient-specific variants. In contrast, tumor-informed approaches achieve substantially higher sensitivity (VAF of ∼0.001%-0.01%) by integrating signal across multiple patient-specific mutations, which is particularly advantageous for MRD detection. This increased sensitivity comes at the cost of requiring tumor tissue for assay design, longer turnaround times, and higher costs. Hybrid strategies that combine elements of both approaches are emerging to balance sensitivity and practicality. Genotyping pretreatment ctDNA using hybrid-capture NGS serves 2 key purposes: identifying mutations with diagnostic, prognostic, or therapeutic relevance, and establishing a personalized panel of tumor-specific mutations to be tracked for tumor-informed MRD assessment in subsequent longitudinal cfDNA samples.^{41,61,93,95,96} This initial genotyping step is therefore essential, although particularly challenging at low VAF typical of ctDNA, and variant callers differ substantially in their sensitivity and specificity. Using multiple variant callers in parallel offers a more conservative analytical approach, reducing the risk of false-positive mutations but increasing the need for expert review. For these reasons, manual inspection of candidate somatic variants remains a critical component of current genotyping workflows, in which the goal is to curate a patient-specific mutation profile from tumor or baseline sequencing data. Expert curators typically review each variant using the Integrative Genomics Viewer, evaluating read support, strand balance, mapping quality, and local alignment context. Variants are often graded on a standardized 5-point scale, in which 1 reflects minimal evidence (eg, a single supporting molecule or clear artifactual features), and 5 represents variants with robust, high-quality support, fully consistent with a true somatic event. To ensure consistency and reproducibility across analysts and laboratories, many institutions implement a formal standard operating procedure for the manual review and refinement of somatic variants.^97,98 Of note, for MRD monitoring, in which hundreds to thousands of variants may be tracked simultaneously across serial samples, individual variant-level manual review is neither practical nor required. Instead, MRD assays use automated, aggregate statistical frameworks that integrate signal across all monitored loci to generate a composite detection call, substantially improving scalability and reproducibility. Machine learning–based approaches for automating variant curation in the genotyping setting are also under active development and may further reduce reliance on manual review as these tools undergo formal clinical validation.

PhasED-seq

PhasED-seq refers to the detection and tracking of multiple somatic mutations (phased variants [PVs]) that are physically close together on the same DNA fragment in ctDNA using hybrid-capture targeted NGS. In B-cell lymphomas, these PVs arise due to activation-induced cytidine deaminase (AID)–mediated somatic hypermutation, resulting in clusters of mutations within short genomic regions. By identifying and sequencing these PVs together on individual cfDNA molecules, PhasED-seq enables highly specific tracking of tumor-derived DNA fragments.⁴⁷ For error suppression, PhasED-seq leverages the extremely low probability that technical or biological sequencing errors would independently generate the same pattern of multiple, closely spaced mutations on a single DNA molecule (ie, observed within a single read or read pair). This co-occurrence of variants on the same molecule enables substantial reduction of background error rates, distinct from duplex sequencing approaches, which instead rely on concordant mutation detection across both strands of the original DNA duplex (Figure 1).

Figure 1.

Conceptual comparison of variant units and error-suppression strategies in cfDNA sequencing approaches. Schematic overview of increasing analytical sensitivity with progressively stricter variant definitions. SNV-based approaches (left) detect single mutations using UMI consensus (∼10⁻⁵ sensitivity). DCVs (middle) require concordant mutations on both DNA strands (in trans), improving sensitivity to ∼10⁻⁶. PVs (right) detect multiple mutations on the same DNA fragment (in cis), achieving sensitivities approaching ∼10⁻⁷. Increasing stringency reduces background error and enhances true signal. DCV, duplex consensus variant; SNV, single-nucleotide variants.

Assays based on PVs can be influenced by false-positive signals arising from biological background. These background sources include occult indolent B-cell neoplasms, low-count monoclonal B-cell lymphocytosis, monoclonal gammopathy of undetermined significance, as well as other processes that release immunoglobulin gene rearrangements into the circulation. Additionally, AID–mediated somatic hypermutation may aberrantly introduce PVs into oncogenes in normal B-cells, particularly within circulating memory B-cells. As a result, ultra-sensitive ctDNA assays may detect these background PVs, potentially leading to false-positive calls.⁹⁹

WES

WES of ctDNA in lymphomas has been investigated in a limited number of studies due to technical and sensitivity constraints related to the costs of ultra-deep sequencing the entire exome.¹⁶

Analytical validation of assays for lymphoma ctDNA qualification and quantification

The Clinical and Laboratory Standards Institute (CLSI) guidance EP17-A2 is used to define assay specificity, or the limit of blank (LOB), and analytical sensitivity, or the LOD. The LOB is the highest value expected to be observed from a series of measurements on a sample that contains no analyte (blank samples). To determine the LOB, multiple replicates (typically ≥60) of blank samples (eg, plasma from healthy donors with no detectable ctDNA) are measured using the ctDNA assay. Then the LOB is calculated using the formula LOB = meanblank + 1.645 × SDblank, in which meanblank is the mean value of the blank replicates and SDblank is the standard deviation (SD) of those replicates. The factor 1.645 corresponds to the 95th percentile of the normal distribution. The LOD is the lowest analyte concentration that can be reliably distinguished from the LOB and at which detection is feasible with a 95% probability. To determine the LOD, low-level ctDNA samples near the expected LOB have to be prepared and multiple replicates (≥60) for each concentration measured. The LOD is calculated as LOD = LOB + 1.645 × SDlowconc_sample, in which SDlowconc_sample is the SD of the low-level ctDNA sample measurements. This approach ensures that the LOD reflects both the blank and low-level sample variability. These procedures are directly applicable to ctDNA assays, including ddPCR and NGS-based methods.^73,100, 101, 102, 103, 104 AMP and CAP recommend reporting LOB and LOD using these standardized approaches for clinical ctDNA assay validation.⁵⁸

Comprehensive multisite evaluations of leading pancancer ctDNA platforms have shown that reliable detection is achieved above 0.5% allele frequency, with high sensitivity and reproducibility. Below this threshold, detection becomes less reliable, particularly when input DNA is limited, underscoring the importance of LOD determination for clinical decision-making.^105,106 ddPCR assays validated per CLSI guidelines report LOBs of 4 mutant copies and LODs of 12 to 22 copies and detection of ctDNA mutations at very low allele frequencies (as low as 0.00125%-0.005%).^73,107 Among ctDNA platforms specifically designed for lymphomas, hybrid-capture NGS assays (eg, CAPP-seq) show a LOD that is generally in the range of 0.002% to 0.01% allele frequencies (20-100 ctDNA particles per million cfDNA particles or ppm), depending on input DNA quantity and error-suppression methods.⁹⁴ PCR-based assays and IG clonotype sequencing methods have higher LODs, typically 0.01% to 0.1% allele frequency (100-1000 ppm). For the PhasED-seq assay, the LOB is defined by a background error rate of 1.95 × 10⁻⁸ (ie, 0.0000000195%) and a false-positive rate of 0.24% in blank plasma samples. Using 120 ng of input DNA under standardized analytical conditions, the LOD is 0.7 ppm, which corresponds to VAF of 0.00007%.¹⁰⁰ However, in clinical practice, the achievable LOD varies across patients depending on the number of trackable PVs identified from the individual tumor and the quantity of input cfDNA available. Recent prospective data demonstrate that LOD95 (limit of detection at 95%) can vary by >2 orders of magnitude across patients, underscoring the importance of reporting patient-specific LOD with each assay result to enable accurate interpretation of negative MRD findings. Key aspects of the studies that quantified lymphoma ctDNA-MRD are summarized in Table 2., Table 3., Table 4., Table 5., Table 6..

Table 2.

Studies that quantified lymphoma ctDNA using ddPCR

References	Cohort	Number of genes assessed by the platform	Sensitivity of the assay	Lymphoma type	Results
Figaredo et al11	Prospective observational	56 genes	0.02%	Follicular lymphoma and LBCL	Baseline: 70 patients tested; ctDNA detected in 50 (71.4%) patients; EoT ctDNA-MRD: 31 samples tested; 14 (45.2%) with persistent ctDNA-MRD
Haider et al108	Prospective observational	252 genes	0.0025%-0.02%	DLBCL and follicular lymphoma	Baseline: 8 patients tested; ctDNA detected in 7 (88%) patients; Interim ctDNA-MRD: 6 samples tested; 0 (0%) with persistent ctDNA-MRD; ctDNA-MRD−: 36-mo PFS, 16.7%; EoT ctDNA-MRD: 6 samples tested; 0 (0%) with persistent ctDNA-MRD; ctDNA-MRD−: 36-mo PFS, 33%
Drandi et al74	Prospective observational	1 (MYD88 L265P)	NA	Waldenström macroglobulinemia, IgM-MGUS, other lymphomas	NA
Song et al109	Prospective observational	112 genes	0.0001%	DLBCL	Baseline: 52 samples tested; ctDNA detected in 51 (98.1%) patients; Interim ctDNA-MRD: 51 samples tested; 14 (26.9%) patients with persistent ctDNA-MRD; ctDNA-MRD+: mPFS, NR; ctDNA-MRD−: mPFS, 23.6-mo (P = .286); EoT ctDNA-MRD: 48 samples tested; persistent ctDNA-MRD, NA; PFS, NA
Heger et al103	Retrospective observational	NA	NA	DLBCL	Baseline: 18 patients tested (12 evaluable samples); ctDNA detected in 10 (83.3%) patients; Interim ctDNA-MRD: 12 samples tested; 8 (66.7%) with persistent ctDNA-MRD; ctDNA-MRD+: 2-y PFS, 0%; ctDNA-MRD−: 2-y PFS, 75%; EoT ctDNA-MRD: NA
Camus et al110	Retrospective observational	9 genes	0.5%	Primary mediastinal LBCL vs cHL	Baseline: 40 PMBL and 60 cHL tested; ctDNA detected in 32 (80%) and 41 (68.3%) patients, respectively
Camus et al111	Prospective observational	9 genes	0.5%	cHL	Baseline: 60 patients tested; ctDNA detected in 41 (68.8%); Interim ctDNA-MRD: 55 patients tested; 0 (0%) with persistent ctDNA-MRD; ctDNA-MRD−: 1-y PFS, 89.8%; EoT ctDNA-MRD: NA
Bohers et al112	Prospective observational	34 genes	NA	DLBCL	Baseline: 30 patients tested; ctDNA detected in19 (63%); Interim ctDNA-MRD: 16 patients tested; 4 (25%) with persistent ctDNA-MRD; EoT ctDNA-MRD: 20 patients tested
Drandi et al113	Retrospective observational	2 genes (MYD88 and CXCR4)	0.005%	Waldenström macroglobulinemia	Baseline: 148 patients (194 samples) tested; ctDNA detected in 60 (41%); Interim ctDNA-MRD: 27 patients tested; 23 (85%) with persistent ctDNA-MRD
Camus et al114	Retrospective observational	1 gene (XPO1)	0.1%	HL	Baseline: 50 patients tested; ctDNA detected in 8 (16%); EoT ctDNA-MRD: 28 patients tested; 7 (25%) with persistent ctDNA-MRD; ctDNA-MRD+: 2-y PFS, 57.1% (95% CI, 30.1-100); ctDNA-MRD−: 2-y PFS, 90.5% (95% CI, 78.8-100; P = .0601)
Alcaide et al115	NA	4 genes (EZH2, STAT6, MYD88, and CCND3)	NA	DLBCL and follicular lymphoma	NA
Mitobe et al116	Retrospective observational	1 gene (MYD88)	67.5%	PCNSL	Baseline: 119 patients tested; ctDNA detected in 40 (33.6%); Interim ctDNA-MRD: 1 patient tested (no mutant droplets after 5-mo); EoT ctDNA-MRD: serial monitoring for MRD was not systematically performed

A systematic literature search for studies evaluating ctDNA-MRD in lymphoma was conducted on 4 March 2026. The search was performed in PubMed using the following strategy: ("lymphoma"[MeSH Terms] OR lymphoma[Title/Abstract]) AND ("liquid biopsy"[Title/Abstract] OR "cell-free DNA"[Title/Abstract] OR cfDNA[Title/Abstract] OR "circulating tumor DNA"[Title/Abstract] OR ctDNA[Title/Abstract])). The search was limited to original research articles published in English. Retrospective and prospective studies including >20 patients in any lymphoma subtype were considered for inclusion.

cHL, classic Hodgkin lymphoma; DLBCL, diffuse LBCL; EoT, end of treatment; HL, Hodgkin lymphoma; CI, confidence interval; IgM, immunoglobulin M; MeSH, medical subject headings; MGUS, monoclonal gammopathy of undetermined significance; NA, not assessable; NR, not reached; PCNSL, primary CNS lymphoma; PFS, progression-free survival; PMBL, primary mediastinal B-cell lymphoma.

Table 3.

Studies that quantified lymphoma ctDNA using immunoglobulin sequencing

References	Cohort	Sensitivity of the assay	Lymphoma type	Results
Khouja et al43	Prospective observational	0.02%	DLBCL	Baseline: 52 patients tested; 49 (94%) patients with ctDNA detected; Interim ctDNA-MRD at 2 time points: at cycle 2 day 1, 48 patients tested; 23 (47.9%) patients with persistent ctDNA-MRD; ctDNA-MRD−: 2-y PFS, 68%; ctDNA-MRD+: 2-y PFS, 91% (HR, 4.51); at cycle 4 day 1, 47 patients tested; 8 (17.0%) patients with persistent ctDNA-MRD; ctDNA-MRD−: 1.5-y PFS, 60%; ctDNA-MRD+: 1.5-y PFS, 84% (HR, 3.34; 95% CI, 0.83-13.48; P = .07); EoT ctDNA-MRD: 44 patients tested; 8 (18.2%) patients with persistent ctDNA-MRD; ctDNA-MRD−: 18-mo PFS, 60%; ctDNA-MRD+: 18-mo PFS, 91% (HR, 6.38; 95% CI, 1.27-32.01; P = .01)
Kurtz et al60	Prospective observational	0.0001%	DLBCL	Baseline: 22 samples tested; 18 (82%) patients with ctDNA detected; Interim ctDNA-MRD: 8 patients tested; 0 (0%) patients with persistent ctDNA-MRD; EoT ctDNA-MRD: 25 patients (51 samples) tested; 3 (12%) patients with persistent ctDNA-MRD
Pott et al79	NA	NA	BCL	NA
Kurtz et al41	Retrospective observational	NA	DLBCL	Baseline: 217 samples tested; 212 (98%) patients with ctDNA detected; Interim ctDNA-MRD: 67 patients tested; 16 (23.8%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 2-y PFS, 50%; ctDNA-MRD−: 2-y PFS, 83% (P = .0015); EoT ctDNA-MRD: 52 patients tested; 11 (21.2%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 2-y PFS, 46%; ctDNA-MRD−: 2-y PFS, 82% (P = .001)
Dean et al117	Retrospective observational	0.0001%	LBCL	Baseline: 57 patients tested; 57 (100%) patients with ctDNA detected; Interim ctDNA-MRD: 53 samples tested; 24 (45.3%) with persistent ctDNA-MRD; EoT ctDNA-MRD: 42 samples tested; <50% with persistent ctDNA-MRD
Soscia et al118	Retrospective observational	NA	DLBCL	Baseline: 73 patients tested; 68 (93.2%) patients with ctDNA detected; Interim ctDNA-MRD: 45 samples tested; 13 (28.9%) with persistent ctDNA-MRD; ctDNA-MRD+: mPFS, 20.7-mo; ctDNA-MRD−: mPFS, NR (HR, 4.2; 95% CI 1.3-13.4; P = .0001); EoT ctDNA-MRD: 47 samples tested; 10 (21.3%) with persistent ctDNA-MRD; ctDNA-MRD+: PFS, 0%; ctDNA-MRD−: PFS, 83.8% (HR, 24.5; 95% CI, 5.9-100.8; P = .0001)
Miljkovic et al119	Clinical trial	NA	Peripheral T-cell lymphoma	Baseline: 45 patients tested; 34 (76%) patients with ctDNA detected; Interim ctDNA-MRD: 24 samples tested; 15 (63%) patients with persistent ctDNA-MRD; ctDNA-MRD+: mPFS, 2.0-y (95% CI, 0.1 to NR); ctDNA-MRD−: mPFS, 8.4-y (95% CI, 0.1 to NR) (P = .32); EoT ctDNA-MRD: 24 samples tested; 11 (46%) patients with persistent ctDNA-MRD; ctDNA-MRD+: mPFS, 0.7-y (95% CI, 0.05-7.7); ctDNA-MRD−: mPFS, NR (95% CI, 0.4 to NR) (P = .08)
Frank et al120	Prospective observational	NA	DLBCL	Baseline: 72 patients tested; 57 (79%) patients with ctDNA detected; Interim ctDNA-MRD: 60 patients tested; 27 (45%) patients with persistent ctDNA-MRD; ctDNA-MRD+: mPFS, 3.03-mo (P = .0001); ctDNA-MRD−: mPFS, NR; EoT ctDNA-MRD: NA
Wang et al121	Retrospective observational	NA	DLBCL	Baseline: 33 patients tested; 26 (78.8%) patients with ctDNA detected; Interim ctDNA-MRD: 28 samples tested; 8 (28.6%) with persistent ctDNA-MRD; ctDNA-MRD+: 2-y PFS, 12.5%; ctDNA-MRD−: 2-y PFS, 45% (P = .003); EoT ctDNA-MRD: 14 samples tested; 3 (21.4%) with persistent ctDNA-MRD; ctDNA-MRD+: 2-y PFS, NR; ctDNA-MRD−: 2-y PFS, 41.7% (P = .017)
Melani et al122	Clinical trial	NA	DLBCL	Baseline: 50 patients tested; 45 (90%) patients with ctDNA detected; Interim and EoT ctDNA-MRD: NA
Taranto et al123	Retrospective-prospective observational	NA	cHL	Baseline: 12 patients (51 samples) tested; 1 (8.3%) patient with ctDNA detected; EoT ctDNA-MRD: 12 patients tested; 3 (33%) patients with persistent ctDNA-MRD; ctDNA-MRD−: 3-y PFS, 0%
Lakhotia et al124	Prospective observational	NA	Mantle cell lymphoma	Baseline: 48 patients tested; 47 (98%) patients with ctDNA detected; Interim ctDNA-MRD: 50 samples tested; 27 (54%) patients with persistent ctDNA-MRD; ctDNA-MRD+: mPFS, 1.8-y (95% CI, 1.2-2.9); ctDNA-MRD−: mPFS, 2.7-y (95% CI, 1.7-8.1); EoT ctDNA-MRD: 50 samples tested; 10 (20%) patients with persistent ctDNA-MRD; ctDNA-MRD+: mPFS, 1.5-y (95% CI, 1.7-3.9); ctDNA-MRD−: mPFS, 2.6-y (95% CI, 0.4-2.0) (P = .003)
Qi et al125	Prospective observational	NA	Extranodal natural killer/T-cell lymphoma	Baseline: 21 patients tested; 12 (57.1%) patients with ctDNA detected; EoT ctDNA-MRD: 20 patients tested; 5 (25%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 1-y PFS, 28.6%; ctDNA-MRD−: 1-y PFS, 83.9 (P = .002)
Zhang et al126	Clinical trial	NA	Peripheral T-cell lymphoma	Baseline: 23 samples tested; 18 (78.3%) patients with ctDNA detected; Interim and EoT ctDNA-MRD: NA
Ruan et al127	Clinica trial	0.0001%	Mantle cell lymphoma	Baseline: 34 patients tested; 100% patients with ctDNA detected; Interim ctDNA-MRD: 21 patients tested; 3 (14.3%) patients with persistent ctDNA-MRD; ctDNA-MRD−: 4-y PFS, 76%

A systematic literature search for studies evaluating ctDNA-MRD in lymphoma was conducted on 4 March 2026. The search was performed in PubMed using the following strategy: ("lymphoma"[MeSH Terms] OR lymphoma[Title/Abstract]) AND ("liquid biopsy"[Title/Abstract] OR "cell-free DNA"[Title/Abstract] OR cfDNA[Title/Abstract] OR "circulating tumor DNA"[Title/Abstract] OR ctDNA[Title/Abstract])). The search was limited to original research articles published in English. Retrospective and prospective studies including >20 patients in any lymphoma subtype were considered for inclusion.

cHL, classic Hodgkin lymphoma; DLBCL, diffuse LBCL; EoT, end of treatment; HR, hazard ratio; CI, confidence interval; MeSH, medical subject headings; mPFS, median progression-free survival; NA, not assessable; NR, not reached; PFS, progression-free survival.

Table 4.

Studies that quantified and qualified lymphoma ctDNA using hybrid-capture NGS

References	Cohort	Number of genes assessed by the platform	Sensitivity of the assay	Lymphoma type	Results
Alcoceba et al13	Prospective observational	344 Kb, 340 SNVs distributed among 50 genes	0.5%	DLBCL	Baseline: 68 patients tested; 61 (90%) patients with ctDNA detected; Interim ctDNA-MRD: 64 patients tested (44 evaluable samples); 6 (13.6%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 2-y PFS, 0%; ctDNA-MRD−: 2-y PFS, 76% (P < .001); EoT ctDNA-MRD: NA
Pirosa et al14	Retrospective observational	344 kb, 155 genes	0.1%	cHL	Baseline: 297 patients tested; 252/280 (90%) patients with ctDNA detected based on somatic mutations; Interim ctDNA-MRD: 6% of cases with persistent ctDNA-MRD; inferior PFS in patients with interim ctDNA-MRD+;EoT ctDNA-MRD: 16% of cases with persistent ctDNA-MRD; all patients with EoT ctDNA persistence showed no response or rapid relapse; EoT ctDNA persistence was associated with inferior PFS compared with ctDNA clearance
Heger et al15	Prospective observational	1516 Mb, 279 genes	NA	CNS lymphoma	Baseline: 67 patients (146 samples) tested; 56 (83.6%) patients with ctDNA detected; Interim ctDNA-MRD: 30 patients tested; 9 (30%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 2-y FFS, 55.6%; ctDNA-MRD−: 2-y FFS, 73.3% (P = .0610); EoT ctDNA-MRD: 21 patients tested; 4 (19%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 2-y FFS, 0%; ctDNA-MRD−: 2-y FFS, 84.9% (P = .0001)
Alig et al16	Prospective observational	151 genes	0.5%	HL	Baseline: 366 patients tested; 293 (80.1%) patients with ctDNA detected; Interim ctDNA-MRD: 109 patients (310 samples) tested; 11 (10%) patients with persistent ctDNA-MRD; EoT ctDNA-MRD in ctDNA-MRD+: 2-y PFS, 0%; ctDNA-MRD−: 2-y PFS, 29.0% (P = 2.7 × 10−7)
Zou et al27	Prospective observational	188 genes	>0.1%	DLBCL	Baseline: 23 patients tested; 21 (91.3%) patients with ctDNA detected; Interim ctDNA-MRD: 23 patients tested; 11 (47.8%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 1-y PFS, 27.3%; ctDNA-MRD−: 1-y PFS, 90.9% (P < .015); EoT ctDNA-MRD: NA
Moia et al26	Prospective observational	109 genes	0.1%	DLBCL	Baseline: 166 patients tested; 100% patients with ctDNA detected (99 low level, 68 high level); Low ctDNA level: 4-y PFS, 71.7%; high ctDNA level: 4-y PFS, 50.3% (P = .0018)
Vimalathas et al22	Prospective observational	53 genes	0.1%	LBCL	Baseline: 14 patients tested; 11 (79%) patients with ctDNA detected; Interim ctDNA-MRD: 14 samples tested; 4 (28%) patients with persistent ctDNA-MRD; EoT ctDNA-MRD: 14 patients tested; 4/5 patients with R/R disease had persistent ctDNA-MRD; PFS, NA
Bruscaggin et al61	Clinical trial	LyV4.0 CAPP-seq, 344 kb, 155 genes	0.1%	DLBCL	Baseline: 67 patients (172 samples) tested; 15 (22.4%) patients with high levels ctDNA detected; Interim ctDNA-MRD: 21 patients tested; 15 (71.4%) patients with persistent ctDNA-MRD; ctDNA-MRD+: mPFS, 2.2-mo; ctDNA-MRD−: mPFS, 15.2-mo (P = .045); EoT ctDNA-MRD: NA
Meriranta et al88	Clinical trial	748 kb, 345 genes	0.0083%	Mantle cell lymphoma	Baseline: 41 patients tested; 39 (95%) patients with ctDNA detected; Interim ctDNA-MRD: 29 patients tested; 4 (14%) patients with persistent ctDNA-MRD; ctDNA-MRD+: mPFS, 126 d; 9-mo PFS, 100%; ctDNA-MRD−: mPFS, 241 d; EoT ctDNA-MRD: NA
Shin et al89	NA	66 genes	88%	Non-HL	NA
Claudel et al128	Clinical trial	130 kb, 111 genes	NA	Follicular lymphoma	Baseline: 141 patients tested; 140 (99.3%) patients with ctDNA detected; EoT ctDNA-MRD: 124 patients tested; 12 (9.7%) patients with persistent ctDNA-MRD; ctDNA-MRD+: mPFS, 17.7-mo (1.4 to NR); ctDNA-MRD−: mPFS, NR
Zhang et al129	Retrospective observational	NA	NA	Non-DLBCL NHL: follicular lymphomas, mantle cell lymphomas, chronic lymphocytic leukemia/small lymphocytic lymphomas, marginal zone lymphomas	Baseline: 203 patients tested; 108 (88.7%) patients with ctDNA detected; Interim and EoT ctDNA-MRD: NA
Yoon et al130	Prospective observational	61 genes	NA	Follicular lymphoma	Baseline: 40 patients tested; 40 (100%) patients with ctDNA detected; Interim ctDNA-MRD: 35 samples tested; EoT ctDNA-MRD: 14 samples tested; no significant difference in PFS based on baseline ctDNA concentration (P = .240)
Jiménez-Ubieto et al131	Prospective observational	NA	0.02%	Follicular lymphoma	Baseline: 84 patients (151 samples) tested; 67 (79.8%) patients with ctDNA detected; Interim and EoT ctDNA-MRD: ctDNA-MRD+: inferior PFS (HR, 19.1; 95% CI 4.10-89.4; P < .001)
Heger et al132	Clinical trial	0.1-35.7 Mb, 275 genes	0.1%	HL	Baseline: 213 patients tested; 165 (77.5%) patients with ctDNA detected; Interim ctDNA-MRD: 64 patients tested; inflammatory immune-escape HL: 3-y PFS, 67% vs 17.9% in MRD− vs MRD+, respectively; virally driven HL: 3-y PFS, 78.6% vs 27.2% in MRD− vs MRD+, respectively
Meriranta et al133	Clinical trial	NA	NA	DLBCL	Baseline: 100 patients tested; 97 (97%) patients with ctDNA detected; Interim ctDNA-MRD: 58 patients tested; 21 (36%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 5-y PFS, 52%; ctDNA-MRD−: 5-y PFS, 54% (P = .22); EoT ctDNA-MRD: 71 patients tested; 10 (14%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 5-y PFS, 10%; ctDNA-MRD−: 5-y PFS, 52% (P = .00011)
Li et al134	Retrospective observational	188 genes	0.1%	DLBCL	Baseline: 52 patients tested; 51 (98%) patients with ctDNA detected; Interim ctDNA-MRD: 38 patients tested; 10 (26.3%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 20-mo PFS, 10%; ctDNA-MRD−: 20-mo PFS, 57.1% (HR, 3.65; 95% CI, 1.452-9.178; P = 0.033); EoT ctDNA-MRD: 25 patients tested; 7 (57.1%) patients with persistent ctDNA-MRD
Herrera et al135	Clinical trial	320 kb	NA	DLBCL	Baseline: 43 patients tested; 43 (100%) patients with ctDNA detected; EoT ctDNA-MRD: 23 patients tested; 12 (52.2%) patients with persistent ctDNA-MRD
Sobesky et al136	Prospective observational	43 genes	0.00945%	HL	Baseline: 121 patients tested; 109 (90%) patients with ctDNA detected; Interim ctDNA-MRD: 35 patients tested; ctDNA-MRD+: inferior PFS (P = .0341); EoT ctDNA-MRD: 22 patients tested
Xia et al137	Clinical trial	112 genes	16 000× (for plasmatic ctDNA)	DLBCL, non–germinal center	Baseline: 26 patients tested; N/A patients with ctDNA detected; Interim ctDNA-MRD: 23 samples tested; 6 (26%) patients with persistent ctDNA-MRD; ctDNA-MRD−: 2-y PFS, 95%; EoT ctDNA-MRD: 21 patients tested; 4 (19%) patients with persistent ctDNA-MRD; in ctDNA-MRD+, only 1/4 patients relapsed 16-mo after treatment
Liang et al138	Prospective observational	475 genes	0.3%	Peripheral T-cell lymphoma	Baseline: 64 patients tested; 25 (96.2%) patients with ctDNA detected; EoT ctDNA-MRD: 58 patients tested; 43 (74.1%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 24-mo PFS, 13.95%; ctDNA-MRD−: 24-mo PFS, 20%
Krasnow et al139	Retrospective observational	143-5000 SNV	<0.01%	LBCL	Baseline: 33 patients tested; 28 (84%) patients with ctDNA detected; EoT ctDNA-MRD: 28 samples tested; 13 (46%) nonresponders with persistent ctDNA-MRD; PFS, NA
Leppä et al140	Clinical trial	NA	NA	LBCL	Baseline: 123 patients tested; patients with ctDNA detected, NA; Interim ctDNA-MRD: 112 patients tested; 37 (33%) patients with persistent ctDNA-MRD; EoT ctDNA-MRD: 99 patients tested; 11 (11%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 20-mo PFS, 18%; ctDNA-MRD−: 20-mo PFS, 80% (P < .0001)
Sworder et al141	Prospective observational	186 genes	NA	LBCL	Baseline: 138 patients tested; 77 (55.8%) patients with ctDNA detected; EoT ctDNA-MRD: 96 patients tested; 14 (14.6%) patients with persistent ctDNA-MRD; ctDNA-MRD+: PFS, significantly inferior (P = .028)
Westin et al142	Clinical trial	37 genes	NA	DLBCL	Baseline: 29 patients tested; 27 (93%) patients with ctDNA detected; Interim ctDNA-MRD: 22 patients tested; 10 (45.5%) patients with persistent ctDNA-MRD; EoT ctDNA-MRD: NA
Desch et al143	Prospective observational	NA	0.6%	Pediatric HL	Baseline: 96 patients tested; 72 (75%) patients with ctDNA detected; Interim ctDNA-MRD: 49 patients tested; 5 (10.2%) patients with persistent ctDNA-MRD; EoT ctDNA-MRD: NA
Spina et al144	Retrospective observational	77 genes	0.1%	cHL	Baseline: 80 patients tested; 65 (81.2%) patients with ctDNA detected; Interim ctDNA-MRD: 24 patients tested; 18 (75%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 2-y PFS, 0%; ctDNA-MRD−: 2-y PFS, 83% (P = .001); EoT ctDNA-MRD: NA
Assouline et al145	Clinical trial	13 genes	NA	DLBCL	Baseline: 26 patients tested; 25 (96.2%) patients with ctDNA detected; Interim ctDNA-MRD: 26 patients tested; 20 (76.9%) patients with persistent ctDNA-MRD; ctDNA-MRD+ associated with significantly inferior PFS (P = .0049)
Chamba et al146	Prospective observational	20 genes	74.1%	EBV+ Burkitt lymphoma	Baseline: 308 patients tested; 268 (87%) patients with ctDNA detected; Interim and EoT ctDNA-MRD: NA
Liang et al147	Prospective observational	475 genes	73.5%	DLBCL	Baseline: 164 patients tested; 100% with ctDNA detected; Interim ctDNA-MRD: NA EoT ctDNA-MRD: 164 patients tested; 61 (37.2%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 36/61 patients (59%) presented progression of disease, with a PFS rate of 41%; ctDNA-MRD−: 13/103 patients (12.6%) presented progression of disease, with a PFS rate of 87.4%
Wang et al148	Retrospective observational	425 genes, 3 mutations (PIM1, MYD88 L265P, and CD79B)	NA	Intravascular LBCL	Baseline: 17 patients tested; ctDNA detected: PIM1 in 12/17 (71%) patients, MYD88 L265P in 11/17 (65%) patients, CD79B in 7/17 (41%) patients; Interim and EoT ctDNA-MRD: NA
Shi et al149,150	Retrospective observational	152 genes	78.78%	DLBCL	Baseline: 90 patients tested; 82 (91.1%) with ctDNA detected; Interim and EoT ctDNA-MRD: NA
Zhong et al151	Retrospective observational	115 genes	93%	R/R DLBCL (primary DLBCL and transformed follicular lymphoma)	Baseline: 29 patients tested; 27 (93.1%) with ctDNA detected Interim ctDNA-MRD: 24 patients tested; 16 (66.7%) patients with persistent ctDNA-MRD; EoT ctDNA-MRD: 20 patients tested; 11 (55%) patients with persistent ctDNA-MRD
Khouja et al44	Clinical trial	73 genes	≥0.02%	Mantle cell lymphoma	Baseline: 57 patients tested; 100% with ctDNA detected; Interim ctDNA-MRD: 55 patients tested; 30 (55%) patients with persistent ctDNA-MRD; ctDNA-MRD+: inferior FFS (HR, 2.52; 95% CI, 0.79-8.06; P = .11); EoT ctDNA-MRD: 47 patients tested; 20 (43%) patients with persistent ctDNA-MRD; ctDNA-MRD+: inferior FFS (HR, 2.04; 95% CI, 0.55-7.63; P = .28)
Hori et al152	Prospective observational	33 genes	100%	Follicular lymphoma	Baseline: 30 patients tested; 100% with ctDNA detected; Interim and EoT ctDNA-MRD: patients with low baseline cfDNA (<16.56 μg/mL) had a 2-y PFS of 75.6%; patients with high baseline cfDNA (≥16.56 μg/mL) had a 2-y PFS of 47.6%
Zhang et al153	Prospective observational	106 genes	71.4%	DLBCL	Baseline: 57 patients tested; 43 (75.4%) with ctDNA detected; Interim and EoT ctDNA-MRD: NA; patients with high pretreatment ctDNA concentrations (≥2.13 log hGE/mL) had significantly worse PFS (P = .026); patients who were CSF ctDNA-positive demonstrated significantly shorter PFS (P = .036)

A systematic literature search for studies evaluating ctDNA-MRD in lymphoma was conducted on 4 March 2026. The search was performed in PubMed using the following strategy: ("lymphoma"[MeSH Terms] OR lymphoma[Title/Abstract]) AND ("liquid biopsy"[Title/Abstract] OR "cell-free DNA"[Title/Abstract] OR cfDNA[Title/Abstract] OR "circulating tumor DNA"[Title/Abstract] OR ctDNA[Title/Abstract])). The search was limited to original research articles published in English. Retrospective and prospective studies including >20 patients in any lymphoma subtype were considered for inclusion.

95% CI, 95% confidence interval; cHL, classic Hodgkin lymphoma; CSF, cerebrospinal fluid; DLBCL, diffuse LBCL; EBV, Epstein-Barr virus; EoT, end of treatment; FFS, failure-free survival; hGE, human genomic equivalent; HL, Hodgkin lymphoma; HR, hazard ratio; kb, kilobases; Mb, megabases; MeSH, medical subject headings; mPFS, median progression-free survival; NA, not assessable; NHL, non-Hodgkin lymphoma; NR, not reached; PFS, progression-free survival; R/R, relapsed/refractory; SNV, single-nucleotide variants.

Table 5.

Studies that quantified lymphoma ctDNA using PhasED-seq

References	Cohort	Number of genes assessed by the platform	Sensitivity of the assay	Lymphoma type	Results
Roschewski et al12	Clinical trial	NA	0.0001%	LBCL, including DLBCL, HGBCL, and PMBL	Baseline: 126 patients tested; 125 (99.2%) patients with ctDNA detected; Interim ctDNA-MRD: 94 patients tested; 42 (45%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 2-y PFS, 67%; ctDNA-MRD−: 2-y PFS, 96% (HR, 28.7; 95% CI, 9.7-85; P < .001); EoT ctDNA-MRD: 126 patients tested; 28 (22%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 2-y PFS, 29%; ctDNA-MRD−: 2-y PFS, 97% (HR, 54.8; 95% CI, 12.6-238.2; P < .001)
Arnason et al23	Clinical trial	>400 regions enriched	NA	DLBCL	Baseline: 86 patients tested; 84 (98%) patients with ctDNA detected; Interim ctDNA-MRD: 77 patients tested; 50 (64.9%) patients with persistent ctDNA-MRD; ctDNA-MRD+: mPFS, 3.2-mo (95% CI, 2.5-4.0); ctDNA-MRD−: mPFS, 19.4-mo (HR, 0.42 [0.24-0.75]; 95% CI, 9.2-33.7); EoT ctDNA-MRD: NA
				Relapsed/refractory follicular lymphoma	Baseline: 65 patients tested; 64 (98%) patients with ctDNA detected; Interim ctDNA-MRD: 60 patients tested; 28 (46.7%) patients with persistent ctDNA-MRD; ctDNA-MRD+: mPFS, 21.6-mo (95% CI, 5.9-32.1); ctDNA-MRD−: mPFS, 42.4-mo (HR, 0.31 [0.14-0.67]; 95% CI, 35.2 to NE); EoT ctDNA-MRD: NA
Kurtz et al47	Prospective observational	315 kb, 468 genes	NA	DLBCL	Baseline: 107 patients tested; 105 (98%) patients with ctDNA detected; Interim ctDNA-MRD: 88 samples tested; 49 (56%) patients with persistent ctDNA-MRD; ctDNA-MRD+ associated with significantly inferior PFS (P = .003); EoT ctDNA-MRD: 19 patients tested; 5 (26%) patients with persistent ctDNA; ctDNA-MRD+ associated with significantly inferior PFS (P = .022)
Vodicka et al154	Prospective observational	521 genes	NA	DLBCL	Baseline: 44 patients tested; 32 (78%) patients with ctDNA detected; Interim ctDNA-MRD: after 2 y, 20 patients tested; 10 (50%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 24-mo PFS, 46%; ctDNA-MRD−: 24-mo PFS, 90% (HR, 0.26; 95% CI, 0.07-0.95; P = .065)
Mutter et al155	Prospective observational	NA	NA	CNS lymphoma	Baseline: 202 patients tested (78 plasma samples + 24 CSF samples); ctDNA detected in 61 (78.2%) plasma samples and 24 (100%) CSF samples; Interim ctDNA-MRD: 28 patients (49 plasma samples) tested; 14 (50%) patients with persistent ctDNA-MRD; ctDNA-MRD+ associated with significantly inferior PFS (HR, 6.2; 95% CI, 2.3-16.7; P = .0002); EoT ctDNA-MRD: NA
Fernández-Miranda et al156	Prospective observational	78 genes	NA	Follicular lymphoma	Baseline: 36 patients tested; 28 (77.8%) patients with ctDNA detected; Interim ctDNA-MRD: 25 samples tested; ctDNA-MRD+: 2-y PFS, 50%; ctDNA-MRD−: 2-y PFS, 90% (P = .008); EoT ctDNA-MRD: 22 patients tested; ctDNA-MRD+: 30-mo PFS, 20%; ctDNA-MRD− 2-y PFS, 72.7% (P = .13)
Dondolin et al157	Prospective observational	109 genes	0.1%	DLBCL	Baseline: 120 patients tested with ctDNA detected (43 ctDNA-high and 77 ctDNA-low); Interim and EoT ctDNA-MRD: NA
Lynch et al158	Clinical trial	NA	0.0001%	HL	Baseline: 29 patients tested; 26 (89.7%) patients with ctDNA detected; Interim ctDNA-MRD: 26 patients tested; 5 (19.2%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 2-y PFS, 0%; ctDNA-MRD−: 2-y PFS, 14.3% (P = .025); EoT ctDNA-MRD: 24 patients tested; 2 (8.3%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 2-y PFS, 0%; ctDNA-MRD−: 2-y PFS, 13.0% (P = .0012)
Lynch et al159	Clinical trial	NA	NA	Aggressive non-Hodgkin B-cell lymphoma	Baseline: 18 patients tested; EoT ctDNA-MRD: ctDNA-MRD+ associated with significantly inferior PFS (P = .0015)
Krupka et al45	Prospective observational	559 kb, 242 genes	0.0001%	LBCL: DLBCL, HGBCL, PMBL, and transformed follicular lymphoma	Baseline: 184 patients tested; 100% with ctDNA detected; EoT ctDNA-MRD: 155 patients tested; 38 (21%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 2-y PFS, 38.0% (95% CI, 24.9-57.9); ctDNA-MRD−: 2-y PFS, 90.2% (95% CI, 84.4-96.3) (P < .001)
Wang et al48	Prospective observational	NA	0.01%	LBCL, including DLBCL and HGBCL	Baseline: 136 patients tested; 134 (99%) with ctDNA detected; EoT ctDNA-MRD: 134 patients tested; 23 (17%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 3-y PFS, 17%; ctDNA-MRD−: 3-y PFS, 85% (HR, 9.3; 95% CI, 4.8-18.1; P = 5.19e−11)

A systematic literature search for studies evaluating ctDNA-MRD in lymphoma was conducted on 4 March 2026. The search was performed in PubMed using the following strategy: ("lymphoma"[MeSH Terms] OR lymphoma[Title/Abstract]) AND ("liquid biopsy"[Title/Abstract] OR "cell-free DNA"[Title/Abstract] OR cfDNA[Title/Abstract] OR "circulating tumor DNA"[Title/Abstract] OR ctDNA[Title/Abstract])). The search was limited to original research articles published in English. Retrospective and prospective studies including >20 patients in any lymphoma subtype were considered for inclusion.

95% CI, 95% confidence interval; CSF, cerebrospinal fluid; DLBCL, diffuse LBCL; EoT, end of treatment; HGBCL, high-grade B-cell lymphoma; HL, Hodgkin lymphoma; HR, hazard ratio; CI, confidence interval; kb, kilobases; MeSH, medical subject headings; mPFS, median progression-free survival; NA, not assessable; NE, not estimable; PFS, progression-free survival; PMBL, primary mediastinal B-cell lymphoma.

Table 6.

Studies that quantified lymphoma ctDNA using WES

References	Cohort	Number of genes assessed by the platform	Sensitivity of the assay	Lymphoma type	Results
Narkhede et al81	Retrospective observational	16 SNV	NA	DLBCL	Baseline: 41 patients tested; 39 (95%) patients with ctDNA detected; EoT ctDNA-MRD: 36 patients tested; 9 (25%) patients with persistent ctDNA-MRD; ctDNA-MRD+: 10-mo EFS, 22%; ctDNA-MRD−: 10-mo PFS, 63% (HR, 6.5; 95% CI, 1.9-22; P < .001)
Mehta et al160	Clinical trial	16 SNV	NA	Indolent non-HL	Baseline: 7 patients tested; patients with ctDNA detected, NA; Interim and EoT ctDNA-MRD: NA
Collins et al161	Clinical trial	600 genes	NA	DLBCL	Baseline: 24 patients tested; Interim and EoT ctDNA-MRD: NA data
Nieto et al162	Retrospective observational	>9982 genes	76.9%	PCNSL, brain metastases, and glioblastoma	Baseline: 17 patients tested; 11 (64.7%) patients with ctDNA detected; among patients with hematologic diseases, 5 patients tested with 100%; Interim and EoT ctDNA-MRD: 1 patient with lymphoma tested (ctDNA-MRD− at T2 and T3); PFS, NA
Shen et al163	Clinical trial	475 genes	100% (tumor tissue), 76.5% (CSF), 68.4% (plasma)	PCNSL	Baseline: 44 samples tested; ctDNA detected in 8/8 patients (100%) in tumor tissue, 13/17 patients (76.5%) in CSF, 13/19 patients (68.4%) in plasma; EoT ctDNA-MRD: 14 patients tested (13 achieving CR, 1 achieving PR); 6 (42.9%) patients with persistent ctDNA-MRD (5 achieving CR, 1 achieving PR); ctDNA-MRD+ associated with worst PFS (P = .04)

A systematic literature search for studies evaluating ctDNA-MRD in lymphoma was conducted on 4 March 2026. The search was performed in PubMed using the following strategy: ("lymphoma"[MeSH Terms] OR lymphoma[Title/Abstract]) AND ("liquid biopsy"[Title/Abstract] OR "cell-free DNA"[Title/Abstract] OR cfDNA[Title/Abstract] OR "circulating tumor DNA"[Title/Abstract] OR ctDNA[Title/Abstract])). The search was limited to original research articles published in English. Retrospective and prospective studies including >20 patients in any lymphoma subtype were considered for inclusion.

CR, complete response; CSF, cerebrospinal fluid; DLBCL, diffuse LBCL; EFS, event-free survival; EoT, end of treatment; HL, Hodgkin lymphoma; HR, hazard ratio; MeSH, medical subject headings; NA, not assessable; PCSNL, primary CNS lymphoma; PFS, progression-free survival; PR, partial response; SNV, single-nucleotide variants; T2, 1 week after initiation of methotrexate.

Clinical applications of ctDNA

Although a mutation profile derived from ctDNA cannot substitute for a pathological diagnosis based on tissue examination, which remains the gold standard, noninvasive molecular testing can serve as a valuable complementary tool in selected clinical contexts. This is particularly relevant when biopsy material is insufficient for molecular profiling or when lesions are anatomically difficult to access. Examples include primary central nervous system (CNS) lymphoma, in which cerebrospinal fluid ctDNA detection of MYD88 L265P may support diagnosis when stereotactic biopsy carries substantial risk, and vitreoretinal lymphoma, in which aqueous humor or vitreous fluid cfDNA sequencing can aid diagnosis when cytology is inconclusive.^155,164, 165, 166 In systemic lymphomas, liquid biopsy can identify clinically meaningful mutations when tissue is exhausted or unavailable for genomic profiling.^{20,128,167,168}

Some lymphoma subtypes exemplify these challenges. Intravascular large B-cell lymphoma (LBCL) often presents with nonspecific systemic symptoms and lacks discrete tumor masses, making biopsies difficult or nondiagnostic, whereas in primary CNS lymphoma lesions are located within the CNS, in which stereotactic biopsies carry procedural risks or may be contraindicated. In both cases, detection of the MYD88 L265P mutation in plasma or cerebrospinal fluid–derived cfDNA can provide a valuable diagnostic clue and guide early clinical decision-making.^155,169

Beyond intravascular LBCL and primary CNS lymphoma, several additional clinical scenarios might benefit from ctDNA-based genotyping and require further investigations.¹⁷⁰ These include suspected relapse in deep or surgically inaccessible anatomical sites, in which repeated biopsies are risky or impractical (eg, patients with poor performance status or comorbidities that preclude invasive procedures)^81,117,129, 130, 131, 132; tumors with high necrotic content or low cellularity, in which conventional biopsies frequently yield insufficient DNA for sequencing^{81,129,130,132,156}; and clinical situations with discordant or equivocal histopathological findings, in which ctDNA mutational signatures may support or refine the provisional diagnosis.^{81,117,129,132,156} Current guidelines and consensus in the medical literature do not support ctDNA as a standalone diagnostic tool for transformation because of the lack of validated molecular signatures of transformation; histopathology is required to confirm high grade.¹⁷⁰

Integrating positron emission tomography (PET) with ctDNA-MRD testing is a logical approach. PET imaging generally requires macroscopic lesions of at least 7 to 8 mm in diameter for reliable detection, and it cannot detect microscopic residual disease.¹⁷¹ PET also has limitations in specificity, with ∼30% of biopsy specimens from ¹⁸fluorodeoxyglucose-avid lesions after lymphoma treatment turning out to be false positives, likely due to therapy-induced inflammatory changes.¹⁷² A growing body of evidence consistently supports ctDNA-MRD at the end of treatment as a prognostic marker in LBCL. Three recent large prospective studies using phase variant–based ctDNA-MRD technology, collectively encompassing >425 patients, have demonstrated that end-of-treatment ctDNA-MRD status is strongly and independently prognostic for progression-free survival and overall survival, outperforming conventional PET-based response assessment.^12,45,48 These findings, together with earlier studies using diverse ctDNA methodologies, provide robust evidence.^{12,26,41,45,47,48,118,133,173}

The National Comprehensive Cancer Network Clinical Practice Guidelines in Oncology for B-Cell Lymphomas version 2.2025, have been updated to include ctDNA-MRD assessment for patients with diffuse LBCL who exhibit partial metabolic response (ie, residual ¹⁸fluorodeoxyglucose uptake with a Deauville score of 4 or 5 with reduced uptake compared with baseline) on PET at the conclusion of frontline therapy. The guidelines recommend ctDNA-MRD assessment with a specific restriction, serving as an alternative to biopsy for evaluating positive PET findings when biopsy is not feasible. Patients with partial metabolic response, but ctDNA-MRD–negative test results, can follow the complete metabolic response pathway, which involves imaging and clinical follow-ups (category 2B recommendation).^{134,135,157,174}

Open questions and next steps

Until recently, the evaluation of ctDNA-MRD in lymphoma had been limited to a few hundred highly selected patients, with a notable lack of prospective data in broader, unselected populations. However, 3 recent large prospective studies have substantially expanded the evidence base, collectively encompassing >425 patients with end-of-treatment ctDNA-MRD data.^12,45,48 These studies demonstrate that systematic prospective sample collection across multicenter networks achieves high feasibility: in the HOVON-902 real-world cohort spanning >50 centers, PV identification was successful in 99% of patients using either tumor tissue or pretreatment plasma, and ctDNA-MRD was prognostic regardless of baseline sample source. Similarly, the DIRECT study demonstrated feasibility using an independently developed, open-source assay across multiple sites. These findings indicate that the practical challenges of missing samples and out-of-qualification patients, which affected roughly half of patients in earlier studies, can be substantially mitigated through standardized prospective collection procedures. Nevertheless, important questions remain regarding the optimal implementation of ctDNA-MRD in routine clinical practice, including the clinical actionability of positive MRD results, the need for serial monitoring vs single time point assessment, and the significance of very low-level residual ctDNA.

The increasing use of core needle biopsies, reaching up to 60% to 80% in some institutions, exacerbates this issue. Core biopsies are often entirely consumed during routine diagnostic workup, leaving no residual tissue available for MRD marker discovery. An alternative strategy is to derive MRD markers from pretreatment plasma samples. This approach has the added benefit of confirming that the tumor sheds ctDNA into bloodstream at detectable levels, but it requires preemptive collection and storage of sufficient plasma volumes from all patients with newly diagnosed diffuse LBCL, which would demand substantial investment in biobanking infrastructure.¹¹⁹

Moreover, the ability of ctDNA-MRD assays to provide results within clinically relevant timeframes in real-world settings has yet to be demonstrated. Turnaround times vary by methodology: ddPCR can provide results within 1 to 3 days^175,176; hybrid-capture NGS (eg, CAPP-seq, PhasED-seq) typically requires 7 to 14 days, depending on sequencing and bioinformatic processing, with 9-day turnaround demonstrated in the SAKK39/19 trial¹⁷⁷; immunoglobulin-NGS platforms generally require 2 to 3 weeks due to centralized laboratory processing and batching requirements.¹⁷⁸ Although commercial providers, such as clonoSEQ, LymphoVista, CLARITY, and the KAPA HyperCap Design Share (DS) non-Hodgkin lymphoma (NHL) panel, are entering the field, and some academic laboratories have developed certified in-house tests, access to ctDNA-MRD remains limited. Ongoing efforts to standardize and harmonize assays, such as those led by the EuroClonality network, are essential for establishing a broader laboratory network, which currently represents a major bottleneck to patient access.^120,136

Several ongoing clinical trials are expected to generate higher-quality and more extensive evidence on the feasibility and diagnostic performance of ctDNA-MRD in LBCL (eg, ClinicalTrials.gov identifiers: NCT06693830, NCT06828991, NCT06902012, NCT06594939, NCT05675982, NCT04604067, NCT04980222, and NCT03758989).

Conflict-of-interest disclosure: D.R. has received honoraria from AstraZeneca, BeOne, Lilly, and Takeda; research grants from AbbVie, AstraZeneca, and Janssen; and travel grants from AstraZeneca, Janssen, and Roche. The remaining authors declare no competing financial interests.