ctDNA for MRD assessment is prognostic and predictive in newly diagnosed and relapsed/refractory lymphoma

Key Points

• •ctDNA clearance during first-line therapy strongly predicts EFS and is more prognostic than standard imaging across lymphoma subtypes.
• •Longitudinal ctDNA monitoring refines risk stratification, enables treatment individualization, and may improve clinical outcomes.

Visual Abstract

Abstract

Circulating tumor DNA (ctDNA) is a novel biomarker to assess tumor burden, molecular residual disease (MRD), and treatment response in various cancers. To our knowledge, this study reports the largest real-world cohort to date evaluating ctDNA testing, ctDNA-MRD detection, and ctDNA clearance dynamics in newly diagnosed and relapsed/refractory lymphoma. We analyzed plasma from 144 patients with stage I to IV primary lymphoma (aggressive B-cell lymphoma, n = 110; aggressive T-cell lymphoma, n = 13; indolent B-cell lymphoma, n = 15; indolent T-cell lymphoma, n = 6; median, 7 samples per patient). ctDNA was evaluated using a personalized, tumor-informed, multiplex polymerase chain reaction assay (Signatera). Pretreatment ctDNA was detectable in 92% of the evaluable patients. End-of-treatment (EOT) ctDNA-MRD positivity was prognostic of inferior event-free survival (EFS; adjusted hazard ratio [HR], 22.43; 95% confidence interval [CI], 6.76-74.45; P < .0001). In multivariate analysis, EOT ctDNA-MRD positivity was the most powerful independent predictor of EFS (HR, 37.10; 95% CI, 6.20-221.8; P < .001), exceeding the prognostic value of EOT positron emission tomography (PET) positivity (HR, 5.88; 95% CI, 1.35-25.6). Early ctDNA clearance during first-line therapy was associated with significantly improved EFS (adjusted HR, 8.57; 95% CI, 2.55-28.81; P = .005). In the relapsed setting, following CART, ctDNA-MRD negativity predicted prolonged event-free intervals. ctDNA-MRD positivity during posttherapy surveillance was associated with significantly inferior EFS (adjusted HR, 33.74; 95% CI, 9.34-121.82; P < .0001). These data establish ctDNA as a robust biomarker for prognosis, treatment monitoring, and early relapse detection across B- and T-cell lymphoma subtypes, with superior prognostic value over conventional imaging, supporting its integration into response assessment.

Introduction

Lymphomas are a diverse group of immune system cancers that include Hodgkin lymphoma and non-Hodgkin lymphoma.¹ In the United States, approximately 20 000 deaths and 90 000 new cases of lymphoma are expected in 2025.² These cancers vary in cell of origin, genomic alterations, histology, aggressive or indolent behavior, and other risk factors and clinical features.3, 4, 5 Early diagnosis of relapsed/refractory disease can lead to timely intervention, improve the likelihood of long-term response, and extend survival.6, 7, 8, 9, 10, 11, 12

Despite advances in treatment and improved patient survival, many patients with lymphoma still face a high risk of primary treatment failure or relapse. Therefore, it is essential to identify these individuals and develop better prognostic and predictive tools to guide personalized treatment strategies. Circulating tumor DNA (ctDNA) contains genetic and molecular information about the tumor and can be found in the plasma of patients with cancer, including those with hematological malignancies.^13,14 ctDNA assessment for molecular residual disease (MRD; ctDNA-MRD) has been validated in both solid tumors and lymphoma to measure treatment response and predict outcomes.15, 16, 17 In addition, ctDNA serves as a prognostic indicator for disease progression and is a minimally invasive method for monitoring residual disease, with the potential to inform clinical decisions and affect disease management.^18,19

The Signatera assay (Natera, Inc) is a personalized, tumor-informed ctDNA test that detects tumor-specific somatic variants found in plasma. In diffuse large B-cell lymphoma (DLBCL), tumor-informed ctDNA testing reveals that 95% of the patients are ctDNA-MRD positive at baseline (before therapy), with levels positively correlating to Revised International Prognostic Index (R-IPI) score and disease stage.¹⁷ Changes in ctDNA levels (dynamics) predict treatment response and have long-term prognostic implications.¹⁷ Specifically, ctDNA clearance during front-line therapy predicts improved response and better outcomes, providing a superior prognostic indicator for event-free survival (EFS) compared to the R-IPI score and predicting complete response (CR) or no evidence of disease (NED) well before standard imaging and confirmatory biopsy.¹⁷ This highlights the value of tumor-informed ctDNA testing in assessing treatment response, improving risk-stratification, and monitoring recurrence in patients with DLBCL. However, the prognostic utility of ctDNA across different lymphoma subtypes (aggressive and indolent) remains understudied.

To our knowledge, this study presents the largest set reported to date of prospective real-world data on tumor-informed ctDNA testing, ctDNA-MRD detection, and ctDNA clearance dynamics in patients with both newly diagnosed and relapsed/refractory lymphoma, including a subset of patients after chimeric antigen receptor T-cell (CAR-T) therapy, in routine practice.

Materials and methods

Study population

This study analyzed data from commercial ctDNA testing prospectively collected in the real-world setting from September 2020 to August 2025 in patients with stage I to IV lymphoma (N = 144; aggressive B-cell lymphoma, n = 110; aggressive T-cell lymphoma, n = 13; indolent B-cell lymphoma, n = 15; indolent T-cell lymphoma, n = 6; Figure 1A). ctDNA testing was performed at baseline (before treatment) and longitudinally during first-line (1L) therapy, during consolidation therapy, at the end of treatment (EOT; 21 days before or until 3 months after treatment completion), and during posttreatment surveillance until the last clinical follow-up. Patient and disease characteristics, including stage, Eastern Cooperative Oncology Group performance status, R-IPI, 1L therapy, follow-up time, and survival outcomes, are summarized in Table 1. Complete information on patient subtype and associated ctDNA positivity rate is provided in Figure 1B. Patients were included in the ctDNA analysis if they had completed clinical annotation and had a confirmed absence of any concomitant malignancy. Treatment and follow-up, including the cadence of ctDNA testing, were at the discretion of the treating physician and were conducted in accordance with standard clinical practice. Clinicopathologic information was obtained for all analyzed patients. A subset of patients with DLBCL evaluated as part of this cohort was included in a previous report.¹⁷ This study incorporates additional data maturation. This study was conducted in compliance with Natera protocol 21-058 and in accordance with the Declaration of Helsinki. A waiver of consent and of the requirement for documentation of informed consent was granted according to the Code of Federal Regulations 45 CFR 46.116(d) and 45 CFR 46.117(c)(2), respectively.

Figure 1.

CONSORT diagram and baseline ctDNA levels. (A) Patient cohort showing the total number of patients and plasma samples acquired, stratified by lymphoma subtype. The number of ctDNA tests evaluated in subsequent analyses is shown. (B) Baseline ctDNA-MRD detection rates in patients with lymphoma, stratified by lymphoma subtype. ALCL, anaplastic large cell lymphoma; CTCL, cutaneous T-cell lymphoma; FL, follicular lymphoma; NA, not applicable; PCNSL, primary central nervous system lymphoma; PTCL, peripheral T-cell lymphoma; tFL, transformed follicular lymphoma; TCL, T-cell lymphoma.

Table 1.

Cohort demographics and characteristics

Characteristic	Overall (N = 144)	Aggressive (n= 123)	Indolent (n = 21)
Age, median (range), y	61 (18-84)	59 (18-84)	64 (20-83)
Sex, n (%)
Female	67 (47)	59 (48)	8 (38.1)
Male	77 (53)	64 (52)	13 (61.9)
Stage
I, n (%)	15 (11)	9 (7.3)	6 (24.6)
II, n (%)	29 (21)	26 (21.1)	3 (14.3)
III, n (%)	16 (12)	12 (9.8)	4 (19)
IV, n (%)	75 (56)	69 (56.1)	6 (24.6)
Not reported, n	9	7	2
ECOG
0, n (%)	50 (46)	42 (34.1)	8 (38.1)
1, n (%)	35 (32)	32 (26)	3 (14.3)
2, n (%)	16 (15)	14 (11.4)	2 (9.5)
3, n (%)	6 (5.6)	5 (4.1)	1 (4.8)
4, n (%)	1 (0.9)	1 (0.8)	0 (0)
Not reported, n	36	29	7
R-IPI
0-2; very good, n (%)	31 (25.2)	31 (25.2)	NA
3-5; poor, n (%)	33 (26.8)	33 (26.8)	NA
Not reported, n	80	59	NA
1L therapy
Other, n (%)	31 (22)	23 (18.7)	8 (38.1)
Other rituximab, n (%)	17 (12)	11 (8.9)	6 (24.6)
POLA-R-CHOP, n (%)	6 (4.3)	6 (4.9)	0 (0)
R-CHOP, n (%)	71 (49)	67 (54.4)	4 (19)
R-EPOCH, n (%)	15 (10)	15 (12.2)	0 (0)
Not reported, n	4	1	3
Follow-up EFS, median (range), mo	20 (1-108)	20 (1-108)	22 (1-72)
Follow-up OS, median (range), mo	21 (1-108)	20 (1-108)	22 (5-72)
ctDNA-MRD time points, median (range)	7 (1-32)	8 (1-32)	4 (1-14)
Deaths, n	24	23	1

ECOG, Eastern Cooperative Oncology Group; OS, overall survival; POLA-R-CHOP, polatuzumab vedotin with R-CHOP; R-CHOP, rituximab, cyclophosphamide, hydroxydaunorubicin, oncovin (vincristine), and prednisone; NA, not applicable; R-EPOCH, etoposide with R-CHOP.

Biospecimen collection and processing

Biospecimens were obtained and processed as part of routine clinical testing using a tumor-informed ctDNA assay. Tumor DNA was isolated from formalin-fixed, paraffin-embedded biopsy samples for all participants. For germ line analysis, a single 6-mL blood sample was drawn into an EDTA tube from each patient. Whole blood samples (n = 1142; median, 7 per patient; [range, 1-32]) for ctDNA analysis were collected in 2 Streck tubes (10 mL each) at baseline, during therapy, at EOT, and during follow-up, at the physician’s discretion.

Personalized multiplex polymerase chain reaction–based next-generation sequencing assay for ctDNA detection

ctDNA was detected and quantified using a personalized, tumor-informed, multiplex polymerase chain reaction assay (Signatera, Natera, Inc.), as previously described.²⁰ Whole-exome sequencing was performed on formalin-fixed, paraffin-embedded tumor tissue and matched whole blood samples to identify up to 16 patient-specific somatic single-nucleotide variants for incorporation into a personalized assay for ctDNA detection. Cell-free DNA was extracted from a median of 9.4 mL of plasma (range, 1.2-14.2). Plasma samples with at least 2 variants detected were defined as ctDNA-MRD positive/detected. ctDNA concentration was reported as mean tumor molecules per milliliter of plasma.

Statistical analysis

Clinical outcomes were correlated with EOT ctDNA-MRD status at the end of 1L therapy (or 1L with consolidation or maintenance therapy, if given), ctDNA clearance dynamics during 1L induction, and ctDNA-MRD status during surveillance. For EOT analysis, ctDNA time points from 21 days preceding to 90 days following completion of 1L therapy were considered. ctDNA clearance was defined as conversion from positive to negative ctDNA result during 1L therapy (with positivity documented at baseline or before the first negative result), encompassing both transient and sustained clearances. Events were defined as relapse, progression in patients with refractory disease, or death from any cause; patients were classified as relapsed if they achieved CR (Lugano classification Deauville score, 1-3) but subsequently experienced recurrence, defined by initiation of second-line therapy (excluding maintenance or consolidation immediately after 1L), positive confirmatory biopsy results, or imaging (eg, Lugano classification Deauville score, 4-5), whereas patients with primary refractory disease never achieved CR and had a persistently positive EOT imaging or biopsy results. Positive positron emission tomography (PET) imaging was not considered a relapse event if a confirmatory biopsy of the [¹⁸F] fluorodeoxyglucose (FDG)–avid lesion was negative. EFS, the primary outcome, was defined as the time from initiation of 1L therapy without progression, relapse, death, or salvage treatment. EFS was censored at the last date of follow-up if the patient was alive and had no events. Lead time was defined as the time interval between a positive ctDNA test result, indicating molecular recurrence, and confirmation of relapse by radiographic imaging or biopsy.

Statistical significance was assessed using the χ² test and the Fisher exact test to compare categorical variables. Survival analyses were performed using the Kaplan-Meier estimator and Cox methods (R version 4.2.2 using the survminer, survival, and coxph packages). Cox models were adjusted for age, sex, stage, and histology (except for histology-specific survival analyses) to account for potential confounding factors. For certain subgroup analyses with a smaller sample size and a low number of observed events, the Firth penalized likelihood Cox regression (coxphf package) was used. To minimize immortal time bias, the survival analysis included only patients who remained alive and event free through completion of 1L therapy, thereby excluding those who could not complete treatment or who experienced an event before the relevant ctDNA time point. All P values were based on 2-sided testing, and differences were considered significant if P ≤ .05. Concordance between ctDNA-MRD testing and imaging events (positive: progressive disease, partial response, stable disease, or Lugano 4-5; negative: complete respons [CR], no evidence of disease [NED], or Lugano 1-3) was evaluated using the Cohen κ statistic to evaluate interrater agreement.

Results

Patient cohort

This study included 144 patients with lymphoma with 1142 plasma time points (aggressive, n = 123; indolent, n = 21; Table 1), with a median age of 61 years (range, 18-84) and a median clinical follow-up of 14 months (range, 1-32). Overview plots indicating timing of ctDNA testing, treatments, imaging, outcomes, and other clinical management for individual patients are presented in supplemental Figure 1. Patients underwent a median of 7 ctDNA tests (range, 1-32) per patient, with a median cadence of 3 months (range, 0-17.4); patients with aggressive lymphoma had more testing time points (median, 8 [range, 1-32]) than those with indolent lymphoma (median, 4 [range, 1-14]; Table 1). ctDNA test results were available for 52.8% (65/123) of the patients with aggressive lymphoma and 23.8% (5/21) of the patients with indolent lymphoma before initiation of 1L therapy. Overall baseline ctDNA positivity was 92% (68/74) and was high in both aggressive [94% (62/66)] and indolent cases [75% (6/8)] (Figure 1B). Baseline ctDNA levels were significantly higher in aggressive lymphoma than in indolent lymphoma (median, 123 MTM/mL [range, 0-5958] vs median 22 MTM/mL [range, 0-59], respectively; P = .045). B-cell lymphomas generally exhibited higher baseline ctDNA MTM per milliliter than T-cell lymphomas.

Association of ctDNA status at EOT with patient outcomes

ctDNA testing time points were available at EOT (21 days before to 3 months after completion of 1L [n = 63] or 1L with maintenance/consolidation [n = 5; supplemental Table 1]) for 68 patients. Of these, 29.4% (20/68) were ctDNA-MRD positive and demonstrated a significantly inferior EFS compared with patients who were ctDNA-MRD negative (adjusted HR [adjHR], 28.03; 95% confidence interval [CI], 8.16-96.29; P < .0001; Figure 2A). Similar trends were observed in a subset analysis that consisted of patients with aggressive lymphoma and DLBCL (Figure 2B-C). EOT ctDNA-MRD positivity rate in these subsets remained consistent (aggressive lymphoma, 29.5% [18/61]; DLBCL, 26.3% [10/38]). In both subsets, 100% of the patients who were EOT ctDNA-MRD positive experienced recurrence (aggressive lymphoma, 18/18; DLBCL, 10/10) and demonstrated a significantly inferior EFS (aggressive lymphoma: adjHR, 25.29; 95% CI, 7.35-86.99; P < .0001; Figure 2B; DLBCL: adjHR, 16.58; 95% CI, 3.64-75.48; P = .0003; Figure 2C). Notably, among patients with DLBCL who were ctDNA-MRD negative at EOT (n = 5) and relapsed, all turned ctDNA-MRD positive during follow-up before or proximal to the time of relapse, with a median time to MRD detection of 4.24 months (range, 3.84-13.14) from the EOT MRD-negative result.

Figure 2.

EOT ctDNA-MRD status is prognostic for EFS. Kaplan-Meier estimates of EFS stratified by EOT ctDNA-MRD status in all patients with lymphoma (A), the subset with aggressive lymphoma (B), and the subset with DLBCL (C). AdjHRs were determined using Cox methods. AdjHR, adjusted Hazard Ratio; NR, not reached.

Among patients with indolent lymphoma who completed 1L therapy (with or without maintenance immediately after 1L) and who had EOT ctDNA-MRD results available (n = 8), 37.5% (3/8) were ctDNA-MRD positive, and 62.5% (5/8) were ctDNA-MRD negative. Both patients who were ctDNA-MRD positive experienced relapse during follow-up, whereas none of the patients who were ctDNA-MRD negative experienced relapse at the time of analysis. Among patients with indolent lymphoma, we then evaluated time to next treatment or last follow-up, defined as the interval from EOT (1L ± maintenance) to initiation of subsequent treatment; patients without additional treatment were censored at last follow-up. Median interval from the negative ctDNA-MRD test result was 17.5 months (range, 9.89-24.1) vs 1.22 months (range, 0.7-7.0) for patients with a ctDNA-MRD positive result (log-rank P = .002).

Prognostic value of EOT ctDNA-MRD status compared to PET/CT imaging

Among the 59 patients in the comparator cohort who had both EOT ctDNA-MRD and EOT PET or computed tomography (CT) results, EOT ctDNA-MRD positivity (representing ctDNA testing at a single time point after 1L therapy or 1L + maintenance or consolidation) conferred a significantly higher risk of inferior EFS than with ctDNA-MRD negativity (adjHR, 45.33; 95% CI, 10.19-201.65; P < .0001), a stronger association than that observed with EOT PET/CT positivity vs negativity (adjHR, 9.98; 95% CI, 2.75-36.15; P = .0005; Figure 3A-B). EOT ctDNA-MRD status demonstrated superior positive predictive value (PPV) for EFS compared to EOT PET/CT status (PPV, 100% vs 62%, respectively; P < .0001) and comparable negative predictive value (91% vs 89%, respectively; P > .05). Furthermore, EOT ctDNA-MRD positivity conferred a significantly higher risk of inferior EFS in both the EOT PET/CT-negative (adjHR, 48.45; 95% CI, 5.89-850.34; P < .001) and EOT PET/CT-positive (adjHR, 20.36; 95% CI, 3.3-125.46; P = .0012) subgroups (Figure 3C-D). In multivariate analysis, EOT ctDNA-MRD positivity remained the strongest independent predictor of inferior EFS (HR, 80.83; 95% CI, 12.07-541.3; P < .001), with EOT PET/CT positivity demonstrating a more modest effect (HR, 5.18; 95% CI, 1.26-21.4; P < .05; Figure 3E).

Figure 3.

Prognostic value of EOT ctDNA-MRD status compared to PET/CT imaging. (A-B) Kaplan-Meier estimates of EFS in all patients with lymphoma stratified by (A) PET/CT status at EOT and (B) ctDNA-MRD status at EOT. (C-D) Kaplan-Meier estimates of EFS in all patients with lymphoma who were PET CT negative, stratified by ctDNA-MRD status at EOT (C), and PET/CT positive, stratified by ctDNA-MRD status at EOT (D). (E) Multivariate analysis of prognostic factors and their association with EFS analyzed across patients with all lymphoma subtypes who had both PET/CT and ctDNA-MRD assessments performed at EOT. (F) Prognostic value of ctDNA-MRD status as a time-varying covariate during postdefinitive treatment surveillance. The Cox proportional hazards model was adjusted for subtype, stage, age, and sex and accounted for changes in ctDNA result status over time (positive ↔ negative). AdjHR, adjusted Hazard Ratio; NR, not reached. ∗, P < .05; ∗∗∗, P < .001.

ctDNA assessment during post-definitive treatment surveillance enables early detection of molecular relapse before imaging- or biopsy-confirmed recurrence. Among the 6 patients who achieved a CR after 1L (± consolidation) therapy but later experienced relapse, all had ctDNA-MRD–positive test result before confirmed relapse, with a median lead interval of 5.3 months (range, 0.7-7.1). When ctDNA-MRD detection is used to indicate a clinical relapse event, relapse is detected earlier than when using imaging/biopsy alone (all patients: median EFS, 6.31 months for MRD + clinical event vs 14.59 months for clinical event alone; supplemental Figure 2). For 61 patients, including 48 with EOT ctDNA testing and EOT PET/CT imaging, surveillance ctDNA testing was available (median, 9 months [range, 0.329-34.6]), with 251 ctDNA time points (median, 3 [range, 1-10]). Of these, 32.8% (23/61) had detectable ctDNA at any time during surveillance, and 62.3% (38/61) remained serially ctDNA-MRD negative. In a time-varying covariate analysis, detection of ctDNA during post-treatment surveillance was associated with a significantly higher risk of experiencing an EFS event (adjHR, 33.74; 95% CI, 9.34-121.82; P < .0001; PPV, 87%; negative predictive value, 95%; sensitivity, 91%; specificity, 92%; Figure 3F).

Association of ctDNA clearance on 1L therapy with patient outcomes

Dynamic changes in ctDNA during treatment, especially ctDNA clearance (converting from ctDNA positive to ctDNA negative), can indicate treatment response, whereas failure to clear ctDNA can indicate treatment resistance. We evaluated associations between ctDNA clearance during 1L therapy and EFS in a subset of 60 patients who were ctDNA-MRD positive at baseline. ctDNA clearance during 1L therapy was associated with significantly improved EFS. Of the 60 patients who were ctDNA-MRD positive at baseline with serial ctDNA testing, 80% (48/60) achieved clearance during 1L therapy and had significantly improved EFS (adjHR, 8.57; 95% CI, 2.55-28.81; P = .005; Figure 4A). Considering early ctDNA clearance (cycle 1; inclusive of testing between cycle 1 day 1 and cycle 1 day 22; n = 14) and later ctDNA clearance (ctDNA-MRD positive at baseline and during cycle 1 and ctDNA-MRD negative in cycle 2 and beyond; n = 21), early clearance was associated with the most favorable prognosis compared to no ctDNA clearance (early: adjHR, 20.95; 95% CI, 2.09-28.81; P = .0097; later: adjHR, 11.74; 95% CI, 2.61-52.86; P = .0013; Figure 4B), although there was no statistically significant difference between early and later clearance.

Figure 4.

Prognostic implications of ctDNA clearance on EFS. Kaplan-Meier estimates of EFS stratified by ctDNA clearance during 1L therapy (A) and early clearance (C1), ctDNA clearance after cycle 1 (later clearance), and no clearance (B). AdjHRs were determined using Cox methods. Adj. HR, adjusted Hazard Ratio; C1, cycle 1; NR, not reached.

Similarly, significant trends were observed in subsets of patients with aggressive lymphoma (n = 44) and patients with DLBCL (n = 30), with ctDNA clearance demonstrating significantly improved outcomes compared to no clearance (aggressive lymphoma: adjHR, 10.87; 95% CI, 2.64-44.78; P = .0010; DLBCL: adjHR, 5.35; 95% CI, 1.19-24.18; P = .0292; supplemental Figure 3A-B). Early clearances consistently resulted in most favorable outcomes within these subsets (aggressive lymphoma, early: adjHR, 23.2; 95% CI, 2.75-195.53; P = .0038; aggressive lymphoma, late: adjHR, 9.74; 95% CI, 2.85-33.22; P = .0003; DLBCL, early: adjHR, 19.21; 95% CI, 1.53-240.73; P = .022; DLBCL, late: adjHR, 5.64; 95% CI, 1.04-30.45; P = .0444; supplemental Figure 4C-D), although the difference between early and late clearance was not significant.

Concordance of ctDNA-MRD status with imaging assessments

We evaluated the agreement or discordance of ctDNA-MRD testing and PET/CT imaging assessment across the entire patient journey. Of the 252 instances where imaging and ctDNA-MRD assessments were performed within 30 days of each other (without treatment changes), 79% (199/252) showed statistically significant concordance, with an overall moderate agreement (κ = .58; P < .0001; Figure 5A). We identified 53 discordant instances (21%) in which PET/CT and ctDNA test results were not consistent.

Figure 5.

Concordance of ctDNA-MRD status with imaging assessments. (A) Frequency of concordant (imaging positive/MRD positive and imaging negative/MRD negative) and discordant (imaging negative/MRD positive and imaging positive/MRD negative) instances of imaging and proximate ctDNA testing. (B) ctDNA corroboration of outcomes for imaging-positive/MRD-negative discordant instances as adjudicated by a clinician within the immediate treatment and follow-up setting. (C) ctDNA corroboration of outcomes for imaging-negative/MRD-positive discordant instances as adjudicated by the clinician within the immediate treatment and follow-up setting. Two instances did not have clinical adjudication available. Nota bene: A single patient may have had multiple discordant events during the clinical history of the disease, as shown in cases where there are more discordant instances indicated than the number of patients.

PET positive/ctDNA-MRD negative (46 instances among 36 patients)

Within the immediate treatment and follow-up setting, for 46 of the 46 discrepancies, ctDNA corroborated the adjudicated clinical outcome for 87% (40/46) of these discordant instances. Of these 40 instances, 31 were adjudicated as CR/NED (10 confirmed to be negative by biopsy, 12 confirmed to be negative by follow-up serial imaging, and 9 followed by no additional therapy). Nine discordant instances (imaging positive or ctDNA-MRD negative), representing 6 patients, eventually became ctDNA-MRD positive (median, 3.96 months [range, 0.95-9.49]), consistent with relapse/refractory disease on PET. Importantly, this demonstrates that among discordances, most were attributable to false-positive PET findings rather than missed early relapses by ctDNA testing. In the remaining 6 instances (13%), ctDNA-MRD testing did not corroborate the adjudicated relapse or refractory outcome (Figure 5B). Of note, cases adjudicated as “relapse/refractory-additional therapy given” represent cases adjudicated as relapse because of treatment initiation not because of radiographic or biopsy confirmation of relapse; the case could be made that additional therapy might not have been necessary, given the negative ctDNA results. One case was confirmed as a bona fide relapse by subsequent imaging; this represents the single discordant instance in which positive PET/CT results were consistent with the adjudicated outcome, and the ctDNA result was a false negative.

PET negative/ctDNA-MRD positive (7 instances)

Within the immediate treatment and follow-up setting, 5 discordances were adjudicated as relapse or refractory disease (2 instances were unavailable for adjudication) by disease-related death (n = 1), follow-up imaging (n = 2), and additional therapy given (n = 2), corroborating the ctDNA-MRD–positive results in 100% (5/5) of these adjudicated outcomes (Figure 5C).

Overall, ctDNA-MRD results were consistently corroborated by clinically adjudicated outcomes in 85% (45/53) of divergent instances with imaging, correctly identifying the presence or absence of disease and yielding a high overall agreement of 96.8% (244/252).

Association of ctDNA dynamics in response to CAR-T therapy

In this cohort, 16 patients with relapsed/refractory disease received CAR-T therapy after 1L. Of these, 4 patients had ctDNA data missing either before or after CAR-T treatment. Among the remaining 12 patients, 3 were ctDNA-MRD negative before initiating CAR-T and remained relapse-free with continued ctDNA-MRD negativity after treatment. Of the 9 patients who were ctDNA-MRD positive before CAR-T, 6 of the 9 patients (66.7%) achieved ctDNA clearance after therapy. Among these 6, a total of 4 remained relapse free, 1 remained relapse free but died from a non–cancer-related cause 2 years after CAR-T, and 1 (patient 122) achieved immediate CR but became ctDNA-MRD positive 6.8 months after CAR-T and clinically relapsed 13 months after CAR-T (6.2 months after first positive ctDNA-MRD test after CAR-T; see case example 7, supplemental Figure 5). Patients who remained ctDNA-MRD positive after CAR-T generally had worse EFS than those who achieved and maintained ctDNA clearance (Figure 6A). Persistent ctDNA-MRD positivity after CAR-T therapy was associated with a significantly shorter duration post-treatment interval (median, 1.97 months) than sustained clearance (median, 16.1 months; P = .0091; Figure 6B).

Figure 6.

ctDNA dynamics are predictive of CAR-T response and prognostic of outcomes at 12 months. (A) Most patients who achieve ctDNA clearance on CAR-T therapy have favorable long-term clinical outcomes, whereas all patients who did not achieve ctDNA clearance on CAR-T therapy experienced either disease relapse or disease-related death. (B) ctDNA clearance on CAR-T is associated with a longer event-free interval after CAR-T. The dotted line indicates the post-CAR-T MRD testing time point.

Discussion

Various novel biomarkers have had limited impact on improving outcomes in lymphoma despite advancements in genomic and molecular testing,21, 22, 23 highlighting the challenge in accurately identifying treatment failure and residual disease after 1L induction therapy.

We show that tumor-informed ctDNA-MRD testing is a strong prognostic and predictive tool that can complement and refine current risk stratification methods. In our cohort, 92% of lymphomas were ctDNA-MRD positive at baseline, with both pre- and on-treatment ctDNA levels correlating with disease burden and treatment response. Specifically, ctDNA status at EOT was highly prognostic of EFS (adjHR, 28.03; P < .0001), and ctDNA clearance during 1L therapy predicted durable response to therapy (adjHR, 8.57; P = .005). Although early clearance trended toward longer EFS compared to later clearance, the differences were not statistically significant due to small sample size. Thus, our findings are hypothesis generating and warrant further evaluation of the impact of treatment deescalation strategies in a subset of patients with low tumor burden who may have exceptionally favorable responses. Moreover, we show that EOT ctDNA-MRD status is a superior prognostic indicator of EFS as compared with EOT PET/CT (adjHR, 45.33 vs 9.98). In multivariate analysis, ctDNA positivity remained the strongest independent risk factor for inferior EFS (HR, 80.83; P < .001), offering prognostic clarification in both PET/CT-positive and PET/CT-negative groups. Finally, changes in ctDNA levels were predictive of response to CAR-T therapy, with clearance forecasting durable responses and improved EFS in patients with relapsed/refractory disease.

Our results are substantiated by other studies on ctDNA-MRD testing in various lymphoma subtypes. In relapsed/refractory indolent non-Hodgkin lymphoma, ctDNA clearance or reduction was associated with a positive response to magrolimab with R-GemOx.¹⁶ Similarly, in relapsed/refractory DLCBL, ctDNA clearance or sustained reduction was associated with favorable outcomes when treated with magrolimab with rituximab/gemcitabine/oxaliplatin, whereas patients with progressive disease consistently remained ctDNA positive.¹⁵ In the pretreatment setting, ctDNA was a reliable indicator of tumor burden in DLBCL, demonstrating strong prognostic value for relapsed/refractory disease with a sensitivity of 88.8% and a specificity of 94.1%.¹⁷ Furthermore, ctDNA clearance during 1L therapy predicted improved treatment responses and better outcomes (EFS: HR, 6.5; P = .003; overall survival: HR, 22; P = .005). Patients who achieved ctDNA clearance during treatment, regardless of R-IPI score, had superior outcomes compared to those who remained ctDNA positive.¹⁷ Finally, ctDNA-MRD negativity at EOT was a strong predictor of superior progression-free survival in DLBCL²⁴ and in advanced-stage follicular lymphoma.²⁵

PET/CT remains the standard-of-care assessment for defining EOT remission in lymphoma, but it lacks disease specificity, cannot detect residual disease at the molecular level, and is limited by both false positives and false negatives.²⁶ Notably, up to 20% of the patients with negative EOT PET/CT scans experience relapse, whereas many PET-positive findings do not correspond to true recurrence, leading to potential overtreatment, unnecessary procedures and financial burden. Although repeated PET/CT imaging may help distinguish durable responders, it carries the drawback of additional radiation exposure, scheduling delays, and various payor-dependent restrictions (number of scans per year) outside the treatment period. Tumor-informed ctDNA-MRD assessments can offer enhanced sensitivity and specificity compared to traditional imaging. A negative ctDNA test result alongside a positive PET scan result can help identify false-positive imaging results, potentially allowing patients to avoid further treatment or consider deescalation. In contrast, a positive ctDNA test with a negative PET scan result might indicate false-negative imaging and could help stratify patients according to risk of treatment failure and guide individualized therapy selection. The phase 2 ALPHA-3 study (ClinicalTrials.gov identifier: NCT06500273) exemplifies the emerging clinical utility of MRD by using ctDNA status to guide therapy selection, randomizing patients with DLBCL who achieve complete metabolic response by PET but remain MRD positive to CAR-T consolidation vs observation.²⁷ This approach supports MRD as a decisive biomarker for therapeutic escalation in lymphoma. In our study, initial assessment demonstrated an overall concordance of 79% between ctDNA-MRD detection and clinical imaging. In 21% of the instances in which initial assessments diverged, our ctDNA-MRD results consistently corroborated the clinically adjudicated outcome in the vast majority of the cases (45/53 [85%]), thereby correctly identifying relapsed/recurrent events when the ctDNA test result remains/becomes positive and correctly identifying the absence of disease when the ctDNA test result remains negative, yielding an overall agreement of 96.8%. This further demonstrates the ability of tumor-informed ctDNA-MRD testing to provide actionable information when imaging is uncertain or ambiguous. Furthermore, ctDNA-MRD trajectories in our data set reflect disease status and treatment response; in the 6 cases in which patients achieved complete remission (radiographic and molecular) on 1L therapy and eventually experienced relapse, surveillance ctDNA-MRD testing detected molecular relapse before radiographic relapse, with a median lead interval of 5.3 months (range, 0.7-7.1).

Our study was conducted through prospective analyses of a heterogeneous patient population with various disease histologies receiving a variety of therapies in real-world clinical practice. ctDNA results were derived from samples analyzed by commercial ctDNA testing and were collected at the physician’s discretion rather than according to a standardized timing schedule; these variables limit the scope of our analysis. A small percentage of patients (n = 4) could not be evaluated because of insufficient variants for MRD monitoring. This limitation was observed in 4 DLBCL cases with favorable disease biology in which no tumor DNA shedding occurred at any point during the disease course. All 4 of those patients achieved CR. This study was also limited by heterogeneity within the aggressive and indolent groupings, which were defined by clinical management rather than shared biology or curability outcomes. A median follow-up of 14 months may be adequate for some aggressive histologies but is likely insufficient for many indolent subtypes with prolonged disease courses. In addition, most patients in the aggressive cohort had DLBCL, in which the prognostic values of PET and ctDNA testing are best characterized, limiting generalizability to less common histologies with lower or variable [¹⁸F] fluorodeoxyglucose avidity. The prognostic performance of ctDNA during treatment and surveillance may therefore vary by subtype and may require longer follow-up and histology-specific analyses. Nevertheless, the tumor-informed ctDNA approach evaluated here is histology agnostic and broadly applicable across lymphoma subtypes, in contrast to other assays with a more limited scope.

In summary, tumor-informed ctDNA-MRD testing provides crucial prognostic and predictive insights in the management of lymphoma. Our study captured real-world clinical experiences and provides a generalized view of the utility of ctDNA assessment in a histology-agnostic, therapy-independent fashion, demonstrating that ctDNA-MRD testing can be applied across various treatment strategies in the evolving therapeutic landscape of B-cell and T-cell lymphoma. Incorporating ctDNA testing early in a patient’s care plan can improve risk stratification through minimally invasive monitoring of treatment response and identify patients who are at high vs low risk of relapse after 1L therapy. Importantly, we show that the benefits of ctDNA-MRD analysis in lymphoma extend beyond baseline risk assessment and 1L treatment. In the surveillance setting for lymphoma, ctDNA-MRD assessment can identify relapse and clarify ambiguous imaging results, and in relapsed/refractory disease, on-treatment ctDNA dynamics can predict response to CAR-T therapy. Taken together, these findings highlight tumor-informed ctDNA-MRD testing as a transformative tool that complements and may even surpass conventional imaging by enabling earlier and more precise detection of residual disease, thereby further empowering personalized, biologically-informed management of lymphoma patients.

Conflict-of-interest disclosure: N.G. reports receiving honoraria for lectures, presentations, speaker’s bureaus, manuscript writing, or educational events from Clinical Care Solutions, LLC. M.I. reports consulting fees from Sanofi US, Bristol Myers Squibb, and ADC Therapeutics and advisory board participation for Sanofi US, Bristol Myers Squibb, and ADC Therapeutics. S.T. reports research support from Rigel, Merck, Novartis, Seagen, Sanofi, Pfizer, Principia, and Kartos; consulting fees from Novartis; and stock or stock options in Procter & Gamble and Eli Lilly. D.N. is an employee of Natera, Inc, receives a salary, and may own stock or stock options. A.T. is an employee of Natera, Inc, receives a salary, and may own stock or stock options. H.Y. is an employee of Natera, Inc, receives a salary, and may own stock or stock options. P.M. is an employee of Natera, Inc, receives a salary, and may own stock or stock options. M.M. is an employee of Natera, Inc, receives a salary, and may own stock or stock options. H.W. is an employee of Natera, Inc, receives a salary, and may own stock or stock options. A.R. is an employee of Natera, Inc, receives a salary, and may own stock or stock options. M.C.L. is an employee of Natera, Inc, receives a salary, and may own stock or stock options. M.N. reports grants or contracts from EUSA/Recordati Rare Diseases and ADC Therapeutics; consulting fees from AbbVie, Genentech-Roche, and Kite; honoraria for lectures, presentations, speaker’s bureaus, manuscript writing, or educational events from BeOne Medicines, AbbVie, Lilly, and Genentech-Roche; and stock or stock options in Syndax Pharmaceuticals and Pfizer. B.M.W. reports consulting fees from Guidepoint Global; honoraria for lectures, presentations, or speakers bureaus from Kyowa Kirin; advisory board participation for Incyte, PPD/Orca, ADC Therapeutics, Genmab, Bristol Myers Squibb, Pfizer, Tempus, and AstraZeneca; and unpaid leadership or fiduciary roles on multiple committees for the Foundation for the Accreditation of Cellular Therapy. C.P.G. declares no competing financial interests.