Peripheral blood cfDNA detection using clonoSEQ assay in multiple myeloma is concordant with bone marrow and clinical biomarkers

Dear Editor,

Minimal residual disease (MRD) assessment is an important component of response evaluation and risk stratification in multiple myeloma (MM) [1, 2]. The need for frequent bone marrow (BM) biopsy specimens has limitations, including procedural burden. Peripheral blood (PB) based approaches provide a less invasive and feasible approach for disease monitoring, including cases of heterogenous bone marrow involvement or extramedullary disease (EMD).

Several methodologies have been studied for PB MRD detection in MM, including mass spectrometry (MS), cell free DNA (cfDNA) whole genome sequencing, and flow cytometry based approaches [3, 4]. Each has distinct advantages and limitations. For example, MS based assessments are limited by the long half-life of the monoclonal protein [3] and flow cytometry may require large blood volumes [4]. clonoSEQ® is a next-generation sequencing (NGS) assay that measures MRD by detecting unique rearranged B-cell receptor (BCR) sequences from DNA in BM. It is FDA-approved for use on BM samples in MM and has demonstrated prognostic value when MRD negativity is achieved [2, 5]. Its use in PB could offer a patient-friendly alternative and allows for more frequent monitoring [3].

To understand whether the clonoSEQ assay from the PB can be used to track disease burden and MRD status in the BM, we correlated BCR sequences measured in the PB, specifically the plasma component, in patients with MM and paired BM and PB samples at all stages of disease (newly diagnosed, relapse, etc).

We conducted a retrospective analysis of 149 PB plasma samples from 100 patients with MM between 2016 and 2024. The mean plasma amount for each sample was 2.8 mL. All patients had previously identified dominant clonotypes using clonoSEQ in the BM. Testing was performed on PB plasma using the clonoSEQ cell free DNA (cfDNA) assay using plasma within one week of the corresponding BM sample. There were eighteen corresponding BM samples that did not have MRD testing performed. Patients were categorized by response status at the time of sample collection per International Myeloma Working Group (IMWG) criteria. Positive M-proteins by mass spectrometry (MS) or SPEP due to the presence of daratumumab were considered negative if patients were receiving daratumumab at sample collection date or within the previous 3 months.

Diagnostic ‘ID’ samples from the BM were identified using the clonoSEQ Assay. gDNA was amplified using locus-specific, two-step, multiplex PCR with a master mix of primers targeting V, D, and J genes of the IgH, IgK, IgL, BCL1/IgH and BCL2/IgH loci, followed by NGS. Sequencing data were processed using a custom bioinformatics pipeline. Sequences are considered acceptable for tracking if it comprises at least 3% of all BCR sequences at a given locus, represents ≥0.2% of all nucleated cells in the sample, there are ≥40 copies of that sequence present, and the sequence has a discontinuous distribution (i.e., it is well separated from the background with <5 other sequences present in the next lower decile of frequency). Once suitable disease-associated sequence(s) are identified, these ID/clonality sequences are compared with those found in successive MRD sample(s) for tracking. cfDNA is extracted from up to 10 mL of plasma using the QiaSymphony. The abundance of each of the disease associated sequences in a sample is measured and reported as a sample-level malignant cell count within the total of all nucleated cells assessed, or for cfDNA, per mL plasma. We ensure that at least 100 housekeeping genes are assessed as a means of ruling out samples in which the cfDNA may have been heavily degraded.

Performance metrics (sensitivity, specificity, PPV, NPV) were calculated for PB cfDNA using BM MRD, M-protein, and free light chain (FLC) ratio as reference standards. Associations between MRD status and clinical variables were assessed using Chi-Square or Wilcoxon rank sum tests. Longitudinal PB cfDNA dynamics were evaluated in patients with ≥3 serial samples. Generalized estimating equations (GEE) with an exchangeable correlation structure, were used to account for multiple samples for some subjects. SAS version 9.4 was used for statistical analysis and p values < 0.05 were considered significant.

Samples from 100 patients with MM were included in the analysis (see Supplemental Table 1 for patient demographics). Other sample characteristics are summarized in Supplemental Table 2. Among 131 paired samples with PB cfDNA and BM MRD assessments (using BM MRD as the reference standard), PB cfDNA demonstrated a specificity of 90.1% (95% CI: 78.9–100) and a positive predictive value (PPV) of 94.9% (95% CI: 88.0–100), indicating strong concordance to BM MRD positivity when PB is positive (Fig. 1A). However, sensitivity was limited at 33.9% (95% CI: 25.1–42.8), and the negative predictive value (NPV) was 21.7% (95% CI: 13.3–30.2), suggesting that a negative PB result does not reliably exclude disease. These results were consistent when using a GEE model (Supplementary Table 3).

Fig. 1.

Performance metrics of PB cfDNA. Compared to BM MRD + status, monoclonal protein (M-protein) detection and abnormal FLC ratio (A). PB cfDNA samples and clinical status (B). Patients with three or more PB cfDNA samples and clinical status category. Nine patients had three or more sequential samples. PB cfDNA is displayed as the logarithmic value on the y-axis. Clinical status is displayed per time (months, x-axis) from first sample with the relevant clinical status denoted by the legend color (C).

When comparing PB cfDNA against the presence of an M-protein, sensitivity was 49.4% (95% CI: 38.3–60.4) and specificity was strong at 83.3% (95% CI: 74.3–92.3) (Fig. 1A). PPV was 78.0% (95% CI: 66.5–89.5) and NPV of 57.9% (95% CI: 48.0–67.8) compared to M-protein presence. When this analysis was performed on a detectable M-protein by MS only, there was a modest improvement in specificity and NPV (Fig. 1A). When comparing PB cfDNA positive samples to an abnormal FLC ratio, sensitivity was 56.9% (95% CI: 44.9–69.0) and specificity was strong at 82.2% (95% CI: 73.4–91.0). PPV was 74.0% (95% CI: 61.8–86.2) and NPV was 68.2% (95% CI: 58.5–77.9). Results using GEE models were similar for both M-protein and FLC ratio. PB cfDNA was also compared to the combination of M-spike presence and an abnormal FLC ratio. Sensitivity was 63.6% (95% CI: 49.4–77.9) and specificity was 78.9% (95% CI 70.5–87.3). PPV was 59.6% (95% CI 45.5–73.6) and NPV 81.6% (95% CI 73.5–89.8).

PB cfDNA status was strongly associated with traditional markers of tumor burden, including a detectable M-protein (p < 0.001) and an abnormal FLC ratio (p < 0.001) (Table 1). PB cfDNA negative status was significantly associated with deeper clinical responses. Among patients in complete response or stringent complete response (CR or sCR, respectively) 83 of 96 samples (86.5%) were PB cfDNA negative (Table 1, Fig. 1B). See Supplementary Table 4 regarding clinical status at time of sample collection for patients in CR/sCR. Among patients in a very good partial response (VGPR) or better, PB cfDNA was negative in 90.6% of samples, with a PPV of 93.6% (data not shown). Patients categorized as VGPR were MRD positive in the BM (three did not have MRD performed) and all but one patient in VGPR still had a detectable M-protein. Conversely, PB cfDNA positive samples were significantly enriched in samples taken from patients with active or residual disease; all 19 samples from patients with progression were PB cfDNA positive. Among samples taken from patients at diagnosis, 8 of 9 were PB cfDNA positive.

Table 1.

Associations with MRD cfDNA status.

	cfDNA result (PB)
	Negative (N = 96)	Positive (N = 53)	Total (N = 149)	P value
CR + , n (%)				<.001b
No	13 (13.5%)	43 (81.1%)	56 (37.6%)
Yes	83 (86.5%)	10 (18.9%)	93 (62.4%)
VGPR + , n (%)				<.001b
No	9 (9.4%)	35 (66.0%)	44 (29.5%)
Yes	87 (90.6%)	18 (34.0%)	105 (70.5%)
Bone marrow % plasma cells				<.001a
N (Missing)	91 (5)	39 (14)	130 (19)
Mean (SD)	0.7 (2.73)	7.8 (16.84)	2.8 (9.98)
Median	0.0	1.0	0.0
Range	0.0, 24.0	0.0, 90.0	0.0, 90.0
FLC Ratio Normal/Abnormal, n (%)				<.001b
Abnormal	28 (31.8%)	37 (74.0%)	65 (47.1%)
Normal	60 (68.2%)	13 (26.0%)	73 (52.9%)
Missing	8	3	11
M-spike OVERALL, n (%)				<.001b
Negative	55 (57.9%)	11 (22.0%)	66 (45.5%)
Positive	40 (42.1%)	39 (78.0%)	79 (54.5%)
Missing	1	3	4
M spike (positive) and abnormal Ratio, n (%)				<.001b
Negative	71 (81.6%)	19 (40.4%)	90 (67.2%)
Positive	16 (18.4%)	28 (59.6%)	44 (32.8%)
Missing	9	6	15

^aANOVA F-test p value.

^bChi-Square p value.

Longitudinal analysis of sequential samples revealed that all samples collected during diagnosis or clinical relapses were PB cfDNA positive (Fig. 1C). All samples collected during CR were PB cfDNA negative. MRD trajectories closely paralleled clinical response, with conversion from PB cfDNA positive to PB cfDNA negative corresponding to treatment response.

In this study, we demonstrate that PB cfDNA MRD testing using the clonoSEQ assay is feasible and may be a clinically informative approach for disease monitoring in patients with MM, but results are still preliminary. Our findings indicate that PB cfDNA status was associated with established markers of disease burden including BM tumor burden, presence of an M-protein, an abnormal FLC ratio, and EMD. PB cfDNA tracked closely with IMWG response status. Nearly all samples taken at the time of high tumor burden (diagnosis, relapse/progression, or in the presence of EMD) were PB cfDNA positive. Although conventional disease biomarkers are largely positive in these cases, PB cfDNA may be particularly important for patients with oligo- or non-secretory disease and/or patients with EMD with negative BMs. Indeed, immunoglobulin clonal rearrangements should remain detectable at the DNA level even in non-secretory disease making PB cfDNA particularly appealing for this clinical scenario [6]. Conversely, PB cfDNA negative was highly enriched in patients who had achieved deep responses.

Although the sensitivity and NPV of cfDNA detection were limited, this likely reflects the characteristics of our cohort, which was enriched for samples collected in deep response (CR/VGPR) and MRD negativity in the BM (82%), where minimal tumor burden reduces cfDNA shedding below assay detection thresholds. This contrasts with prior cfDNA studies in MM, where many samples were collected in the setting of >CR and MRD positivity in the BM, thereby likely increasing detection rates [7–10]. Enriching for patients in deeper responses is particularly important in the current treatment era, as newer combination regimens achieve higher rates of MRD negativity, making it critical to evaluate assay performance in this setting [1, 11, 12]. Importantly, the methodology underlying the clonoSEQ assay enables accurate estimation of the assay’s limit of detection (LOD) and supports standardized measurement of cfDNA using BCR sequencing. In contrast, mutation-based cfDNA approaches require independent LOD estimates for each mutation tested, and sensitivity is influenced by the number and selection of mutations included as well as the technical performance of the assays, creating variability across studies [9].

Our data suggests that PB cfDNA via clonoSEQ could complement BM-based MRD assessments, rather than replace them. In cases where BM sampling is not feasible or frequent monitoring is desired, PB cfDNA may provide a surrogate marker of disease burden and response. Limitations of this study include its retrospective design, small sample size, and the fact that in approximately 10% of patients, a dominant clone by clonoSEQ cannot be identified. In addition, samples were not collected at standardized timepoints across patients. Limitations will be addressed in the prospective study, which will include the cfDNA component from plasma and the cellular component (PBMCs), thus potentially increasing the sensitivity of the peripheral blood disease monitoring in MM.

Author contributions

JWN contributed to study conceptualization, oversaw and guided statistical analyses, interpreted data, and led manuscript drafting and revision. HK performed all statistical analyses. MA contributed to study conceptualization and clinical data collection. HS and AJ contributed to study conceptulization and performed peripheral blood sample processing and analyses. RF and SSB aided in clinical data collection. LB and GA contributed to sample collection and processing. UY, SC, LB assisted with manuscript preparation. RF contributed to study conceptualization.

Conflict of interest

Adaptive Biotechnologies processed all peripheral blood samples. RF Consulting: AbbVie, Adaptive, Amgen, Apple, BMS/Celgene, GSK, Janssen, Karyopharm, Pfizer, RA Capital, Regeneron, Sanofi; Scientific Advisory Board: Caris Life Sciences; Board of Directors: AntengenePatent for FISH in MM - ~$2000/year. HS reports stock ownership in Adaptive Biotechnologies. AJ is an employee and stockholder of Adaptive Biotechnologies. SC: Honorarium: Sanofi, Sobi, Ascentage Pharma, BMS, Pfizer, AstraZeneca, Legend Biotech; Institutional Research Funding: C4 Therapeutics, CARSgen, Cynata Therapeutics, Johnson & Johnson, Ascentage Pharma, AstraZeneca, Incyte, Abbvie, Simcere, Cullinan Therapeutics, National Marrow Donor Program; LB: Consulting: Oncopeptides, Salarius Pharmaceuticals, Radmetrix International, Omeros Corporation, CellCentric Limited, Abbvie, Pfizer, Cancer Network; Licensed Intellectual Property: Genetically Engineered Mouse Models of Multiple Myeloma (including IMiD-responsive models), Active Licensing agreements involving Novartis, Pfizer, Opna Bio, including royalty-bearing arrangements.

Supplementary information

The online version contains supplementary material available at 10.1038/s41408-026-01508-9.