Learn
How MRD tests differ
Liquid-biopsy CGP and MRD tests both analyze ctDNA, but they answer different questions, are used at different times, and are built to different sensitivity targets. A negative MRD result does not prove cancer is gone. A negative CGP liquid biopsy does not prove an actionable alteration is absent.
Liquid biopsy CGP versus an MRD test
81.6% of cfDNA mutations in people without cancer and 53.2% in people with cancer had features consistent with clonal hematopoiesis (Razavi Nat Med 2019). Razavi et al., Nat Med 2019.
Therapy-selection CGP
In advanced or metastatic cancer, identify genomic alterations that can help choose a targeted drug or immunotherapy.
Diagnosis of advanced disease or progression. ASCO 2026: may be used when tissue is challenging, delayed, or risky, or when a drug indication allows or requires ctDNA testing. Confirm negatives with tissue.
FoundationOne Liquid CDx (an FDA-approved plasma CGP assay, not an MRD test): 95% LoD 0.40% VAF for select substitutions/indels in the enhanced-sensitivity region (0.82% in the standard region) and 0.37% VAF for select rearrangements (Woodhouse 2020, PMID 32976510).
A valid actionable finding can support targeted therapy when the indication allows it. Does not prove the alteration is absent (low shedding, fusions/CNVs, quality). Reflex to tissue (ESMO 2022, ASCO 2026, IASLC 2021).
An MRD test
Detect residual or recurrent tumor DNA after curative-intent treatment, or during surveillance, at much lower ctDNA fractions than CGP.
Landmark after surgery, during/after adjuvant therapy, surveillance. Clinical validity is strong in many cancers (ESMO 2022). Clinical utility: DYNAMIC (Safe-SeqS research assay, stage II colon) and IMvigor011 (Signatera-guided atezolizumab in MIBC, NEJM 2025). ASCO 2026 guideline (search through Feb 2025, before IMvigor011 publication) is cautious and anticipates future MRD-specific updates.
ESMO 2022: advanced-cancer genotyping assays are not sensitive enough for MRD; landmark clinical sensitivity often about 50% or less; assays able to detect below 0.01% VAF remain a research goal. Commercial MRD analytic LODs in this draft range from about 0.01% VAF to a few PPM, and are not interchangeable.
10 ng cfDNA is on the order of 3,000 haploid genome equivalents (Wan 2017). At 0.01% tumor fraction, expected mutant molecules at a single site can be below one.
High risk of recurrence; clinical specificity often 90% or higher if no further treatment is given (ESMO 2022). Not itself a radiographic diagnosis.
Lower risk, but does not rule out residual disease or future recurrence. GALAXY: 9.5% (81/852) of 4-week ctDNA-negative patients still recurred (Kotani 2023).
- DYNAMIC. Safe-SeqS research assay, not a commercial kit. Stage II colon: adjuvant chemo 15% vs 28%; 2-year RFS 93.5% vs 92.4%, noninferior.
- IMvigor011. MIBC, serial tumor-informed ctDNA. DFS 9.9 vs 4.8 months (HR 0.64); OS 32.8 vs 21.1 months (HR 0.59). Persistently negative (n=357): DFS 95% at 1 year, 88% at 2 years.
Tumor-informed versus tumor-naive
Tumor-informed
Sequence the tumor (and usually matched normal), identify tumor-specific mutations, build a personalized plasma test.
Deep sequencing of known sites; CHIP/germline filtering. In a 2026 CRC meta-analysis, higher serial sensitivity than tumor-agnostic assays (0.88 vs 0.59, p=0.001), no significant difference in false-positive rates; landmark differences not significant (Camblor PMID 41401645). Pooled, not a named-test head-to-head.
Needs adequate tissue. Invitae PCM: panels for 100% of 63 samples with >=20% tumor content but only 80% of 30 with >=10% (PMID 37632661). Longer time to first result. May miss later-emerging mutations.
Tumor-naive
Fixed plasma test: shared mutation panel and/or methylation and fragmentomics. No tumor map required.
No tissue; potentially faster first result; methylation can add detection (Reveal evaluates >20,000 epigenomic regions, PMID 39110016).
Fixed genomic panels may miss this patient's mutations; CHIP risk unless filtered by methylation, fragment length, or WBC sequencing. Lower pooled serial sensitivity in the Camblor CRC meta-analysis.
What methylation means
Methylation is a small chemical tag cells attach to DNA, usually at CpG sites, to help turn genes on or off. It does not change the genetic letters. Cancer cells often have a different tagging pattern. A blood test can look for that pattern without knowing the patient's personal mutations and without a tumor sample. Methylation is less likely to be confused with CHIP, which is a sequence-mutation problem in blood cells.
How the technologies work
Example tests are listed so the chart can link here. Listing a test is not an endorsement, and it is not a ranking.
A3.1
Tumor-informed multiplex PCR
Tumor and matched-normal DNA are sequenced (WES for Signatera exome). Clonal somatic variants are selected and custom PCR primers are built for those sites. Subsequent plasma is amplified at those sites and sequenced at very high depth. A positive call often requires more than one tracked variant.
Strengths. Deep sequencing of known patient-specific sites; CHIP filtered by matched normal; large clinical literature for Signatera, including IMvigor011 utility data.
Limits. Requires tissue and a design step. Relatively few sites (16 for published Signatera exome). Formal peer-reviewed Signatera analytic validation paper not identified (0.01% VAF from AACR 2018 abstract and vendor materials).
Reinert JAMA Oncol 2019Coombes CCR 2019Powles NEJM 2025 IMvigor011Invitae PCM analytic 2023
A3.2
Tumor-informed hybrid-capture / WGS-informed panels
Tumor and matched-normal WGS (or WES) identifies a large somatic-variant set. A custom hybridization-capture panel pulls those fragments from plasma cfDNA for sequencing. Panels often track hundreds to about 1,800 variants.
Strengths. More independent sites improve the chance of seeing tumor fragments at very low tumor fraction. NeXT Personal has a peer-reviewed analytic validation (detection threshold 1.67 PPM, LOD95 3.45 PPM).
Limits. Requires tissue and a longer design step. LOD figures are assay-specific and not a ranking.
A3.2
Tumor-informed personalized targeted panel (chemistry not forced)
Tumor sequencing (often WES) selects a personalized set of variants (up to 48 for RaDaR) that are then tracked in plasma. RaDaR's plasma chemistry was not clearly classified as hybrid capture versus multiplex PCR in a full peer-reviewed methods paper found in this review.
Strengths. Personalized tumor-informed tracking with a moderate number of sites; substantial clinical literature for RaDaR in NSCLC, breast, and HNSCC.
Limits. Requires tissue. RaDaR LOD95 of 11 ppm is from a press release, conference abstract, and clinical-paper detections, not a full analytic-validation paper.
A3.3
Tumor-informed whole-genome approaches
Tumor WGS builds a large mutation fingerprint. Plasma is then tested either with a larger personalized variant set or by integrating signal across thousands of mutations in plasma WGS (MRDetect).
Strengths. Access to many more somatic events, including structural and phased variants in some designs; useful when exonic mutation burden is low. Zviran et al. reported tumor-fraction sensitivity as low as 1e-5.
Limits. Requires tumor (and usually normal) WGS. Some commercial LOD claims are not in full peer-reviewed analytic papers.
A3.4
Phased variants (PhasED-seq)
A phased variant is two or more mutations on the same DNA fragment. Requiring co-occurrence on one molecule suppresses background error that can mimic a single mutation.
Strengths. Biological error suppression. Peer-reviewed analytic validation and prospective MRD data in B-cell malignancies.
Limits. Requires enough clustered mutations. Most clinical MRD data to date are in lymphoma, not a broad solid-tumor menu.
Kurtz Nat Biotechnol 2021PhasED-seq analytic 2025Prospective LBCL MRD
A3.5
Tumor-naive methylation
Plasma cfDNA is assayed for cancer-associated methylation at many regions, sometimes combined with a genomic variant classifier, without tumor tissue. A model calls detected or not detected.
Strengths. Tissue-free; potentially faster first result; methylation is complementary to SNVs and is a different error process from CHIP SNVs. Serial testing improves clinical sensitivity versus a single landmark draw.
Limits. Performance depends on cancer type and metastatic site (e.g., Reveal CRC: liver 100%, lung 53%, peritoneal 40%). Some analytic LODs are conference abstracts only. Northstar Clear had no peer-reviewed validation at review.
Parikh CCR 2021Parikh 2024 Reveal CRCNakamura CCR 2025 Tempus xM naive
A3.6
Fragmentomics
Tumor-derived cfDNA often differs in fragment length and cut sites from normal cfDNA. Assays use length and end motifs as a cancer signal, alone or with mutations or methylation.
Strengths. Does not depend on a specific mutation; can be tissue-free; combinable with other signals.
Limits. Requires careful validation. Standalone fragmentomics-only solid-tumor MRD menus are less mature than mutation- or methylation-based tests.
A3.7
Duplex and error-corrected sequencing
Unique molecular identifiers tag each starting molecule. Duplex sequencing requires a mutation to be consistent on both complementary strands, distinguishing true mutations from PCR or sequencing error.
Strengths. Lowers the technical noise floor. iDES improved CAPP-Seq about 15-fold, to 4 in 1e5 cfDNA molecules (Newman 2016). Safe-SeqS (DYNAMIC) is in this lineage; Haystack MRD uses background-noise suppression in a tumor-informed assay (Clark 2026).
Limits. Cannot create molecules that were never in the tube. Finite genome equivalents set a biological floor. Do not equate DYNAMIC Safe-SeqS with commercial Haystack.
Newman Nat Biotechnol 2016Tie NEJM 2022 DYNAMICClark J Mol Diagn 2026 Haystack
A3.6
Virus-informed HPV ctDNA (NavDx)
Quantifies circulating tumor HPV DNA (types 16, 18, 31, 33, 35) without sequencing the patient's tumor genome. Applies only to HPV-driven cancers.
Strengths. Tissue-free; high NPV when serially negative in a prospective oropharyngeal surveillance study (Chera 2020).
Limits. Not a general solid-tumor MRD test.
B clonoSEQ
Immunoglobulin / TCR sequencing (hematologic MRD)
High-depth sequencing of immunoglobulin and T-cell receptor gene rearrangements. Not a solid-tumor ctDNA SNV assay. Included for context.
Strengths. FDA-authorized in specified heme indications; analytic LOD supporting 1e-6 when input is sufficient (Ching 2020).
Limits. Different biology and sample types (often bone marrow) from solid-tumor ctDNA MRD.
A3.8 / Pathlight
Tumor-informed structural-variant tracking
WGS-level discovery of unique rearrangement junctions in the tumor, then tracking of those structural variants in plasma. Junctions are unlikely to arise from random sequencing error.
Strengths. Low-background unique junctions; peer-reviewed 2026 breast-cancer papers for Pathlight.
Limits. Requires tumor WGS-level SV discovery. Lock table-level LOD and variant counts from PMIDs 42242004 and 42691549 before publish.
A3.7
Tumor-informed error-corrected targeted sequencing
Patient-specific somatic mutations are tracked with background-noise suppression (Haystack JMD 2026). Related scientifically to Safe-SeqS, the DYNAMIC research assay. Quest acquired Haystack Oncology in April 2023.
Strengths. Peer-reviewed 2026 analytic and clinical validation paper (Clark et al., J Mol Diagn).
Limits. Variant count and WES vs WGS basis not in the abstract; full text was not open. Do not present DYNAMIC results as Haystack results.
A3.8
Whole-genome versus whole-exome versus a targeted panel
Tumor-informed assays differ in what is sequenced from the tumor (targeted CGP panel vs WES vs WGS), how many variants the plasma test then tracks, and which variant types are used. More independent sites improve the statistical chance of observing mutant molecules at low tumor fraction, given finite cfDNA genome equivalents (often a few thousand in a typical draw; Wan 2017). This does not by itself rank commercial tests.
targeted panel
Tissue CGP of a few hundred genes. Trackable mutations are often few, especially in low-mutation tumors.
- Determined by mutations on the FoundationOne tissue CGP report; no single product-wide N published as a fixed design target.PMID 38753705
WES
Protein-coding genome (about 1-2%). Typically enough clonal SNVs/indels for a modest personalized set.
- 16 clonal SNVs/indelsPMID 30992300
- 18-50 variantsPMID 37632661
- Up to 48 tumor-specific variants (LIONESS panels were 34-52 in 17 patients)PMID 35132238
WGS
Coding and non-coding DNA; can yield thousands to tens of thousands of somatic events including non-coding SNVs, more indels, structural variants, and phased-variant sites. Assays then pick a customized subset or integrate genome-wide.
- Up to ~1,800 somatic variantsPMID 38484152
- Panels of up to 1,000 targetsAbstract onlyAACR 2026 abstract 2598
- 64 high-quality variants (SNVs, phased variants, structural variants)Abstract onlyJCO 2025 abstract 3142; SABCS 2025 PS2-07-26
- Genome-wide integration of thousands of mutations (research MRDetect), not a small fixed panelPMID 32483360
One haploid human genome is about 3.3 pg; 10 ng cfDNA is on the order of 3,000 haploid genome equivalents. At 0.01% tumor fraction, expected mutant molecules at a single site can be below one (Poisson limit).
A Myriad JCO 2023 conference abstract (e15022) reported that expanding a panel from 16 to 300 target variants increased sensitivity at 0.005% tumor fraction from about 20% to more than 95%. Conference abstract, not a full paper. Abstract only
- SNVs. Most abundant; most affected by PCR/sequencing error.
- indels. Usually fewer; Myriad AACR 2025 abstract 6641 reported lower technical error for many indels (conference abstract).
- structural variants. Unique junctions with low random-error background; Pathlight; Signatera Genome conference materials.
- CNVs. Harder at very low tumor fraction; not the primary MRD signal in personalized SNV assays reviewed.
- phased variants. Two or more mutations on the same molecule (Kurtz 2021).
- methylation. Epigenetic signal, mainly tumor-naive assays, not a tumor-informed SNV panel count.
A short animation
about 45-60 seconds, 8 scenes. No numeric LODs, percentages, or brand rankings unless a cited source is on screen. Brand names, if used, should be a neutral list, not a comparison.
Two ways to find leftover cancer DNA — and what methylation means
0-6s
A clear scan does not always mean every cancer cell is gone.
CaptionAfter treatment, scans can look clear. Small amounts of cancer can still remain.
Scene 1 of 8. About 60 seconds.
Comparison chart
Medicare cells were checked against the CMS page for A58454 on November 13, 2026. The Noridian article now in effect has revision effective date March 12, 2026, and no ending date. LCD L38814 remains revision effective February 5, 2026. The draft had used an A58454 window of August 6 to September 30, 2026, which is not the version CMS marks in effect.
Performance figures come from different studies, assays, cancer types, stages, time points (landmark versus serial), follow-up lengths, and definitions of sensitivity, specificity, and limit of detection. They are not head-to-head comparable and should not be used to rank tests. 'Not published' means a peer-reviewed source for that cell was not found. Conference abstracts and vendor-only figures are labeled and are weaker evidence than full papers.
17 tests. Sort by a column heading. On a phone, scroll sideways. The assay name stays pinned.
| Tumor sequencing | Variants tracked | Variant types | Methylation | Tissue needed | Analytic LOD | Clinical examples | Medicare | ||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| clonoSEQ | Adaptive Biotechnologies | immune-receptor (not solid-tumor ctDNA SNV) | Immunoglobulin / TCR sequencing (hematologic MRD) | identifies patient-specific Ig/TCR rearrangements from a diagnostic sample | Patient-specific immune-receptor rearrangements, not a fixed SNV countChing BMC Cancer 2020 | IGH/IGK/IGL and TCR rearrangements | No | Bone marrow (and, for some indications, blood) — not a plasma-only solid-tumor test | LOD 1.903 malignant cells, corresponding to 6.77e-7 at 20 ug DNA input, supporting 1e-6 when input is sufficientChing BMC Cancer 2020 | Included for heme context only. Clinical performance is indication-specific (MM, B-ALL, CLL) and is not charted against solid-tumor ctDNA assays. | Separate MolDX article for lymphoid malignancies (historically A56307), not A58454. Current Noridian article A58454, revision effective March 12, 2026, no ending date.FDA De Novo authorization for MRD in multiple myeloma and B-cell ALL; later expanded to CLL (verify current label at publish). |
| Foresight CLARITY (PhasED-seq) | Foresight Diagnostics | tumor-informed phased-variant sequencing (primarily lymphoma) | Phased variants (PhasED-seq) | tumor sequencing to identify phased variants | Phased variants; not a single solid-tumor SNV count — see PMIDs 40353795 and 34294911 | phased variants | No | Yes | See PMID 40353795 analytical validation for B-cell malignancies; lock exact LoD from that paper at publish. | large B-cell lymphoma. Prospective validation PMID 41385760. Included for technology context, not as a general solid-tumor MRD test. | Not summarized from A58454 (solid-tumor article). Current Noridian article A58454, revision effective March 12, 2026, no ending date.Verify current regulatory status at publish; primarily a lymphoma MRD test. |
| FoundationOne Tracker | Foundation Medicine / Natera | tumor-informed (CGP-informed, not dedicated WES/WGS) | Tumor-informed multiplex PCR | FoundationOne tissue CGP panel | Determined by mutations on the tissue CGP report; no single product-wide fixed N extractedPLOS One 2024 | SNVs/indels present on the CGP profile | No | Yes | See PLOS One 2024 (PMID 38753705) tables; lock exact LoD figures from that paper before going live.PLOS One 2024 PMC11098318 | Clinical monitoring literature exists; extract indication-specific endpoints at publish if used on the public page. | A 2023 Foundation Medicine and Natera press release announced Medicare coverage for immune-checkpoint inhibitor response monitoring. The current A58454 (Noridian, revision effective March 12, 2026) names Signatera, not FoundationOne Tracker, for that indication.LDT. |
| Guardant Reveal | Guardant Health | tumor-naive (tissue-free) | Tumor-naive methylation | none | Not a personalized SNV panel; evaluates >20,000 epigenomic regions (2024 paper); some versions also use genomic variantsParikh 2024 | methylation, SNVs/indels (in described multiomic versions) | Yes | No | AACR 2025 abstract 692: >=95% LoD down to an order of 1e-5 to 1e-4 at 3-4 ng and 15 ng cfDNA; 100% specificity in 102 age-matched healthy donors. Do not mix with 2020 JCO abstract e15549 LoD (0.3% at 10 ng) for a prior version.Abstract onlyAACR 2025 abstract 692 | colorectal. landmark ~1 month after definitive therapy; longitudinal. landmark 55.6%; longitudinal 69%; 91% for draws within 4 months of recurrence. 100% (in this 2021 study, patients with >=1 year follow-up)Parikh CCR 2021colorectal (stage II or higher colon for the 81% figure). longitudinal postoperative, 1977 plasma samples from 342 patients. 81% longitudinal in stage II or higher colon cancer; liver 100%, lung 53%, peritoneal 40%; rectal 60%. 98.2% among 1461 post-treatment samples. median 5.3 months. n=342Parikh 2024 | A58454 current Noridian article lists Guardant Reveal for advanced colorectal (progression, recurrence, or relapse). Guardant360 Response is a distinct test listed for ICI monitoring. Current Noridian article A58454, revision effective March 12, 2026, no ending date.LDT, not FDA-approved as MRD.CMS article |
| Haystack MRD | Quest Diagnostics / Haystack Oncology | tumor-informed | Tumor-informed error-corrected targeted sequencing | not published in the 2026 J Mol Diagn abstract (full text not open at review) | not publishedClark J Mol Diagn 2026 abstract does not state a count | patient-specific somatic mutations (types not detailed in the abstract) | No | Yes | 95% LOD 0.063 mean ctDNA molecules/mL; analytical measurement range 0.25-1000 mean ctDNA molecules/mL; quantitative R-squared 0.9998 vs orthogonal methodClark J Mol Diagn 2026 | stage II-III colorectal, post-treatment. 100% (95% CI 72.3-100%) at 3-year follow-up; 83.3% (95% CI 55.2-97.0%) at 5-year follow-up. 100% (95% CI 92.9-100%) at 3 years; 100% (95% CI 92.6-100%) at 5 years. n=not stated in the abstract; CIs imply a modest number of events — do not invent nClark J Mol Diagn 2026 | Not listed in the A58454 current Noridian article. Current Noridian article A58454, revision effective March 12, 2026, no ending date.LDT. |
| Invitae Personalized Cancer Monitoring (PCM) | Invitae; Labcorp finalized acquisition of select Invitae assets on 2024-08-05. Current Labcorp menu name should be verified before publish. | tumor-informed | Tumor-informed multiplex PCR | WES (tumor and germline) | 18-50 variantsInvitae PCM analytic 2023 | SNVs (patient-specific panels; anchored multiplex PCR with UMIs) | No | Yes | LOD ranged from 0.008% allele frequency with 60 ng cfDNA and 18-50 variants (>99.9% sensitivity) at a baseline threshold. Sensitivity 96.3% at 10 ng, 0.008% AF, 50 variants. Specificity >99.9%. WES component sensitivity 99.8%, specificity >99.9%. Panels designed for 100% (63/63) of samples with >=20% tumor content and 80% (24/30) with >=10%.PMID 37632661 PMC10590345 | early breast cancer, longitudinal. PMID 39424680 used this assay; extract exact sensitivity/specificity from PMC11785695 at publish.2024 early breast longitudinal | Not listed by the Invitae name in the A58454 current Noridian article. Current Noridian article A58454, revision effective March 12, 2026, no ending date.LDT. |
| Labcorp Plasma Detect (Genome) | Labcorp | tumor-informed | Tumor-informed whole-genome approaches | WGS (company product description) | not published as a single peer-reviewed product-wide count | tumor-informed WGS somatic variants (company materials) | No | Yes | AACR abstract B028 (2024) reported 0.005% ctDNA-content sensitivity and 99.4% specificity for tumor-informed WGS MRD analyses — conference abstract; confirm it is the commercial product.Abstract onlyAACR abstract B028 | Peer-reviewed clinical performance of the commercial assay not identified as a full paper. Do not automatically equate PMID 35674097 (PGDx fragment-length research) with this product. | Not listed in the A58454 current Noridian article. Distinct from Invitae PCM. Current Noridian article A58454, revision effective March 12, 2026, no ending date.LDT. |
| NavDx | Naveris | virus-informed (tissue-free; not a general solid-tumor SNV MRD test) | Virus-informed HPV ctDNA (NavDx) | none (HPV types 16, 18, 31, 33, 35) | HPV genotypes 16, 18, 31, 33, 35 — not a tumor SNV panelDiagnostics 2023 | circulating tumor HPV DNA, fragmentomic profiling | No | No | LOBs 0-0.32 copies/uL; LODs 0-1.10 copies/uL; LOQs <1.20-4.11 copies/uL (ranges across genotypes)Diagnostics 2023 PMC9955790 | HPV-associated oropharyngeal, post-treatment surveillance. two consecutive positives: PPV 94% (95% CI 70-99%). 87 patients undetectable at all post-treatment time points had no recurrences (NPV 100%). median 3.9 months (range 0.37-12.9). n=115 patients, 1006 blood samples, median follow-up 23 months, 15 recurrencesChera JCO 2020 | A58454 current Noridian article lists NavDx for HPV-driven oropharyngeal and anal squamous cell cancers (recurrence/relapse with personal history of cancer). Current Noridian article A58454, revision effective March 12, 2026, no ending date.LDT. |
| NeXT Personal | Personalis | tumor-informed | Tumor-informed hybrid-capture / WGS-informed panels | WGS | Up to ~1,800 somatic variantsIn the analytic cohort, nearly half of 118 panels had >1,800 targets; 70% had >1,000.Northcott Oncotarget 2024; also Mayadev 40500687 | SNVs (tumor-normal WGS somatic calls; CHIP and germline filtered) | No | Yes | Analytical validation (Northcott, Oncotarget 2024): published. Detection threshold 1.67 PPM (range 1.47–1.87 PPM). LOD95 3.45 PPM. Limit of blank 0.719 PPM. Analytical sensitivity 100% (95% CI 98.5–100%). Analytical specificity 100% (95% CI 96.0–100%) in the accuracy study; the validation measured 100% specificity, with an in silico confidence interval of 99.92–100%. Up to ~1,800 somatic variants. cfDNA input 2–30 ng. Linearity 0.8 to 300,000 PPM.PMID 38484152 | lung adenocarcinoma (TRACERx), preoperative. 81% of LUAD; 53% of pTNM stage I LUAD. Patients with <80 PPM preoperative ctDNA still had reduced OS vs ctDNA-negative.Black Nat Med 2025locally advanced cervical cancer (CALLA). ctDNA detected in 98.9% (183/185) at baseline. Detection at C3D1 (post-CRT) preceded clinical progression by median 164 days (95% CI 85-250).. n=185 baseline samplesMayadev Ann Oncol 2025 | Current A58454 (Noridian), revision effective March 12, 2026: NeXT Personal is named for recurrence or relapse in non-small cell lung cancer and in advanced breast cancer (personal history of cancer). It is not named in that article for neoadjuvant breast cancer or for immune-checkpoint inhibitor response monitoring.LDT. |
| Northstar Clear | BillionToOne | tumor-naive (tissue-free) | Tumor-naive methylation | none | not published | not published for Clear; do not apply Northstar Response or Select figures | Yes | No | not publishedLaunch announced for 2026-12-15. Distinguish from Northstar Response (vendor LOD 0.01% tumor fraction, treatment monitoring) and Northstar Select (liquid CGP, peer-reviewed LOD 0.15% VAF for SNV/indels).Vendor only | not published. Company has said initial CRC data will compare head-to-head with a tumor-informed assay; paper not found. | Not listed in the A58454 current Noridian article. Current Noridian article A58454, revision effective March 12, 2026, no ending date.Announced as LDT. Not FDA-approved. |
| Oncodetect | Exact Sciences | tumor-informed | Tumor-informed personalized targeted panel (chemistry not forced) | not locked to a peer-reviewed methods paper in this pass (vendor materials describe a bespoke tumor-informed test)Vendor only | not published in a peer-reviewed paper identified in this pass | not published in a peer-reviewed paper identified in this pass | No | Yes | Vendor white papers: 0.005% tumor fraction (50 ppm); later materials 15 ppm with observed detection down to 2 ppm. Vendor only. Full peer-reviewed analytic paper not identified.Vendor onlyExact Sciences vendor materials | Peer-reviewed clinical endpoints should be pulled from Exact Sciences' cited studies at publish; not locked to a single full paper in this pass. | A58454 current Noridian article lists Oncodetect, with Signatera and Guardant Reveal, for advanced colorectal cancer (progression, recurrence, or relapse). Current Noridian article A58454, revision effective March 12, 2026, no ending date.LDT. |
| Pathlight | SAGA Diagnostics | tumor-informed | Tumor-informed structural-variant tracking | WGS-level structural-variant discovery (confirm exact basis in PMIDs 42242004 and 42691549 at publish) | Lock from 2026 papers at publish (PMC full texts retrieved)2026 Pathlight papers also 42691549 | structural variants | No | Yes | Lock table-level numbers from PMIDs 42242004 and 42691549 before going live. | early and recurrent breast cancer. PMID 42242004 and 42691549; extract sensitivity, specificity, and lead time from full texts at publish.also 42242004 | A58454 current Noridian article lists Pathlight for advanced breast cancer (recurrence/relapse with personal history of cancer). Current Noridian article A58454, revision effective March 12, 2026, no ending date.LDT. |
| Precise MRD | Myriad Genetics | tumor-informed | Tumor-informed hybrid-capture / WGS-informed panels | WGS (tumor, including FFPE/core needle, and matched normal) — AACR 2026 abstract 2598 | Panels of up to 1,000 targetsAbstract onlyAACR 2026 abstract 2598; AACR 2025 abstract 6641 | SNVs, indels (AACR 2025 abstract 6641) | No | Yes | AACR 2026 abstract 2598: LoD95 <5 ppm in genome-equivalent units (described as comparable to <2.5 ppm VAF) under typical conditions. Myriad website cites 1.45 ppm LOD95 — vendor only.No full peer-reviewed analytical-validation paper identified.Abstract onlyVendor onlyAACR 2026 abstract 2598 | AACR 2026 abstract 2598 states clinical sensitivity >85% and specificity >99% across tested indications, and 100% sensitivity />99% specificity in colorectal, ovarian, endometrial, and lung — conference abstract without n or time-point definitions. MONITOR-Breast PMID 42324681: confirm commercial Precise MRD was used, then extract endpoints from the full paper.Abstract onlyAACR 2026 abstract 2598 | Not listed in the A58454 current Noridian article. Current Noridian article A58454, revision effective March 12, 2026, no ending date.LDT. |
| RaDaR / RaDaR ST | NeoGenomics (formerly Inivata) | tumor-informed | Tumor-informed personalized targeted panel (chemistry not forced) | WES (manufacturer descriptions of RaDaR ST; clinical papers use tumor sequencing to select variants) | Up to 48 tumor-specific variants (LIONESS: panels of 34-52 in 17 patients)Flach 2022; manufacturer and Gale 35306155 | SNVs | No | Yes | 0.0011% VAF (11 ppm) stated as LOD95 in a 2021 NeoGenomics press release and AACR 2023 abstract 3384. Gale et al. report detections from 0.0011% (11 ppm) upward. Manufacturer 'as low as 1 ppm under study-specific conditions, data on file' is vendor-only.Full peer-reviewed analytical-validation paper not identified.Data on fileGale Ann Oncol 2022 detections; press release / AACR 3384 for LOD95 | early-stage NSCLC. after treatment, ctDNA detected in 18/28 (64.3%) who had clinical recurrence of the primary. >98.5%. median 212.5 days. n=88 patients, 363 serial plasma samplesGale Ann Oncol 2022high-risk HR+/HER2- breast, late recurrence (CHiRP). enrolled >5 years after diagnosis, no clinical recurrence at enrollment. 6 patients (7.2%) developed distant metastatic recurrence, all MRD-positive beforehand; median lead time 12.4 months; MRD not identified in 1 patient with local recurrence.. n=103Lipsyc-Sharf JCO 2022HPV-negative HNSCC (LIONESS). Baseline ctDNA 17/17. Post-surgery detection as low as 0.0006% VAF. All clinical recurrences had prior ctDNA, lead times 108-253 days. Small sample.. n=17Flach BJC 2022 | A58454 current Noridian article lists RaDaR for HPV-negative HNSCC and for advanced breast cancer (recurrence/relapse with personal history of cancer). Current Noridian article A58454, revision effective March 12, 2026, no ending date.LDT. |
| Signatera (exome) | Natera | tumor-informed | Tumor-informed multiplex PCR | WES | 16 clonal SNVs/indelsCoombes CCR 2019; also Reinert 31070691 | SNVs, indels | No | Yes | 0.01% VAF (estimated, by targeting 16 mutations)AACR 2018 abstract 4542: variant-level sensitivity ~60% at 0.03% spiked tumor DNA; assay specificity 99.9% in wild-type DNAA full peer-reviewed analytical-validation paper was not identified. Do not present 0.01% as a journal-validated LOD95.Abstract onlyVendor onlyAACR 2018 abstract 4542 and vendor white papers | colorectal stages I-III. serial surveillance after definitive treatment. 88%. 98%. mean 8.7 months (range 0.8-16.5). 14/15 ctDNA-positive relapsed vs 2/60 ctDNA-negative. Preoperative detection 108/122 (88.5%).. n=75 patients with longitudinal plasma (122 preoperative in the same paper)Reinert JAMA Oncol 2019colorectal (GALAXY / CIRCULATE-Japan). 4 weeks after surgery, stages include II-III analyses. not framed as a single sensitivity in the abstract; recurrence 61.4% (115/187) if ctDNA-positive vs 9.5% (81/852) if negative. not a single specificity percentage in the extracted results; 81 of 852 ctDNA-negative patients still recurred (9.5%). 18.0% (187/1039) ctDNA-positive at 4 weeks. Prognostic HR 10.82 in stage II-III. ACT benefit among ctDNA-positive HR 6.59.. n=1039Kotani Nat Med 2023breast. after surgery and adjuvant therapy, every 6 months up to 4 years. 89% (16/18 relapsed patients). 100% (0/31 non-relapsing; 156 plasma samples). median 8.9 months (range 0.5-24). n=49Coombes CCR 2019muscle-invasive bladder (IMvigor011). serial ctDNA after cystectomy, no radiographic disease; ctDNA-positive randomized to atezolizumab vs placebo. Clinical utility trial. Median DFS 9.9 vs 4.8 months (HR 0.64, 95% CI 0.47-0.87). Median OS 32.8 vs 21.1 months (HR 0.59, 95% CI 0.39-0.90). Persistently negative: DFS 95% at 1 year, 88% at 2 years. Signatera used for selection (trial acknowledgments / Natera collaboration).. n=761 enrolled; 250 ctDNA-positive randomized (167 atezolizumab, 83 placebo); 357 persistently ctDNA-negativePowles NEJM 2025 | A58454 current Noridian article lists Signatera for advanced colorectal; muscle-invasive bladder and ovarian; neoadjuvant breast; NSCLC; advanced breast; and ICI-response monitoring in advanced solid tumors. Confirm current article. Current Noridian article A58454, revision effective March 12, 2026, no ending date.Not FDA-approved as an MRD diagnostic (LDT).CMS article |
| Signatera Genome | Natera | tumor-informed | Tumor-informed multiplex PCR | WGS | 64 high-quality tumor-specific variantsConference abstracts, not a full analytic paper.Abstract onlyJCO 2025 abstract 3142; SABCS 2025 PS2-07-26 | SNVs, phased variants, structural variants | No | Yes | Vendor: LOD95 <1 PPM and observed LOD <0.1 PPM in a cell-line dilution study (Natera data on file). JCO 2025 abstract: 99.8% analytical sample-level specificity in healthy subjects.LOD figures not found in a peer-reviewed analytical-validation paper.Abstract onlyVendor onlyData on fileNatera website, data on file; JCO abstract 3142 | pan-solid (breast, NSCLC, melanoma, RCC, CRC). WGS-based tumor-informed monitoring paper PMID 42886530: 62% (72/116) of MRD-positive patients had ctDNA <100 ppm. Confirm at publish that this is the Genome product.. n=3862026 WGS tumor-informed monitoring paperbreast, surveillance. SABCS 2025 abstract PS2-07-26: longitudinal ctDNA positivity associated with worse DRFS (HR 22.2, 95% CI 11.0-44.7); stated 100% sensitivity and specificity for preceding clinical recurrence — conference abstract.. n=228Abstract onlySABCS 2025 PS2-07-26 | Not listed as a separately named test in the A58454 current Noridian article. Current Noridian article A58454, revision effective March 12, 2026, no ending date.Not FDA-approved as an MRD diagnostic (LDT). |
| TrueMRD (MRDetect-based) | C2i Genomics / Veracyte | tumor-informed | Tumor-informed whole-genome approaches | WGS (MRDetect framework) | Genome-wide integration of thousands of somatic mutations (research paper), not a small fixed panelZviran Nat Med 2020 | genome-wide SNVs (mutational integration) | No | Yes | Research MRDetect: tumor-fraction sensitivity as low as 1e-5 (Zviran 2020). Commercial lab-validation figures are in conference abstracts (e.g., JCO 2022 e13582), not full papers identified here.Zviran Nat Med 2020 | Zviran 2020 reported postoperative residual-disease detection associated with adverse outcome in the research framework. Commercial clinical performance should be cited only from product-specific papers if used on the public page. | Not named in the current Noridian article A58454 (revision effective March 12, 2026, no ending date). A separate contractor article, A58468 (Wisconsin Physicians Service), revision effective October 1, 2026, lists TrueMRD for muscle-invasive bladder cancer (recurrence or relapse with a personal history of cancer).LDT (commercial). |
Footnotes
Personalis and Tempus announced an exclusive U.S. commercialization collaboration for NeXT Personal in November 2023. Tempus markets that tumor-informed assay as xM (NeXT Personal Dx). Separately, Tempus has a tumor-naive plasma-only MRD assay, also called xM (Nakamura CCR 2025, PMID 39513962). These are different tests that share a brand name. The tumor-naive assay is described in this footnote. No preference is expressed.
Tempus xM (tumor-naive) — different assay from xM (NeXT Personal Dx)
- Vendor
- Tempus
- Approach
- tumor-naive
- Technology
- Tumor-naive methylation
- Tumor sequencing
- none
- Variants tracked
- Not a personalized panel; methylation classifier plus genomic-variant classifier, binary MRD callNakamura CCR 2025
- Variant types
- methylation, genomic variants (CRC classifier)
- Methylation
- Yes
- Tissue needed
- No
- Analytic LOD
- not publishedNot in the Clin Cancer Res clinical paper. Do not apply NeXT Personal 1.67/3.45 PPM figures to this assay.
- Clinical examples
- colorectal stage II-III, GALAXY subset, enriched for 50% recurrence (not natural-history prevalence). landmark 4 weeks and longitudinal. landmark 61.1% (22/36 recurrent); longitudinal 83.3%. landmark 87.9% (29/33 nonrecurrent); longitudinal 89.5%. median 4.77 months overall; 5.30 months with no adjuvant treatment. 12-week MRD status associated with DFS, adjusted HR 9.69, vs CEA HR 2.13.. n=80 selected; 69 evaluable at landmark; 74 longitudinalNakamura CCR 2025
- Medicare
- Not listed by this name as a distinct covered test in the A58454 current Noridian article. Do not infer NeXT Personal indications apply. Current Noridian article A58454, revision effective March 12, 2026, no ending date.LDT.
Performance figures come from different studies, assays, cancer types, stages, time points (landmark versus serial), follow-up lengths, and definitions of sensitivity, specificity, and limit of detection. They are not head-to-head comparable and should not be used to rank tests. 'Not published' means a peer-reviewed source for that cell was not found. Conference abstracts and vendor-only figures are labeled and are weaker evidence than full papers.