Reference products and why people stop using them — ctDNA MRD assay validation

1. Summary

knowledge-base/domains/ctdna-mrd-assay-validation.md §5 already names and sources three commercial contrived-reference-material vendors (LGC/Seraseq, Horizon Discovery/Revvity, Twist Bioscience) and records that none has a located discontinuation, recall, or independent commutability study. This pass does not re-survey those three; it works four items the dossier's own gap list left open. The downstream market this reference material serves shifted during 2026: a tumor-informed ctDNA MRD test moved, for one indication, from a CLIA-validated laboratory-developed test to an FDA Class III premarket-approved companion diagnostic, which raises the evidentiary bar for at least part of this market beyond the daily-control requirement the dossier already documented. An explicit federal-registry recall-database query for the three named vendors, run fresh for this report, returned no matching record — moving that item from "not searched" to "not found." Two separate published lines of work — one academic, one from a national metrology institute — are independently pursuing exactly the two limitations the dossier flagged in the three named commercial products: a synthetic or "plasma-like" matrix rather than native ctDNA fragmentation, and unassessed isolation-recovery bias upstream of any limit-of-detection claim. Neither is a commercial product with adoption evidence yet.

2. What Changed

_Baseline (v1). No prior version; this establishes the starting point for future diffs._

3. Details

The downstream market's evidentiary floor moved for one product

Natera's Signatera CDx received FDA premarket approval (PMA) as a companion diagnostic, decision date 2026-05-15, product code PQP, device class 3 [1] [single-source]. The approval order's stated intended use:

"Signatera™ CDx is a personalized, tumor-informed, multiplex-PCR and next-generation sequencing (NGS)-based assay that detects circulating tumor DNA (ctDNA) molecular residual disease (MRD) from plasma of peripheral whole blood collected in Streck Cell-Free DNA Blood Collection Tubes (BCTs)... intended to identify patients with ctDNA MRD positive status who may benefit from treatment... Muscle Invasive Bladder Cancer (MIBC)... TECENTRIQ® (atezolizumab), TECENTRIQ HYBREZA® (atezolizumab and hyaluronidase-tqjs)." — openFDA PMA record P260004 [1]

This is distinct from the CLIA daily-control requirement the dossier already established at 42 CFR 493.1256. A PMA application must include, per 21 CFR 814.20(b)(2), "separate sections on nonclinical laboratory studies and on clinical investigations involving human subjects" [2] — a submission-level analytical-validation requirement reviewed by FDA, not only a CLIA laboratory's own daily quality-control obligation. The dossier's §4b assays (Haystack MRD, CancerDetect, NeXT Personal, Invitae PCM, GeneBits) remain, on the evidence located in this domain, laboratory-developed tests reporting their own internally-run analytical validation; Signatera's MIBC indication is now the one instance in this evidence base of a tumor-informed MRD assay's analytical package having passed an FDA premarket review rather than only a CLIA inspection. [single-source]

The recall record was checked, not only assumed silent

The dossier's §5 records that no discontinuation or recall was located for the three named vendors and flags that "no vendor press archive was searched specifically." This pass instead queried FDA's public device-recall database directly, by recalling firm, for "LGC," "SeraCare," "Horizon Discovery," and "Twist Bioscience" over a ten-year window; each query returned zero matching records. FDA's recall database implements the reporting duty at 21 CFR 806.10, which requires a manufacturer or importer to report "any correction or removal of a device... initiated... to reduce a risk to health" [3]. A queried and empty federal registry is a different fact than an unsearched one: this is not found, not not searched, for these four name variants of the three vendors, over the window queried. [single-source]

A published alternative to a synthetic or "plasma-like" matrix

A 2025 paper reports a preparation method for ctDNA reference material built by digesting nucleosomal DNA from cultured cancer cell lines with micrococcal nuclease, rather than the "plasma-like matrix" or "synthetic plasma" formulations the dossier documents for the three named commercial vendors:

"By digesting nucleosomal DNA derived from cancer cell lines with micrococcal nuclease, this method can closely mimic the properties of clinical ctDNA." — PMID 40428398 [4]

The accompanying dPCR assays reported a limit of detection of 0.143% VAF (TP53 R175H) and 0.092% VAF (TP53 R248W), with repeatability (RSD) of 0.16%-7.65% across a 50%-0.1% VAF range [4]. This sits above the bottom of idea-003's 0.001%-0.5% target range and above the three named vendors' own lowest labeled levels (0.1%), so it does not itself close the sub-0.1% white space the dossier's §5 and §7 already discuss — its contribution is a published, alternative manufacturing method aimed at the matrix-fidelity gap, not at the sensitivity floor. [single-source]

A metrology-institute alternative aimed at a different QC problem: isolation recovery

A separate 2026 paper from a national metrology institute addresses a pre-analytical failure mode the dossier's §5 does not cover at all — recovery bias during cfDNA isolation, upstream of any limit-of-detection claim:

"Fragment-length-dependent recovery during isolation represents a major source of pre-analytical bias... certified reference materials (CRMs) specifically designed to evaluate cfDNA isolation recovery are currently lacking." — PMID 41936819 [5]

The paper reports two new CRMs (UME CRM 3022: 80- and 240-bp fragments; UME CRM 3024: 80-, 120-, 160- and 240-bp fragments), gravimetrically prepared and characterized by droplet digital PCR, with demonstrated "applicability in a plasma matrix" [5]. This is a different QC dimension than the LOD/VAF characterization the three named commercial vendors and the dossier's §4b assays compete on, and no adoption or commercial-availability evidence for these two specific CRMs was located in this pass. [single-source]

An earlier, independent interlaboratory reference-material study

A 2020 study, run outside the three named vendors, prepared an SI-traceable "ctDNA" reference material for BRAF V600E by gravimetrically mixing a synthetic amplicon with sonicated wild-type genomic DNA, and used it in an interlaboratory ddPCR assessment:

"The limit of blank (LoB), detection (LoD) and quantification (LoQ) of ddPCR assay were determined to be 0.01%, 0.02% and 0.1%, respectively... a systematic error caused by uncorrected droplet volume in Naica Crystal ddPCR platform was found by using the ctDNA RM." — PMID 31594564 [6]

This is a second, independent example — on a different chemistry (ddPCR, not NGS) and a different variant than the dossier's cross-vendor NGS finding (its source 12) — of a reference material surfacing a platform-specific systematic error during interlaboratory use, rather than only a smooth precision-versus-VAF curve. [single-source]

4. Sources

[1] Premarket Approval (PMA) database record — Signatera CDx, applicant Natera, Inc. — U.S. Food and Drug Administration, openFDA device/pma dataset (decision date 2026-05-15; accessed 2026-09-01). PMA number P260004 — https://api.fda.gov/device/pma.json?search=pma_number:P260004 [federal-registry] [2] 21 CFR 814.20, "Application" — retrieved via python3 -m connectors regulation 814.20 --title 21 --expect "valid scientific evidence" --jurisdiction US, eCFR versioner title-21, issue 2026-08-27 (accessed 2026-09-01). 21 CFR 814.20 — https://www.ecfr.gov/current/title-21/chapter-I/subchapter-H/part-814/subpart-B/section-814.20 [federal-registry] [3] 21 CFR 806.10, "Reports of corrections and removals" — retrieved via python3 -m connectors regulation 806.10 --title 21 --expect "recall" --jurisdiction US, eCFR versioner title-21, issue 2026-08-27 (accessed 2026-09-01); openFDA device recall database queried by recalling firm for LGC, SeraCare, Horizon Discovery and Twist Bioscience, 2016-09-03 to 2026-09-01, zero matching records for each. 21 CFR 806.10 — https://www.ecfr.gov/current/title-21/chapter-I/subchapter-H/part-806/section-806.10 [federal-registry] [4] Establishment of Reference Measurement Procedure for TP53 R175H/R248W Detection and a Novel Preparation Method for ctDNA Reference Material — Genes (published 2025-05-14; accessed 2026-09-01). PMID 40428398 — https://pubmed.ncbi.nlm.nih.gov/40428398/ [peer-reviewed] [5] Certified Reference Materials for Standardization of Cell-Free DNA Isolation Recovery in Liquid Biopsy — The Journal of Molecular Diagnostics (published 2026-06, day not given in the PubMed record; accessed 2026-09-01). PMID 41936819 — https://pubmed.ncbi.nlm.nih.gov/41936819/ [peer-reviewed] [6] Interlaboratory assessment of droplet digital PCR for quantification of BRAF V600E mutation using a novel DNA reference material — Talanta (published 2020, month not given in the PubMed record; accessed 2026-09-01). PMID 31594564 — https://pubmed.ncbi.nlm.nih.gov/31594564/ [peer-reviewed]

5. Sourcing & Gaps

Well established: nothing in this report rests on two independent non-affiliated primary sources; every substantive line here is [single-source] by design, because each fills a distinct gap the dossier had already recorded as un-investigated rather than re-confirming a line the dossier already established from multiple sources.

Thin: all six sources are marked [single-source] or, for the recall-database check, treated as one federal-registry query result. The PMA approval fact [1] was independently corroborated during this pass by a directly-fetched trade-press article (medtechdive.com, "Natera wins FDA approval for companion diagnostic") that agrees with the openFDA record, but that article carries no resolvable identifier of the kind this report's format requires (no DOI, docket number, or other registry identifier was present on the page), so it is named here rather than cited as a numbered source, and the underlying finding is carried as [single-source] on the openFDA record alone.

Rescoped from class 3: none newly performed in this report. The dossier's own §9 already records a class-3 substitution for this domain (would labs prefer a third-party panel to in-house controls, twinned to a discontinuation/complaint search that came up empty) and this report does not reopen it; the recall-database query above is an extension of that same already-recorded twin, run against a different registry (FDA recalls rather than a literature or press search) with the same empty result.

Out of scope: CPT/HCPCS billing codes (not applicable to this RUO reagent domain, as the dossier's §6 already states); fda.gov guidance documents (intermittent bot mitigation, not attempted); any EU or non-US recall, approval, or reference-material registry (no connector in this repo reaches one).

Not searched vs. not found: Not found — a PubMed search for "College of American Pathologists proficiency testing circulating tumor DNA survey" returned zero records, confirming rather than merely assuming the dossier's flagged CAP/CLIA paywall gap is not closed by an open-access proxy; a search for "commercial cell-free DNA reference standard comparison NGS platform variant allele frequency accuracy" also returned zero records, so no head-to-head comparison among the three named commercial vendors was found by this pass either. Not searched: SEC EDGAR 10-Q filings for an absolute (non-growth-rate) Signatera or Reveal MRD-specific unit count, which the dossier already flags as reachable and unattempted; this report did not pursue it because it bears on §1 (testing volume), not §5.

[inference] The reading that Signatera CDx's PMA approval raises the evidentiary bar for "at least part of this market" is this report's own synthesis from comparing 21 CFR 814.20's submission content requirement against the CLIA daily-control requirement already on record in the dossier; no source states that comparison directly, and no source establishes that this premarket-approved indication changes what any of the three named reference-material vendors must supply for it, which was not searched.

6. Claim Candidates