Tuesday, November 25, 2025

Big News: FDA to Down-Classify Many Companion Diagnostics as Class II (510k)

Last year, when FDA was sparring with stakeholders and courts over its LDT regulations, FDA promised to downclassify many types of diagnostics from Class III to Class II.   They went radio-silent from April to November, but now, the regulation is in print.

This is big news because it changes the landscape of how hard it is to get an FDA label as a companion diagnostic.  It also means that new ranges of tests will qualify for Medicare benefits.  NCD 90.2, for NGS testing in cancer, automatically covers NGS tests that are "cleared or approved" as CDx.   And sole-source tests (run from one lab) are eligible for ADLT pricing rules if they are "cleared or approved."   Now the range of "cleared" tests will be larger.

See an early essay at Linked In by Karin Hughes PhD here.

See the Fed Reg regulation proposal here.  Comment in 60 days until January 26, 2026.

The rule runs 14 pages and covers many considerations and details.  The regulation for 510K aka Class II CDx will be at 21 CFR 866.6075, as “Nucleic Acid-Based Test Systems for Use with a Corresponding Approved Oncology Therapeutic Product.”  Within the 14-page publication, the actual regulation at 21 CFR will be about 680 words long (two full columns of the Federal Register).


###

AI CORNER

###

FDA Down-Classifies 
Key Oncology Companion Diagnostics:
A Policy-Level Summary

In a significant regulatory shift, FDA has proposed reclassifying a cluster of oncology companion diagnostics (CDx) and CDx-adjacent molecular tests—from Class III (PMA) to Class II with special controls. The 14-page notice marks FDA’s first broad structural change to CDx oversight in more than a decade and reflects the agency’s conclusion that these technologies are now mature, well-characterized, and manageable within the 510(k) framework.

Rationale for Reclassification

FDA’s justification rests on two main pillars:

1. A Large, Stable PMA Evidence Base

FDA points to 17 prior PMA approvals and a panel-track supplement covering tests such as NGS panels, PCR hotspot assays, MSI/LOH signatures, and gene-specific mutation assays. Across these submissions, the agency notes:

  • Highly consistent technology (PCR and NGS platforms with known performance boundaries).

  • Longstanding and predictable risk modes.

  • No unique or recurring post-market safety events.

  • Extensive analytic validation data accumulated over more than a decade.

In effect, FDA signals that this body of PMA experience now functions as the predicate scientific foundation for safe 510(k) benchmarking.

2. Technological Maturity and Standardized Methods

The agency emphasizes that these oncology molecular devices no longer present the novelty or uncertainty that once justified PMA status. Their design, specimen types, analytic characteristics, and error profiles are sufficiently well understood that FDA can now define the controls needed to mitigate risks without PMA-level scrutiny.

The agency also implicitly acknowledges that analytic performance expectations for CDx have converged across platforms, making differentiated regulatory treatment (PMA vs. 510(k)) less scientifically justified than it once appeared.

Special Controls Framework

FDA proposes a structured set of special controls that anchor the new Class II classification. These include:

  • Analytic validity requirements: Defined performance characteristics for accuracy, precision, LoD, reportable range, reference materials, and quality system parameters.

  • Clinical validity expectations: A requirement that sponsors demonstrate the test’s ability to identify the biomarker for which the associated therapeutic product is indicated.

  • Design and labeling requirements: Clear statements on intended use, limitations, specimen types, cutoff justification, and the relationship between biomarker detection and associated therapies.

  • Quality system-oriented controls: Ensuring calibration, reagent stability, and software verification meet standards previously demonstrated in PMAs.

  • Post-market controls where appropriate, although FDA explicitly states that historical PMA experience shows minimal adverse trends.

Importantly, these special controls are structured such that substantial equivalence to a well-characterized archetype is possible, eliminating the need for clinical trials for each new test if analytic comparability is demonstrated.

FDA’s Regulatory Method: A Retrospective Meta-Analysis of PMAs

The agency’s method is unusually transparent: it explicitly “looks back” across the entire set of oncology CDx PMAs, extracts shared risk profiles, and evaluates the sufficiency of accumulated scientific knowledge. This retrospective analysis provides the risk-benefit justification required under the statute’s reclassification provisions (including the six-year rule for PMA vintage).

In short, FDA is codifying what many regulators have long concluded informally—that these tests behave like a well-bounded technological family, and that PMA is no longer proportionate to risk.

Summary of the Actual Regulatory Text (Final Pages)

In the final codified section, FDA formally:

  1. Amends the classification regulations to place these oncology CDx molecular assays into Class II.

  2. Defines the device type as tests intended to detect specified genetic alterations for the purpose of selecting patients for targeted therapies.

  3. Lists the special controls, which function as binding performance, design, labeling, and verification requirements necessary for reasonable assurance of safety and effectiveness.

  4. Establishes that future devices in this category will follow the 510(k) pathway, provided they meet the special controls and demonstrate substantial equivalence.

This codified amendment is concise—essentially a legal wrapper around the analytic and risk-based reasoning laid out in the earlier narrative. But its practical impact is substantial: it opens a 510(k) pathway for a major class of oncology molecular diagnostics that, until now, were locked into PMA.

####

####

With this change, how does this future FDA rule and method for CDx compare to the current recently revised European union rule and method for CDx? Would you say this change brings the two regimes into closer alignment, or would you say the alignment WAS close, and now the FDA is pulling away? Asking because of the harmonization thing.

####

Here is a more concise, paragraph-style summary that preserves the key policy and practical points.

Concise Practical Comparison: FDA’s New CDx Pathway vs. IVDR

Under the EU IVDR, all companion diagnostics are locked into Class C, requiring a full technical file, Notified Body review, and a mandatory EMA (or national authority) consultation on the drug–test linkage. In practice, this produces long, multi-step reviews often running 12–24 months, driven partly by Notified Body capacity constraints. Costs are substantial: NB technical-file reviews, EMA consultation fees, and heavy internal preparation. Change control also remains strict; many modifications trigger new formal reviews. IVDR has therefore become a high-friction, high-burden system for CDx, even when the underlying assay technology is mature.

Historically, the FDA PMA pathway for oncology CDx was similarly burdensome: sponsors routinely submitted massive analytic and clinical datasets, endured multi-year PMA cycles, and managed ongoing PMA supplement obligations. The process could be exacting and expensive, with user fees in the hundreds of thousands of dollars and multi-year internal program costs.

The new FDA proposal fundamentally changes this equation for nucleic-acid based oncology CDx. After reviewing 17 PMAs and a decade of experience, FDA concludes that these technologies are now well understood, with risks fully characterizable through special controls. As a result, these CDx will move from Class III (PMA) to Class II (510(k)). The required evidence—robust analytical validation and a clear demonstration of clinical relevance—remains rigorous, but the submission is now structured against a defined regulatory template with narrower debates than PMA. User fees fall dramatically, and review timelines align with the 90-day 510(k) process, even if oncology CDx will receive close scrutiny.

Practically, this means that for follow-on EGFR, BRAF, MSI, LOH, or similar oncology NAAT/NGS CDx, the U.S. will become materially faster, cheaper, and more predictable than Europe. The scientific expectations of FDA and IVDR remain broadly aligned—both still require strong analytical and clinical performance, drug-test linkage, and clear labeling—but the operational burden diverges sharply. IVDR continues to function as a high-cost, multi-agency process; FDA is reclassifying mature CDx technology into a streamlined, mid-tier regulatory track.

The net effect is that scientific harmonization persists, but regulatory workload is no longer harmonized. FDA’s move aligns the evidence standards but pulls away from Europe in efficiency, creating a lighter U.S. pathway for many oncology CDx than the IVDR can presently offer.

###

Sections (iii) and (iv) at the end of the rule seem to require any CLAIMS FOR THE DIAGNOSTIC LABEL to match to SIMILAR CLAIMS ON THE DRUG LABEL.   Let's say I investigate GENE A, which the biopharma sponsor declined to study.  I find 20% of patients have GENE A and they have 0% chance of drug response.  I go to FDA with my data, my claim, and my check for 510K review.  But FDA won't pass my test unless the FDA drug label also discusses GENE A and its link to drug failure.  The Biopharma is disinterested in this labeling update, as it says that it sees some problem with the data

###

Here is a clear, expert-level explanation of the regulatory tension you’ve identified, and why it is not just theoretical—it is built into the structure of the new rule. This can go directly on your blog.

FDA’s New Rule Creates a Required Link Between Test Claims and Drug Labeling — and That Creates a Real-World Stalemate Scenario

The FDA special controls for the new Class II oncology test category (sections (iii) and (iv)) impose a strict alignment requirement:

  • Any clinical claim made by the test about predicted drug benefit, lack of benefit, or risk must map directly to language already present in the FDA-approved drug label.

  • The test manufacturer may not imply new drug-response information that is not already articulated in the drug labeling.

  • The test’s labeling cannot exceed, reinterpret, or contradict the drug’s own efficacy or risk information.

This means, functionally:

The test’s claims about drug response must “foot to” the drug’s label; the drug’s label becomes the limiting reagent for the test’s claims.

This is exactly the issue you are raising in the hypothetical scenario.

The Hypothetical You Gave: Why It Raises a Regulatory Collision

Scenario

You develop a test for GENE A.
You have strong evidence:

  • GENE A → 0% response to Drug ABC

  • Non-GENE A → 50% response

You apply for 510(k) clearance for your GENE A test as a drug-benefit predictor (not essential, but providing clinically meaningful information about drug benefit).

Problem

The Drug ABC label does not discuss GENE A at all.
Drug ABC’s sponsor (Big Pharma) has no interest in introducing such a biomarker, because it narrows the eligible market.
The label is silent.

Under the new FDA rule

Your test claims cannot exceed the information in the drug label.
Since the drug label does not mention GENE A, FDA will not allow you to make any drug-benefit claims—or even cautious “strongly associated” language—about Drug ABC in your test labeling.

In other words:

You cannot get labeling-level recognition of your GENE A → no benefit finding unless the drug manufacturer cooperates.

And the drug manufacturer has no incentive.

Will this create face-offs or stalemates?

In practical terms: yes, and in some cases it will block clearance.

This is the same structural bottleneck that has existed for true companion diagnostics for 15 years, but now it is extended to the newly created “non-essential but drug-informative” tests.

Why it’s a stalemate:

  1. The test manufacturer needs the drug label to say something—anything—about the biomarker.
    Without that, FDA cannot allow drug-response language in the test’s labeling.

  2. The drug manufacturer may strongly prefer silence.
    Adding biomarker-negative non-responders to the drug label shrinks the indicated population, invites payer restrictions, and may trigger postmarket regulatory requirements (analytic validity, postmarketing studies, REMS-like scrutiny).

  3. FDA cannot compel the drug sponsor to update labeling based on an IVD sponsor’s evidence—unless FDA reopens the drug application, which drug sponsors fiercely resist.

  4. Therefore, the IVD sponsor is stuck:
    The test is scientifically actionable but regulatorily ineligible to make the clinically useful claim.

This has happened before under PMA, and now becomes relevant for Class II / 510(k) tests as well.

The deeper regulatory logic: FDA is avoiding the “shadow CDx” problem

FDA’s concern is that a test could effectively redefine a drug’s benefit–risk profile without the drug sponsor ever submitting supportive data.
Hence the rule:

If the test says something about drug benefit or risk,
the drug label must already say it.

This is meant to:

  • Prevent IVDs from implicitly narrowing drug indications

  • Prevent “off-label drug labeling by proxy”

  • Keep drug–IVD biological claims inside the drug’s regulatory authorization

  • Avoid contradictory or clinically destabilizing claims

But it creates the regulatory stalemate you described.

So what will happen in real life?

1. Some scientifically valid tests will be blocked from meaningful labeling.

Tests could still be cleared as purely analytical devices,
but without any mention of drug response.
This neuters commercial utility.

2. Some test sponsors will need to negotiate with pharma.

This becomes a business-development negotiation, not a scientific one.
Pharma may demand:

  • Co-development agreements

  • Exclusivity

  • Shared IP

  • Favorable economics

  • Or simply refuse

3. FDA may see industry pressure to define a path for evidence-based third-party biomarker claims

—but historically FDA has been extremely reluctant to create such a pathway.

4. In oncology, where independent prognostic markers are common, this could chill innovation

unless the test-maker positions the assay as, say:

  • a prognostic test,

  • a stratification biomarker,

  • a therapy-agnostic classifier,
    but not as a drug-benefit predictor, even if scientifically true.

This undermines clinical reality but aligns with FDA’s legal constraints.

Bottom-line sentence for your blog

The new Class II/CDx-adjacent category still hardwires the rule that test claims about drug benefit or risk must map exactly to the drug’s FDA labeling. As a result, test makers can discover real drug-response markers—but unless the drug sponsor cooperates and updates its own label, FDA will not allow the test to make those claims. This creates a structural stalemate in cases where pharma has no incentive to acknowledge biomarker-negative non-responders.

If you want, I can extend this into a short sidebar on historical precedents (KRAS, UGT1A1, CYP2D6, PD-L1) where FDA faced exactly this cross-pressure.

Sidebar: Historical Precedents for the “Labeling Stalemate” Problem

FDA has run into this drug–test alignment conflict before. A few examples illustrate the pattern:

KRAS and anti-EGFR therapy (circa 2008–2012)

Independent academic groups showed that KRAS-mutant colorectal cancer patients had essentially 0% response to cetuximab or panitumumab.
For years, FDA could not allow IVD manufacturers to make this claim because the drug labels had not yet been updated.
Only after strong clinical consensus—and eventually cooperation from drug sponsors—did FDA revise the labels and allow KRAS testing to be formally recognized.

UGT1A1 & irinotecan toxicity

Robust data linked UGT1A1*28 genotype to heightened neutropenia risk.
Many labs developed tests, but IVDs could not claim drug-related risk until the irinotecan label incorporated genotype-specific dosing language.
Some assay submissions were effectively stalled until the drug label caught up.

CYP2D6 & tamoxifen metabolism

By 2009, dozens of studies linked CYP2D6 variants to altered tamoxifen activation.
But FDA never incorporated CYP2D6 into the tamoxifen label, citing inconsistent data, so no IVD could claim predictive information for tamoxifen efficacy even though many clinicians used the information informally.
This remains a canonical example of a real biomarker that FDA kept out of test labeling because it was absent from drug labeling.

PD-L1 assays

PD-L1 is the opposite scenario: the drug sponsors wanted the biomarker in the label.
Because pharma cooperated, multiple PD-L1 assays moved smoothly through PMA with explicit drug-linked claims.
This underscores that drug-sponsor alignment, not analytical science, often determines whether a biomarker becomes “real” in FDA’s regulatory universe.

####

KRAS and anti-EGFR therapy (what actually happened)

The KRAS non-responder finding did not originate from diagnostic companies. It emerged when investigators performed retrospective analyses of banked samples from large cetuximab/panitumumab trials. These academic groups—working outside the IVD industry—showed that KRAS-mutant metastatic colorectal cancer patients had virtually zero likelihood of benefit.

For several years, however, the drug labels did not yet include KRAS, even though the science was increasingly definitive. During that interval, no diagnostic manufacturer could obtain FDA-recognized drug-response labeling, even if their KRAS assay was analytically excellent.

Once EMA and oncology societies (ASCO/ESMO) endorsed KRAS as a true “negative predictor” and the drug sponsors cooperated in revising labeling, FDA updated the cetuximab/panitumumab labels—and only then could IVDs claim KRAS as an anti-EGFR non-responder marker.

KRAS is the archetype of a biomarker discovered academically that remained “invisible” in FDA IVD labeling until the corresponding drug label was amended.

If you’d like, I can similarly refine UGT1A1 and CYP2D6 to emphasize where the science came from (usually academia and cooperative groups, not diagnostics firms) to make the sidebar fully precise.

You said:
The academics would have even less time and money to get a generic KRAS test approved, or pursue the multiple device-drug regulatory changes with small armies of regulatory experts and lawayers.
ChatGPT said:

Exactly — and that point deserves to be made explicitly, because it exposes the structural asymmetry baked into FDA’s “test labeling must match drug labeling” rule.

Here is a clean, polished paragraph you can drop directly into your blog to make this clear:

Why academics cannot fix this problem

In cases like KRAS, UGT1A1, or similar biomarkers discovered through retrospective analyses of banked biospecimens, the scientists who discover the predictive effect have no realistic path to turning that discovery into FDA-recognized test claims. Academic groups lack the multi-million-dollar budgets, regulatory lawyers, full-time submissions teams, and multi-year program management needed to pursue:

  • a stand-alone IVD submission (even now a 510(k) with substantial special-controls burden),

  • a labeling change to a drug they do not own, and

  • simultaneous coordination across CDRH and the drug review center.

Since FDA will not allow a test to make drug-response claims unless the drug’s own labeling speaks to the biomarker, academics are structurally excluded. They cannot initiate drug-label revisions; they cannot file supplements for a biologic they do not sponsor; and they cannot force a pharmaceutical company to acknowledge a biomarker that shrinks its indicated population.

Thus, the very parties who might generate the strongest biomarker evidence (academic cooperative groups) are the least able to convert that evidence into regulated drug-linked test claims.
The system functionally depends on the drug sponsor’s willingness to take on the regulatory burden — and when that willingness is absent, the evidence cannot migrate into FDA-recognized IVD labeling.


###

See a deep-dive sidebar about EU vs FDA

https://brucedocumentblog.blogspot.com/2025/11/cdx-usa-vs-eu-and-changes.html