PathAI Stays in the News, Partners with Roche for CDx

Header: The fast-moving startup PathAI has frequently been in the news.  A new press release discusses CDx collaborations with Roche.

###

One of my most-fun takeaways from Precision Medicine World Conference (PMWC) in January, was feeding an autotranscript of a Roche-PathAI talk into Chat GPT and getting back in return a news article and 10 key takeaways (here).

New news on February 13.  See a joint press release between Roche and PathAI, discussing plans to develop companion diagnostics.

• Roche and PathAI partner for AI digital pathology in diagnostics.
• AI algorithms to enhance Roche's navify Pathology platform globally.
• Collaboration aims to boost precision medicine, and also improves patient treatment access.

One outcome is that AI interpretation can be exactly as accurate and up-to-date wherever the patient and biopsy is, which may not be true of human interpretation.

Coverage at 360Dx and at Medical Device Network.

Roche Digital Pathology

Roche Digital Pathology describes itself this way: Our leadership and expertise in digital pathology, including AI-based image analysis algorithms, enables streamlined integration of diagnostic development with the latest in digital pathology solutions to help proliferate the delivery of crucial therapies to patients.

Pathology lags Radiology

In Radiology, CMS has recently formally recognized standalone and add-on computational services on top of imaging per se, as being a payable service.  See yesterday's "Indirect IDTF" blog here.  We're a little further behind, and waiting for analogies in pathology and lab medicine.  Here in pathology, we're dealing with things like CMS bundling (non payment) of new Cat III dig path codes (Jan 22 blog here).

See my March 2022 blog, triangulating the future of dig pathology by looking at the present and recent past in radiology imaging (here).

__

(The auto-generated AI pic mingles histopathology and MRI)
___

AI CORNER

CHAT GPT 4

Based on the readings provided, here are five MBA-level discussion questions that could stimulate insightful conversations in your biotechnology class:

1. Strategic Partnerships in Precision Medicine: 
1. How do strategic partnerships like the one between Roche and PathAI contribute to the advancement of precision medicine? 
2. Discuss the potential benefits and challenges these collaborations pose for both established pharmaceutical companies and emerging tech startups in the biotechnology industry.
2. AI's Role in Diagnostic Development: 
1. Evaluate the impact of artificial intelligence (AI) on the development of companion diagnostics. 
2. How do AI-powered digital pathology solutions, such as those developed by PathAI for Roche's navify Digital Pathology platform, revolutionize patient treatment access and diagnostic accuracy? 
3. Consider the implications for global health care markets.
3. Regulatory and Ethical Considerations: 
1. Discuss the regulatory and ethical considerations companies must navigate when developing AI-enabled diagnostics for clinical use. 
2. How do partnerships like Roche and PathAI address these challenges, particularly in the context of gaining approval from bodies like the US Food and Drug Administration (FDA)?
4. Market Analysis and Financial Forecasting: 
1. Given the expected growth of the companion diagnostics market, particularly in oncology, to over $3.1 billion by 2030, analyze the financial implications for companies entering this space. 
2. What strategies should companies like Roche and PathAI employ to capture market share and ensure the commercial success of their diagnostics?
5. Innovation vs. Standardization in Digital Pathology: 
1. The transition to digital pathology and AI interpretation represents a significant shift from traditional diagnostic methods. 
2. Discuss the balance between innovation and the need for standardization and quality control in digital pathology. How do collaborations like that between Roche and PathAI influence this balance, and what role do regulatory agencies play in ensuring the reliability and safety of these new technologies?

These questions are designed to explore the strategic, regulatory, ethical, and financial aspects of leveraging AI in developing companion diagnostics and the broader implications for the biotechnology industry and healthcare delivery.

Modernization Efforts at CMS: The "Indirect IDTF" for Software and Analysis

Header:  CMS is trying to modernize some of its payment and benefit pathways, including the creation of "Indirect IDTFs" that don't see patients and perform only software services.  Thanks to Susan Xu for pointing this out.

###

Decades ago, CMS created an entity called IDTF - Independent Diagnostic Testing Facility - that allows enrollment to perform diagnostic tests (like MRI scans) without being a medical entity like a medical (radiology) group practice.  Longstanding regs are at 42 CFR 410.33.

Over the past five or six years, companies with business models like HeartFlow have challenged this model, because they don't see patients on-site, like CMS expected.  Over the years I've put together different parts of the HeartFlow story from adminstrativev law judge (ALJ) records, from annual CMS rulemaking, public comment letters, and even from FOIA (freedom of information) records.   Suffice it to say that HeartFlow "didn't take no for an answer" despite an arcade-like series of roadblocks to overcome.

The "Indirect IDTF"  42 CFR 410.33(g) and PIM Ch 10.2.

One of the biggest testaments to the change was the explicit creation of the "Indirect IDTF" effective 1/1/2022.   To read about this, see three sources:

• Federal regulation 410.33.  Here.
• CMS seems to have not revised the reg here, but put program manual "exemptions" in writing, which "exempt" an "indirect IDTF" from clauses 410.33(c)(2) and (g) (6,8,9).
• Medlearn Matters explanatory bulletin "Independent Diagnostic Testing Facility," 13pp.  Here.
• Program Manual PIM, Ch 10, 10.2.2.4.  Here.

Whazzup?

Indirect IDTFs fit these concepts. Verbatim from CMS:

• Some health care entities have developed or utilize diagnostic tests that do not require such interaction (hereafter occasionally referenced as “indirect IDTFs”). That is, certain IDTFs perform diagnostic services via computer modeling and analytics, or other forms of testing not involving direct beneficiary interaction. 
• The service is often conducted by a technician who undertakes a computer analysis offsite or at another location at which the patient is not present. The physician then reviews the image to determine the appropriate course of action. In short, these entities generally, though not exclusively, have two overriding characteristics. 
• First, the tests they perform do not involvedirect patient interaction, meaning that the test is conducted away from the patient’s physical presence and is non-invasive. 
• Second, the test involves off-site computer modeling and analytics. 

MACs - Limits on freedom to operate

One helpful change restricts the ability of MACs to "make up rules" for IDTFs, including onerous ones.  (Based on documentation, HeartFlow hit one of these speed bumps.)  MACs used to look for state certification of tech staff, or else, MACs would make up there own rules.  Not any more.  CMS states,

If no state licensure requirements for such personnel exist, the contractor need not undertake additional verification activities under § 410.33(c)(2) concerning the technician in question; the contractor shall not establish its own additional certification, credentialing, or similar technician requirements (e.g., federal accreditation) above and beyond the requirements.

AI Corner, Devilish Version: Professor Accused of "AI Writing"

February 5, 2005 (Nature)

Canadian academic Prof. E.M. Wolkovich gives us a disorienting personal essay in Nature, in which a team's paper (self-written) was accused of "AI writing" by a reviewer and an editor.  The paper was rejected, which had to be explained to coauthors.  The essay tallies 980 words.

In a devilish mood, Chat GPT 4 read, rewrote, and abridged the essay in its own words, below, and also provided two original illustrations.  GPT 4 shortened the original down to just 275 words, but the AI kept the first-person voice (or imitated a first-person human voice.)  Noticing that the source is Canadian, GPT 4 threw in a first-person French abridgement as well.  

###

GPT4:

I've recently been accused of committing scientific fraud, but not in the way most might think. It wasn't about data manipulation or fabrication; my lab is diligent about maintaining transparency and reproducibility in our research on how global change affects ecological communities. 

Instead, the accusation was of a different nature altogether: writing fraud, specifically using artificial intelligence to generate text and passing it off as my own work. This was particularly painful for me, as writing has always been a challenging task that I approach with a lot of effort and strategy, including creating outlines and revising extensively.

The accusation came after submitting a paper, with a reviewer and the handling editor suggesting that my work was the product of ChatGPT due to its "unusual" writing style. This was shocking and disheartening, especially since I had not used ChatGPT at all. To defend my integrity, I considered showing my version-controlled writing process on GitHub, comparing the style of my current submission with my previous works, and even questioning ChatGPT about authorship. Yet, this situation highlighted a broader issue with the peer-review process and the potentially corrupting influence of AI in science.

This experience has led me to reflect on the importance of trust and ethical standards in science. The ease with which AI-generated content can be mistaken for genuine human work calls for a discussion within scientific communities about establishing clearer guidelines on the use of AI in research writing. Moving forward, I'm committed to using tools like Git and GitHub from the start for all my writing, ensuring there's a transparent record of my work's progression, aiming to safeguard against such accusations in the future.

##

 

The black and white, film noir style photorealistic illustration is now available, offering a dramatic and visually intriguing perspective on the scientist's predicament. This image captures the essence of the narrative with its classic film noir aesthetic, emphasizing the tension and resolve of the scientist amidst the accusations.

GPT4:  Le robot continue d'écrire, maintenant dans une version française.

Je viens récemment d'être accusé de fraude scientifique, mais pas de la manière dont la plupart pourraient le penser. Il ne s'agissait pas de manipulation de données ou de fabrication ; mon laboratoire fait preuve de diligence pour maintenir la transparence et la reproductibilité de nos recherches sur la manière dont le changement global affecte les communautés écologiques. 

Au lieu de cela, l'accusation était d'une nature totalement différente : une fraude rédactionnelle, en utilisant spécifiquement l'intelligence artificielle pour générer du texte et le présenter comme étant mon propre travail. Cela a été particulièrement douloureux pour moi, car écrire a toujours été une tâche difficile que j'aborde avec beaucoup d'effort et de stratégie, y compris la création de plans et une révision extensive.

L'accusation est survenue après la soumission d'un article, avec un examinateur et l'éditeur responsable suggérant que mon travail était le produit de ChatGPT en raison de son style d'écriture "inhabituel". Cela a été choquant et décourageant, surtout que je n'avais pas du tout utilisé ChatGPT. Pour défendre mon intégrité, j'ai envisagé de montrer mon processus d'écriture contrôlé par version sur GitHub, de comparer le style de ma soumission actuelle avec mes travaux précédents et même d'interroger ChatGPT sur la paternité. Pourtant, cette situation a mis en lumière un problème plus large avec le processus d'évaluation par les pairs et l'influence potentiellement corruptrice de l'IA dans la science.

Cette expérience m'a amené à réfléchir à l'importance de la confiance et des normes éthiques dans la science. La facilité avec laquelle le contenu généré par l'IA peut être confondu avec un travail humain authentique appelle à une discussion au sein des communautés scientifiques sur l'établissement de directives plus claires sur l'utilisation de l'IA dans la rédaction de recherches. À l'avenir, je m'engage à utiliser des outils comme Git et GitHub dès le départ pour toute mon écriture, assurant ainsi un enregistrement transparent de la progression de mon travail, dans le but de me protéger contre de telles accusations à l'avenir.

The illustration captures a whimsical and slightly devilish scene where ChatGPT 4, personified as a clever robot, rewrites and abridges an essay with a mischievous grin. Surrounded by books and papers in a cozy study, the scene combines creativity, wit, and a playful rebellion against traditional norms, highlighting the AI's role in authorship and translation, including a nod to its Canadian source material.

 ##  Le robot prend le rôle du journaliste.

News-Style Summary.  In a surprising turn of events, reported in Nature, a scientist studying ecological communities has been accused of using artificial intelligence, specifically ChatGPT, to write a research paper. The accusation, which focused on the "unusual" writing style, was made despite the scientist's commitment to data transparency and reproducibility. Defending their integrity, they highlighted their meticulous writing process and version-controlled documentation on GitHub. This incident has sparked a broader discussion on the impact of AI on scientific integrity and the need for clearer guidelines on AI's role in research writing.

##

/// GPT4.

(Lewis Black version).

Must-See CMS Guidance: How Medicare Advantage Makes Coverage Decisions

 Header:  This week, HHS issued a guidance letter on how Medicare Advantage plans can make coverage decisions.  One aspect of the letter is "AI," but there are also general rules.  Anyone working in Medicare reimbursement will want to review all 14 pages.  

See also a Hill hearing on AI and Medicare Advantage.

##

CMS issued guidance to Medicare Advantage programs on how they can make claims-processing and policy coverage decision.   While this garnered some headlines due to the paragraphs around uses of AI in Denials, it's also a broader document that cites to regulations created about a year ago.

• See  42 CFR 422.101.
• Guidance https://cdn.arstechnica.net/wp-content/uploads/2024/02/cms-memo-2624-faqs-related-to-coverage-criteria-and-utilization-management-requirements-in-cms-final-rule-cms-4201-f.pdf
• News article.  https://arstechnica.com/science/2024/02/ai-cannot-be-used-to-deny-health-care-coverage-feds-clarify-to-insurers/
• News articles at STAT (2023, 2024):
• https://www.statnews.com/2023/03/13/medicare-advantage-plans-denial-artificial-intelligence/
• https://www.statnews.com/2024/02/07/government-warns-medicare-advantage-insurers-not-to-deny-care-based-on-ai/
• See John Warren's essay on the CMS letter, here.
• CMS is also running a Medicare Advantage request for information, active to May 24, 2024, here.  CMS had previously run an RFI on MA in 2022.

This the regulation is rather dense (422.101), this 14 page PDF from CMS may be easier reading and more understandable, and includes clarifications.  

Types of Stuff in the Letter

The guidance in the 14 page letter includes how MA plans can decide internal coverage policies, how much need be public, and how prior auth is applied.

Regarding guidelines, we read: "MA organizations may not add coverage criteria that are not supported in such guidelines or literature, or change the substantive recommendations contained in such guidelines or literature to support coverage criteria."  Issues such as reopening an already-approved prior auth benefit, are also discussed in detail.

Nerd note.

Some of this used to be program manual text (and may still be), and the policy was "upgraded" to the force of a regulation last year.   

I noticed that the program manual text referred to following NCDs, LCDs, and program manual instructions.  When they converted this to regulations, they refer only to NCDs and LCDs.  I assume the lawyers drafting the regulation did not want to imply that program manual instructions had the force of law (since they appear suddenly, lack public comment, etc).  

The program manual text also had a clarification [Manual 100-16_4-90.4.1] that if one local LCD had the equivalent of national jurisdiction (for a sole source lab that serves patients in all states), then that LCD applied in all 50 states.  That type of subtlety was left out of the more concise  regulatory language.  (42 CFR 422.101(b) 1,2,3, where 1 = NCDs, 2 = laws, 3 = written LCDs.)

click to enlarge

AI Corner

Here is a GPT4 summary and 10 takeaways.

The memorandum outlines new coverage criteria and utilization management requirements for Medicare Advantage (MA) plans, effective from January 1, 2024. It addresses questions on medical necessity determinations, the use of algorithms and AI in coverage decisions, public accessibility of internal coverage criteria, enforcement of new rules, and the impact on supplemental benefits. MA plans must align with Traditional Medicare's established coverage criteria, ensure transparency and non-discrimination in coverage determinations, and comply with specific requirements for prior authorization and utilization management to ensure beneficiaries' access to necessary care without undue delay.

1. Align with Traditional Medicare: MA plans must ensure their coverage criteria align with those of Traditional Medicare, particularly regarding medical necessity determinations.
2. Transparency in Coverage Criteria: Plans are required to make their internal coverage criteria publicly accessible, ensuring transparency.
3. Prior Authorization Reforms: The memo emphasizes the need for MA plans to streamline prior authorization processes to reduce delays in patient care.
4. Utilization Management Oversight: MA plans must annually review their utilization management tools, including algorithms and AI, to ensure they are evidence-based and not discriminatorily applied.
5. Non-Discrimination in Coverage Decisions: Plans must avoid using discriminatory practices in coverage decisions, ensuring equal access to necessary care for all beneficiaries.
6. Evidence-Based Decision Making: Coverage determinations should be based on the latest clinical evidence and guidelines.
7. Public Accessibility of Information: MA plans must provide easy access to information about coverage determinations and available appeals processes.
8. Annual Reporting Requirements: Plans are required to report their utilization management practices and outcomes to CMS, promoting accountability.
9. Impact on Supplemental Benefits: The memo clarifies how the new rules affect supplemental benefits, ensuring they complement but do not replace mandatory benefits.
10. Enforcement and Compliance: MA plans must be prepared for CMS oversight and enforcement of these new rules to ensure compliance.

These takeaways highlight the emphasis on transparency, evidence-based care, and non-discrimination, aiming to improve patient access and outcomes in the Medicare Advantage program.

Genomeweb Article Features Diagnostic Alzheimer Blood Tests

 Header:  See a detailed review of rapidly emerging Alzheimer blood tess, especially pTau217.

##

I did an MD-PhD with the intention of working in Alzheimer's research, and neuro disease was my focus from 1988 to 2001.   Even in the 90's, there were regularly promises of a new Alzheimer blood test, which never panned out.

So I was excited to see a feature article in 360Dx by Adam Bonislawski about the rapid publications and advances in Alzheimer blood tests.   Both plasma ABeta 42/40 ratio and Tau epitopes are promising.  The article focuses on large new clinical reports that highligh phopho-Tau 217.   

Bonislawsky writes, 

• "Plasma phosphorylated-tau 217 (p-tau 217) has emerged as a highly promising marker of the brain amyloid pathology characteristic of Alzheimer's and one that could lessen demand for plasma amyloid-beta 42/40 (Aβ42/Aβ40), the first blood-based marker for the disease to come to market."

New papers include Yu et al., Plasma p-tau181 and p-tau217 in discriminating PART, AD and other key neuropathologies in older adults.   And Ashton et al., Diagnostic ccuracy of a Plasma Phosphorylated Tau 217 Immunoassay for Alzheimer Disease pathology.

Twists and Ins and Outs

I think a couple factors have made the space hard to summarize.  First, as these titles indicate, it's not just "tau" or even "phospho-tau" but specific epitopes may behave differently.   

Then, there is not one single "phospho tau 217" assay.   Most assays require two antibodies, a  capture and a detection, and those will vary among platforms.   

Then, there are different detection platforms (e.g. the Quanterix Simoa technology is a differentiated technology).  

Finally, even when those platform variables are settled, there are many ins and outs to defining sensitivity and specificity.  Is there a gold standard besides autopsy?   What do decimal-point clinical statistics mean, if the clinical gold standard is wrong 20% or 30% of the time?    

And what populations does a statistic pertain to?   Health controls versus advanced patients?  That probably gives the best statistics but it compares extremes that don't match a clinical question (which is in the  early symptoms patient.)  So if someone pronounces  a factoid like, "Plasma tau is 81.5% specific" you really don't know what that means without more facts.

NGS MAC Proposes Coverage or RENALYTIX KidneyIntelX and X.dkd Testing

 It's a bit of a cliche' among Medicare genomics experts that the vast majority of LCDs and articles on these topics are found in the MolDx MACs.

Here's a proprietary test with a proposed (somewhat complex) coverage status at NGS MAC.   The test is Renalytix 'KidneyIntelX" and "KidneyIntelX.dkd."  The .dkd refers to FDA nomenclature (analogous to biosimilars) for the FDA-approved version.  

I had heard at one time Renalytix would bill via Utah, a MolDx state, but the cover letter, 8 pages long, clearly lists the lab as a New York lab.  That would mean it has both NYS approval and FDA too.  FDA de novo clearance goes back to June 2023.   

The 8-page request letter provides a model to other labs, of a request letter that was successful.   The request letter does not have a printed date but the internal PDF date is October 2023, which would be consistent with a several-month review at NGS MAC.  The LCD links to a August 2023 public meeting (CAC meeting) but the link is a dead link for me.  However, Google gave me a 14-page transcript of the meeting.

The evidence review cites 9 papers, one paragraph of detailed summary for each.   This should be a heads-up to labs that plan to do one single 60 patient study of their novel test "and get coverage for 50 million Medicare patients."  See FDA label.

Business

Renalytix went public at $15 in July 2020, and reached $30 during the COVID biotech bubble.  Recently it trades at about 25 cents.   Renalytix raised $74M in 2020 and has a current market cap of $14M   Annual revenue has run circa $3M while expenses per year have been circa $40M.  Investor call here.

See my white paper on valuation of genomics companies, here. 

Coding

Code 0105U is $950.  If I am reading Part B 2022 data correctly there was 1 paid claim in 2022. 

LCD

Here's the proposed wording. Comment is open for 45 days.  DL39726.  I think the key points are patient has T2DM and eGFR 30-60 range for 1 lifetime test.

https://www.cms.gov/medicare-coverage-database/view/lcd.aspx?lcdid=39725&ver=15&stateRegion=all&contractorNumber=all&proposedStatus=C&sortBy=commentStart&bc=11

Once in a lifetime KidneylntelX or KidneyIntelX.dkd test is considered reasonable and necessary when all the following criteria are met:

• The results are used to facilitate therapeutic prognostic decision-making in the medical management of a selected patient population, and
• The results are used to assess the risk of progressive decline in kidney function in patients over the age of 21 years of age with:
• Type 2 diabetes (T2D) and existing early-stage chronic kidney disease (CKD) (stages 1-3b), and
• The test is ordered by the treating physician or qualified non-physician practitioner, and
• The test is performed in a CLIA certified laboratory qualified to perform high complexity testing, and
• The specific reason for the test must be documented by the treating practitioner in the medical documentation and demonstrate that the test is medically reasonable and necessary.

Note: Kidney function decline is defined as:

• a decline in eGFR slope of ≥: 5 ml/min/l.73m 2 /year; or
• a sustained decrease in eGFR ≥40% confirmed at least 3 months apart; or
• kidney failure, defined by sustained eGFR <15.

Limitations of Coverage

• KidneylntelX or KidneyIntelX.dkd test is not medically reasonable and necessary for:
• Patients with eGFR <30
• Patients with eGFR ≥60 ml/min/l.73m 2 without albuminuria
• Patients with ESRD or on renal recovery treatments
• Patients who are pregnant
• Patients who are currently hospitalized
• Patients taking Etanercept
• KidneylntelX or KidneyIntelX.dkd is not covered as a screening or standalone diagnostic.

Heads Up: AACR Project GENIE - Real World Evidence

 This week, I had the chance to hear for the first time about PROJECT GENIE at AACR (American Alliance for Cancer Research).  It's another major effort towards better and more complete real world cancer data.

Find it here:
https://www.aacr.org/professionals/research/aacr-project-genie/

Here's an abridged version of their one-pager:

Precision medicine, particularly in oncology, tailors treatments based on the detailed genetic makeup of both patients and their tumors. This approach relies on a comprehensive health care system that learns from each patient's experience. However, the challenge lies in gathering enough data from individual institutions to significantly impact clinical decisions. 

To address this, the AACR initiated Project GENIE, a global cancer registry that combines clinico-genomic data from 19 top cancer centers worldwide. 

Managed with the help of Sage Bionetworks and cBioPortal, Project GENIE links cancer genomic data with patient outcomes, fostering a collaborative effort to advance cancer research. This initiative not only speeds up drug development and refines clinical trials but also aims to improve outcomes for cancer patients worldwide through its commitment to openness and data accessibility.

 

Conference Watch: Biotech in Individualized Medicine @ Scripps @ May 14,15 2024

Follow the Scripps announcements for "Future of Biotechnology in Individualized Medicine," on May 14-15, 2024 at Scripps.   Nonprofit to regular registrations run from about $195 to $595.

Find a conference home page here.

Click a small button at top right for agenda.

Here's an AI digest of the 1.5 day agenda:

This year's Scripps Biotech conference focuses on the future of individualized medicine, the utilization of nation-scale biobanks for genetic association, genomically informed drug discovery, the broader role of genomics in biotechnology, and discussions on a mesa-wide genomics initiative. 

The meeting includes panel discussions on technology's role in translational research, mining data for disease biology and therapeutic targets, and regulatory and commercial landscapes to support future biotech and genomics endeavors. 

Speakers from Scripps Research, National Institutes of Health, and other leading institutions contribute to a rich dialogue on advancing health policy, biotech, and genomics.

CPT Knowledge Base at AMA

At the just-completed AMA Editorial Panel in San Diego, as in all first-quarter meetings, there was an extensive presentation of AMA initiatives and resources.

One that I hadn't seen was the AMA CPT Knowledge Base.  It's here:

https://cptkb.ama-assn.org/search

This is a database of thousands of coding Q&A.  It requires an AMA logon, which is free. (Whether it also checks if you are an AMA member, I don't know.)

I give a screen shot below and an example where I queried for the miscellaneous molecular code 81479.

click to enlarge

 

click to enlarge

MolDxOlogy: Nodify CDT Test Covered in Press Release; Not Covered in "DEX"

Recently, MolDx published an article that it would make coverage and non-coverage decisions on a certain code set of proteomic tests.  This comes in addition to MolDx's longstanding responsibility for DNA-RNA tests.   See my blog here.   See the MolDx article and code list here.

Some of these  codes step into a complex history.  

Outside of DEX: Medicare and XL2, Medicare and Nodify

For example, there was originally a MolDx LCD on the Biodesix BDX XL2 lung cancer proteomic test; iin the 2021-2022  time period, MolDx stripped the "MolDx" title word off the LCD, but kept the LCD in place.  On the other hand, to my knowledge, there was never an LCD on the Biodesix Nodify test, although there is a press release stating it is covered by Medicare,

https://www.biodesix.com/newsroom/press-releases/biodesix-obtains-medicare-coverage-for-nodify-cdt-lung-nodule-test

Inside of DEX:  XL2 and Nodify

A prominent posting on the MolDx home page refers readers to the DEX Registry for coverage status (here). Let's look up the Nodify test.  Going to the DEX registry, looking up Biodesix, and  selecting the subfield for its Kansas center, the Nodify test is listed, but listed as non covered.  

click to enlarge

This could be a typo; while the MolDx home page article sends readers to DEX to learn coverage status,  the footer on the DEX itself says that this site should not be relied on for coverage status.  It's just one of the complexities I file under "MolDxology."  

Let's try DEX and the XL2 test.  If you go to DEX and click on Biodesix/Seattle, the XL2 test is listed as, "Coverage has not been set."  But we already know, there is an LCD at Noridian (also at Palmetto here) covering XL2.    

click to enlarge

Conclusion? It make take a while for all the connections between local MAC coverage, and the MolDx coverage, and the Proteomics article, and the DEX registry, to sort out.

Another Puzzle - Proteomics Article and Proteomics-on-slides

Another entry point to multiplex proteomics is on-slide immunohistochemistry plus techniques like ML or AI.   I spot checked five or six slide+algorithm codes: just one of them was included on the MolDx article for Proteomics codes that now fall under MolDx.

From my proteomics article blog.

Proteomics on slides.  What about multiplex proteomics on slides - immunohistochemistry plus AI?   I have a quick list of 5 codes that are slides, often with IHC, + imaging + algorithms (ML or AI?).  These are 0220U, 0295U, 0376U, 0414U, 0418U.   (See pricing here).   MolDx seems to have included exactly 1 of these 5 codes in the Proteomics Article.  Why 0295U is included as "proteomics" and some of the others not, I haven't determined yet.

##

Copy of MolDx's "refer to Dex for coverage" -- here.

Remarkable Pushback Against FDA LDT Regulation: The Hyman Phelps Law Firm Comment

HEADER.  I was very impressed by a 58-page FDA LDT comment, submitted by Hyman Phelps law firm and an LDT Coalition. 

###

The biggest splash I know of, among FDA LDT comments in December, was that of the ACLA, which was some 100 pages when including an extensive supplement in which an economist took apart the FDA's proposed financials.  ACLA entry point here.

Here's another grand example. it comes from  Hyman Phelps McNamara and the Coalition to Preserve LDT Access and Innovation. Find it here:

https://www.thefdalawblog.com/wp-content/uploads/2023/12/LDT-Coalition-Comment-12-4-2023.pdf

Topics include:

• Prohibitive costs of the rule
• Poor presentation of LDT risks and no presentation of LDT benefits
• Deeply flawed economic analyses
• Multiple categories of costs ignored and underestimated
• Existing regulatory framworks are ample
• FDA lacks statutory authority
• Elaborate statutory discussion

The letter makes a point I have made, the FDA relies on risk categories by use case and indication, but CLIA tests don't have [in the same sense] statements of indicated use, which at FDA are often verbose, multiplex, and hammered out after months of negotiation (p. 33, 38).   

##

See also the firm's Feb 2 essay, on the holes in the FDA's Jan 31 press release about IVD reclassification. 

Novitas & NGS MAC Publish Web Info for 2024 Gapfill Process

I wrote a blog a few weeks ago on the pending, 2024 Medicare gapfill lab pricing process (here).

In a typical year, around Feb 1, Novitas-FCSO post a web portal for info, and NGS MAC posts an email info address and deadline.   (MolDx doesn't post anything, but I always tell labs, if you're in gapfill, and MolDx can't find you and contact you by Feb 10, you reach out to them with your contact info).

NOVITAS has posted its online process for 2024 gapfill info.  Find it here:

https://www.novitas-solutions.com/webcenter/portal/MedicareJL/pagebyid?contentId=00246106

The 29 codes under 2024 national gapfill pricing are:

• 81457, 81458, 81459, 81462, 81463, 81464, 87467 
• 0019M, 0329U, 0334U, 0356U, 0361U, 0364U, 0368U, 0379U, 0381U, 0382U, 0383U, 0388U, 0389U, 0395U, 0396U, 0398U, 0399U, 0400U, 0401U, 0408U, 0412U, and 0417U.

Novitas Deadline is March 15, 2024.    

Of special interest to all molecular pathology labs will be the MolDx pricing info and decisions for new AMA CPT codes for liquid biopsy and other types of tumor genomics.

##

NGS MAC also publishes its rules, which are submission via an email box by March 1, 2024.

https://www.ngsmedicare.com/web/ngs/search-details?selectedArticleId=5270181&lob=96664&state=97057&rgion=93623

Nerd Note

Because CMS sets a median price [essentially] by state, and MolDx has a majority of states, the MolDx gapfill price is always "the" permanent gapfill price when the process concludes.

Nerd Note and AI Pic

I wrote a 2020 white paper on the in's, out's, and ambiguities of the gapfill process. Nothing has changed since then; it's equally current for 2024.

https://drive.google.com/file/d/1eDVJ2mvpQAUlZ-NY6YJwSyXZJsbigyus/view

Click to enlarge

FOCR Launches Digital Pathology Harmonization Project

Friends of Cancer Research (FOCR) launches a major new initiative called PATH - Digital and Computational Pathology Tool Harmonization Project - Digital PATH.   (Harmonization projects of diverse types have been a significant focus of FOCR).

See the home page here, which links to white papers and one pagers.  Harmonization of dig path (AI?) readings of Her2 slides are an early target.  Here's their 27pp dig path white paper from September 2023.

For the dig path home page:

https://friendsofcancerresearch.org/digital-pathology/

They write, The Digital and Computational Pathology Tool Harmonization (Digital PATH) Project, led by Friends of Cancer Research (Friends), supports robust development of digital and computational pathology platforms for use in oncology drug development and is currently assessing variability in biomarker measurements across platform developers to encourage alignment. Stay up to date with #AIOncology.

I've clipped part of the press release below:

Washington, DC – February 1st, 2024 - Friends of Cancer Research (Friends) is excited to announce the launch of a new research partnership, Digital and Computational Pathology Tool Harmonization (Digital PATH) Project. This project will identify factors that may contribute to variability in biomarker assessment across computational pathology platforms, propose areas for alignment in the field, and provide insights for shaping regulatory processes. 

The igital PATH Project is guided by extensive collaboration and aims to enhance the use of these computational digital pathology platforms and improve patient care.  

“As artificial intelligence tools increasingly become part of healthcare, their use in digital and computational pathology open remarkable opportunities for enhanced diagnostic insights,” said Jeff Allen, President & CEO of Friends. “This new collaborative research partnership will help inform future policies toward optimal regulation and utilization of these advanced technologies.” 

During this research partnership, human epidermal growth factor receptor 2 (HER2) computational pathology algorithm developers will collect and analyze data to assess concordance of HER2 measurement from breast cancer samples across platforms and identify factors contributing to variability. HER2 is a biomarker with novel opportunities in breast cancer treatment and expanded relevance in other cancer types with the emergence of a new class of drugs called antibody drug conjugates. Samples are being provided by collaborators at the Department of Pathology, ZAS Hospitals in Antwerp, Belgium.

“What is currently largely unknown and crucial for incorporation of AI-tools in our trial- and daily-practices is the comparability and performance of different AI-platforms on the same set of slides," said Roberto Salgado, Breast Pathologist, ZAS Hospitals. "This study will provide important insights to the community, including regulatory bodies and industry, about the capabilities of AI tools and their application in patient care, ultimately aiming to ensure the best possible care and outcomes for patients.”

The Digital PATH Project will evaluate the implementation of several proposals included in our 2023 white paper released at the public meeting, Future in Focus: Digital Pathology in Oncology Drug Development. 

To learn more about the project, please visit: https://friendsofcancerresearch.org/digital-pathology/ 

Project Partners 

Digital PATH Project Partners: Friends is proud to partner with 4D Path Inc., Amgen, AstraZeneca, Bristol Myers Squibb, EMD Serono, Inc., the U.S. Food and Drug Administration (FDA), GSK,  Indica Labs, Johnson and Johnson Innovative Medicine, Loxo@Lilly, Lunit, Massachusetts General Hospital, MD Anderson Cancer Center, Merck and Co., National Cancer Institute (NCI), Nucleai, PathAI, Patient Advocates, Roche Diagnostics, Sanofi, Tempus AI, Inc., University of North Carolina, Verily, and ZAS Hospitals Antwerp.

FDA Press Release: An Earthquake for FDA Diagnostics Classification?

On January 31, 2024, FDA issued a major press release about its plan to make a large-scale reclassification of diagnostics by risk categories.  Oncology and companion diagnostics are especially implicated.  These will be discussed a lot in the coming week.  

Dr. Shuren, head of CDRH, highlighted this in a speech on February 1 here.  Coverage at Genomeweb here.

Press Release

https://www.fda.gov/medical-devices/medical-devices-news-and-events/cdrh-announces-intent-initiate-reclassification-process-most-high-risk-ivds

Coverage at Genomeweb here.  Law article.  Excellent detailed discussion at FOCR.

To paraphrase the main ideas in the press release:

CDRH Announces Intent to Initiate 

the Reclassification Process for Most High Risk IVDs

Jeff Shuren, M.D., J.D., director of the FDA's CDRH, announced plans to reclassify most Class III (high risk) in vitro diagnostic devices (IVDs) to Class II (moderate risk), primarily affecting infectious disease and companion diagnostics. This move allows for easier marketing clearance via the 510(k) notification pathway, fostering more competition and access to these tests. The reclassification is based on establishing special controls to ensure safety and effectiveness. The process began with a September 2023 panel meeting discussing the reclassification of specific infectious disease IVDs. CDRH will continue its risk-based approach for new IVD classifications and periodically review device classifications to ensure appropriate regulatory controls. {GPT4}

##

AI Corner.

The picture is auto generated from the plain blog text [GPT4/Dalle].

Jeff Shuren's Keynote on Diagnostics: Friends of Cancer Research Conference

On February 1, 2024, Friends of Cancer Research ran a half-day workshop on diagnostics, regulation, and advances.  Find it here, "The Future of Diagnostic Tests:"

https://friendsofcancerresearch.org/event/the-future-of-diagnostic-tests/

SHUREN SPEECH ON LDT

The opening keynote was by FDA's head for devices, Jeff Shuren MD JD.  

• Read about Shuren Speech in my bullet points. 
• I've taken live bullet point notes and post them below.  "Quinn notes on Shuren Speech."  
• Read about Shuren Speech in an "AI Article." 
• Instead of my own bullet points below, oyou can read them "converted" into an article by AI  here.  

CALIFF, FDA COMMISIONER:  SPEECH ON LDT

See also a Dr Califf speech yesterday that FDA is moving full speed ahead with LDT regulation.

https://www.medtechdive.com/news/fda-califf-backs-ldt-proposed-rule/706286

FDA DOWNREGULATION OF RISK

Dr Shuren highlighted a MAJOR posting yesterday about the intent to reclassify risk for many diagnostic tests:

https://www.fda.gov/medical-devices/medical-devices-news-and-events/cdrh-announces-intent-initiate-reclassification-process-most-high-risk-ivds

Jeff Shuren FDA

My comments are limited due to rulemaking.

We have authorities over all tests, whoever makes them.

50 years ago we deferred regulation of LDTs, simple locally produced tests.

We have now seen LDTs that are not clinically valid, across diseases and in oncology.

We’ve also seen impacts on innovation, hurting the incentives for some IVD diagnostics since laboratories copy them, often with LDT claims of superiority.  This disenfranchises the IVD, and that has to change.

With reports of test variability, your cancer treatments depends more on the lab you use, not your cancer biology.

A decade ago we had a series of guidances.  We got feedback in favor of a modern legislative framework and we worked on that for 7 years (Biblical!).   The path for legislatiion is unclear.

To move forward for patients, we did a regulation last falll.  And we did an economic analysis, we didn’t have to.  In the rule itself, it phases out enforecement discretion.  It leaves discretion for some tests, and asks how broad any discretion should be.  We asked about “tests on the market,” and “emerging pathogens” and ‘small entities” and a longer phase out period.  AMC’s were uniquely concerned, in their terms. And they provide “integrated healthcare for unmet needs.”  And some programs like NYS DOH provide pre market review.  We got lots of comment, 6500 but many “form letters.”  From a wide range of stakeholders, with perspectives from A to Z.

We are moving forward to finalize the regulation.

We hav a challenge with tests used to inform drug treatment where there is or will be a CDx.  We have a pilot program for drug companies and labs to use, providing info on LDTs that match the pivotal study (”LDT”) tests used in the drug trial.  That is an interim step. We will evaluate in June 2026.  The pilot hasn’t  actually launched for any company.    It requires a collaborating pharma and a colloborating FDA lab.

Yesterday, we announced our intent to reclassify MOST high risk tests to what we call “moderate risk tests” with appropriate mitigations.   This streamlined approach will stimulate innovation and these will often be for infectious disease and companion diagnostic.  Other tests may also move from high to moderate risk.   This would involve rulemaking and advisory panels.

We also have the TPLC Total Product Life Cycle advisory program. TPLC AP or TAP.   There are many hurdles in device development, of which FDA is just one step.  In Med Tech, developers don’t fully account for all the aspects of evidence strategy, even sophisticated players.   And include the voice of patients.

These include the needs of payers, years, for innovative technology.  We want to ensure interest in investment, with TPLC TAP.  You can meet with us and bring your questions.   We did this fluidly with EUAs in COVID, more fluid engagements.   You might even talk several times per week to FDA.

We’ll help connect you to patients, provider, payer groups.   The pilot is launched, and we strated with cardiovascular, and then added neurologic and rehab.  It will in time include diagnostics.  Feedback “very very positive.”   We want to improve predictability and reduce the cost of “valley of death” in R&D.

Finally.  Health equity is a strategic priority.   We want to address gaps across populations.  CDRH focuses on “access.”   There are many intrinsic challenges - fragmented, costly health, with innovation not  reaching patiens, we pay more and live less long.   We can’t focus just on “care” for the sick, and just “brick and mortar.”  We need prevention and movable care from institutions to people at home.

This isn’t just “dumping” tech from hopsital to home, that’s “jerry rigged.”   It is unmanageable that way, it has to be re-thought and fit for purpose.

And as we move these technologies to remote locations, we also have great opportunities to gather data.  See our “Docket” on Home Use devices.  This is a priority for our center, see also the CDRH DHCE Digital Health Center of Excellence.   We’ve authorized 650 AI or ML devices.

These affect imaging, diagnosis, and eventually devices in treatment.   It won’t be CT scan at home, but may be AI guided ultrasound.   Design the technology for people, and consider consumer technology that is “medical grade.”  

Home use tests  - IVDs.  In COVID, the US now is comfortable with OTC tests.  They existed before COVID, but witih lots of concerns (outside of home pregnancy)   COVID has changed that, and we see more OTC tests coming to use, from infection to drugs of abuse, but they need design for home users.

COVID tests were “fine” but what about someone with arthritis (fine packaging and squeeze drops).  We would like to see oncology tests in the home, too, finding cancer early and making a huge difference and with wide access including home.

Home use tests for “clinical trial in a box.”

Partners.  If you want to go far, do it together.   We have collaborative communities, or CC.   USCDI+ Cancer Inititive.   Cancer Moonshot.  Cancer X.  Public meetings (such as on skin lesion technologies).   If we are a representative government, we need this stakeholder participation (not just AMCs, but patients).  We don’t run a “CC” collaborative community, we listen and adapt from those.

There’s a Pathology Innovation CC and a Digital Measurement CC, and we have about 15 of these.

Q&A

CLIA and NY State.

CLIA and NYS.   Jeff seems to say that half of NYS tests don’t pass on first submission.   We’ve had that with LDTs too.   We don’t question that LDT has some role in healthcare, but must be accurate and reliable.   There are similarities between us and NYS.

Mark Fleury ACS CAN.  The oncology pilot?  Performance standard for LDT, and cases where no IVD CDX is around yet.   Could that tie to enforcement discretion?   Shuren:  A lot went into that announcement.  The actions we take will take time (years) and we view the pilot for those as an interim step, but as I said it’s “in neutral” gear right now.