Friday, September 29, 2023

Sidebar 1: FDA Legal Justification for LDT Regulation

On September 29, 2023, FDA released proposed regulations that would allow it to regulate lab developed tests.   See my discussion in a previous blog:

https://www.discoveriesinhealthpolicy.com/2023/09/fda-releases-proposed-regulation-to.html

The FDA provides a 5000-word legal justification (Part IV, legal justification, is only a paragraph to restate statutory authorities, but other parts of III and V have extensive legal expositions, often with numerous court cases cited).  

  • But recall that there are other arguments FDA just can't do this - See Genzen; see Clement and Tribe.
  • Recall that the Trump administration forbid FDA to regulate LDTs ...and the Trump administration could be back.  

For one exampole of FDA's legal adventures (p 43), FDA says that "few have argued" that a "modern" interpretation of the interstate commerce clause allows the FDA to regulate purely intra-state activities. Then FDA embarks on an affirmative discussion of the commerce clause in deference to those few naysayers.  FDA says the FDA act as a whole doesn't require interstate commerce explicitly, but only a "constitutional minimum" applies to most of it. FDA Says are very few, and specific points, where the FDA act remarks on interstate commerce "elements," but these are generally not germane to devices.

Some of the FDA's arguments depend on precedents that are subject to change at the Supreme Court.  For example, yes, FDA writes that few doubt that interstate commerce can include FDA's authority to regulate solely intra-state activities.  But a near-future SCOTUS might say that "interstate commerce" means what it would have meant to Ben Franklin, something put in a crate and shipped  over a state line.  And while FDA states that "interstate commerce" actually has nothing to do with its power to regulate LDTs, this seems a little shakier when the FDA Act itself repeatedly refers to, "interstate commerce."     

With that said...

 I am not an attorney, and neither is Chat GPT, but here is a GPT summary of the legal section.

___________________________

The FDA's legal justifications for its lab test policy can be analyzed in detail as follows:

IV.  Regulatory Framework for Lab-Developed Tests (LDTs):

The FDA argues that LDTs are subject to its regulatory authority based on the existing legal framework. The key points supporting this argument include:

ONE

Existing Law

a. Statutory Authority: The FDA asserts that its authority to regulate LDTs is derived from the Federal Food, Drug, and Cosmetic Act (FD&C Act), which grants the agency jurisdiction over medical devices. It contends that LDTs qualify as medical devices, as they are designed and used for medical purposes, and therefore, fall within the scope of FDA regulation.

b. Historical Perspective: The FDA highlights that it has exercised jurisdiction over in vitro diagnostic (IVD) devices, including LDTs, for many decades. This historical practice is seen as consistent with the agency's interpretation of its statutory authority.

TWO

Functional Equivalence:

The FDA argues that LDTs are functionally equivalent to commercial IVDs. This equivalence is a key aspect of the FDA's justification for regulating LDTs. The main points include:

a. Similar Technology: The FDA emphasizes that LDTs often employ the same technology, instruments, and software as commercially available IVDs. It contends that the core function of both LDTs and commercial IVDs is to analyze patient samples and provide diagnostic information.

b. Development Teams: The FDA highlights that even LDTs developed in laboratories often involve specialized teams with expertise in bioinformatics, software development, and medical genetics, similar to commercial IVD development teams. This suggests that the technical aspects of LDTs are on par with commercial IVDs.

c. Clinical Validation: The FDA argues that clinical validation and design activities are critical aspects of ensuring the safety and effectiveness of both LDTs and commercial IVDs. These activities, it asserts, fall under its regulatory purview.


THREE

CLIA and Laboratory Role:

The FDA acknowledges that laboratories play a distinct role in diagnostic testing, particularly in the performance of tests. However, it argues that this role does not exempt laboratories from FDA jurisdiction when they are involved in manufacturing test systems. Key points in this regard include:

a. CLIA Certification: While laboratories certified under the Clinical Laboratory Improvement Amendments (CLIA) are authorized to perform high-complexity testing, the FDA contends that this does not absolve them of their obligations under the FD&C Act when manufacturing test systems. It asserts that CLIA governs the testing process, while the FD&C Act governs the manufacturing of test systems.

b. Practice of Medicine Provision: The FDA addresses the "practice of medicine" provision in the FD&C Act and argues that it is limited in scope. Section 1006 of the FD&C Act is seen as protecting the authority of healthcare practitioners to prescribe or administer legally marketed devices within a legitimate practitioner-patient relationship. It is emphasized that this provision does not extend to the manufacturing of new test systems.

FOUR

Impact of Other Legislation:

The FDA refutes the notion that other legislation, such as CLIA, supersedes its authority over LDTs. It argues that CLIA and the FD&C Act have distinct focuses and purposes and are intended to complement each other. This perspective is based on the premise that CLIA primarily addresses proficiency in clinical testing, while the FD&C Act covers various aspects, including clinical validation and design activities.

FIVE

Interstate Commerce and Commercial Distribution:

The FDA addresses the arguments related to interstate commerce and commercial distribution, asserting that these elements do not significantly affect its jurisdiction over LDTs. Key points include:

a. Interstate Commerce: The FDA points out that the FD&C Act does not require all devices to have a specific nexus with interstate commerce beyond the constitutional minimum. While some provisions include interstate commerce elements, they are not universal, and the absence of such elements does not diminish the FDA's jurisdiction over devices, including LDTs.

b. Commercial Distribution: The FDA contends that LDTs are indeed intended for commercial distribution, and the phrase "for commercial distribution" is interpreted to mean "on the market." It asserts that LDTs, like other medical devices, are available for commercial use and are therefore subject to FDA regulation.

In summary, the FDA justifies its regulatory authority over LDTs based on its interpretation of the FD&C Act, historical practices, functional equivalence with commercial IVDs, and the distinct roles of CLIA and the FD&C Act. It also addresses arguments related to interstate commerce and commercial distribution, asserting that they do not alter its jurisdiction. These legal justifications form the foundation of the FDA's policy regarding LDTs and its efforts to ensure their safety and effectiveness.

###

###

The above summarized the FDA's writings.

Next, we ask Chat GPT to analyze rather than just summarize.

###

Are any of these arguments weaker than others?   For example it regulates devices "introduced into commerce" and the "Device" (the lab developed test in its lab) is hardly "shipped" into interstate commerce.   It says that the lab's authority to offer LDTs under CLIA "does not absolve" it of FDA regulation, but maybe it does.   Discuss and use other issues and examples as appropriate.


ChatGPT

Certainly, within the FDA's legal justifications for its lab test policy, some arguments may be perceived as weaker than others. It's important to note that the strength of these arguments can be a matter of interpretation and legal analysis. Here are some aspects where certain arguments might be considered weaker:

Interstate Commerce Argument:

The argument that the FDA's jurisdiction hinges on devices being "introduced into commerce" can be seen as relatively weaker in the context of LDTs. This is because, as you mentioned, LDTs developed and used within a single laboratory may not involve the same level of commercial activity or "shipment" as traditional medical devices.

Counterargument: Opponents may argue that the "introduced into commerce" requirement implies a more significant connection with interstate commerce than what is typically associated with LDTs developed and used within the confines of a single laboratory. They might contend that LDTs should be treated differently due to their localized nature and the absence of traditional commercial distribution.

CLIA Authority Argument:

The argument that CLIA certification does not absolve laboratories of FDA regulation could be subject to debate. While the FDA asserts that CLIA and the FD&C Act serve distinct purposes, some may argue that the authority granted under CLIA should carry more weight, especially in cases where CLIA-certified laboratories are operating within the scope of their certification.

Counterargument: Opponents might point out that the CLIA framework is specific to clinical laboratory practices and standards. They may argue that laboratories operating within the confines of their CLIA certification should have some degree of exemption or flexibility from FDA regulation, particularly for LDTs that are developed and used exclusively in a clinical laboratory setting.

Commercial Distribution Interpretation:

The FDA's interpretation of "commercial distribution" as meaning "on the market" may face scrutiny. Some may argue that this interpretation could be too broad, potentially encompassing activities that do not involve traditional commercialization.

Counterargument: Critics could contend that the FDA's interpretation of "commercial distribution" should be more narrowly defined to exclude LDTs developed and used within a single laboratory. They might argue that the mere availability of a test within a laboratory's own practice should not trigger FDA regulation based on a broad interpretation of commercial distribution.

Historical Practice:

The FDA's reliance on historical practice to assert jurisdiction over LDTs may be challenged. Critics might argue that historical practice should not necessarily dictate current regulatory decisions, especially as technology and healthcare practices evolve.

Counterargument: The FDA could counter this by asserting that historical practice reflects a longstanding interpretation of the law and that the agency's actions are consistent with the FD&C Act's original intent. They may argue that established regulatory frameworks provide stability and predictability in the industry.

In summary, the perceived strength or weakness of these arguments depends on one's legal interpretation and perspective. Opponents of the FDA's regulatory approach to LDTs are likely to focus on areas where they believe the agency's rationale is less compelling, such as the interpretation of interstate commerce, the relationship between CLIA and the FD&C Act, or the definition of commercial distribution. Legal challenges and ongoing debates in this area will continue to shape the regulatory landscape for LDTs.



FDA Releases Proposed Regulation to Control Lab Developed Tests

On September 29, 2023, the FDA released its long awaited regulation to control the lab-developed test industry.   The rule was released in typescript form, with the Federal Register publication on October 3 (88 Fed Reg 68006-68031).  The rule had been imminent, since it was under review at the Office of Management and Budget since August.  In addition, the FDA's leadership has been stating clearly, particularly since last fall, that they would produce regulation if Congress does not pass law, such as the VALID Act.

See FDA press release here:

https://www.fda.gov/news-events/press-announcements/fda-proposes-rule-aimed-helping-ensure-safety-and-effectiveness-laboratory-developed-tests

See the Federal Register rule page here:

https://www.federalregister.gov/public-inspection/2023-21662/medical-devices-laboratory-developed-tests

Red Flag: You can't full understand the published 26-page rule without its online 127-page financial supplement, which also has some novel facts and novel arguments besides the finances - here.  While the "averaged" cost to industry over 20 years is around $5B, some $40B costs occur suddenly in years 4/5.   Those are tangible costs; tables project hundreds of billions of health QALY intangible benefits (millions per patient) project out decades into the future.

Altogether, the materials include over 150 pages, over 40 tables, some 90 footnotes and 150 citations.

Comment until December 4.

News

Early news at Genomeweb here. Follow up by Bonislawski, here.  Fierce Biotech here. AdvaMed here. NYT here.   PBS here.  Stat here.  Medpage here.   RAPS (regulatory professionals) here.  Akin Gump law firm here.  Skadden law firm here.  Goodwin law firm here (pt 1).  CAP requests 120 days' comment.  AMP is "disappointed" and states that lab procedures are not "IVDs" boxed and shipped.   >> See my sidebar on how FDA won the press battle.

Financials

On Day 1, Quest and Labcorp were off 1%, Natera off 3%.   (These small slippages took several hours to sink in.)

Trump 1, Trump 2

Recall that  shortly after the Trump election, FDA canceled an Obama plan to regulate LDTs. Recall that ;ater during the Trump administration, HHS forbid FDA to regulate LDTs and (deja vu) the Trump administration could be back.  

Sidebars to This Blog

Sidebar 1 - Legal Excerpts ["Can FDA Do This?"].  See Genzen; see Clement and Tribe.

Sidebar 2 - FDA Press Release, Deconstructed

Sidebar 3 - The Crucial Financial Supplement, 127pp

On a humorous note, a fourth sidebar asks ChatGPT to describe the FDA proposal in the voice and talking points of several diverse stakeholders. I also asked Chat GPT to summarize all of my published materials on the rule - here.

With that said...dive in.

A Micro Regulatory Change!

The FDA's press release does not summarize the rule, leaving readers to go through the whole 83-page rule to understand it.  However, the "core" part of the rulemaking is only a few words long, by "making explicit" that a test is an in vitro diagnostic device (IVD) "when its manufacturer is a laboratory."  At its core, the regulation does nothing except add "^when the manufacturer of these products is a laboratory" to the definition of diagnostic test devices at 21 CFR 809.3 (last page of PDF document). 

  1. In vitro diagnostic products are those reagents, instruments, and systems intended for use in the diagnosis of disease or other conditions, including a determination of the state of health, in order to cure, mitigate, treat, or prevent disease or its sequelae.  [INSERT] ^These products are devices as defined in section 201(h)(1) [Act] and may also be biological products subject to section 351 of the Public Health Service Act, including when the manufacturer of these products is a laboratory.

FDA controls both device performance and marketing claims, and FDA gives substantial weight to both issues in this document.    As to problems with marketing claims, to show these may be contentious or disputed, FDA cites lawsuits over marketing claims in areas such as pharmacogenetics.

Comments run to about November 29.  The FDA notes that it expects "interest" and "new parties" in third-party review of LDTs.

One way to think about all this, you can voluntarily go to the FDA  today for 510(k) authorization of a Class II test (Mammaprint) or a Class III test (Foundation F1 CDx).  Start by striking the word "voluntarily."    As proposed, the zillions of 510K or de novo or PMA applications must be submitted in a couple years, although each one will act to 'stop the clock' on enforcement until that document is reviewed.  

FDA Summary (Sic)

The change is styled as a rollback of "enforcement discretion" of LDTs that FDA has allowed, since the Device Amendments of 1976.  However, the upfront "Summary" section of the FDA proposal is not really a summary, but simply restates that FDA is rolling back its prior discretion.  

Costs and benefits are summarized, in terms highly favorable to FDA: Financial benefits "averaged per year over 20 years" will be $2B to $86B, with a best estimate of benefits of $22B.   These includes avoidance of... health damage being caused by LDTs and lawsuits.  

Here's what you don't read in the summary: Most of the benefits are virtual ($26B per year forever) based on millions of dollars of life per patient not harmed by LDTs.  Costs of compliance regulation will be $6B ($2B-$20B), again, "annualized", but industry costs are highly front-loaded, with about $40B in years 4-5.  These are buried in table 35 & 37 of the online docket aka Supplement.

Click to enlarge, Supplement, Table 37.  Up to $83B in costs, then divide by 20.

Justification

This section "need for the rule" is extensive.  FDA  sights news coverage that allegedly raise concerns about under-regulation of NIPS testing and "IVDs offered as LDTs are ubiquitous."   Further, "FDA has witnessed an explosion in the volume, complexity, and scope of IVDs offfers as LDTs for use in determining cancer treatments...used in some of the most complex areas of medicine."  FDA asserts that "current information raises serious questions whether patients can rely on IVDs offered as LDTs."  In addtion to citations to marketing-focused lawsuits, FDA writes that "four different studies have documented high variability in performance among [LDTs]."

As to the value of self-regulation, FDA writes, "We are aware of arguments that [requirements under CLIA] already provide adequate oversight...Our review of the evidence indicates otherwise."  They also dismiss the regulatory option of adding more authority to CMS/CLIA ("enhanced CLIA.")  

There is an extensive section regarding the history of this rulemaking and the FDA's legal authorities.  I provide a sidebar blog where I give a detailed Chat GPT summary of the FDA's 5000-word legal justification.

FDA has no problem cherry-picking references that fit its needs best for a particular paragraph.  There's no attempt to portray the citations as systematic or reflecting averages from meta-analyses.


Description of Policy (Section VI, page 68021ff)

Below, I provide an AI-assisted discussion of the actual implementation rules (which are about 7000 words long).

Proposed Phased Approach: The FDA proposes a five-stage phased approach to regulate LDTs:

Stage 1 (1 Year): End the general enforcement discretion approach with respect to Medical Device Reporting (MDR) requirements and correction and removal reporting requirements. Compliance with these requirements would help the FDA monitor adverse events and safety issues associated with LDTs.

Stage 2 (2 Years): End the general enforcement discretion approach for requirements other than MDR, correction and removal reporting, Quality System (QS) requirements, and premarket review requirements. This includes registration and listing, labeling, and investigational use requirements. Compliance with these requirements will provide better oversight of LDTs.

Stage 3 (3 Years): End the general enforcement discretion approach with respect to QS requirements. For LDTs manufactured entirely within a single Clinical Laboratory Improvement Amendments (CLIA)-certified laboratory and not distributed outside that lab, some QS requirements may continue to be enforced. Compliance with QS requirements is crucial for ensuring the quality of LDTs.

Stage 4 (3.5 Years): End the general enforcement discretion approach with respect to premarket review requirements for high-risk IVDs offered as LDTs. This applies to IVDs eligible for classification into class III. The timeline aligns with the start of fiscal year 2028 and the beginning of a new user fee cycle, providing stability to the marketplace.

Stage 5 (4 Years): End the general enforcement discretion approach with respect to premarket review requirements for moderate and low-risk IVDs offered as LDTs, requiring premarket submissions. This includes 510(k) submissions and De Novo requests. The proposed timeline is extended by 6 months to prioritize high-risk IVDs' review.

Use of Third Party Review Program: Laboratories may seek to use FDA's Third Party review program for their submissions, potentially streamlining the review process.

  •  "FDA also anticipates that laboratories may seek to utilize Third Party Review...we anticipate interest [!] ...as well as new Third Party Review organizations." (p 68027)

Effective Date: The proposed rule, if finalized, would become effective 60 days after publication in the Federal Register.

Economic Impact: The rule includes an [extensive] Preliminary Economic Analysis of Impacts, which assesses the potential costs and benefits associated with the proposed regulation.

Public Health Rationale: The FDA's proposed rule aims to ensure the safety and effectiveness of LDTs, addressing concerns about their quality and performance. It is designed to provide better oversight, enhance patient safety, and align LDTs with established FDA regulations for IVDs.

Risk Categories

The proposed FDA rule for Laboratory Developed Tests (LDTs) describes device risk categories for these tests, particularly in relation to premarket review requirements. In the context of this rule, there is a distinction between high-risk (Type 3) and moderate to low-risk (Type 2) LDTs, which corresponds to the regulatory pathways typically associated with them:

High-Risk (Type 3) LDTs: These are LDTs that may be eligible for classification into Class III devices. In the proposed rule, the FDA plans to phase out the general enforcement discretion approach for premarket review requirements for high-risk LDTs offered as LDTs (Type 3) 3.5 years after the finalization of the rule but not before October 1, 2027

Class III devices typically require a Premarket Approval (PMA) application, which is the most stringent regulatory pathway for medical devices. [E.g. current PMA for a few lab tests like FMI F1 CDx].

Moderate and Low-Risk (Type 2) LDTs: These are LDTs that may be eligible for classification into Class II or Class I devices. The proposed rule suggests ending the general enforcement discretion approach for premarket review requirements for moderate and low-risk LDTs (Type 2) that require premarket submissions (e.g., 510(k) submissions and De Novo requests) 4 years after the finalization of the rule but not before April 1, 2028. Class II devices typically require a 510(k) submission, which is a less stringent regulatory pathway compared to PMAs.

By categorizing LDTs into these risk categories (Type 2 and Type 3), the FDA is proposing different timelines for phasing out the enforcement discretion approach, reflecting the level of risk associated with these tests. High-risk LDTs (Type 3) are subject to a longer phase-out period, aligning with PMA requirements, while moderate and low-risk LDTs (Type 2) have a shorter phase-out period, aligning with 510(k) and De Novo requirements. This approach allows the FDA to prioritize high-risk tests while providing additional time for laboratories to comply with regulatory requirements for moderate to low-risk tests.

It's only mentioned once or twice, but critically, applying for a 510k or PMA will "stop the clock" on enforcement against the LDT involved.

##

Defining the risk category (and thus, whether 2027 or 2028 deadlines apply) includes defining a clinical use statement (that is, FDA intended use), which itself is a complex process and which has no close predicate for LDTs today.  For example, devices introduced after 1976 are Class III by definition unless they have a legal 1976 predicate or a post-1976 cleared or approved predicate.  The rule makes several references to De Novo reclassification. 


Thoughts and Next Steps?

Note that the FDA is not reclassifying LDTs as IVDs.  Rather, it says that LDTs have always been IVDs, but LDTs have been a type of LDT under enforcement discretion, until now.

For "risks of LDTs," FDA cites poor submissions it receives, literature reporting discrepancies, news articles, and lawsuits.  (Submissions included faulty LDTs for COVID).  These include e.g. ref 33, a "memorandum to file" that is filed within an FDA office.

It's not entirely clear where the borders of LDTs are.  For example, paraffin block to H&E slides are not made as a "kit" for paraffin blocks and H&E slides; so are slides an "LDT IVD?"  We tend to think not, but it's opaque from the regulation.  For example, a pathologist has three slides on a case: H&E, Alk Phos, and Her2.  The latter is a PMA test (e.g. PMA980024).  Where do you draw lines and why?

Some HLA and some forensic tests would continue to have FDA enforcement discretion (p. 68022). But FDA is doubtful that e.g. even when test methodology is "well understood," it ought to be exempt from FDA review (p. 68023).  

NIPT/NIPS tests have never reviewed by the NDA, but are used by the hundreds of thousands.  It's unclear to me how FDA would  frame the risk/benefit of a positive result (to the mother? to the fetus??) or if that could become too-hot-to-handle politically.   Plus, the amniocentesis tests that occur as a reflex are surely not "fully manual" tests A to Z, therefore, are also illegal LDTs. 

CAP now has complex policies for  outsourcing parts of genomic tests, where both parties to the test are themselves CLIA labs.  It's unclear what happens to this under the FDA IVD concept (and FDA seems to touch on this on Supplement, page 13-14, new LDTs that are "not manufactured and used in a single laboratory."  One entity has a slide deck about the CAP Distributive Model; while the FDA online financial assessment supporting the LDT proposal (FDA ref 34) states, "FDA is aware of some entities that claim a connection to laboratories and offer IVDs as LDTs...software developers offering software...with LDTs through laboratory partnerships."

FDA has a somewhat challenging discussion of "grandfathering," I think flagging fo readers that FDA is doubtful it is a good idea, but FDA does foresee comments (p. 68023).  They also ask for comments if academic medical centers (AMCs) should be handled in a different way (p. 68023ff).   FDA seeks comments on leveraging the NYSDOH program and whether it could trigger enforcement discretion (p 68024).  

Expect legal challenges, one reason the FDA has "lawyered up" with pages and pages of legal justificaitons in its proposal including interstate commerce under the Constitution, etc.  (See also sidebar blog.)  If finalized, it would cause a frenzied boom of demand for the limited pool of experienced FDA IVD consultants.  (Vs a complementary boom in staff needed by FDA or by its third-party reviewers).   FDA expects $500M in new FDA costs; at $100,000 per person-year, that would be 5,000 FDA person-years of time.

Minimal summary financials are stated in the proposal (e.g. financial benefits might reach $80B...).  For details see a 127 page FDA website summary of the actual financial guesswork (FDA ref 34, here.)  For example, they estimate that 600 to 1200 to 2400 labs will be affected by this rule (p 24).  LDTs to be regulated may be 40,000, 80,000, or 160,000 (p 25; Suppl. Table 2).  New tests per year may be 3600, or 7200, or 14000 (p 25).  

If finalized expect a boom in third-party review.  For those whoo have said the 1976 PMA/510k regulation is a bad fit for LDTs - well, this is exactly what you get if the only change you make is to apply that 1976 law to LDTs (rather than statutory change).  On the Hill, stakeholders might rally around some kind of statutory change (VALID or other) rather than  the crude rolling-over of old 510k/PMA law onto LDTs, which is what we get in today's announcement. 

If finalized...expect delays running several years longer that the FDA's proposals.

The PR Game: Well played, sir.  The FDA released the 80 page rule on a Friday morning.   It came with an FDA press release that didn't really describe the plan, and focused on the bad actors in the lab industry and the urgent need for public safety.   This FDA press release was the only easy entry point for journalists.  Few would be qualified to read and absorb the 80 page rule (or the 127 page supplement), much of it arcane or legalese.  They'd have only the FDA's own press release to work from.   The weekend would soon come, then a new news week starting Monday, badly aging any Monday interviews leading to Tuesday stories (buried on page 6, or the internet equivalent).   Had the FDA wanted to bury the news further, it could have released at 5 pm Friday.  But it didn't do that: It wanted some news on Friday, but little would come out besides the FDA's own press release spin.  (The NYT only got out a first story by 3 pm.)  In fact, nearly all the stories on Friday and Saturday hewed close to the FDA's highly self-congratulatory press release because little else was available.  Mission accomplished, I would say.

Policy Breakdown

1, Executive Summary, 3pp
2. Abbreviations, 1p
3. Background, 25p
4. Legal authority, 1p.  But see Legal topics in III-C and V-B
5. Regulation proposed, 14p
6. Enforcement policy and roll-in, 21p

See also the 127-page online supplement.

The PR Game: Well played, sir.


Thursday, September 28, 2023

Journal Club: NEJM Compares AI and Traditional Statistics

 A new paper in NEJM from London by Hunter & Holmes provides a deep dive comparison of data analysis by traditional statistics and by AI.   Find it here:

https://www.nejm.org/doi/full/10.1056/NEJMra2212850  [subscription]



The article has no abstract.   Below, I provide an AI short summary and then an AI long summary.  I also ask Chat GPT to imagine what the paper is, from the citations alone.  Finally, I ask Chat GPT for weak points or missing points in the paper.

##

This article by Hunter & Holmes in the New England Journal explores the evolving synergy between Artificial Intelligence (AI) and traditional statistical methods in medical science. While AI offers powerful tools for data analysis and risk prediction, it can present challenges related to transparency and reproducibility. The text underscores the value of combining AI's capabilities with human expertise to ensure responsible and effective advancements in medical research and practice. It highlights the importance of robust validation and the need for a collaborative "human-in-the-loop" approach to maximize the benefits of AI while maintaining critical oversight.


##

Writing in the NEJM, Hunter and Holmes explores the dynamic intersection of Artificial Intelligence (AI) and traditional statistical methods within the domain of medical science. It presents a nuanced perspective on how AI is reshaping the landscape of medical research and clinical practice, while also highlighting the inherent challenges and opportunities it brings to the table.

The text begins by introducing the concept of AI in medical science and underscores its potential to revolutionize the analysis of vast and intricate datasets. It acknowledges AI's strengths in tasks such as automated feature extraction and risk prediction. However, it also emphasizes that the complexity of AI algorithms can make it challenging to audit and reproduce their conclusions, thus raising concerns about the transparency and verifiability of AI-based findings.

Moving on to the comparison between statistical methods and AI models, the text sheds light on their fundamental differences. Traditional statistical approaches are characterized by hypothesis-driven analysis, where predefined plans guide the research process, and reproducibility is a central tenet. In contrast, AI often takes an exploratory route, seeking patterns in data without a priori hypotheses. This exploratory nature can render AI models opaque, potentially leading to overfitting and false positives, highlighting the importance of clear reporting and the availability of code for external validation.

The text recognizes AI's utility in reducing the dimensionality of complex datasets, making them more amenable to traditional statistical analysis. It also draws attention to the challenge of biases in AI models, particularly when trained on biased data, and the critical issue of interpretability in AI-driven research and decision-making.

Furthermore, stability and statistical guarantees in AI analyses are addressed. The text stresses the need for robust validation, either through independent datasets or in-sample partitioning, to mitigate the risk of erroneous positive results. It emphasizes that human expertise remains indispensable in designing studies, tailoring models to specific research questions, and interpreting results, aspects that extend beyond the capabilities of data-driven AI algorithms.

In conclusion, the text advocates for a collaborative approach that combines the strengths of AI and human expertise. It suggests that AI can complement the work of statisticians and researchers, but it should not replace them. This interdisciplinary collaboration, often referred to as "human-in-the-loop" development, is seen as pivotal for advancing medical science responsibly and effectively.

Overall, the text offers a comprehensive overview of the multifaceted relationship between AI and statistics in the realm of medical science. It underscores the importance of careful consideration, validation, and transparency when integrating AI into healthcare practices, while also emphasizing the indispensable role of human expertise in guiding AI-driven research and clinical decision-making.

###

I ask Chat GPT what it could surmise or impute about the publication, if it had ONLY access to the reference list.

From the titles in the detailed bibliography, we can infer that the overarching paper likely discusses the intersection of medical statistics and artificial intelligence (AI) within the field of healthcare and biomedical research. It appears to provide a comprehensive exploration of how AI and machine learning techniques are being integrated into various aspects of medicine and medical research. The references cover a wide range of topics related to AI in healthcare, including its applications, challenges, and ethical considerations. Additionally, the paper seems to emphasize the importance of transparency, reporting standards, and the mitigation of bias in AI algorithms used in the medical domain. It may also touch upon the need for human expertise and domain knowledge in conjunction with AI for responsible and effective healthcare applications. Overall, the paper likely aims to provide insights into the evolving landscape where medical statistics meets AI in the context of modern healthcare.

I ask Chat GPT for any weaknesses or missing points in the paper.

In the full paper, the authors provide a comprehensive discussion of the intersection of statistics and AI in the field of healthcare and medical research. However, there are a few potential weak points or missing points that could be considered:

Lack of Specific Examples: While the paper discusses the integration of AI into healthcare and the importance of statistical thinking, it could benefit from more specific examples or case studies to illustrate how AI and statistics have been effectively combined to address real-world medical challenges. Providing concrete examples would help readers better understand the practical applications of these concepts.

Ethical and Privacy Concerns: While the paper touches on the need for transparency and addressing bias in AI algorithms, it could delve deeper into the ethical and privacy considerations associated with AI in healthcare. Issues related to patient data privacy, informed consent, and potential biases in AI models are critical aspects of AI adoption in medicine.

Interdisciplinary Collaboration: The paper emphasizes the importance of collaboration between statisticians and AI experts but does not delve deeply into how such interdisciplinary collaboration can be fostered and optimized. Exploring strategies for effective collaboration and knowledge sharing between these domains could be valuable.

Regulatory and Legal Frameworks: The discussion could include more insights into the evolving regulatory and legal frameworks governing the use of AI in healthcare. Addressing issues related to FDA approvals, liability, and compliance with healthcare regulations would be relevant, as these factors play a crucial role in the adoption of AI in the medical field.

Patient-Centeredness: While the paper mentions the potential benefits of AI in healthcare, it could emphasize the importance of patient-centered care. Discussing how AI can enhance patient outcomes, improve diagnoses, and contribute to personalized medicine would strengthen the argument for its integration into healthcare.

Challenges in Validation: The paper briefly mentions the importance of validating AI findings but could explore the specific challenges and methodologies involved in the rigorous validation of AI models in healthcare settings. Addressing issues of data quality and reproducibility is crucial.

Future Directions: While the paper discusses the current state of AI and statistics in healthcare, it could provide more insights into future directions and emerging trends in this field. Predictive analytics, causal inference, and the integration of multi-modal data sources are areas that could be explored in greater detail.

In summary, while the paper offers a solid foundation for understanding the convergence of statistics and AI in healthcare, it could be enhanced by providing more concrete examples, delving deeper into ethical and regulatory considerations, and offering insights into the evolving landscape of AI in medicine. Additionally, discussing challenges and future directions would further enrich the discussion.


Wednesday, September 27, 2023

CMS Releases Proposed Prices for 85 New 2024 Lab Test Codes

On September 27, 2023, after market close, CMS released proposed prices for 85 new laboratory test codes.   

This triggers a comment period through October 27, 2023.   The proposed prices follow a public comment meeting in June and a public expert panel meeting in July.

The biggest surprise, AMA introduced a set of new CPT codes for genomic procedures like tumor gene panels with TMB or MSI.  CMS offers to price these at only the lowest price of a small tumor panel (about $600, CW 81445, 81450).  This bears out the fears of some stakeholders last spring, that omitting gene panel numbers could trigger low pricing by CMS.   AMP, CAP, and others will be commenting on this.   One result would be gapfill by Moldx in 2024, but that would leave the codes formally unpriced all year long (for CMS and for payers that reference CMS).

Here's the CMS CLFS page:

https://www.cms.gov/medicare/payment/fee-schedules/clinical-laboratory-fee-schedule-clfs/annual-public-meetings

Scross down to Meeting Notice, Agenda, and Other Important Materials and click to open this section of the webpage.  Scroll down further to a zip file of "2024 proposed prices."

Numerology

I counted 18/85 codes where "CMS disagreed with the majority of the panel" and CMS proposed something different.   So this result is not rare, but not common.   There were 3 other codes where CMS used the term "agreed with the majority of the title."  

At least one code flipped on appeal, 0329U, where CMS writes it now understands the code and crosswalk better and will flip from "gapfill" (which is currently in process) to CW 0242U based on this year's appeal.

I count 22 codes proposed for "gapfill," about the same as last September.  But more codes may flip into the gapfill column by the final decision, as  some stakeholders will complain about a low price crosswalk and get gapfill as a consolation prize. About 32 codes were actually gapfilled in 2023.

Nearly all crosswalk codes are crosswalked to 1 single code at exactly its value.  There's a fractional crosswalk for 0387U, a melanoma IHC code.  There's a two-code crosswalk for 0367U.  

There were several codes with slide-based imaging and AI or ML, 0376U, x084U, X088U.   Like the expert panel, CMS crosswalks these to 0220U.   That's $706, the Precise Dx breast cancer test.  Now, a similar class of test, the Prelude DCISioniRT, 0295U, won ADLT status last spring and thus a price at $5435.   (This triggered the same $5435 for a test that was crosswalked to 0295U at its original price).  This means that tests under the same paradigm - several slides and ML - are priced almost 10X apart.

Further, at least one of these codes, 0376U, is classed by CMS as not payable in the hospital outpatient setting, where a lot of biopsies originate.  (It gets status Q4, bundled).  The same problem impairs the new AMA Category III codes for whole slide imaging (status N, not paid separately).  Possibly these codes would meet an unbundling exception for proteomic cancer MAAA tests found in the date of service rule.  42 CFR 414.510.b.5.  CMS has ruled that these slide-based tests are "proteomic" tests because they have to be to be eligble for ADLT status, which a few get.

AI Corner.

Below, government accounts visit the clinical laboratory.




Calendar Watch: AdvaMed Dx Summit: October 11, 2023 (Anaheim)

AdvaMed's big fall conference is coming in October in Anaheim, and this year will feature a day for AdvaMedDx and the diagnostics industry.

Find full information here for Diagnostics Summit on October 11, 2023.

https://themedtechconference.com/dx-summit/




##

In the afternoon, I'm moderating a panel on digital diagnostics and AI, with Robert Jarrin (Resmed) and Paul Gerrard (Artera).


Journal Club: New Review of HEOR of Liquid Biopsy

Liquid Biopsies (LBx) offer a promising alternative to tissue biopsies for cancer diagnosis and treatment, and promise to open a new era in cancer prevention for cancers like ovarian and pancreatic.

 A new systematic review by Fagery et al., in Pharmacoeconomics, collects 24 studies in order to assess LBx health economics and modeling challenges in precision medicine. 

Open access here.

Most studies (80%) conducted economic evaluations, with 75% demonstrating LB cost-effectiveness across diverse cancer types and stages. LBx were reported as beneficial for screening (2/3) and treatment selection (1/4).  Budget impact varied, with many suggesting cost savings or modest impact. LBx could be cost-effective for lung cancer treatment selection and early detection of various cancers.  However, "further research is needed," and the authors felt some publications under-rate modeling challenges, such as uncertainty.




Monday, September 25, 2023

Unprecedented CMS Press Release Insists No Change in a MAC LCD Policy (Transplants)

Close readers of the Wall Street Journal will have noted three articles in the past week on coverage for molecular transplant-monitoring tests (such molecular tests are offered, in different forms, by CareDx, Natera, and Transplant Genomics, perhaps others).

These articles included an Op Ed, and a Letter, and a second Op Ed on the heels of the first one.  

  • September 10, Op Ed, "Medicare's Bad Call on Transplants," here.
  • September 17, Letter, "Medicare is Penny Wise, Pound Foolish," here.
  • September 22, Op Ed again, "Medicare's Organ Transplant Face Plant," here.
    • Some of this covered June 27 at Genomeweb, Forest Ray, here.
While MolDx does not generally (or ever?) directly comment on newspaper articles, CMS has issued a national press release on the topic today, September 25.


The press releases is 7 paragraphs long but packages the main information in the first paragraph, here:

The Centers for Medicare & Medicaid Services (CMS) confirms that neither CMS nor the Medicare Administrative Contractors (MACs) have made changes that affect patients’ ability to have blood tests used to monitor for organ transplantation rejection covered when ordered by their physicians in medically appropriate circumstances. Patients with transplanted hearts, lungs, or kidneys who meet the coverage criteria for these tests, as laid out in the Local Coverage Determination (LCD), can continue to access these tests in place of a more invasive biopsy.

See the link above for the full CMS response. 



##

AI Corner 1:

I asked Chat GPT to read the WSJ op eds, read the CMS press release, and then discuss, compare, contrast.

https://brucedocumentblog.blogspot.com/2023/09/chat-gpt-discusses-wsj-op-eds-vs-cms.html

AI Corner 2:

Chat GPT summarizes the 540 word press release in 90 words.  Less impressive to the eye this time, in that CMS basically made its own 90-word summary in its own first paragraph.

The Centers for Medicare & Medicaid Services (CMS) confirms no changes in the coverage policy for blood tests used for monitoring organ transplantation rejection. The tests remain covered for heart, lung, or kidney transplant patients meeting specific criteria, offering a less invasive alternative to biopsy. Despite concerns over improper billing and overutilization, and subsequent revisions to billing instructions, the core coverage remains unchanged. A new proposed Local Coverage Determination (LCD), released for public comment, maintains existing coverage criteria, ensuring continued patient access to these essential tests when ordered by physicians.

And the Chat GPT haiku:

Tests still covered well
Transplants monitored with care
Biopsy spared now.

Saturday, September 23, 2023

Lecture Notes: German Health Minister Sets Goal for Fast Digital Advances in German Healthcare

Some readers notice that I am a lifelong German-language hobbyist, and I know more about the health system in Germany than any other European country.  Sometimes this pays off, like this week, when I attended a fascinating lecture by Germany's Health Minister on dramatic new goals for a digital push forward in the German healthcare system.   The event was a joint Q&A with Dr Karl Lauterbach and Dr Micky Tripathi, the National Coordinator for Health Information Technology at the US HHS.  It was held at the Kennedy School at Harvard on September 22.

Dr. Lauterbach, who holds a PhD in health policy from Harvard, became well-known in Germany during the COVID crisis but did not ascend to the position of Health Minister until the change of  government in late 2021.   As preparation, I ran across a 30 minute interview that Lauterbach gave a few days ago with the German newspaper DIE ZEIT, on AI in healthcare (German; here).   He recapitulated many of the same points in Boston this week.

Lauterbach focused on 4 or more pieces of legislation just now moving through the writing, Cabinet-approval, and parliamentary process in Berlin.   The results would connect doctors' offices, hospitals, and pharmacies throughout German with a unified electronic health record.   

Moreover, there would be an "outer ring" and an "inner ring," per Lauterbach.  The outer ring would be encrypted but patient-specific, and patients (who could opt out entirely) could easily allow transfer of clinic visits and history to new providers.   The inner ring would be anonymized and constitute a regulated secure "Federal Research Space" as a learning health system and for research.  AI could help create retrospective narrative structures from original free text.  

Other features mentioned by Lauterbach included a reform of the research approval process so that high standards of patient safety could be maintained while shrinking the burocractic approval process to 30 days.   Both at DIE ZEIT and at Harvard, he referred several times to 2025 as the goal for putting major pieces in place (this seemed remarkable to me).  

In the DIE ZEIT interview, countries like Israel and Sweden were used as models and goals for electronic health records for the great-leap-forward in Germany.   In addition, online, the interviewer noted that Germany was not only working with or learning from nations, like Israel and Sweden, but Microsoft and Google were at the forefront of digital health and AI.   Echoing this exactly, at Harvard, Dr. Tripathi closed by noting in the last few days he hsd keynoted at meetings where he was preceded or followed by Microsoft or Google.   

##

Context.  It's difficult for me to entirely place this "new plan" in context, as a quick Google search the last five years shows a plethora of articles about prior legislation and prior goals and deadlinees to help Germany "leap forward" from its backward position in digital health records and interoperability ("Digitalisierung.")  E.g. this page is dated 2020.  Compare a 2023 essay here (also McKinsey 2022 here, ICTH 2023 here, 44-p German ministry depiction of its goals here).  Other articles talk about the shaky status of German hospital economics, and Covington just published an article updating us on pending major inpatient DRG reforms in Germany, here.


AI Corner

I left the Lauterbach-Tripathi session with a roughly-typed  page of fragmented phrases and keywords.  I dumped them into Chat GPT and asked it to edit them into a smooth journalist-style essay.   Here.

Bonus Book

Health care change, let alone health care reform, has always been messy and chaotic and confusing.  See a 2021 book by a Yale professor, Peter Swenson, covering 150 years of zig-zagging US health policy - Disorder.

November 9-10, Berlin

Robert Koch Institute, 9-10 November, Berlin, AI in Public Health, here.

44p German 2023 guide to digital reform and adoption

ONC Rulemaking on AI and EHR 3/2023

A few links to more information about Spring 2023 ONC rulemaking on uses of AI esp. in EHR, here.


Journal Club: New JAMA Essay on Rolling Out AI in Healthcare (Gottlieb & Silvis)

A new, open-access JAMA Health Policy essay, "How to Safely Integrate Large Language Models Into Health Care," two authors with both industry and FDA experience discuss pathways for integration of Large Language Models (LLMs) in healthcare. Dr. Scott Gottlieb  served as FDA commissioner and Lauren Silvis is a former senior FDA regulator now at TEMPUS.

The essay highlights the potential of LLM's to enhance patient interactions and healthcare delivery.  Three takeaways:

LLM Potential: The essay underscores LLMs' transformative potential in healthcare, enabling human-like text generation, aiding in diagnosis, and providing ongoing patient support.

Staged Integration: The authors advocate a cautious and phased approach, starting with well-understood conditions, to help ensure safe LLM integration and boost confidence.

Data and Bias: Effective LLM use in healthcare demands improved data sharing, bias detection, and inter-institutional collaboration for accurate and unbiased results.



Wednesday, September 20, 2023

Very Brief Note: 153 Comments Posted on "TCET" Proposal

 CMS has posted (after a few weeks' delay) the 153 comments on its TCET, emerging technology proposal:

https://www.regulations.gov/docket/CMS-2023-0107

TCET was discussed multiple times at a September 19, 2023, House Energy & Commerce hearing on Medicare and innovation.

https://www.discoveriesinhealthpolicy.com/2023/09/september-19-2023-congress-holds.html


Tuesday, September 19, 2023

September 19, 2023: Congress Holds Hearing on CMS and Innovation

On September 19, 2023, the House Energy and Congress committee held a hearing on, "Examining policies to improve seniors’ access to innovative drugs, medical devices, and technology."

Find the home page here.  The agenda includes Dr Dora Hughes, chief medical officer for CMS. The E&C website links to the YouTube archive video stream.  Her ten page testimony here.  (Most of it, was a fairly dry review of LCD and NCD and CED processes).

Chair's remarks here.  

The most interesting document is the 6 page hearing memo here.  Remember, for Dora Hughes, this is essentially a hostile audience - the Republican House vs the Democratic HHS administration.  The hearing memo reviews 10 or 12 "legislative fixes" that have been proposed to improve the NCD and other CMS processes.



AI CORNER

I've posted an unofficial auto transcript beginning with an unofficial detailed summary, in the cloud here.



MolDxology: DEX Registry Allows Categorical Browsing

I'm not sure when this started, but I'm pretty sure it didn't used to be here.   The MolDx Z code online registry, DEX (which shows all about a registered test, EXCEPT its actual secret Z code) has always allowed searching by test name and lab name.   

(Test name can be a real wild west, and lab name can even be complicated in case of acquisitions and subsidiaries.)

Now Palmetto DEX allows searching by five different drop-down categories.  At the left-hand side, these are diseases, FDA status, medical specialty (e.g. cardiology), method, and test type.  (Test type is "diagnostic, confirmatory, predictive," etc).  Find DEX here:  https://app.dexzcodes.com/

There appears to be a built-in "and" logic, for example, here I've searched for the Test Type category "DIAGNOSTIC" while I happened to have let the test name "RaDaR" in the name search field.  This excluded quite a few variants of the RaDaR test, but showing one that is classed as "diagnostic."   

The DEX registry is available with free email registration.


##

Bonus

The New York State "Wadsworth" system allows searching for lab tests endorsed by New York State, here.  Article in 360dx.

##

AI Corner:


This blog in French.

ACLA, Others, Comment on Verbose and Confusing Novitas LCD for Oncology

 Last year, Novitas issued a draft LCD for oncology that primarily deferred coverage judgements to compendia like NCCN.  They finalized a far longer, more confusing version of the LCD this past June, withdrew it in July, and reissued it as "draft" in July.  It's accompanied by an extremely long and confusing billing article.   See my July blog about the "re-issue" here.

ACLA, in a press release, flags some of the public comments, including one by a multi-stakeholder group and one by ACLA itself.   Stakeholders quarrel with the deferral of coverage decisions (unless made in a one-off way for a few tests), to the NCCN and other guidelines.

See the ACLA press release here.  It has a number of links in it. These include the multi stakholder letter, the ACLA letter, and its supplement.

The comment letters are worth reading but frankly, I'm surprised they all just take the very long, verbose, repetitive, and confusing policy and article "on its face" and try to respond rationally to it.

For the AMP response, not yet posted, check here.

##

Open Meeting Transcripts Available.

See the LCD open meeting page here.  See the transcript (and audio) listing here. 30,000 words.

##

NCCN Compendia for Drugs - As a Model

If you wanted to use compendia as a guide to coverage (more a floor than a ceiling, though), there's an excellent precedent.   SSA 1861(t)(2) refers to cancer compendia, and CMS rulemaking provides process guidance.  See the CMS webpage here.  The regulatory discussion, in 2009, was just a few pages long here, The process in regulation at 414.930, is here, just a few paragraphs long.  They're written in a few paragraphs of normal, clear professional English.    (For NCCN's request letter to be endorsed, 10pp, here. For CMS's review of it, here.)


AI Corner #1.

Here is a summary by ChatGPT of the multi-stakeholder letter.

The letter, dated September 8, 2023, is addressed to Dr. Patrick Mann and Dr. Alicia Campbell regarding concerns about the draft local coverage determinations (LCDs) for "Genetic Testing for Oncology" by Novitas and First Coast Service Options. The letter highlights fears that the LCDs, if finalized as drafted, would limit Medicare beneficiaries' access to genetic testing for cancer, potentially violating Social Security Act and Medicare Program Integrity Manual requirements. Concerns include default non-coverage for tests not in knowledgebases, ICD-10 code issues, documentation requests, and restrictions on hereditary cancer syndrome testing. 

The letter requests collaboration with stakeholders to address these concerns and ensure access to essential genetic tests for cancer diagnosis and management. Signed by numerous healthcare organizations.

AI Corner #2

A summary by ChatGPT in the style of Lewis Black.  Here.


Friday, September 15, 2023

Brief Blog: Medicare Fiasco News: Patient Bowled Down by Obscure SAD List

BULLET.  Patient trapped with sudden $176,000 drug bill due to an obscure Medicare MAC policy change. +Lawsuit.

####

Background

Medicare policy makes extensive use of "incident to" services.   For example, there is a benefit for durable medical equipment like drug pumps and the drugs they pump are actually secondary tag-alongs to the DME benefit for the pump.  (Weird!).   Drugs that are administered in physician offices, like chemotherapy infusions, are covered "incident to" the service of a physician.

There are some rules around this.   The drug must be NOT self-administered HALF the time or more.   MACs are required to keep and update lists of injectible drugs that are NOT self administered, and thus eligible for office payment.   CMS defines this across all patients; if one patient is quadriplegic, for example, that doesn't matter if 51% of all patients self administer the drug (such as insulin).

MACs don't always agree (SAD lists can differ) and big debates sometimes occur.  See a sample "excluded" list here.

Today's News

MEDPAGE TODAY has a detailed article (by Cheryl Clark) about a patient who faces a major crisis because his drug was switch to "self administered, not payable in office" status by the MAC where he and his doctor live.   (Article may require email registration).  He got a bill for $176,000, whereas up to that point, his 80% payments were covered by CMS and his 20% payments were covered by his medigap plan.  A lawsuit by the Center for Medicare Advocacy is in flight.

Find the article here:

https://www.medpagetoday.com/special-reports/exclusives/106338



###

Another example of a 50% rule separates physician services and other services.   Medicare classifies a physician service IF AND ONLY IF a physician signs performs the service more than 50% of the time.   (42 CFR 415.102 . The rule doesn't state 51% numerically but this is how CMS implements the word "ordinarily" done by a physician.)   

The pathology rule is even tougher, the test/service must REQUIRE (quote - unquote) the service of a pathologist (415.130).  Back in 2012/2013, some stakeholders wanted the then-new genetic test codes to be on the physician fee schedule, but CMS determine that genetic tests did not "require" physician signout (the lab director can be a PhD).  (My 2012 white paper still downloadable here.)

###

Article summary.

Medicare unexpectedly changed its policy on the drug Stelara, classifying it as "self-administered" (SAD) on October 15, 2021. This led to retirees like George Beitzel, suffering from Crohn's and Parkinson's diseases, facing unexpected bills of up to $176,000 for previously covered injections. The Center for Medicare Advocacy filed a class-action lawsuit against this policy shift, arguing for notice, cost waivers, and professional administration options for patients unable to self-administer. Thousands of Medicare beneficiaries may be affected, raising concerns about the impact of such changes on patients' health and financial well-being.

As Haiku:

Medicare's surprise,
Stelara reclassified,
Burdens patients' lives.

##

AI Corner.

I fed a CMS MAC SAD Article to Chat GPT and asked it to figure it out and explain it.  Here.

Very Brief Blog: MAC CAC on BOTULINUM

 While we think of botulinum toxin first for cosmetic uses, it has a range of medical useless as well, such as in neuromuscular conditions.  E.g. blepharospasm.  An example of a MAC LCD is here.

The several MACs, all of them, have a "multi jurisdictional" CAC or public expert advisors meeting on botulinum, coming up on Thusday, October 19.  

FCSO link here (entry point).

NGS MAC runs the show here.

A 17-page background and question list is here:

https://www.ngsmedicare.com/documents/d/ngs/2118_0923_questions_for_botulinum_toxins_sme_panel_508-pdf

Click to enlarge.  Sample.


Very Brief Blog: Interesting Posts from Center for Genomic Interpretation (CGI)

While I don't agree with every position that the CGI takes (Center for Genomic Interpretation), they held my attention for a half hour today looking through their blog posts and Linked In posts from recent months.   

Find their blog posts here:

https://www.genomicinterpretation.org/blog/

And find their Linked In article feed here:

https://www.linkedin.com/company/center-for-genomic-interpretation/posts/?feedView=all

For example, a recent blog here points to a peer reviewed article here.  This links to an August 2023 paper, Patel et al., titled, Genomic Data Heterogeneity across Molecular Diagnostic Laboratories: A Real-World Connect Myeloid Disease Registry Perspective on Variabilities in Genomic Assay Methodology and Reporting.

For another recent example, they linked to the CLIA Advisory Committee CLIAC with regard to an August workgroup on NGS in CLIA labs.   Find the blog here.   You can link through to a 29 page CLIAC summary document here.  Or the direct PDF at CDC here.

See a CGI blog several months ago on what they see as weak points of flaws in recent state level biomarker legislation - here.

Fun fact - the laws vary across the 10 states where they have been passed.  There's a website that details the differences, but not that the table is *very* large and long and hard to read.  (I didn't even realize the table was there for a couple minutes, scroll down and down.)

See a link to Pfeifer et al. 2022 who discuss reference samples for inter lab comparisons of NGS, here.



Thursday, September 14, 2023

Multiple Articles: Concerns about Quest LDT Alzheimer Blood Test?

FDA-approved Amyvid (PET Scan) advertised next to LDT article

Multiple news articles this week quote experts as expressing concern about an LDT (lab-developed test) being marketed as an Alzheimer diagnostic.   The test is a blood test; the only FDA-validated tests for Alzheimer's so far have been PET scans and CSF tests.

See an article from Reuters here.  With remarks such as, "Dr. Sarah Kremen, a neurologist at Cedars-Sinai in Los Angeles, was concerned that people who test positive but have no symptoms will come in seeking further testing."

See an article at MedPage here.  With remarks such as, "There are no large-scale, long-term clinical trials that support the idea that the AD-Detect test can predict whether a cognitively unimpaired person will transition to cognitively impaired," said Rebecca Edelmayer, PhD, senior director of scientific engagement at the Alzheimer's Association in Chicago. "As a result, it is unclear what the results of this test may mean about your Alzheimer's risk or your health status/"

The longest article, by Adam Bonislawski, is by subscription at 360Dx.  With remarks such as, "mass spec-based assays for Aβ 42/40 ratio have higher performance ... than do immunoassays, but "you have to be very rigorous about" your measurements, said Suzanne Schindler, associate professor of neurology at the Washington University School of Medicine. "There's only about a 10 percent difference between positives and negatives, and so if you are off by a little bit or your assay drifts, then you can really misclassify a lot of people," she said, adding that Quest has released little data on the analytical and clinical performance of its test. [The article continues, "At the 2022 Alzheimer's Association International Conference, Quest presented a poster on AD-Detect, and said it aims to publish data on the test in a peer-reviewed publication."]

##

The test is NY State approved; search for analyte amyloid and facility Quest, here.

The test is stated in an article above to be 71% specific, which could mean about 1 in 3 would get a false positive, although the negative and positive predictive values are highly dependent on the test population (population spectrum).   One historical problem - not necessarily relevant here - is Alzheimer tests over the decades that were validated on 100 perfect controls and 100 perfect Alzheimer cases, but then perform much worse on the real world and borderline patients who actually need testing.  

##

AI Corner:

ChatGPT reviews the three articles, then describes them in the satirical voice of Louis Black, here.


Wednesday, September 13, 2023

Journal Club: (1) Scenarios for Whole Genome Seq Adoption, (2) Scenarios for Any MedTech Adoption

This week I ran across two excellent papers on medtech adoption.  

One is recent, by van de Ven 2021, on adoption of whole genome sequencing.  The other (from a citation within Ven) turns out to be a classic, with 1200 citations to it, which is Greenhalgh 2017 on principles for the adoption of any med tech.


van de Ven 2021: 

Whole genome sequencing in oncology: Using scenario drafting to explore future developments.  BMC Cancer 2021.

The Van de Ven paper explores the use of scenario drafting and expert elicitation to anticipate future developments in the implementation of Whole Genome Sequencing (WGS) in clinical oncology. 

It identifies potential barriers and facilitators, highlighting the importance of factors such as price, clinical utility, and turnaround time in determining the likelihood of WGS adoption, offering valuable insights for policymakers and stakeholders in genomics.

Open access here.

For further reading see Ellis 2023 or Fleck 2023.


Greenhalgh 2017:

Beyond Adoption:  A new framework for theorizing and evaluating non adoption, abandonment, and challenges to scale-up, spread, and sustainability of healthcare technologies.  J Med Internet Res 2017.

The Greenhalgh paper introduces the NASSS framework, which offers a comprehensive and nuanced approach to understanding the adoption, scale-up, spread, and sustainability of complex medical technologies. The authors emphasize the importance of recognizing the non-linear and context-dependent nature of technology adoption and highlight the need to address factors related to the technology, the organization, the wider system, and the individual in healthcare settings.

Open access here.

For more about NASSS, see an implementation for a cardio health tracking device, open access, Abimbola 2019, here.  See a shorter overview of NASSS ("Cliff notes"), Greenhalgh & Abimbola 2019, here.  See a 2020 paper that introduces a "NASSS Toolkit" here.


Bonus 1:

There's a new 2023 paper by Greenhalgh et al. on the complexity of figuring out differences in values, among stakeholders in healthcare.   Find it open access at Milbank Quarterly, here.    

Bonus 2:

Greenhalgh et al. 2023 writes a letter to the editor of Annals of Internal Medicine about a review of N95 masks.   It contains this great quote:  

  • "Some members of the evidence-based medicine community seem to assume that a randomized controlled trial, however imperfectly and illogically designed, is necessarily superior to other forms of evidence. This is not the case."
  • Find it open access here.
  • See a lively video by Greenhalgh on flaws of evidence based medicine here.

AI Corner x 3

Find a ChatGPT discussion of Greenhalgh 2017, here.  For example, Chat applies the framework to genomics, and to digital pathology.

Find a ChatGPT discussion of van de Ven 2021, futurism in WGS, here.   

Find a ChatGPT discussion of Greenhalgh 2023, on complex values in healthcare, here.