Monday, January 25, 2021

Trump-Era HHS Decisions Hit Three, Not One, FDA/CMS Positions. (1) LDTs, (2) CED, (3) Digital Health.

Executive Summary:  There's wide awareness of an August 2020 HHS ruling that FDA can't regulate LDTs.  But in the past couple weeks, add two more Trump-era HHS rulings to the list.  

One affects CMS use of "Coverage with Evidence Development (CED; see heading #2), one affects CMS rules that depend on "clearance or approval" of devices before Medicare policies kick in (see heading #3).



Everyone who follows policy at the FDA, and/or in diagnostics, recalls the week in August 2020 when HHS released a short paragraph on its website, announcing that FDA could not enforce controls on lab-developed tests, whether of the COVID type or other LDTs.   (Here).   

The decision in August 2020 built on a lengthy memo by General Counsel of HHS in June 2020, which basically accepted years of arguments from industry that FDA could not regulate LDTs, at least not without formal notice and comment regulations.  (Zip file here).  The June 22 memo, which includes a full exposition of FDA law for diagnostics, came from Robert Charrow, General Counsel (GC) of HHS under President Trump.

But Wait, There's More.

#2. HHS Nixes "CED"

See also HHS Advisory Opinion 21-03 that appears to "cancel" Coverage with Evidence Development, used in the National Coverage Decision process for years and (previously) enshrined in CMS policy documents and even discussed in Federal Register policymaking.  

The January 14, 2021 opinion is online at HHS here and I've put a cloud copy here.

Basically, Medicare law blogs Medicare from covering things that are less than "reasonable and necessary," and services under an NCD CED clinical trial are less than reasonable and necessary, hence, under CED.   CMS has stated its authority is a clause of statue allowing CMS to pay for services in support of AHRQ studies.   Therefore, every CED opinion (web page here) makes a nod to supporting AHRQ studies.   As signed by Robert Charrow, he acknowledges that while the statute does not define "AHRQ supported studies" the links between AHRQ and CED NCDs have been too thin to support a reasonable interpretation of "Supported by AHRQ."  

Do CMS Staff Tell Charrow, "Thanks for your input."?  

Note that in a decision posted January 19, after Charrow departed but before the Biden inaugural (head-spinning), CMS issued a new decision (on mitral valves) that includes CED, and this occurred despite the January 14 HHS memo.   I'm sure, thereby hangs a tale.   

CMS has a webpage for CED projects here.  Whatever the success or fizzle of CED in the past, potentially the future would be brighter with new capabilities in pragmatic trials, machine learning, real world big data, and so on.

#3.  FDA Posts Guidance on Digital Health & AI, 
Even While HHS Proposes to Nix FDA Review

For this topic, which got submerged in the Capitol protests and election news, I'll defer to an excellent article at MedCityNews by Elise Reuter (here).

Basically, in the last year, FDA has created a new Digital Health Center of Excellence - here.  And FDA has just released a major guidance and goal document on regulating digital health and AI - here, dated January 2021.

But on January 15, HHS published in the Federal Register (86 FR 4088) a lengthy list of device categories, including many in digital health, that HHS proposed to withdraw from FDA regulation (e.g. move to Class I).   

See Federal Register here, comment open 60 days until circa March 14.

The main argument that HHS uses - and the rule is signed by the head of HHS, Alex Azar, not the head of FDA - is that the device categories in question (like AI-assisted radiology) have not had injury and flaw complaints in the FDA's MAUDE database of device malfunctions.  Dozens of categories are proposed for de-listing from FDA review.

OK.  I'm all in favor of deregulation, but a few thoughts.

  1. Were reviews really pointless?  The assertion is that the current reviews are regulating nothing, since there are no complaints.  Isn't this a little like saying this intersection, with a stoplight, hasn't had any accidents in 3 years, so we can remove the stoplight?
  2. Authorization regulates claims.  Regulated devices, both 510(k) and PMA, have restrictions on advertising and labeling claims.  Class I devices don't; nobody worries about the labeling claims of a tongue depressor.
  3. FDA-CMS Interactions.  CMS has been proposing rules that are predicated on FDA clearance or approval, like special coverage for breakthrough devices (acronym MCIT) or special pricing rules for Advanced Diagnostic Laboratory Tests that are "FDA cleared or approved."   
Pulling categories from review could effect these and other tangential policies.


See a November 2020 GAO report (106pp) on the growing value of AI in healthcare - report here, JAMA note here.  The lengthy document has two separate parts, Part One jointly produced by GAO and National Academy of Medicine, Part Two (pp 38ff) produced by National Academy of Medicine.


See a pathology & AI alliance and their public website developing response to the FDA proposal here.  They refer to the FDA proposal as "86 FR 4088," appropriately enough.