Friday, March 16, 2018


Note:  This article has been revised several times on a rolling basis as the implications of the NCD have become more clear.

CMS released the final NCD for NGS testing in cancer.
Online at CMS here.
Also a PDF cloud copy here.

CMS also issued a press release here.

Coverage at Genomeweb here. Genomeweb follow up here.  MedCityNews here.  Forbes here. Reuters here.  Wired here.  Medscape here.  Targeted Oncology here.  Regulatory Focus/RAPS, here.

My best current effort at a comprehensive logic flow chart for the NCD is here.

The NCD provides a great measure of forward-looking certainty for investors and developers, because "Tests that gain FDA approval or clearance as an in vitro companion diagnostic will automatically receive full coverage under this final NCD, provided other coverage criteria are also met." [from press release]

Highlights as discussed further below:

  • Difference #1:  All FDA CDx at Parity.  Approved and Cleared FDA CDx are treated at parity.  (But there are almost no "cleared" CDx.)
  • Difference #2:  MSK IMPACT loses its CED based coverage.  The Draft had clear coverage (though with CED) for tests like 510(k) MSKCC IMPACT.  No such coverage for MSK IMPACT in the final.
  • Difference #3:  FDA CDx tests + disease type.  CMS looks to tests BOTH being cleared/approved as IVD, AND as companion diagnostic, AND with an indication in each patient's cancer (e.g. "lung cancer.")
  • Difference #4:  LBx LDTs better off.  There was essentially no coverage for advanced LDT liquid biopsy panel tests as they awaited FDA approval; now they can be covered by LCDs.
  • Difference #5:  Nix Huge National CED Program. The Draft proposed a massive national CED program; there's no CED at all in the final.
  • Difference #6:  LCD Coverage Not Killed.  There was no LCD coverage under the Draft, it was voided.  Now, LCDs can cover NGS tests with some stipulations.
  • Side Issue #1:  Safe Harbor for Germline Risk Tests.  Germline tests with NGS (such as Lynch panel or BRCA-related panel) seem exempt from the NCD due to a sentence on page 4.
  • Side Issue #2:  Some "Excess coverage" scenarios.  NCD could lead to "excessive coverage" in some possible scenarios.
  • Side Issue #3: Paradoxically discourage migration to NGS?  Since the NCD controls NGS methodology so closely, it could discourage migration of some tests to NGS platforms, since they'd require revisions to the NCD even if PMA approved but outside the NCD rules.  Labs can do Sanger and PCR to their heart's content since the NCD doesn't touch them.
  • Side Issue #4:  Oncologist must order.  CMS clarifies that while Medicare allows pathologists to order some tests themsleves ("a stain needed to complete a case"), only treating physicians can order NGS tests.
  • Side Issue #5:  What about FDA and 510(k) CDx?  FDA has cracked the door open for 510(k) "companion diagnostics" with a 12/2017 BCR ABL test; can this be a wider door for developers?
  • Medicare Nerd Points include that: (A) there's no clue how MACs will handle the potential opening up of off label gene results and off label drug options with tens of thousands of imminent, real-world patients.  (B) LDTs might dodge the "one test per patient" rule by having a 300-gene test, a 305-gene test, a 310-gene test.   (C) CMS covers a 400-gene panel, based on its CDx indications, even if only one gene was CDx-relevant to a patient's cancer, and that gene is already known to be positive or negative.  

Differences from the draft include:
  • (Difference One)  Approved and Cleared (meaning PMA and 510k) tests treated at parity - but only for patients that have a "companion diagnostic" indication.  And so far, FDA has almost never granted the "companion diagnostic" status to a 510(k) test.
    • Is this sharp difference between 99% accurate tests a good policy idea?
    • Both PMA and 510(k) classes of tests might be, for example, 99% accurate (Kim et al 2018; or Shin eta la. 2017; or Strom et al. 2015), run on essentially similar platforms and chemistry, and thus produce nearly identical reports on the same tumor.  Diagnostic lab tests measure things, which is different than a therapeutic device or drug.  They don't measure or cause outcomes; they measure things.
      • PMA tests are always validated on "clinical data."   FDA allows either (1) de novo clinical studies, or (2) briding studies aka concordance studies in paraffin blocks on a lab bench to an existing test.  Most of the FMI genes were PMA approved via concordance [bridging] studies in about 200 paraffin blocks (here)...e.g. these are benchtop analytical accuracy studies.  Accuracy studies are equally available to 510(k) tests.
      • 510(k) tests require at a minimum "paper dossiers" collating clinical trial data, guidelines, and I suppose FDA labeling in order to list gene test results as clinically validated.   (This is tier 2 of the FDA's "food pyramid" of precision medicine test results).   However, at least one 510(k) test for BCR-ABL includes some clinical studies in its paperwork, meriting classification as a companion diagnostic.  
      • For a short deck on the FDA's PMA and 510(k) CDx approaches, see here.
      • FDA's approach to giving out test indications is evolving.  
      • FDA's historic approvals of single-gene tests (like cobas BRAF) were for one drug based on one clinical trial of one gene.  
      • But when the FMI test was validated against one BRAF diagnostic that was validated for one drug, to my eye FMI gets on label indications for all BRAF drugs.  See here.  
      • This indicates FDA is shifting to test accuracy, not one-by-one pairings of genes with different drugs, for indications.  
      • FDA's own paperwork reviewing the FMI test (58pp) and the MSK test (57pp) show extremely high NGS read accuracy for both tests.  I'm a board certified pathologist; if both tests are 99%+ accurate at reading sequences from paraffin blocks, then their reports are the same, and what's left to debate?
        • The FDA required studies showing that the FMI test is >90% concordant with prior PMA single gene tests.  See online.
        • The FDA required studies showing that the MSK test is extremely reliable and accurate for sequencing, and its reported genes are clinically valid based on literature (including FDA labeling).   See online.
      • Nutshell:  Private payer policies and MAC LCDs may not see as much difference as CMS does between accurately-measured gene reports and accurately-measured gene reports labeled as CDx. 
  • But wait! There's more.
  • (Difference Two)   The draft had CED-based NCD coverage for MSK IMPACT.  The final has no automatic coverage for MSK IMPACT (since it's not released as a CDx so-labeled by FDA).   
    • FDA allows 510(k) gene panel tests based on accuracy + clinical literature.  That's the road IMPACT took.   However, and experts tell me this is new, FDA has also begun clearly 510(k) tests if they have high accuracy + some clinical study data.  See the December 2017 approval of the MolecularMD BCR-ABL test, which is a 510(k) test supplemented with a few pages of clinical data. It's K173492.
    • FLAG.  Don't try to interpret the NCD without referring back and forth to the FDA Companion Diagnostics Page.   
    • See Medicare Nerd Question 1, below.
  • (Difference Three) For NGS, companion diagnostic tests, CMS looks to the test being cleared or approved as a companion diagnostic, "AND," the patient for whom the test is ordered must have an on-label indication (e.g. "lung cancer.)   
      • This means the NCD covers the Thermo Fisher Oncomine test for lung cancer; the FMI F1 test for lung cancer, colon cancer, and a few others.  
      • However, my reading is that CMS only looks to the test labeling ("lung cancer.")  So such a test is usable in all advanced lung cancer patients, even if a particular on label drug has narrower labeling (like only "second line lung cancer patients that have failed cisplatin.")   
  • (Difference Four) Liquid Biopsy Tests.
    • The draft had no coverage for LDT gene panel LBx tests.
    • In the final, prior to FDA approval,  NGS LBx panels can be covered under MAC review.
    • As they become available, FDA authorized "CDx" NGS panel liquid biopsy tests will be covered under the same conditions as other NGS tests.   
    • CMS states later in the NCD discussion that it doesn't distinguish NGS tests by sample type. 
    • Note that FDA defines the recently created NYS/510(k) accelerated pathway as being for tissue tests, not LBx tests.
  • (Difference Five) There's no "Coverage with Evidence Development" / CED. Period.  
    • Pretty massive CED had been in the draft.
  • (Difference Six) LDT NGS tests can get MAC LCD coverage.  
    • Local MACs aren't obligated to cover them, however.
    • This LCD escape clause isn't a free ride.  
    • (A) NGS LDTs that are in the 28 MolDX states get extremely close technical review which can take hundreds of pages and months of time.
    • (B) Other MACs have been pretty frugal about covering gene panel tests at the LCD level.
    • When a test like FMI CDx is used "off-label," for example in kidney cancer, it can be covered by this local LCD option.
    • If I were MSK IMPACT, I'd quickly publish a study that I matched with FMI F1 99% of the time, if I could, and bring that to MACs.
    • There's one restriction, the NGS test must be used in a patient with advanced cancer, even under the LCD pathway.
A few interesting side issues include:
  • (Side Issue One) Germline NGS tests are "OK" - when you read the fine print!
    • The full NCD needs to be read, not just the coverage section.
    • First, be aware that today, all states cover LCD based NGS testing for germline cancer risk in patients with personal history of breast cancer and personal history of colon cancer.  
    • But, the opening section of the NCD warns that it allows NGS testing by NCD or LCD "only" for patients with is advanced cancer.
    • However:  In the body of the NCD we read:
    • "We do note this decision is not applicable to all diagnostic laboratory tests using NGS, but rather to a unique diagnostic laboratory test that uses NGS for patients with cancer to manage the patient’s cancer by identifying either targeted therapies with known efficacy or in some cases, eligibility for a cancer clinical trial."   
    • Since BRCA panels and Lynch panels aren't for "targeted therapies" they seem to be exempt from review under the NCD.  Assuming the sentence I've just quoted is operative and interpreted the same way by CMS and MACs. 
    • Bad drafting!  I would have put this clause in the opening sentences and not buried in the 60 page text. The sentence appears on page 4 of my PDF version.
click to enlarge (p4 of 60p NCD)

  • (Side Issue Two)  Could the NCD lead to "Excess Coverage?" 
    • Serial testing by different tests?  The NCD seems to cover multiple serial, different FDA reviewed tests, without much restraint, as long as the same one isn't used twice.
    • This is easier for LDTs, which could make a 350 gene version, a 351-gene version, a 352-gene version.
    • CDx Tests Covered Even If CDx Genes are "known positive" or "known negative."  For example, the multi-thousand dollar FMI F1 CDx test is covered in melanoma, even though it only has one approved gene in melanoma (BRAF).  Even if a patient's cancer is already known to be BRAF positive or negative, CMS seems to cover the FMI test.  This undercut the high importance given by CMS to pivotal on label CDx results as a justification for testing.
    • No "value based purchasing" - everything is equally covered..  For example, if Test A has three genes in lung cancer, each at 10% frequency, and a Test B has one gene in lung cancer, at 1/1000 frequency, both are covered.  It's hardly "value based purchasing..."
    • No limits on block age.   From time to time, a patient relapses after 5-10 years.  The NCD has no restrictions on age of blocks.  CMS probably assumes clinicians will use common sense.
    • Duplicate testing.  Medicare's Correct Coding pathology manual has the commonsense policy that CMS won't pay for testing the same analyte twice.  So if you run the FMI panel, you probably shouldn't later run individual ALK, EGFR, etc., tests.  However, if you run the single gene tests first, the NCD does protect payment for the FMI test, as far as I read it.  (I'm not saying I'd actually try this.)
    • The MSI elephant in the room.   Keytruda is approved for all advanced cancers, so an NGS test with MSI at the "FDA Compansion Diagnostic" indication level  would be covered for all advanced cancers.    
      • However, the test would have to be "approved or cleared as a companion diagnostic for MSI."  
      • The Medicare-covered population would probably be "all advanced cancers" if the MSI test is so approved, because Medicare doesn't plan to read through to the actual more restrictive Keytruda drug labeling (advanced cancers that have failed lines of therapy and have no options).  
  • (Side Issue Three)  Discourage migration to NGS platforms for non advanced cancers?
    • The NCD certainly encourages migration of tests onto FDA/NGS platforms if they are tests for therapies in advanced cancers.  If you move your Sanger or PCR test for advanced cancer to an NGS platform (and it's an FDA companion Dx) it's auto-covered by the NCD.
    • But, the final could discourage migration of diverse tests to NGS platforms for indications for non advanced cancers, since they can't be covered, since cancer is non advanced and it's a test that employs or deploys NGS.   
    • For example, if the Oncotype test for non metastatic breast cancer migrated from PCR to NGS, it would become non-covered, which is weird.  It would fail twice under the NCD: it would be for non advanced cancer, and, it would fail to be a companion diagnostic.
    • Mitigation 1:  This may be muted by the text in the body that says it applies the NCD only to tests intended "to identify targeted therapies."
    • Mitigation 2:  I suspect as soon as someone actually had in hand a real FDA approved NGS test for non advanced cancer, CMS would revise the NCD.
    • Mitigation 3:  The LCD section is not restricted to any particular indication for NGS testing.   But it is restricted to advanced cancers X 1 use per test.
  • (Side Issue Four)  Tests must be ordered by treating physician and that is not a pathologist.  
    • See Q&A discussion inside NCD.  Normally tests must be ordered by "treating physician."  There's a longstanding CMS exemption for a pathologist to order a test in order to complete a report (CMS BPM Ch15:80.6.5).   CMS says don't use the exemption here.   Ordering a $4000 test to complete a tumor report seemed like a stretch; CMS confirms it's a no-go.
  • (Side Issue Five)  Creates a very high incentive to make 510(k) type "Companion Diagnostics."   Even an FDA expert I talked to thought that the "companion diagnostic" label was available only to PMA tests.  PMA devices involve high regulatory costs, fees, annual reports, etc.   With the coverage guarantee from CMS applying to either 510(k) or PMA CDx, developers will be looking closely at how the MolecularMD test got CDx indication on a 510(k) platform in December.   
    • This is another flag to how fast the FDA is changing the options.  
    • In June 2017, the Thermo Fisher Oncomine test was approved for 3 CDx genes, and only allowed to report 20 additional genes most closely tied to lung cancer.  
    • In November 2017, FMI is allowed to report a larger handful of CDx benes, but throw in 350 extra genes of highly variable clinical utility.   
    • Then in December 2017, a BCR-ABL test gets CDx status on the less costly, less regulated 510(k) policy platform.
The NCD is 62,000 words long and prints, for me, at 60 pages.

Here's my graphic of current FDA approaches to CDx labeling.  Full deck here.

click to enlarge

Medicare Nerd Points

Medicare Nerd Question 1.
Are 510(k) tests really not covered at all?

This one is a doozy.

"Tests that gain FDA approval or clearance as an in vitro companion diagnostic will automatically receive full coverage under this final NCD, provided other coverage criteria are also met."

Wait for the rest of CMS's requirements.  The test must be cleared as a companion diagnostic AND have an indication for THAT patient's cancer (e.g. lung cancer).

PMA approved tests are approved "as a companion diagnostic" but 510(k) tests are very rarely "cleared as an in vitro companion diagnostic."  See the FDA's list of "companion diagnostic" tests here.  It lists drugs (NDA/BLA) and then tests (PMA) although the page title says "cleared or approved as CDx" so it is possible for a 510(k) to be a CDx.  One isolated example being MolecularMD BCR-ABL K173492 - the test is a Class II 510(k) clearance but the validation included two explicit clinical studies, which is why it ranks on the FDA CDx webpage.

The labeling for FMI CDx F1 clearly states it is approved as a companion diagnostic for a limited table of genes and drugs (this table doesn't currently include F1's MSI or TMB signatures and certanly doesn't include all the hundreds of genes).

The FDA CDx home page indicates that the pathway may be PMA, 510(k), or Humanitarian Device, but virtually all of the CDx's are PMA.

CMS will depend on the FDA CDx table.   This rolls up older PMA CDx whether or not FDA in older times used the term "CDx" as prominently or consistently in labeling.  PMA CDx are important to list, since you can get a new PMA test by matching the performance of an existing PMA test. 

Another FDA source of information is their "food pyramid" and attending text about CDx claims that FDA published in NOvember.  To be a CDx, you must have Analytical validity + direct correlation with clinical trial OR concordance in paraffin blocks to a prior CDx test

510(k) tests can now be released that report gene panels for cancers, where to  clinically significant mutations, you can have high analytical accuracy, no clinical studies, but compile the relevant clinical literature.   See here

Let me use an example.  So one test, you measure ALK fusion 99% accurately and also show it matches at least 90% of the time to a prior PMA ALK test which correlated with clinical results of Xalkori.   In this box, you'll find the FMI F1 CDx PMA test.  The next test, you measure ALK fusion 99% accurately and know that correlates with clinical results of Xalkori based on this PMA labeling and an NEJM article.  In this box, you'll find the MSKCC IMPACT test. 

Medicare Nerd Point 1.  
Advanced Cancer versus the Diversity of Cancer

NGS tests[for therapeutics] can be covered in advanced cancer, but never covered in non advanced cancers.
  • Is glioblastoma an "advanced cancer?" by CMS's specific definition?   It's not usually recurrent (it doesn't go away) and it's not metastatic and not usually referred to as Stage III or IV (which varies by cancer but for most solid cancers is based on anatomy and distant metasasis.)  
  • Is leukemia an "advanced cancer?"  It's not always "recurrent" and isn't commonly called "metastatic" (leukemia cells aren't found in places foreign to white blood cells, and who hears, see the patient in room 209 with metastatic leukemia?)   There are "staging" systems which really look like "cell classification" systems in which CMS can't find a way to describe any advanced stages and terms like "III" and "IV" aren't used - here.

Medicare Nerd  Point 2. 
On label versus Compendium cancers.

Based on a long history of policymaking, CMS generally covers FDA-labeled and Compendium-endorsed drugs at parity.   However, the NCD is specific to FDA labeling for the test and its drugs. The NCD has no entry point for compendium endorsed drugs, except for the LCD clause at the bottom of the coverage section.

Medicare Nerd Point 3.
Time Needed for Real-World Implementation of the NCD.

NCD's are legally binding on the day they are issued.  However, MACs usually only implement them per CMS coding and implementation instructions that may appear months, even a year, later.   With an NCD of this immediate impact on hundreds of thousands of patients, let's see what happens.

Medicare Nerd Point 4.
How Will CMS Handle "Medicare Advantage" Financials?

Medicare Advantage Plans must cover what is covered by CMS.   However, there is a concept where some large NCDs or other abrupt coverage expansions are not the financial obligation of MA plans for the first year or two.   It remains to be seen if CMS rules this NCD is an immediate coverage and cost expansion for MA plans, or, if it will use the option to run NCD-based claims through regular Part B for the first year or two for MA plan patients.

Another aspect of the NCD is that Medicare Advantage plans seem obligated to cover all PMA approved gene panel tests even if they would rather contract and discount with a few preferred venders or based on the MA plan's assessment of test quality and usefulness.  Again, the NCD doesn't include "value based purchasing."


Medicare Nerd Questions remind me of another project I am working on this year.

In sepsis policy, CMS has a 2015 metric called SEP-1 which requires several steps to be executed within 3-6 hours of diagnosing sepsis.  The key bullets are very brief.  However, CMS has released a 150 page rulebook for implementing this metric, and many implementation issues are still contentious.  (How do you define the minute when sepsis is diagnosed?  If antibiotics must be administered, which drugs count?  All these issues lead to pages of discussion and terrorizes providers.)   Similarly, the coverage section of the NGS NCD is short, but implementation across a wide range of cancers, drugs, labeling variations, and constantly updated FDA tests and indications will require a rulebook.   (Similarly, the CED proposal in the November proposed NGS NCD was short to write and read, but would have been very complex to implement, and as originally written led to some nonsensical implications.)


CMS NCD Final Coverage Text:

A.  Coverage

The Centers for Medicare & Medicaid Services (CMS) has determined that Next Generation Sequencing (NGS) as a diagnostic laboratory test is reasonable and necessary and covered nationally, when performed in a CLIA-certified laboratory, when ordered by a treating physician and when all of the following requirements are met:

Patient has:
  • either recurrent, relapsed, refractory, metastatic, or advanced stages III or IV cancer; and
  • either not been previously tested using the same NGS test for the same primary diagnosis of cancer or repeat testing using the same NGS test only when a new primary cancer diagnosis is made by the treating physician; and
  • decided to seek further cancer treatment (e.g., therapeutic chemotherapy).

The diagnostic laboratory test using NGS must have:
  • FDA approval or clearance as a companion in vitro diagnostic; and
  • an FDA approved or cleared indication for use in that patient’s cancer; and
  • results provided to the treating physician for management of the patient using a report template to specify treatment options.

B.  Other

Medicare Administrative Contractors (MACs) may determine coverage of other Next Generation Sequencing (NGS) as a diagnostic laboratory test for patients with cancer only when the test is performed in a CLIA-certified laboratory, ordered by a treating physician and the patient has:
  • either recurrent, relapsed, refractory, metastatic, or advanced stages III or IV cancer; and
  • either not been previously tested using the same NGS test for the same primary diagnosis of cancer or repeat testing using the same NGS test only when a new primary cancer diagnosis is made by the treating physician; and
  • decided to seek further cancer treatment (e.g., therapeutic chemotherapy).



Endnote One.
Volumes of Data on Test Results Will Have No Tracking

While this effort is closed for now, the testing will generate volumes of genomic data on Medicare patients and will lead to large amounts of drug administration whose outcomes aren't well enough tracked.   See a current March 2018 article in JAMA by Feero, tied to a March 16 National Academies publication on learning healthcare systems in genomics. 

I've argued that if you had test, cancer, test results, and Part B and D drug records, it would be enormously important, directionally correct real world evidence.  The only part CMS is missing is the tests.  Such data would start to enormously improve the work of tumors boards and provide return for CMS's now-finalized billion dollar investment in tumor gene panels; see Patel et al. 2018 and see Perera-Bel et al. 2018.  While groups like TAPUR move in this direction, and Syapse facilities sharing among users, and the FMI-Roche-Flatiron provides a giant-size proprietary footprint in this region, CMS/HHS should be doing something, because important issues like racial health disparities in genomic data are involved (see here, and see also the NAS report).

Endnote Two.
How Will MACs Handle New Opportunities for Off Label Drugs?

Some have argued that the broad sequencing will release a lot of data to clinicians of Medicare patients regardless of whether local MACs or Medicare Advantage will cover off-label drugs, the net risk-benefit of off-label drugs, etc.   CMS sidestepped a Q&A question on this issue by treating it as primarily one of speculative emotional stress rather than speculative medical risk/benefit.  For a paper on potential emotional stress, see JAMA Oncol 2017 here

Endnote Three:  
Why Use a $100 Test If Medicare Covers a $4000 Test?

Regarding the possibility that patients will get a $4000 NGS MSI test (on a panel) when they only "need" a regular $100 MSI test, CMS stated that cost considerations were out of scope of their review.