Friday, December 8, 2017

The Foundation Medicine NGS NCD: Did CMS Accidentally Roll Over Leukemia and Liquid Biopsy?

On November 30, 2017, CMS released a detailed and complex NCD that originated in a Parallel Review assessment of the newly FDA approved Foundation Medicine CDX test based on paraffin block analysis in solid tumors (here).

The NCD, which is under comment until December 29, provides coverage for PMA-approved NGS tests as long as they are used in a patient with an on-label cancer for that test.  It covers the Foundation One PMA CDx test, as well as other PMA-approved NGS tests like the Thermo Fisher OncoMine Target Dx tests.  Tests are defined as covered now and on a rolling basis in the future.

However, the NCD states that it governs use of testing in all advanced cancers when NGS testing is used.   The NCD doesn't mention differences between solid tumors paraffin based tests and liquid biopsy based tests, and it doesn't call out any differences for tests in hematopoeitic cancers (leukemia/lymphoma.)  The NCD would have unexpected results for patients needing a liquid biopsy test under NCCN guidelines, for all patients who have leukemias or lymphomas, and for all Medicare patients with have cancers outside the top 4 or 5 like lung and colon.  (Medicare Advantage patients could get medically necessary tests in these categories, but no available coverage would be possible for patients in Traditional Medicare.)  The NCD works differently for each group, as presented below.


Here is an overview of the landscape.  (FIGURE 1).   Working from the bottom of the chart upwards, there are solid cancers, and hematopoeitic cancers.   Solid tumors can be assessed with paraffin block tissue; or NCCN recommends oncologist consider liquid biopsy when solid tissue isn't available.   Testing has progressed from single gene tests (e.g. BCR ABL alone) to multiple gene separate tests to gene panel tests.   NCCN guidelines cover all of this, and there are pretty tightly written LCDs.   Overall usage of gene panel codes is not high, about 1000-3000 uses for codes 81445, 81450, 81455 in Medicare data for 2016.  This indicates that LCD controls are tight.  As FIGURE 1 shows, in the upper left, the chart culminates in 2017 with the FDA approval for several paraffin block based, solid tumor NGS tests.

click to enlarge

AFTER the FDA had issued PMAs for several paraffin block test, and in coordination with the FDA, CMS created the NCD, as shown in FIGURE 2.

click to enlarge

So far so good.

The tricky part is, CMS waited until there were several paraffin NGS panel tests on the market before releasing the NCD.   But the NCD provides coverage ONLY for PMA approved tests, and the FDA' library of PMA-approved tests is currently very incomplete compared to the wide range of major human cancers.   Over half of cancers have no PMA approved CDx, immediately shunting hundreds of thousands of patients into complex and very costly "CED" immediately, on the day the NCD is finalized.


Yikes.  There aren't yet any PMA approved NGS panel tests for hematopoietic tumors, or for liquid biopsy in refractory scenarios in solid tumors, or for over half of the incident U.S. cancers (pancreatic, leukemia, lymphoma, kidney, etc).  So we get FIGURE 3:

click to enlarge

The text of the NCD provides de facto coverage only for paraffin tumor block NGS tests used in a small number of solid tumors (far left column of Figure 3), because these are the only type of PMA tests approved by FDA so far

Wait, what about the CED coverage in the NCD?   

CMS provides back-up coverage by allowing Coverage with Evidence Development (CED) clinical trials for gene panel tests, if they have NYS and FDA 510(k) clearance.  But there are still two big problems, big enough to make this channel essentially impossible.

Problem 1:  Many gene panel tests are NYS approved, but FDA hasn't even created a 510(k) pathway for leukemia, or lymphoma, or liquid biopsy tests.  Nor has FDA approved any tests in these several broad categories.  That's a pretty big problem.  These are the entry points for the CED channel, and they're not even opened yet. 

click to enlarge

In the preceding pie chart, I estimated incidence of advance cancers based on NCI death rates; mapped the cancers with at least one PMA NGS CDx gene (green); categorized other solid tumors as CED (red); allow about 5% for solid tumor liquid biopsy per guidelines (pink); and separately show the population for leukemia/lymphoma (maroon).   The total pie chart in the US represents 450,000 advanced cancers (deaths); as a rule of thumb you could estimate 1/2 of that is the Medicare, >65 population.  So far circa 100,000 Medicare patients, the NCD covers the $3000 NGS test; for the rest of the circa 100,000 patients, it would cover the $3000 FMI NGS test off label but require ?$10,000 or $20,000 of CED as well.  If it was all paid for, it would be 100,000 PMA NGS patients for $300M (green slice), and circa 100,000 off label CED patients at a similar $300M for testing and perhaps 100,000 x $10,000 or $1B for CED.

Problem 2:  In addition, in all practical terms, CED trials for 510(k) gene panels tests and off label PMA tests will be impossible as proposed.

The most severe issue is that the CED requires RECIST testing (special whole body imaging tests, typically conducted monthly) in order to have accurate ORR and PFS data.  This could mean monthly, $1000 scans surveillance imaging (MRI or CT) - $12,000 in the first year.  Obviously, a lab can't spend $10,000-20,000 on CED research in return for the patient's Medicare payment of $600-$2900 test.  It's a non starter.[*]   This concern - that CED is impossibly costly - also applies directly to all the Medicare cancer patients who are unlucky enough to have each and all of the many solid tumors that aren't approved at the PMA level on Foundation's test.
Cost of Proposed CED?  I peg the cost of the CED section of the NCD at two billion dollars. NCI lists about 450,000 patients dying each year of the 13 major cancers.  Of these, about 250,000 have NGS CDx tests (lung, colon, etc).  About 200,000 do not.   If half of these advanced cancer patients are Medicare patients, it's 100,000.  If the CED with monthly RECIST studies costs $20,000 per year, the cost of the CED research would be  $2B annually.   See my table based on NCI data, here.  
An additional problem, for Medicare patients with leukemia, is that the CED is written to require mandatory RECIST testing, which is impossible for these patients.  Dr. Gottlieb of the FDA strongly supported the NCD in testimony on the Hill on December 7; but today if he was a Medicare patient and had a relapse of his lymphoma or a recurrence as leukemia, he couldn't get any genomic testing.  So either CMS envisions no genetic testing in leukemias, or, would have to admit to substantially mis-writing the basic scope of the NCD and/or major high level requirements of the CED, despite the NCD being the result of a 2-year process.

CMS should defer imposing either [non coverage or $20,000 CED registries] on cancer patients before there are at least several PMA approved test for that patient's category of cancer.  

Issues of quality are over-rated in the NCD.   Recall that FDA now equates NYS approval with FDA clearance and clinical use of tests (e.g. the MSKCC IMPACT test was cleared in this new pathway).  And tests like Foundation One are essentially the same before and after the FDA approval.  And all cancer patients in Europe, Canada, Japan, Australia are getting NGS testing without FDA PMA approval.   And the FDA has regularly approved major on-label drugs on a rolling basis (based on KRAS, ROS1, MSI testing) ahead of the existence of any PMA approved test, suggesting the FDA isn't as concerned as CMS is.  Concern about test quality has a place, but doesn't need to be directed against presently NYS and CAP approved tests that are in NCCN guidelines and are in pipelines for FDA approval.  (CMS also seems to conflate the medical necessity of gene-drug pairs that have been FDA endorsed, like ALK-Xalkori, with the function of analytically accurate tests for the same gene).  For broad categories of patients that have no PMA-approved gene panel test, in the short term, over the couple of years, LCD coverage has already proved to be stringent and frugal and it will be enough to control overuse.

Note on Process.  NCD clearly directly only to solid tumors in its body.  CMS didn't justify "non coverage" in these additional cancer areas (shown in red).  CMS studied the NGS literature in paraffin tissue from solid tumors, using Pubmed search keywords like "lung cancer" or "melanoma" (not leukemia or lymphoma) and assisted by Foundation Medicine publications on its paraffin based test use in solid cancers.

Also, as mentioned above, CMS demands that all patients in CED must have RECIST studies, which makes no sense in leukemia patients, and suggests the NCD authors were not actually thinking of applying the NCD to leukemia/lymphoma.  This implies that CMS may have miswritten the legal scope of the NCD (all cancers) relative to the available body of facts and research inside it (some cancers).

Another Note on Process.  NCD predicated on today's 510(k) & PMA terminology, while FDA is racing ahead to other concepts.  The NCD is based on the current 510(k) and PMA processes.  These have been changing substantially; for years, FDA refused to give 510(k) approval based on accuracy of genetic testing in CDx tests, and then it opened the 510(k) pathway in cancer for analytically accurate tests with NYS approval in November.   The FDA is also opening up important non-PMA, non-510(k) pathways for germline human genetic tests, such as for hereditary recessive genes, an enormous change in policy based  fundamentally on the accuracy of modern NGS testing and sequencing (82 Fed Reg 51560, 51563, 51567, links here).  The NCD doesn't contemplate these recent FDA publications.  In addition, if Congress passed a new diagnostics law, it would create terminology and approval categories other than PMA and 510(k) for diagnostics, which the NCD doesn't contemplate.

A final note on Access and Process.   CED Channel is Burdensome, but FDA itself approves drugs ahead of their CDx tests. Sometimes by a couple years.   The NCD has only a VERY burdensome CED channel, and it's available only to 510(k) or PMA approved tests, but is triggered whenever a cancer is off label (e.g. outside the top 4 cancers).  However, on a regular basis, FDA has released drug approvals ahead of any PMA test.  For example, there was approval for use of KRAS testing a full couple years before a PMA KRAS drug.  There was approval for ROS1 drugs long before  a PMA ROS1 test (the Oncomine test in June 2017).   There is currently approval, right now, for MSI based drug use for Keytruda, but no PMA test.  (MSI is in the FMI test, but it's not PMA CDx).  All of these categories of gene-drug get shunted into burdensome or impossible CED if and when the FDA drug approval precedes the PMA test approval by a year or two.   Another example of problems in this category would be the discovery of important genes that act on the effectiveness of generic drugs, or limit the optimal population for an existing drug.  These aren't going to be funded and researched by wealthy biopharmas and will never reach on-label PMA gene tests, no matter how strong the evidence and how wide the use.   No one is going to spend $50M out of their pocket getting FDA PMA approval for a generic gene on a generic drug.

[*] Technically, CMS can cover $20,000 in RECIST research costs or other costs implied by the rigorous CED.  Congress gives CMS carte blanche to cover costs related to research that furthers any purposes of AHRQ (which is to study and improve the health of the US population).  If it's $20,000 per CED cancer patient x 100,000 patients, that's $2 billion.  I don't believe CMS has ever attempted to cover research costs in CED rather than only medical costs, but CMS can pay for costs that are not medically necessary but are necessary for AHRQ research.