Tuesday, January 9, 2018

The Strangest Ever: FMI CDx at MAC, FDA, and CMS Without Review of Panel Utility

On November 30, 2017, CMS released a draft blanket coverage determination for the Foundation Medicine F1CDx test for all cancer patients with tumors for which it is on-label, expanding on a rolling basis with label expansion. 

There are some unusual follow-on aspects of the decision.  For example, if the F1CDx covers one PMA gene in a certain cancer, and even if that gene is already known positive or negative based on the prior IVD that F1CDx is itself based on, F1CDx is still covered.   (Let's say for a million patients at $3000, or $3B).   This seems to imply F1CDx has some value beyond its on-label genes.   However, neither FDA nor CMS ever broached that conclusion.   Also puzzlingly, there will be some cancers where the F1CDx points only to off-label drugs for 95% of patients (especially if common mutations like EGFR have already been filtered out by prior testing).  Despite requiring CED for all other testing approaches, the NCD doesn't require (or misses the opportunity to require) any form of records, tracking or CED for this huge cancer population, as long as their pathway to off-label drugs was via "failing to have an on-label drug on the F1CDx test."

Everybody "Defined-Away" the Clinical Utility of Large Panels

There has been huge debate in the US and world oncology literature on the value of large gene panel tests, and Vinay Prasad of OHSU has publicly protested that the F1CDx coverage decision is premature (here).

In succession, MolDX, FDA, and CMS defined away this question in their reviews. 

MolDx Decision 
The MOLDX LCD for FMI F1 (L36198) doesn't state that 300- or 400 gene panel tests are medically necessary.   Rather, the LCD discusses that for one or two identified genes, like EGFR, if you do an FDA IVD point mutation test, it will pick up certain common mutations, and if you sequence that known PMA biomarker gene, you will pick up several extra actionable mutations (e.g. 17% more).  What is done with, or why, regarding the other 300 or 400 genes that the test reports simply isn't brought into the question.

FDA Decision
Go to the FDA F1CDx webpage (P170019, here) and look up the safety and effectiveness review document.   In the pivotal risk-benefit determination, page 55, section C, FDA states that "F1CDx assay has demonstrated non inferiority to companion diagnostics [reviewed in Section X] and therefore, does not introduce additional risks above [the] other approved devices."   FDA has been able to define away any actions with any of the additional several hundred genes since they are not on-label claims, so, as judges say, FDA need not reach or speculate towards any conclusion on the large panel, in order to approve F1CDx. 

FDA also comments that the F1CDx was not referred to an advisory panel because all questions or issues in the F1 PMA duplicate information already reviewed (e.g. the prior review of the single EGFR gene for erlotinib.)

CMS Decision
Finally, while CMS's NCD is 40 to 60 pages long, depending how you print it, the pivotal decision for coverage of F1CDx at the end is extremely brief, and merely states because F1CDx includes PMA on label genes, F1CDx is "reasonable and necessary."   Any emergent or associated collateral risks or benefits, or whether the test "meets but does not exceed the patient's need" (a traditional CMS metric), simply don't enter into the final decisional sentences.

Is This Good Or Bad?

I mostly see the NCD as having a lot of weird consequences that people will look back in in 3, 6 or 12 months and say, "Wow."   For example, leukemia seems to be under the NCD (it's oncology), but there are no PMA gene panel tests, and no 510(k) tests, so coverage of genetic testing is available only in NCI trials, and for the most common genes and drugs, those trials won't even exist any more.  (Alternately, leukemia patients must be in C.E.D., for which CMS mandates RECIST imaging testing for participation, which leukemia patients can't do, so they can't get the C.E.D. arm.)   Another "wow" is that very soon, massive numbers of Medicare patients without any on-label genetic results will get lots and lots and lots of off-label genetic results, for which no plan is made.

My main point is that one would think that along the pathway to this decision, CMS or FDA would have "had to have" looked at clinical utility (or net risk benefit) of large gene panel tests, and they didn't, at least not in the pivotal decisional parts of their documents. 


I posted a very informal, freehand summary of the FDA concordance studies for F1 CDx, here.