Header: Everyone is aware that in September, FDA issued rulemaking to take over regulation of all lab-developed tests in about 4 years. Here, I review briefly then add two thoughts.
Last fall, FDA released a regulatory proposal to take control of essentially all lab developed tests (LDTs) in about 4 years, a proposal that it estimated (obscurely) would cost some $40B over a few years, which is 200,000 man-years of regulatory expert time at $200,000 per person per year. See my white paper.
By December 4, many thousands of comments were submitted, which, subtracted form letters, over 1000 were individualized by entities like ACLA, CAP, Mayo Clinic, AHA, and many others. A few days ago, CMS and FDA issued a joint press release confirming their viewpoint that any form of "enhanced CLIA" was off the table. In a few days, Jeff Shuren MD JD, head of devices at FDA, will be giving a talk on the FDA's current thinking at a Friends of Cancer Research conference in DC (also streamed; Feb 1).
Last week, I had the pleasure of attending the Precision Medicine World Conference in the Bay Area. On Thursday, There was an excellent talk on the FDA proposal by Sheila Walkoff, head of Goldbug Strategies. Afterward, I was toying with two ideas.
Idea 1 - FDA and "Clinical Utility"
Several sources have said that FDA won't evaluate "clinical utility" in this effort, but only "analytical validity and clinical validity." I find this unbelievable, and can show why. What FDA does have to show is that benefits exceed risks, and the value of [benefits - risk] pretty much defines the clinical utility. There might be this difference between FDA and payors or other health technology assessments. The latter will try to quantify clinical utility - for example, dollars per QALY, unnecessary surgeries avoided per 10 tests, and so on. FDA doesn't need to "quantify" clinical utility, but, it darn well does need to verify that [benefits - risks] is positive.
And it can be a tortuous process. See the recent FDA approval of the AvertD pharmacogenetics test. Though accurate, reading the genes, the test failed an FDA ad board because of net clinical benefit and harms concerns (e.g. clinical utility, not analytical validity of sequencing). And the FDA fussed with the test for a couple years, even moving around the regulatory category from de novo 510k to PMA for the final. It was all about benefits and harms, and the balance of each, and the net benefit, which I can't distinguish from "clinical utility" for patient care. My blog here.
Idea 2 - FDA and LDT Risk Category
FDA makes much of sorting each of the thousands of LDTs into "risk category." But first, this isn't how the 1976 device law is defined; it's defined by things on the market before and after May 1976. And second, bigger, LDTs don't come with indications for use, as FDA products do. There aren't any uniform, vetted, "indications for use" in CLIA as in an FDA label. And "indications for use" and connected limits and warnings that set risk level at FDA and fought over for months by companies, FDA staff, and ace regulatory lawyers. This idea of defining test risk (by implication, only with intended use or indications for use) for LDTs is far from a simple one or self-implementing.
Auto-generated illustration (!) based on blog text; GPT4 to Dalle3.