On September 23, the FDA’s Molecular and Clinical Genetics Panel (a group of outside experts the agency convenes to review medical devices, including diagnostic tests) will meet to review GRAIL’s premarket approval application (PMA) for Galleri. No other multi-cancer early detection (MCED) test has gotten this far in FDA review. The two pathways a test can take to market, premarket approval and the laboratory-developed test route, are covered in Regulation and Reimbursement.
The panel’s vote is advisory-only. The FDA usually follows its advisory committees, but it is not required to, and the agency’s decision on the application will come later. In the meantime, the test has been sold as a laboratory-developed test since 2021, which means people already have results in hand and clinicians have already been handed them.
I have split this post into three seats: (1) the person who has taken the test or is thinking about it, (2) the clinician who gets asked about it, and (3) the investor or operator watching the vote.
What the Panel Will Be Looking At
Three studies carry most of the weight, and each has a corresponding journal club post (PATHFINDER, PATHFINDER 2, and NHS-Galleri). The FDA’s executive summary, posted two business days before the meeting, reported a different set of numbers for a different version of the test. Both trials ran an earlier version of the assay, which the agency calls MCED-V2, and the version under review is, in the FDA’s words, a different device, with a different assay workflow, classifier, and bioinformatics pipeline. To measure the version under review, stored plasma from a subset of trial participants was re-tested with it. This included everyone whose earlier-version result was positive, everyone diagnosed with cancer, and a weighted sample of the rest, which came to 7,766 of 22,698 eligible PATHFINDER 2 participants and 7,926 of 70,323 in the first round of NHS-Galleri. On those samples, sensitivity for cancers diagnosed within 12 months of the blood draw was 35.0% in PATHFINDER 2 and 31.6% in NHS-Galleri, specificity was 99.85% and 99.74%, and the positive predictive value was 77.0% and 66.2%. For the 12 prespecified cancers, sensitivity was 61.3% and 52.1%. The FDA’s PATHFINDER 2 figures also come from an earlier data cut than the results presented at the ASCO Annual Meeting in June 2026 (25,125 analyzable participants with 12 months of follow-up as of December 31, 2024, against 32,007), so the ASCO figures and the FDA figures differ in the version of the test, the participants included, and the date. The safety data in the review, including the workups that followed a positive result, are from the earlier version, because results from the newer version were never returned to participants.
A note on sourcing: PATHFINDER 2 and NHS-Galleri are conference data (ASCO 2026 late-breaking abstracts LBA10509 and LBA100). Neither has been published as a peer-reviewed manuscript as of this writing, so the full datasets are not yet available for independent review.
Seat One: The Person Holding a Result
The meeting does not change the test. Whatever the panel decides on September 23, the assay run on a blood sample last year is the same assay with the same performance numbers, and the meaning of a result that has already come back does not change with a vote.
For a positive result, the studies above say that about 6 in 10 people with a “cancer signal detected” result were diagnosed with cancer within a year, and about 4 in 10 were not. Sorting out which group someone is in means a workup that starts from the test’s predicted cancer signal origin (the organ the test predicts the signal is coming from), and in PATHFINDER 2 that workup took a median of 48 days. The workup typically involves what is laid out on the After a Positive Result page for each predicted origin.
For a negative result, sensitivity is king. In PATHFINDER 2, the test found roughly 4 in 10 of the cancers diagnosed within a year of testing, and it missed the rest. Every study to date ran the test as an addition to standard screening (mammography, colonoscopy, etc.), and no study has evaluated it as a replacement.
On cost, the test is generally paid for out of pocket (current list prices and the consumer channels are in The Competitive Landscape), and the imaging, procedures, and specialist visits that follow a positive result are presumably billed the way any other diagnostic workup is billed. A favorable vote changes neither of those in the near term, but FDA approval is a prerequisite for Medicare coverage under the law passed earlier this year, and that law sets January 1, 2029, as the earliest date a covered test can be performed.
Seat Two: The Clinician Who Gets the Question
Two things change for clinicians if the panel votes favorably and the FDA follows.
The first is volume. The meeting will be covered widely, and coverage produces questions in the exam room whichever way the vote goes. A test that most clinicians have not been trained to interpret will show up more often, usually either in the hands of a patient who has already paid for it and already has the report or in the form of a question from a patient who is considering the test.
The second is a label. As a laboratory-developed test, Galleri is sold today without an FDA-reviewed statement of who the test is for and how it should be used. An approved PMA comes with an intended-use population and labeling that the agency has reviewed. For a clinician trying to answer “Is this test meant for someone like me?” a defined intended-use population is something concrete to point to that does not exist today. I think clinicians should start developing plans for what happens when they meet a patient with a positive result, and the After a Positive Result page was built for that purpose. It covers each of the 21 predicted origins the test can report.
Then, there is the survivorship version of the negative result question. People who finished cancer treatment more than three years before enrollment were eligible for PATHFINDER 2 and NHS-Galleri, and a survivor with a negative result may reasonably ask whether surveillance imaging or visits can be spaced out. None of the studies evaluated the test as a substitute for surveillance after cancer treatment. So, to me, false reassurance from a negative result could potentially be a harm of the test if a patient is not educated about this.
Seat Three: The Investor Watching the Vote
The panel will vote on three questions, posted with the FDA’s briefing materials on September 21: whether there is reasonable assurance the test is safe for the proposed indication, whether there is reasonable assurance it is effective, and whether its benefits outweigh its risks. The three votes can split, and the discussion around them is important because it previews the conditions the agency may attach to an approval (post-approval studies, or a narrower intended use than the company proposed, among others).
Then there is the precedent. In 2024, the FDA approved Guardant Health’s Shield blood test for colorectal cancer screening after its panel voted 7 to 2 that the benefits outweighed the risks, and the supporting study was a prospective observational study of about 20,000 participants with no randomized controlled trial (RCT). Galleri arrives with an RCT of 142,924 people that missed its primary endpoint and reduced stage IV cancers by 14%, and the FDA has set that result to one side: its executive summary says the proposed indication makes no claim about reducing stage IV cancer and that the agency is not asking the panel whether the test does (the summary reports the first round of screening only, a 3% reduction in stage IV cancers across all cancers and 13% for the 12 prespecified cancers, with no formal statistical testing). What the agency is asking, in its discussion questions, is what the per-cancer numbers mean for patients and clinicians, separately for the five cancers with recommended screening (breast, colorectal, lung, cervical, and prostate) and for the cancers without it; whether the results by stage support the word “early” in the indication (covered in the next section); how the test’s benefits, risks, and limits should be explained to patients and clinicians; and, if the test is approved on the existing data, what data post-approval studies should collect.
The Nancy Gardner Sewell Medicare Multi-Cancer Early Detection Screening Coverage Act was signed into law on February 3, 2026, inside the Consolidated Appropriations Act, 2026, and it added MCED tests to Medicare as a new benefit category. It applies only to tests with FDA marketing authorization. It leaves the coverage decision itself to CMS through a national coverage determination process, and it sets January 1, 2029, as the earliest date a covered test can be performed, at a rate of $509 through 2030. So, approval is necessary for Medicare coverage, and the earliest covered test is still more than two years away regardless of how quickly the approval comes. Until then, the test stays in the channels it is in today: out-of-pocket purchase, the consumer partnerships, and a very limited amount of insurance coverage.
For the rest of the field, a favorable vote sets the bar. Guardant holds a Breakthrough Device Designation for Shield MCD but has not filed a PMA, and Abbott (which acquired Exact Sciences) is selling Cancerguard as a laboratory-developed test while running an FDA-reviewed real-world evidence study. If Galleri is approved on the strength of PATHFINDER 2 and NHS-Galleri, every other test will presumably be measured against that evidence package.
The Word “Early”
The proposed indication says the test is intended “for screening for the early detection of multiple types of cancer, in adults aged 50 years or older,” and the FDA is asking the panel whether the results by stage support the word “early,” and if not, whether they support an indication for detection without it. In PATHFINDER 2, the version under review detected 21.4% of the stage I cancers diagnosed within a year of the blood draw, 33.3% of stage II, 58.1% of stage III, and 63.2% of stage IV. In the first round of NHS-Galleri, the figures were 13.6%, 34.4%, 44.3%, and 60.0%. The FDA flags a bias in figures like these, since a cancer the test missed is staged whenever it is (later) diagnosed, and it asks the panel to keep that in mind.
The numbers by cancer type point in the same direction. For the five cancers with recommended screening, sensitivity was 22.9% in PATHFINDER 2 and 27.5% in NHS-Galleri. For the cancers without recommended screening, it was 47.3% and 37.3%, with liver and bile duct cancers at 81.2% and 80.0% at the top and skin cancers (other than basal and squamous cell), myeloid cancers, and thyroid cancer at or near zero.
These are the stage numbers behind the negative result point in Seat One: in both studies, most stage I cancers diagnosed in the year after the draw were not flagged.
The Case for Deciding Sooner
On September 9, two weeks before the meeting, a Viewpoint was published in eClinicalMedicine by Peter Sasieni of Queen Mary University of London and Stephen John of the University of Cambridge titled “The clinical consequences of excessive caution: rethinking how cancer screening policy is made.” Sasieni is the lead statistician on NHS-Galleri and, per the paper’s declaration of interests, a paid member of GRAIL’s Scientific Advisory Board. The question it takes on (how much evidence a screening test needs before it is offered) is the same question the panel will be answering for Galleri. The Wall Street Journal’s editorial board made the case for approval on September 21 (“A Grail Cancer Test for the FDA”), under the subhead that the test “is proven to increase early detection that can save lives.”
Their argument is that screening committees treat starting a program that turns out to be a mistake as the main risk, and treat waiting as neutral, when waiting has its own cost. The example they build the arithmetic on is bowel screening in England. Two RCTs showed fewer colorectal cancer deaths with stool-based screening in 1996, and the English program did not finish rolling out until 2010. By their estimate, screening prevents about 2,000 colorectal cancer deaths a year in England, so a rollout in 2000 could have averted 20,000 premature deaths. They run the same arithmetic for prostate cancer screening, conditional on an assumed 25% reduction in prostate cancer deaths, and get a similar figure for a 20-year delay.
What they propose in place of the current process is (1) counting the harm of not starting a program that would have done more good than harm alongside the harm of starting one that does not, (2) replacing the yes-or-no decision with a publicly funded pilot designed to generate the missing evidence while the program runs, (3) having expert committees summarize the evidence and its uncertainty without making a recommendation, and leaving the decision to elected representatives and stakeholder groups, and (4) where the tradeoffs depend on personal values, offering the test without encouraging it. On MCED tests specifically, they write that an RCT across a group of cancers needs more than 100,000 people, that testing each cancer type separately would need millions of people, and that by the time one version of a test has been evaluated for clinical utility, the technology will be outdated. I agree with quite a bit of their diagnosis.
In the US, the pilot the authors describe has, in a sense, been running since 2021. Galleri has been sold as a laboratory-developed test for five years with no study required as a condition of sale.
What I Think
I have no prediction for the vote. I do think the number of questions about this test will go up after September 23, whichever way it goes. I will write up the meeting here after it happens, including what the briefing documents say and how the panel votes.