FDA advisers back GRAIL’s Galleri cancer blood test on safety; split on efficacy

An FDA advisory panel voted in favor of GRAIL’s Galleri multi-cancer early detection blood test for adults 50+, unanimously backing safety while splitting on effectiveness; the votes are recommendations, not approval.

HEALTH desk — a blood test that claims to catch many cancers early just cleared a high-stakes FDA panel on safety and benefit-risk, but the split efficacy vote shows the agency still has a hard call before America gets a nationwide multi-cancer screen.

What the three ballots asked, as CancerNetwork prints them. First: is there reasonable assurance Galleri is safe for patients who meet the proposed indication? The panel voted 10 to 0 in favor. Second: is there reasonable assurance Galleri is effective for those patients? The panel voted 6 to 4 in favor. Third: do the benefits outweigh the risk for those patients? The panel voted 7 to 2 in favor, with 1 abstention. Those wordings are CancerNetwork’s. GRAIL’s release states the same three tallies and calls the panel the Molecular and Clinical Genetics Devices Panel. Reuters states the same three tallies and calls them outside experts advising the FDA. The numbers match across the three pages. The panel’s name is not one shared sentence.

What Galleri is, on the pages this desk read. GRAIL calls it a multi-cancer early detection test, shortened to MCED: one blood draw meant to look for many cancers, including cancers that have no recommended screening today. The release says it looks for cancer-specific methylation patterns shared by many cancers, before symptoms appear. Methylation, here, means small chemical tags on DNA. The test reads those tags. CancerNetwork says the tags sit on cell-free DNA in plasma. Cell-free DNA is pieces of DNA floating in the blood, outside a cell. Plasma is the liquid part of blood. When the test reports a cancer signal detected, GRAIL says it is designed to predict the cancer signal origin — the company’s name for where in the body that signal seems to come from — to help guide the next diagnostic test. CancerNetwork adds a third report, returned only after a positive: a supplemental guess of the cancer’s biology, such as cell type. These lines are the pages’. This desk did not process a tube of blood.

How it is supposed to be used. GRAIL says the test is intended in addition to, and not as a replacement for, guideline-recommended single-cancer screenings. A single-cancer screening is a test aimed at one cancer, such as a mammogram or a colonoscopy. Reuters says the committee told doctors to explain the limits so patients do not skip those proven screenings, and that Galleri must be an “addition, not substitution” to standard care. CancerNetwork, citing an FDA executive summary this desk did not open, says a “no cancer signal detected” result does not rule cancer out, a “cancer signal detected” result still needs a medically established diagnostic test, the test can be wrong in either direction, and the benefits and risks of repeat screening with Galleri are still being studied. Those limits are CancerNetwork’s account. They are not a sentence this desk took from an FDA page.

Where the application stands. GRAIL says it submitted the premarket approval application, shortened to PMA, on Jan. 29, 2026. A PMA is the FDA’s toughest path for a medical device. The agency reviews safety and effectiveness before the device can be marketed under that approval. GRAIL says the FDA named Galleri a Breakthrough Device in 2018. CancerNetwork says that designation was August 2018. Breakthrough Device is a program that can speed talks and review. It is not an approval. GRAIL says the FDA is not bound by the committee’s advice, though it considers that advice, and that a final decision on the Galleri PMA is expected in the coming months. Reuters says the agency is not bound but weighs the votes heavily, and that a decision on whether Galleri can be sold nationwide is expected in the coming months. “Sold nationwide” is Reuters’ frame. It is not a sentence on GRAIL’s release, and it is not an approval.

The studies GRAIL names. The release says Galleri is the only MCED test supported by large interventional and randomized studies in the people it is meant for, including studies run under an FDA investigational device exemption. It names PATHFINDER 2 in North America and NHS-Galleri in England. An investigational device exemption, shortened to IDE, is permission to study a device in people before approval. A randomized study assigns people to the test or to usual care by chance, so the groups can be compared. GRAIL also says the program has detected about four to seven times more cancers than standard screening alone, with most of those cancers found earlier, a low false-positive rate, and high accuracy on the origin guess. That multiplier is GRAIL’s claim. This desk did not recompute it.

What Reuters cites, and only Reuters. In a U.S. study, Reuters says, Galleri detected 35% of cancers within 12 months and reached a specificity of 99.85%. Specificity is the share of people who do not have cancer whom the test correctly calls negative. A high specificity means few false alarms. 35% means Reuters says the test found about one cancer in three that was there inside a year. It did not find them all. Reuters does not print the study’s name in that sentence. It says a separate study in Britain was broadly consistent, and that this reinforced a low false-positive rate and the test’s ability to find cancer. “Broadly consistent” is Reuters’ sentence about the British study. It is not a late-stage result, and it is not GRAIL’s four-to-seven-times line.

What CancerNetwork reports from PATHFINDER 2, which is a different set of figures. It says Nature Medicine published the study on 22 Sep 2026, the day before the panel. The study looked at people 50 and older who were not already being checked for a suspected cancer. Of 32,007 patients who finished a 12-month cancer assessment, 173 had a true positive. CancerNetwork’s rates: a detection rate of 0.54%, a positive predictive value of 60.3%, a negative predictive value of 99.2%, a specificity of 99.64%, and a 12-month episode sensitivity of 39.3% for all cancers and 69.8% for a preset group of 12 cancers. A positive predictive value of 60.3% means that, on this page, about 60 in 100 positive tests were real cancers. Sensitivity is the share of cancers the test caught. 39.3% is this page’s figure. It is not Reuters’ 35%. Specificity of 99.64% is this page’s figure. It is not Reuters’ 99.85%. Do not average them, and do not treat them as one sentence. This desk did not read the Nature Medicine PDF.

What CancerNetwork reports from NHS-Galleri, which asked a different question. It says the New England Journal of Medicine published the trial on 22 Sep 2026. The trial asked whether adding Galleri to usual care lowered the rate of stage III or IV cancer — the later stages — across 12 named cancers, after three rounds of screening. It did not show that drop. Of 142,250 people assigned by chance, the later-stage rate was 300.5 per 100,000 person-years with the test and 292.5 without it. The incidence rate ratio was 1.03. The 95% confidence interval was 0.92 to 1.15. CancerNetwork prints the P value as .63. A ratio near 1 means the rates are close. An interval that crosses 1, with that P value, means the study did not show a clear reduction. Those numbers are CancerNetwork’s account of the journal paper. They are not Reuters’ “broadly consistent,” and they are not GRAIL’s four-to-seven-times line. A later look on the same page, which was not the trial’s main question, found more stage I or II diagnoses in the test group, with a relative risk of 1.16. That is not the result the trial was built to answer. This desk did not re-read the journal.

Safety counts on that same CancerNetwork page, filed as the page’s counts and not as this desk’s verdict. In the English trial, events tied to the trial were 0.52% in the test group and 0.45% in usual care. None were called serious. Events tied to the device were 0.04%, in 27 patients. The most common events were bruises at the needle site, fainting, dizziness, a blood blister at the needle site, and anxiety. In PATHFINDER 2, the safety set was 35,335 patients. After a positive test, 0.6% had an invasive procedure, 213 people. Four people with a false positive had surgery, and each time the finding was benign, not cancer. Those counts are CancerNetwork’s. The 10–0 safety vote is the panel’s. One does not replace the other.

What people can already be offered, as Reuters states it. Both GRAIL’s Galleri and Abbott’s Cancerguard are available under Clinical Laboratory Improvement Amendments rules, and neither is FDA-approved. Those amendments, shortened to CLIA, are the U.S. rules for labs that run tests. A lab can offer a test under CLIA without an FDA device approval. Reuters’ sentence means Galleri can already be offered that way. It does not mean Wednesday’s vote approved it. Cancerguard is a different test. It is not this filing.

The quotes, and only the pages they sit on. Josh Ofman, identified as GRAIL’s chief executive, said on the release that the vote reinforces the strength of Galleri’s clinical evidence, that the panel’s recommendations reaffirm a high evidence bar, and that GRAIL will keep working with the FDA as the agency finishes the review. He also called Galleri the first PMA submission for a multi-cancer early detection test. That is his sentence on the release. Reuters quotes Subbu Nambi, an analyst at Guggenheim, saying the vote “increases our confidence even more on FDA approval — almost derisked,” and that he thinks the FDA will approve the test for a wide range of cancers, with some debate about the label. That is an analyst’s forecast. It is not an FDA decision. Reuters also quotes Dr. Marc Matrana, a medical oncologist at Ochsner MD Anderson Cancer Center and a principal investigator on one of GRAIL’s trials, who told Reuters before the meeting, “By far, I think this is a paradigm shift.” Before the meeting means that sentence is not a vote. A principal investigator helps run a study. This desk did not interview them.

Plain English for the rest of the card. PMA is premarket approval, the FDA’s toughest device review. MCED is multi-cancer early detection, one blood draw aimed at many cancers. An advisory vote is advice. It is not approval. Ten voting members cast three ballots: safety 10–0, effectiveness 6–4, benefits over risks 7–2 with one abstention. Efficacy, the word in the headline, is the plain word for whether the test works. The ballot’s word was effectiveness. Methylation means chemical tags on DNA. Cell-free DNA means DNA fragments in the blood. Cancer signal origin is the test’s guess of where a found signal seems to come from. Specificity is how often the test correctly says no in people without cancer. Sensitivity is how often it catches a cancer that is there. 35% and 99.85% are Reuters’ U.S. study figures. 39.3%, 69.8%, and 99.64% are CancerNetwork’s PATHFINDER 2 figures. 1.03, with a P value CancerNetwork prints as .63, is that page’s account of the English trial’s main late-stage result: no clear drop. 18:17 Eastern is GRAIL’s release. 6:05 p.m. Eastern is Reuters. Jan. 29, 2026 is the PMA date GRAIL states. 2018 is the Breakthrough Device year. Coming months is the decision window both GRAIL and Reuters state. This filing is the 23 Sep panel vote. It is not an approval, and it is not a lab manual.

PRIMARY here: GRAIL’s 23 Sep 2026 PR Newswire release at 18:17 Eastern — Tier A PRIMARY, the company’s own announcement of the vote. Reuters, by Padmanabhan Ananthan and Kamal Choudhury, stamped September 23, 2026, 10:05 p.m. UTC, which is 6:05 p.m. Eastern, is same-day Tier B corroboration of the three tallies, the U.S. study’s 35% and 99.85%, the British-study sentence, the CLIA line, and the analyst and physician quotes. CancerNetwork, by Tim Cortese, published September 23, 2026, is the same-day vote write-up that prints the ballot questions and the PATHFINDER 2 and NHS-Galleri figures from papers it dates to 22 Sep 2026. The FDA advisory-committee calendar URL for this meeting redirected this desk to an abuse-detection page, so this filing does not quote an FDA notice. The 10–0, 6–4, and 7–2-with-one-abstention tallies are GRAIL, Reuters, and CancerNetwork. The Jan. 29, 2026 PMA date, the 2018 Breakthrough year, the in-addition-to line, the four-to-seven-times claim, and “coming months” as a PMA decision are GRAIL’s. August 2018, the ballot wording, the PATHFINDER 2 rates, and the NHS-Galleri late-stage ratio are CancerNetwork’s. The 35%, the 99.85%, “broadly consistent,” “addition, not substitution,” the CLIA line, the Nambi quote, and the Matrana quote are Reuters’. NOT claimed: FDA approval, a nationwide on-sale date, that 35% and 39.3% are one figure, that 99.85% and 99.64% are one figure, that the English trial cut late-stage cancer, that this desk watched the hearing or read the journal PDFs, a stock tip, or investment advice. Distinct from the already-filed enveda-311m, basecamp-research-140m, ultrasight-24m, and oracle-health-oncology-ehr.