FDA advisers all called Grail's cancer blood test safe. Only six of ten said it works.

The stories

16 on this beat today.

Every story told the same way: what happened, who is involved, why it matters, how it works, the truth and the good news.

1. 01

   FDA advisers all called Grail's cancer blood test safe. Only six of ten said it works.

   10, 6 and 7 of 10

   advisers who said Galleri is safe, that it works, and that its benefits outweigh its risks

   One adviser abstained on the last question. [1][2]

   35%

   of cancers found within a year that the test flagged in Grail's US study

   It gave a false alarm to about 15 in every 10,000 people without cancer. [1][5]

   $949

   list price of a Galleri test, mostly paid by the buyer

   A law Grail lobbied for lets Medicare pay from 2028 if the FDA approves it. [3]

   117,000

   people who took the test in the first half of 2026, outside research

   It has been sold since 2021 without FDA approval. [3]

   What happened

   • On 23 September, ten outside experts who advise the FDA, the US drug regulator, voted on Galleri, a blood test made by the US company Grail. [1][2]
   • Galleri looks for signs of many kinds of cancer in one blood draw, in adults aged 50 and over. [1]
   • No test that looks for many cancers at once is approved in the US. [1][4]
   • The panel voted three times. [2]
   • All ten said the test is safe. [1]
   • Six of ten said it works, and seven said its benefits outweigh its risks, with one abstaining. [1][2]
   • The FDA does not have to follow the panel, but it weighs its votes heavily. [1]
   • A final decision is expected in the coming months. [1]
   • Galleri has been sold in the US since 2021 through a regulatory exception, without FDA approval. [3]
   • About 117,000 people took it in the first half of this year, outside research studies. [3]
   • Most buyers pay the $949 list price themselves. [3]
   • Grail lobbied for a law that lets Medicare, the US government's health insurance for older people, pay for the test from 2028, if the FDA approves it. [3]
   • In Grail's US study, the test flagged 35% of the cancers found within a year. [1]
   • It raised a false alarm in about 15 of every 10,000 people without cancer. [5]
   • The test caught late cancers far more often than early ones. [5]
   • In the US study it flagged 21% of stage 1 cancers, the earliest, and 63% of stage 4, the most spread. [5]
   • A paper in Nature Medicine on 22 September covered more than 35,000 people. [4]
   • The test found a cancer signal in 287 of them, and about 60% of those were later diagnosed with cancer. [4]
   • When the test did find cancer, it named the right part of the body about 91% of the time. [4]
   • In the US study, 159 of 218 people who were checked after a positive result had at least one invasive procedure. [5]
   • Among those who turned out not to have cancer, 47.8% had one. [5]
   • In February, a big trial in the UK found the test did not cut the number of cancers diagnosed at stage 3 or 4. [3][6]
   • No trial has shown that tests like it reduce cancer deaths. [3]

   Stage 121%

   Stage 233%

   Stage 358%

   Stage 463%

   Who is involved

   • Grail the US company that makes Galleri; it asked the FDA for approval in January
   • The FDA advisory panel ten outside experts the FDA asked for advice; the agency is not bound by their votes
   • The FDA the US regulator for drugs and medical tests; it makes the final decision in the coming months
   • Scott Ramsey a cancer researcher at Fred Hutch Cancer Center in Washington state; he says nobody among cancer doctors is recommending the test

   How it unfolded

   1. 2018 The FDA gives Galleri its breakthrough device label, meant for promising new devices. [7]
   2. June 2021 An earlier version of the test goes on sale in the US by prescription. [5]
   3. January 2026 Grail asks the FDA to approve the test. [6]
   4. February 2026 A big trial in the UK fails to cut late-stage cancer diagnoses. [3][6]
   5. 21 September FDA staff raise no major concerns, and Grail's shares rise 35%. [6]
   6. 23 September The advisers vote 10 to 0, 6 to 4 and 7 to 2. [1]

   Why it matters

   The FDA decides in the coming months; watch whether it follows the panel, and whether it limits the test to people at higher risk of cancer, as one researcher suggested. [1][4]

   The truth

   • What is really going on Grail has sold Galleri in the US since 2021 without FDA approval, and 117,000 people took it in the first half of this year, most of them paying the $949 price themselves. [3]The panel's vote matters most for who pays: a law Grail lobbied for lets Medicare cover the test from 2028, but only if the FDA approves it. [3]
   • Why it works on people Grail's own announcement leads with the panel voting in favour of approval, and one yes is easier to remember than three different counts. [7]
   • Who gains Grail — An FDA approval would let Medicare pay for its test from 2028, under a law it lobbied for, and would bring routine insurance coverage. [3][6]
   • Who pays People without cancer who test positive — In the US study, 47.8% of those checked after a positive result who turned out not to have cancer had an invasive procedure. [5]
   • Who pays People who buy the test before any approval — Most pay the $949 list price themselves. [3]
   • Nobody knows yet

     Whether finding cancer with Galleri means fewer people die of it.

     No randomised trial has shown that Galleri or any test like it reduces cancer deaths. [3] The UK trial did not cut the number of cancers found late. [3][6]

   • Nobody knows yet

     How often a positive Galleri result really means cancer.

     The FDA's summary of the US study puts it at 77% of positive results. [5] A Nature Medicine paper on more than 35,000 people found about 60% of 287 positives were later diagnosed with cancer. [4] The two count different groups over different follow-up, and we could not reconcile them.

   • Nobody knows yet

     What the FDA will decide, and for whom.

     The agency is not bound by the vote, and its decision is due only in the coming months. [1] One researcher said the test may suit people at high risk better than everyone over 50. [4]

   Good news

   • In a study of more than 35,000 people, when Grail's blood test found cancer, it named the right part of the body about nine times in ten.

1. 02

   The FDA nominee was told to go fast

   What happened

   • Heidi Overton, President Trump's pick to run the FDA, goes before the US Senate health committee on 24 September. [9]
   • In an opening statement STAT saw in advance, she says Trump told her to go fast. [8]
   • She says slow FDA rules are pushing drug investment to China. [8]
   • She backs Operation Trailblazer, an FDA effort started this year to speed up clinical trials. [8]
   • The agency has lost more than 3,000 staff in the past year. [9]

   Why it matters

   If the US Senate confirms her, her view of speed will shape decisions like the one on Grail's test.

   The last FDA head, Marty Makary, resigned in May after clashes with officials in the Trump administration. [9]

1. 03

   Stem cells fail to help babies' hearts

   What happened

   • Longeveron's stem-cell treatment, laromestrocel, did not help babies born with hypoplastic left heart syndrome, where the left side of the heart is too small to work. [10]
   • In a mid-stage trial, 40 infants got the cells during their second heart operation. [10]
   • After 12 months, the right side of their hearts pumped no better than in babies who had surgery alone. [10]
   • No side effects were judged to come from the cells. [10]

   Why it matters

   Longeveron's shares fell 60%, and it had $10.1 million in cash at the end of June. [10]

   It is cutting costs and has hired an investment bank to review all its options. [10]

   How it works

   Safe, working and worth it are three separate questions

   Advisers to the FDA voted separately on whether Grail's cancer blood test is safe, whether it works and whether it is worth using, and gave three different answers.

   The twist

   Passing one of the three questions tells you nothing about the other two. Galleri got ten yes votes on safety, but only six on whether it works.

   The picture

   It is safe10

   It works6

   Benefits beat risks7

   Step by step

   1. 1 A regulator splits one big decision into separate questions
   2. 2 Is it safe: does it hurt the people who use it?
   3. 3 Does it work: does it do what it claims?
   4. 4 Is it worth it: do the gains beat the harms and the other choices?
   5. 5 Each question gets its own vote, and a yes on one does not carry to the next

   The same force, elsewhere today

   Where this chain is also running, in today's other stories.

   • Stem cells fail to help babies' hearts

     No side effects were blamed on Longeveron's cells, so the safety question looked fine. The works question failed, because the babies' hearts pumped no better.

   • A skin lupus antibody is dropped

     Immunovant's drug was safe and did not work in skin lupus. The company also weighed worth: rival drugs are further ahead, so it stopped.

   • Roche kidney injection passes a big trial

     Sefaxersen met its main goal and showed no new safety problems. Whether it is worth using will be judged against Fabhalta, a pill for the same disease that is already approved.

   Where you've seen this

   A used car

   The brakes can pass inspection and the engine can start every morning, and the price can still be too high next to the other cars on offer.

   Airport scanners

   A scanner can be harmless to travellers and still miss hidden items, or find them but add hours to every queue.

   A new school rule

   A rule can hurt nobody, change nothing about how pupils behave, or work and still cost teachers more time than it saves.

   The catch

   Splitting the decision makes each vote clearer, but someone still has to add the three answers up, and the FDA can decide differently from its advisers.

   And the whole of it

   The advisers voted for everyone aged 50 and over at once. A person offered the test has to weigh the same questions for themselves: a $949 price and a possible false alarm, against a cancer it might find early.

   The truth

   • Who pays Longeveron's shareholders — The shares fell 60% after the trial failed. [10]
   • Nobody knows yet

     Whether Longeveron can keep going.

     It had $10.1 million in cash at the end of June and has hired a bank to review all its options. [10]

1. 04

   Roche kidney injection passes a big trial

   What happened

   • Roche's monthly injection sefaxersen met its main goal in a late-stage trial in IgA nephropathy, a disease in which the immune system damages the kidneys' filters. [11]
   • After 37 weeks, patients on it leaked less protein into their urine than patients on a dummy injection. [11]
   • Protein in the urine is a sign of kidney damage. [11]
   • The trial stays blinded until week 105 to measure kidney function itself. [11]

   Why it matters

   It would compete with Novartis's Fabhalta, a twice-daily pill that blocks the same protein, factor B, and is already fully approved. [11]

   Roche hopes the early result earns a faster approval. [11]

   How it works

   Sefaxersen met its main goal and showed no new safety problems. Whether it is worth using will be judged against Fabhalta, a pill for the same disease that is already approved.

   The idea, in story 03: Safe, working and worth it are three separate questions →

   The truth

   • Nobody knows yet

     Whether sefaxersen protects the kidneys, not only the urine test.

     Roche's trial stays blinded until week 105 to measure kidney function itself. [11]

   Good news

   • A monthly injection from Roche cut the protein leaking into the urine of people with IgA nephropathy, a kidney disease, in a late-stage trial, with no new safety problems.
2. 05

   A skin lupus antibody is dropped

   What happened

   • Immunovant stopped testing its antibody IMVT-1402 in cutaneous lupus, a form of lupus that inflames the skin. [12]
   • In 57 adults, skin scores after 12 weeks were no better than on a dummy drug. [12]
   • The drug blocks FcRn, a protein that keeps the body's own antibodies in the blood for longer. [12]
   • The company said it stopped because of the result and because rival drugs are further ahead. [12]

   Why it matters

   Immunovant is still testing the same drug in five other diseases, including Graves' disease and myasthenia gravis. [12]

   It caused no safety worries in this trial. [12]

   How it works

   Immunovant's drug was safe and did not work in skin lupus. The company also weighed worth: rival drugs are further ahead, so it stopped.

   The idea, in story 03: Safe, working and worth it are three separate questions →

   The truth

   • Who pays Adults with skin lupus — One drug that was being tested for them has been dropped. [12]
3. 06

   China aims for a quarter of new drug types

   What happened

   • China's government wants a quarter of the world's first-in-class drugs, meaning drugs that work in a way no earlier drug does, to come from China by 2030. [13]
   • The goal is in a five-year plan published on 18 September by China's drug regulator and eight other agencies. [13]
   • It also wants 50 Chinese drugmakers each selling at least $1.5 billion a year. [13]
   • A GSK executive said at a panel that its deals with Chinese companies such as Hansoh are paying off. [14]

   • China's target share25 · 25%
   • Rest of the world75 · 75%

   Why it matters

   Chinese companies are now a major source of new drugs for US and European drugmakers. [13][14]

   The US Treasury is drafting rules that would probably keep most of those deals legal. [13]

   The truth

   • Who gains China's drug companies — Their government's five-year plan backs new kinds of drugs, and GSK says its Chinese deals are paying off. [13][14]
   • Nobody knows yet

     Whether US rules will keep most drug deals with Chinese companies legal.

     The Treasury's draft rules are not final and could change. [13]

4. 07

   A plan to control NIH grants stalls

   What happened

   • A planned Trump administration order that would let political appointees veto grants from the NIH, the US government's medical research agency, is looking dead. [15]
   • That is according to Senator Susan Collins and a person familiar with the decision. [15]
   • Collins, a Republican who leads the US Senate's spending committee, wrote on Tuesday urging officials not to go ahead. [15]
   • Democrats, including Senator Patty Murray, also objected. [15]

   Why it matters

   The panel would have included Russell Vought, President Trump's budget director, who had accused the NIH of funding work that does not match the Trump administration's goals. [15]

   Collins wrote that grants should be decided on scientific merit, not political ideology. [15]

   The truth

   • Who gains Scientists funded by the NIH — A plan to give political appointees a veto over their grants looks dead after senators objected. [15]
   • Nobody knows yet

     Whether the plan to control NIH grants is dead or only paused.

     The word 'dead' comes from Senator Collins and one unnamed person, and the Trump administration did not comment. [15]

   Good news

   • A Trump administration plan to give political appointees a veto over NIH research grants looks dead, after senators from both parties objected.
5. 08

   An AI drug finder raises $311 million

   What happened

   • Enveda, a Colorado company that uses AI to find drug-like molecules in plants and microbes, raised $311 million from investors. [16]
   • It has three drugs in human testing. [16]
   • One, ENV-294, is an eczema pill that shrank the area and severity of the rash by 68% on average after 28 days, in a small early trial. [17]
   • Another, ENV-308, copies a hormone the body releases after exercise. [16]
   • It is being tested for keeping weight off after people stop weight-loss drugs. [16][17]

   Nov 2024$130m

   Sep 2025$150m

   Sep 2026$311m

   Why it matters

   The round is as big as some of this year's largest biotech share sales on the stock market. [16]

   Its drug results so far come from small early trials. [17]

6. 09

   Basecamp raises $140 million for AI-designed DNA

   What happened

   • Basecamp Research raised $140 million to train its EDEN AI models, which design pieces of DNA. [18]
   • It wants to use them to make cell therapies that are simpler and cheaper to produce than today's. [18]
   • In a study not yet checked by other scientists, half of the DNA-inserting enzymes EDEN designed worked in human cells. [18]
   • Investors include Nvidia's venture arm and Anthropic, the company that makes the Claude AI models. [18]

   Why it matters

   It is one of two AI drug-discovery rounds announced on 23 September, alongside Enveda's. [16][18]

   Basecamp's results so far are all in the lab, not in patients. [18]

   The truth

   • Who gains AI drug-discovery companies — Enveda raised $311 million and Basecamp Research $140 million on the same day. [16][18]
7. 10

   US military money for Flagship biotechs

   What happened

   • The US Department of War, the country's defence department, gave Flagship Pioneering a $47 million contract over five years. [19]
   • Flagship starts new biotech companies. [19]
   • The money is for medical countermeasures: vaccines, drugs and tests for use after a biological or chemical attack, or a new natural outbreak. [19]
   • The first programmes are at Apriori Bio, which uses AI to predict how viruses change, and Prologue Medicines, which designs proteins. [19]

   Why it matters

   The contract adds military money to AI biology. [19]

   It comes in the same week that US civilian research money at the NIH was the subject of a fight over who decides grants. [15]

   The truth

   • Who gains Flagship Pioneering — It gets $47 million over five years from the US Department of War for two of its companies. [19]
8. 11

   Boehringer widens a genetics deal

   What happened

   • Boehringer Ingelheim, a German drugmaker, expanded its deal with Variant Bio, a Seattle company, adding more than $250 million in possible payments. [20]
   • Variant uses AI on genetic data from people in many parts of the world to find new drug targets. [20]
   • The deal began in January, worth $120 million, for kidney disease, and now covers heart disease too. [20]
   • It came a day after Boehringer signed a deal worth up to $1 billion with another AI company, Envisagenics. [20]

   Why it matters

   Boehringer already sells Jardiance, a drug for heart failure and kidney disease. [20]

   The companies did not say how much is paid now and how much depends on future results. [20]

9. 12

   A numbing drug that lasts weeks, in rats

   What happened

   • Researchers at Boston Children's Hospital made tiny fat bubbles, called liposomes, that release a drug very slowly. [21]
   • They filled them with tetrodotoxin, a nerve poison from pufferfish, and injected them near a nerve in rats' legs. [21]
   • The leg stayed numb for two to three weeks. [21]
   • A numbing drug sold today lasted four to eight hours. [21]
   • No harm was seen at the injection site or elsewhere in the body. [21]

   4 to 8 hoursNumbing drug sold today

   vsvs

   2 to 3 weeksNew slow-release version

   Why it matters

   The team says it could one day replace opioids for pain after surgery. [21]

   Tetrodotoxin is not approved for patients, and the tests so far are only in rats. [21]

   Good news

   • In rats, a new slow-release numbing drug kept a leg numb for two to three weeks, with no harm at the injection site. The version sold today lasted four to eight hours.
10. 13

    Cleaning and reusing drug-growing liquid

    What happened

    • Many antibody drugs are made by living cells grown in a nutrient liquid, and the liquid is costly. [22]
    • Researchers at MIT used an electric field to pull waste out of used liquid so it can be reused. [22]
    • In tests they recovered 87.5% of it. [22]
    • Factories that recycle the liquid today usually reuse less than half, and reusing it without cleaning can slow the cells down. [22]

    Why it matters

    The method is still at pilot stage. [22]

    The researchers say it still has to be checked for any effect on the quality of the drug. [22]

11. 14

    A smarter controller for drug-making cells

    What happened

    • Researchers at Sartorius built a control system for tanks of hamster-ovary cells, the standard cells used to make antibody drugs. [23]
    • It adjusts three flows at once: fresh food in, old cells out and product out. [23]
    • It raised production by 68% while keeping at least 95% of the cells alive. [23]
    • Much of the testing was done on digital copies of the process. [23]

    Why it matters

    It lets a factory switch between running steadily and squeezing out more product without stopping the run. [23]

    It stayed stable even when the computer model did not match the real process well. [23]

12. 15

    Drugmakers turn back to brittle bones

    What happened

    • Drug companies and investors are returning to osteoporosis, the thinning of bones that makes them break easily in older adults. [24]
    • Lilly, Merck, Novartis and GSK each once had a drug for it. [24]
    • Awkward rules for taking the drugs and safety worries held their use back. [24]
    • Many people who could be treated take none of today's drugs. [24]
    • A long-awaited change in US regulation has renewed interest, and Lilly is hiring scientists for the work. [24]

    Why it matters

    Venture investors are now making bone health one of their priorities, a turn for a field most drug companies had left. [24]

    The truth

    • Nobody knows yet

      What the US regulatory change on osteoporosis drugs actually is.

      STAT reports a long-awaited change has brought drugmakers back, but its details sit behind a paywall we could not read. [24]

13. 16

    Promised safety studies run years late

    What happened

    • Hundreds of studies that companies promised to run after their drugs or devices were approved are listed by the FDA as delayed, an analysis by KFF Health News found. [25]
    • Some are more than a decade late. [25]
    • For some, the maker was still writing a plan for the study. [25]
    • The FDA uses these studies to settle questions left open at approval. [25]

    Why it matters

    Until they report, patients and whoever pays, including Medicare, may be paying for products that do more harm than good. [25]

    Galleri's biggest open question, whether it saves lives, also needs longer studies. [4]

    The truth

    • Who pays Patients and payers relying on promised studies — Hundreds of studies meant to settle questions left open at approval are late, some by more than a decade. [25]
    • Nobody knows yet

      What Heidi Overton's 'go fast' would change at the FDA.

      Her opening statement names speed and China but, in the part we could read, no specific check that would be dropped. [8] Hundreds of safety studies already promised after approval are listed as late. [25]