A blood test that hunts 50 cancers reaches the FDA, and the math that decides what a positive means

The US drug regulator, the FDA, will convene an advisory panel on September 23 to review Galleri [1][2]. It is a blood test that screens for signals of more than 50 cancers from a single draw [1]. Its maker, Grail, filed the application on January 29 [1]. This is the first time a “multi-cancer early detection” test, the field’s big bet, faces a formal FDA vote [2].

The test reads chemical tags on DNA, called methylation patterns, in the cell-free DNA that tumors shed into the blood [1][2]. Software then flags a possible cancer and predicts where in the body it started [1]. It is meant to complement mammograms and colonoscopies, not replace them [1]. Galleri and a rival, Abbott’s Cancerguard, already sell as lab-ordered tests, but neither is FDA-approved yet [1].

Why 99.6% accurate is not the whole story

Galleri is very good at one thing: clearing healthy people. In Grail’s PATHFINDER 2 study of more than 35,000 people, its specificity was 99.6%, a false-alarm rate under 0.4% [3]. But it is far less sensitive at finding cancer that is present: about 39% of all cancers, rising to 70% for the 12 deadliest [3]. And here is the trap. Cancer in any given year is rare in a healthy screened population. Test enough people and the small false-alarm rate, multiplied across millions, produces a pile of scares, workups, and biopsies in people who were fine.

The clinic already saw this. A three-year UK trial in 70,000-plus healthy adults aged 50 to 77 found that adding Galleri to standard care did not significantly improve early detection or cut late-stage diagnoses [1]. Fuller results in May were more hopeful: by the third yearly round, stage IV diagnoses fell 26% for 12 cancers, and total detection rose fourfold versus standard screening alone [1]. Whether that translates into fewer deaths, the number that matters, is still unproven. A January funding bill set up a possible Medicare coverage path for approved tests, for ages 50 to 65, from 2029, only if clinical benefit is shown [1].

The same tension across the screening world

Galleri is one instance of a pattern running through cancer screening this week. In prostate cancer, a blood marker called PSA, followed by a biopsy, is so good at finding disease that it finds too much [4]. Up to 70% of new prostate diagnoses are low-grade and never need treatment. Yet more than half of US men with that harmless disease still get treated [4]. Doctors now debate scanning with MRI first, standard in most Western countries, to spare men needless biopsies [4]. Researchers reviewed the growing use of “liquid biopsies”, blood tests that track tumor DNA to guide treatment [7]. A Spanish study weighed whether an organized stool-based colorectal screening program cut deaths [5].

The commercial edge of all this is the full-body scan. Clinics like Neko Health (a 299-pound scan, 100,000 on its waiting list), Prenuvo, and Ezra sell healthy people a look inside [6]. NHS England warns the boom is driving “over-diagnosis, unnecessary investigations and avoidable anxiety” [6]. The same math applies: scan enough well people and you will find harmless spots that lead to real, sometimes risky, follow-ups.

Elsewhere in the labs

Approvals moved. The FDA cleared Takeda’s narcolepsy drug, a boost for a new class of treatments [8], and approved Replimune’s melanoma therapy after rejecting it twice [9]. Moderna won clearance for the first mRNA flu shot [10]. Jazz agreed to buy Actio in a potentially 1.3-billion-dollar epilepsy deal [11]. Eli Lilly’s weight-loss pill got a UK green light [13].

Not everything landed. The FDA surprised investors by rejecting a radiopharmaceutical, a drug that carries radiation to tumors, that was expected to rival Novartis’s [14]. Sionna’s cystic fibrosis pill failed a mid-stage trial [15]. And in China, another child died in an investigator-led gene-editing trial, reviving hard questions about transparency and safety in fast-moving research [12].