Will one-stop blood tests for cancer save lives?

GRAIL’s Galleri blood test failed to cut late-stage cancer diagnoses, raising fresh questions over multi-cancer early detection

GRAIL IS AN American biotech company with an ambitious goal: to develop a blood test for the early detection of over 50 types of cancer. It will not be easy. On February 19th the company announced that its test, called Galleri, had fallen at its latest hurdle. Over the course of a three-year trial that involved 142,000 people aged between 50 and 77, half were screened with Galleri in addition to the usual tests recommended for their age. The hope was that it would spot enough early-stage cancers to reduce the number of late-stage cancer diagnoses. It did not.

GRAIL IS AN American biotech company with an ambitious goal: to develop a blood test for the early detection of over 50 types of cancer

The news was a big blow for advocates of the technology. But the full results of the trial, expected to be released in May, could paint a more nuanced picture; Galleri may be more useful in spotting certain types of cancer among certain types of people, for instance. At the same time, GRAIL’s competitors are busy developing alternative blood tests that may be more sensitive than Galleri. Whether the early detection they promise can lead to improved survival rates remains a contested question.

The principle behind cancer-spotting blood tests is relatively simple. Some, like Galleri, analyse DNA fragments shed by cancer cells, whereas others look for proteins or metabolites that those cells produce. Cancerguard, a rival test developed by Exact Sciences, an American company, combines multiple approaches.

Galleri, which has been on the market since 2021, has been the most widely used—and studied—of these tests. Its potential was on full display in October 2025, when GRAIL announced the results from a trial with 23,000 people aged 50 or older. Galleri spotted around 40% of the cancers that were diagnosed over the next year (of which half were early-stage) and usually identified the right part of the body. At the same time, it missed 60% of cancers and generated plenty of needless scares. It threw up a cancer “signal” in about 1 in 107 participants, and roughly two in five of those alarms proved to be wrong.

The company’s other trials suggest a more optimistic interpretation of these results: that patients were simply not followed for long enough after they were tested. In a trial of 6,000 patients in England who were referred for various cancer investigations because of worrying symptoms, a third of presumed false alarms were diagnosed with cancer when they were followed for an additional 15 months. This suggests the test was detecting some cancers before they were advanced enough to be spotted by standard methods.

Another challenge facing test designers is that early detection may not necessarily save lives. A crucial factor, and something that is still unknown about the new multi-cancer tests, is whether they predominantly spot tumours that need treatment, rather than the slow-growing, harmless type that is best left alone.

Some prostate cancers, for example, would kill the men who have them only if they live well beyond the age of 100. Finding such harmless cases on tests and scans—whether by accident or through routine screening—is the bane of cancer medicine. Doctors cannot always say with certainty which are in that category, so usually treat them, just in case, with surgery, radiation and so on.

In November 2025, the expert committee that reviews evidence on screening tests for England’s National Health Service (NHS), recommended against universal use of PSA, a blood test for prostate cancer. For every two lives extended thanks to the test, the committee found, around 20 men are likely to be overdiagnosed with harmless tumours and 12 will undergo treatment of a type that often results in harms such as impotence or incontinence (to say nothing about the unnecessary anxiety). Instead, the PSA test was recommended only for men at high risk, such as those with a mutation in the BRCA gene.

Another complication is that, for reasons not well understood, some cancers are more lethal than others even if found early. A trial of a blood test for ovarian cancer called CA125, conducted by England’s NHS from 2001 to 2011, showed how this can play out. Even though the blood test spotted more early cases, the lethality of the cancer meant it did not save lives.

According to GRAIL, this month’s findings suggested a “favourable trend” in the results for some types of cancer later in the trial. If that holds up, the balance of harms and benefits associated with using Galleri is likely to be different for each of the cancers it looks for. That makes the administration of the test a decision to be taken carefully. There are additional dilemmas, too: patients with a fatal cancer may not benefit from an earlier diagnosis, but such knowledge could help researchers improve their knowledge of early-stage cancers.

That will not be the end of the story. “GRAIL’s technology is somewhat outdated now,” says Anna Schuh, who leads molecular diagnostics research at Oxford University’s Department of Oncology. Galleri examines fragments of DNA in the blood for signs that chemical tags have been added to them, altering which genes are switched on and off, a process known as methylation. Galleri’s way of doing so destroys most of the DNA, limiting what can be gleaned from a sample.

Scientific advances reported by several academic groups in the past three years, however, may soon lead to more sophisticated and powerful tests, says Dr Schuh. In addition to using less destructive ways to identify methylation, some new methods scan the entire genome for cancerous changes in the DNA code. New tests also analyse patterns in the physical features of DNA fragments that may hold clues about the presence of cancer. Such multi-pronged approaches, supported by artificial intelligence tools to assist with data analysis, hold promise. If they can be turned into reliable tests, they may well help identify which cancers need prompt action and which do not.