Skepticism about multi‑cancer blood tests

A multi-cancer blood test sounds simple: draw a few tubes of blood, look for stray DNA fragments, and catch a tumor before a scan or symptom ever would. The hard part is that blood also carries DNA from normal cell turnover, so the test has to tell a faint cancer signal from a lot of ordinary biological noise. (grail.com) Grail’s Galleri test tries to do that by reading chemical tags on circulating DNA and using a machine-learning model to guess whether the pattern looks like cancer. If it flags a signal, the company says it can also predict the tissue of origin with 93.4% accuracy, which is meant to help doctors decide what scan or biopsy comes next. (grail.com) The pitch is huge because Galleri is marketed as one blood test for more than 50 cancers, while most standard screening programs each look for just one disease at a time, like mammograms for breast cancer or colonoscopy for colorectal cancer. Grail says the test is aimed at adults with elevated risk, especially people 50 and older. (grail.com) The catch is that Galleri is not cleared or approved by the Food and Drug Administration. Grail sells it as a laboratory-developed test run in a Clinical Laboratory Improvement Amendments regulated lab, which means people can buy it before the Food and Drug Administration has ruled that it improves outcomes the way approved screening tests are usually expected to. (grail.com) That difference matters because a screening test is not judged only by whether it can find cancer DNA. The real bar is whether using it leads to fewer late-stage cancers, fewer cancer deaths, or at least a clear benefit big enough to outweigh extra scans, biopsies, radiation, cost, and anxiety. (medscape.com) The biggest real-world test of that idea has been the National Health Service-Galleri trial in the United Kingdom, which enrolled more than 140,000 people starting in August 2021. That trial was designed to see whether annual blood testing could reduce advanced cancers when added to usual care. (academic.oup.com) In February 2026, Grail said that National Health Service trial missed its primary endpoint, meaning it did not show a statistically significant drop in stage three and stage four cancers at the planned analysis. The company and outside observers said secondary findings may still be useful, but missing the main target is exactly why doctors are still arguing over how ready these tests are for routine screening. (fiercebiotech.com) (bmj.com) Grail points to other studies with stronger-looking performance numbers. In a June 2025 update from its PATHFINDER 2 study, the company said the test had a positive predictive value of 61.6%, meaning about 6 in 10 positive results in that analysis were confirmed as cancer, and a specificity of 99.6%, meaning false alarms were uncommon. (grail.com) (galleri.com) Those numbers still do not answer the biggest screening question. A test can be accurate enough in a study and still fail to prove that giving it to large numbers of healthy people actually helps them live longer or avoid harsher treatment. (medscape.com) (academic.oup.com) That is why doctors keep separating “can detect” from “should screen.” Guardant Health’s Shield blood test did win Food and Drug Administration approval in July 2024, but only for colorectal cancer screening in average-risk adults 45 and older, which is a much narrower claim than saying one blood draw can screen the whole body. (fda.gov) Business Insider’s reporting zeroed in on the consumer version of this uncertainty: Galleri costs about $949, and a positive result can open the door to more imaging and specialist visits without a guaranteed diagnosis. That leaves buyers paying real money for a result that may be reassuring, ambiguous, or the start of a long follow-up workup. (businessinsider.com 1) (businessinsider.com 2) Medscape put the same problem in broader terms this week: whether the signal comes from a smartwatch, a blood biomarker, or circulating tumor DNA, predicting disease before symptoms strike is scientifically possible in some cases and clinically hard in most of them. Medicine has gotten much better at measuring tiny changes in the body than at proving which tiny changes are worth acting on. (medscape.com)