NHS trial shows 84% accuracy for 'holy grail' cancer blood test Galleri

A blood test that hunts for the genetic signatures of more than fifty different cancers has demonstrated an 84 percent accuracy rate in identifying the disease, even among patients who had previously been told they were cancer-free. The test, called Galleri, works by searching for DNA patterns associated with malignancy, often catching the disease before symptoms emerge.

The evidence comes from an NHS trial involving 6,238 patients who presented to their doctors with symptoms suggestive of cancer and were referred for standard diagnostic imaging and testing. Researchers from the University of Oxford and the diagnostics company Grail analyzed blood samples from these patients using the Galleri test, then compared the results against the diagnoses that emerged from conventional NHS care. Crucially, the GPs involved in the study did not have access to the Galleri results and could not use them to decide which scans their patients would receive. This design meant the test was being evaluated against real-world diagnostic outcomes, not in a controlled laboratory setting.

The findings revealed something striking: among the 79 patients initially flagged as false positives—people the test suggested might have cancer but who were subsequently cleared by standard testing—28 were actually diagnosed with cancer within the following two years. That represents 35.4 percent of those initially dismissed cases. In 27 of those 28 people, the Galleri test had correctly identified which organ or tissue harbored the cancer. When researchers factored these cases back into the overall accuracy calculation, treating them as true positives rather than false alarms, the test's positive predictive value climbed to 84 percent.

The implications for clinical practice are substantial. In more than half of the cases where Galleri identified cancer that standard NHS protocols had missed, the actual cancer type did not match what the patient's symptoms and initial scans had suggested. This points to a practical application: the test could help general practitioners direct patients to the right specialist clinics more quickly, potentially accelerating diagnosis and treatment. Sir Harpal Kumar, president of international business and biopharma at Grail and former head of Cancer Research UK, framed the contrast plainly. Currently, when doctors refer patients for urgent cancer investigation based on symptoms alone, only about 6 percent are ultimately diagnosed with the disease. A test that could identify cancer in 84 percent of cases would represent a significant improvement in diagnostic efficiency.

Kumar emphasized that the test's ability to pinpoint the affected organ matters especially for patients presenting with vague, non-specific symptoms—unexplained weight loss, abdominal pain, fatigue—where a GP might struggle to know which clinic to send them to. The Galleri test could serve as a kind of diagnostic compass, pointing toward the right pathway before symptoms become severe. The company proposes that the test could be administered annually as a screening tool.

These results were presented at the Early Detection of Cancer Conference in Oregon but have not yet undergone peer review or been published in a medical journal. That distinction matters. NHS officials, through Prof Peter Johnson, have indicated they are waiting for final trial results before deciding whether to roll out Galleri more widely across the health service. The test is not alone in this evaluation. The NHS is also considering miONCO-Dx, a blood test developed at the University of Southampton that uses artificial intelligence to detect fragments of tumor DNA and can identify ten different cancer types, including bowel, lung, gastric, prostate, ovarian, and pancreatic cancers, as well as certain brain tumors and bone sarcomas.

The path from promising trial results to widespread clinical use remains uncertain. But the Galleri data suggests that a simple blood draw might catch cancers that would otherwise go undetected until symptoms forced patients into the system—and by then, the disease might have progressed further than it needed to.