Galleri screening did not lower advanced cancer incidence after three annual rounds
The NHS-Galleri trial has delivered its clearest answer yet: annual multicancer early detection screening with the Galleri blood test did not reduce the incidence of stage III or IV cancer across 12 prespecified cancer types after three rounds of testing. The randomised comparison, reported in New England Journal of Medicine, found an incidence rate ratio of 1.03 with a 95% confidence interval from 0.92 to 1.14 for the primary outcome. In other words, the trial did not show that adding Galleri to routine care moved the population needle on late-stage diagnosis over the study period.
A blood test can find cancer signals long before symptoms emerge. Translating that signal into fewer patients arriving with advanced disease is a harder test entirely.
What the trial measured and why the result matters
The comparison was straightforward at its core: participants were randomised either to have blood collected and tested with Galleri or to have blood drawn and stored. After three annual testing rounds, investigators looked for the effect on incident stage III or IV cancer among the cancers prespecified in the protocol. The result was neutral on the primary endpoint, even though the screening arm showed some secondary patterns that will attract attention, including fewer cancers diagnosed after emergency presentation and fewer stage IV cancers in the Galleri group.
Those secondary findings fit a plausible biological story. A multicancer assay that detects circulating tumour-derived DNA could, in theory, shift diagnosis away from crisis presentations and toward earlier clinical work-up, especially for cancers that often stay silent until they are advanced. Yet a reduction in emergency diagnoses does not automatically translate into a measurable fall in the overall incidence of late-stage disease within the study window. Screening can change the route to diagnosis without necessarily changing the final stage distribution quickly enough to satisfy a hard clinical endpoint.
Diagnostic performance was strong, but sensitivity remained limited
A companion analysis in Nature Medicine examined the diagnostic performance of the test in more than 140,000 participants aged 50 to 77 years. The blood samples were randomised to Galleri testing or storage, and the assay showed a positive predictive value of 58.0%, a negative predictive value of 99.0%, and a specificity of 99.6%. Episode sensitivity was 37.2%, meaning the test identified a little more than a third of participants who were diagnosed with cancer within 12 months of the first test.
Those figures explain the promise and the constraint of the platform. High specificity and a high negative predictive value matter in screening because false alarms can quickly overwhelm clinics and erode trust. At the same time, a sensitivity of 37.2% means many future cancers were not flagged by the first test, which leaves ample room for interval diagnoses and for cancers that shed too little DNA to be detected reliably at a given moment. In practical terms, a test can be highly selective and still miss too much disease to produce a large immediate shift in advanced-stage incidence.
Fewer emergency presentations do not settle the screening question
The study’s secondary findings are clinically relevant, but they do not rescue the primary endpoint. A lower rate of emergency presentation suggests some participants may have entered the diagnostic pathway in a more controlled setting, rather than through acute illness or decompensation. A lower stage IV burden in the Galleri arm raises the same possibility. Both observations are compatible with earlier detection of some tumours, yet neither proves that population screening with a multicancer assay can reliably prevent advanced disease at scale.
Screening trials live or die on the balance between earlier detection, downstream confirmation, stage shift, and actual patient benefit. Galleri has now shown that a sophisticated blood-based assay can generate useful signals and an acceptable specificity profile, but the central question remains whether those signals can be converted into fewer late-stage cancers often enough to matter clinically.
Blinding kept the behavioral effect hidden
Interpretation is also shaped by one unusual feature of the trial design: participants were not told when they had a negative result, in order to preserve blinding. That decision protected the internal validity of the comparison, but it also limits what can be learned about how a negative result changes behaviour, reassurance, follow-up patterns, or the intensity of subsequent clinical work-up. If a negative multicancer test result reduces the likelihood of further investigation after subtle symptoms, it could alter diagnostic timing in ways the trial could not fully observe.
This is not a trivial issue. Screening affects more than biology. It changes patient expectations, clinician thresholds, and the path a person takes through the health system. A blood test that is reassuring on paper may still reshape decisions in primary care and specialty referral, for better or worse.
Where multicancer early detection now stands
The NHS-Galleri programme has provided a large, carefully controlled dataset for one of the most ambitious ideas in cancer screening. The assay’s performance metrics are encouraging enough to keep interest alive, but the randomised outcome data are harder to ignore. Annual testing over three rounds did not reduce the incidence of stage III or IV cancer in the prespecified set of tumours, even with some evidence of stage and presentation shift in secondary analyses.
For clinicians, the result argues for restraint before treating blood-based multicancer screening as a substitute for established cancer detection pathways. For researchers, the next questions are narrower and more demanding: which tumour types, which risk groups, and which testing intervals might produce a clinically meaningful stage shift without generating a new burden of false reassurance, follow-up testing, or overdiagnosis? Those answers will have to come from longer follow-up and from trials that can show not just detection, but durable benefit.