Summary
An analysis of the NHS-Galleri trial reports Galleri’s performance across three annual screening rounds, including falling episode sensitivity in later rounds. The trial’s primary endpoint—a reduction in stage III/IV cancer diagnoses—was not met.
A performance analysis of the NHS-Galleri trial reports that the Galleri blood test returned positive results for about 0.8–1.0% of evaluable participants in each of three annual screening rounds. The paper, published in Nature Medicine on 22 September 2026, reports these prespecified secondary test-performance measures among participants assigned to screening. It also notes that the trial’s primary endpoint—a reduction in stage III/IV cancer diagnoses in the screening arm compared with usual care—was not met.
NHS-Galleri randomized 142,250 people aged 50–77 to either an intervention arm, in which blood samples were tested, or a control arm. The performance analysis covered evaluable intervention-arm participants: 70,325 in round one, 64,498 in round two and 62,323 in round three. The reported performance analyses were descriptive, with no hypothesis testing.
What the three screening rounds found
Galleri looks for cancer-associated methylation patterns in cell-free DNA from blood. A positive result includes a predicted cancer signal origin (CSO)—the anatomical site or cell lineage most likely to be associated with the signal—to help guide diagnostic evaluation. In the trial, people with positive results were referred through NHS diagnostic pathways.
Positive results were recorded for 722 participants in round one (1.03%), 518 in round two (0.80%) and 561 in round three (0.90%). The paper reports 937 participants with MCED-detected primary cancers across the rounds. The cancer detection rate was 0.60% in round one, 0.40% in round two and 0.41% in round three.
The positive predictive value (PPV)—the share of positive results followed by a primary cancer diagnosis within the specified follow-up period—was 58.0% in the first round, 50.4% in the second and 45.8% in the third. Specificity, the proportion of participants without a cancer diagnosis who had a negative result, remained between 99.50% and 99.60% across rounds. Episode sensitivity measures the share of participants diagnosed with cancer within 12 months of a screening blood draw who had received a positive result. It was 37.2% in round one, 27.1% in round two and 26.7% in round three. For 12 prespecified cancer types, sensitivity was 63.4%, 47.6% and 50.7%, respectively.
Among participants with true-positive results, the first or second CSO prediction matched the diagnosed cancer’s origin in 93.6% of first-round cases, 92.3% of second-round cases and 91.1% of third-round cases. Across all three rounds, 68.3% of MCED-detected cancers were diagnosed at stage I–III.
The first screening round draws on a pool of cancers already present but undiagnosed when screening began. Later rounds instead reflect cancers that became detectable between screens, as well as cancers missed earlier. The paper identifies this difference as relevant to interpreting the lower PPV, detection rate and sensitivity in the second and third rounds.
Performance measures are not the same as screening benefit
The test-performance results describe how often the test returned a signal, how those signals corresponded with diagnoses during follow-up, and how accurately it predicted cancer origin. They do not override the trial’s primary result: the planned reduction in stage III/IV cancer diagnoses between the randomized arms was not achieved. Finding cancers, including cancers diagnosed at earlier stages, is not by itself evidence that a screening programme reduces cancer deaths.
Each round’s performance was assessed against diagnoses within 12 months of the blood draw. The paper says longer follow-up is important for assessing longer-term clinical outcomes, including mortality, and for evaluating potential overdiagnosis. It also stresses that a negative Galleri result should not be taken to rule out a particular cancer. The test was studied as an addition to usual care, not as a replacement for existing cancer screening.