Evidence Trace, complete trial record

In plain terms

In plain terms: Multi-cancer early detection blood test (cell-free DNA) has been tested in human trials aimed at the relevant outcome and did not show benefit at the doses and durations studied. The animal research may be encouraging, but it has not yet translated to humans in these trials.

Multi-cancer early detection blood test (cell-free DNA)

D Not shown: no human trial shows benefit

Route 2 — tested in humans; benefit not shown

Category Devices

Evidence Tested in one large randomized trial; primary endpoint not met

Reviewed October 2026

PMIDs 42776074 · 42773209 · 39898371 · 42217468 · 40953446 · 42185047

The endpoint this grade is based on: Surrogate biomarker endpoint

The evidence as the record states it

Evidence record

PMID 42776074 PMID 42773209 PMID 39898371 PMID 42217468 PMID 40953446 PMID 42185047

Sasieni et al. (New England Journal of Medicine, 2026; PMID 42776074): NHS-Galleri primary results, randomized trial of 142,250 adults aged 50 to 77 in England. The incidence of stage III or IV cancer did not differ between groups (rate ratio 1.03, 95% CI 0.92 to 1.14); stage IV alone had a rate ratio of 0.86 (95% CI 0.74 to 1.00). Funded by GRAIL, which sells the test, with GRAIL employees among the authors. Neal et al. (Nature Medicine, 2026; PMID 42773209): prespecified secondary test-performance results from the same trial, descriptive with no hypothesis testing, also GRAIL-funded with company employees among the authors (vendor-authored); positive predictive value 45.8 to 58.0 percent by round and episode sensitivity 26.7 to 37.2 percent for all cancers. Wade et al. (Health Technology Assessment, 2025; PMID 39898371) - NIHR systematic review of blood-based multi-cancer early detection tests for general population screening in adults aged 50 to 79 without clinical suspicion of cancer. Mann et al. (JAMA, 2026; PMID 42217468) - observational difference-in-differences study across all 21 English cancer alliance regions: diagnostic delay rates for suspected head and neck, lung and upper gastrointestinal cancers rose 3.4 and 4.8 percentage points in participating regions during the trial's first two 6-month periods, with no significant difference after that; the authors judged the effect unlikely to have materially affected the trial's findings. Observational, so it carries no grade. Kahwati et al. (Ann Intern Med, 2025; PMID 40953446) - AHRQ systematic review: no completed controlled study reports a benefit of screening with multicancer detection tests; accuracy varies by test and study design. Batson et al. (BMJ, 2026; PMID 42185047) - UK National Screening Committee position statement: late-stage incidence is a surrogate that cannot yet replace mortality as the definitive outcome in cancer screening trials. Grade D, Route 2.

PMID 42776074 PMID 42773209 PMID 39898371 PMID 42217468 PMID 40953446 PMID 42185047

Trials showing no benefit

Trial 1

PMID 42776074

Endpoint: Incidence of stage III or IV cancer among 12 prespecified cancer types (primary endpoint)
Population: NHS-Galleri: 142,250 adults aged 50 to 77 in England, randomized 1:1 to the test plus usual care or usual care alone, up to three annual screening rounds

No benefit shown

High for the primary endpoint. Incidence rate ratio 1.03 (95% CI 0.92 to 1.14, P = 0.63): no reduction. Stage IV alone, a key secondary endpoint, had a rate ratio of 0.86 (95% CI 0.74 to 1.00), a nonsignificant trend toward fewer stage IV cancers that cannot override the primary result. Trial funded by GRAIL, which sells the test.

Why Grade D Route 2 and not Grade C

Grade D Route 2 applies when a compound has been tested in humans and benefit was not shown. Multi-cancer early detection blood test (cell-free DNA) has been tested in 1 human trial on endpoints relevant to the claim being evaluated, and the primary endpoint was not met.

A randomized, placebo-controlled trial pre-specifying a meaningful clinical or functional endpoint and demonstrating statistically significant benefit would move this to Grade C or higher, depending on effect size, population size, and whether it is independently replicated.

What this grade covers and what it does not

Grade D on Route 2: tested, no benefit demonstrated. This is not an untested technology and the D is not a statement about plausibility. NHS-Galleri was run at scale, randomized, in exactly the population the test is marketed to, and it missed its primary endpoint of reducing late-stage cancer at diagnosis. Stage IV alone showed a nonsignificant trend toward fewer cancers (rate ratio 0.86, 95% CI 0.74 to 1.00), which is the finding proponents cite; a secondary endpoint whose interval reaches 1.00 is not a demonstrated improvement and cannot override the primary. EBL grades against the endpoint the trial itself pre-specified. THE COMPARISON THAT MATTERS: EBL grades fecal immunochemical testing A, because randomized trials show it reduces colorectal cancer mortality. That is what an A-grade screening test looks like. A test detecting a signal is not the same as a test changing an outcome, and the distance between those two claims is this entire grade. A SPILLOVER SIGNAL WORTH READING: the JAMA analysis (PMID 42217468) found diagnostic delay rates for suspected head and neck, lung and upper gastrointestinal cancers rose 3.4 and 4.8 percentage points in trial regions during the first year, then no longer differed; the authors judged it unlikely to affect the trial's interpretation. It is observational and carries no grade. THE ENDPOINT IS A SURROGATE: late-stage incidence stands in for cancer mortality, which no multicancer detection trial has reported, and the UK National Screening Committee has said such surrogates cannot yet replace mortality (PMID 42185047). The AHRQ review found no completed controlled study showing screening benefit (PMID 40953446). One randomized trial exists, so nothing here is replicated. WHAT THIS GRADE DOES NOT COVER: this letter grades the Galleri test used for population screening in asymptomatic adults, judged against the endpoint the trial itself pre-specified. It does not grade the test's analytical performance, which is a separate question about detecting a signal rather than changing an outcome. It does not grade its use in someone with a known cancer or a clinical suspicion of one, which the trial excluded. And a D here is not a claim that no multi-cancer blood test can work. It is a statement about what this one did in the largest trial anyone has run.

What would move this grade

A randomized, placebo-controlled trial pre-specifying a clinical aging endpoint and demonstrating statistically significant benefit on that primary endpoint would be sufficient to move this to Grade C or higher, depending on effect size and population.

How to track whether this is working Nothing to track. A negative result does not mean no cancer, and a positive signal starts a diagnostic workup rather than answering anything.