The cancer blood test finally has a big trial. Read the sensitivity number twice.

A multi-cancer early detection test looks for fragments of tumour DNA circulating in blood and, when it finds them, predicts which organ they came from. The category spent five years being sold on its promise. It now has a registrational trial, and the trial is genuinely good news provided you read the number that matters rather than the one in the headline.

PATHFINDER 2 enrolled 35,878 adults aged 50 and over with no clinical suspicion of cancer, making it the largest interventional MCED study conducted in North America.1 Two results define what the test is. Episode sensitivity across all cancers was 39.3%. Specificity was 99.6%, a false positive rate below 0.4%.1

Why 39.3% is not the failure it looks like

A test that misses six cancers in ten sounds broken. The number beside it changes the reading.

Specificity of 99.6% is exceptional for a screening assay. Run it across 10,000 healthy people and roughly 40 receive a false alarm. Set that against the incidental finding rates attached to whole-body imaging, where about a third of scanned people leave with something critical or unresolved on the report. The blood test trades away sensitivity and buys precision with the proceeds. For an assay meant to sit on top of existing screening rather than replace it, that is the right trade.

Figure 1

Signal-of-origin prediction matters more than it first appears. Galleri identifies the likely tissue of origin with greater than 90% accuracy.1 A positive result therefore arrives attached to a specific diagnostic pathway rather than a whole-body hunt, which is the difference between a workup and an odyssey.

Incremental yield is substantial. Added to USPSTF grade A and B screenings, the test increased cancer detection more than seven-fold.2 Screening four times detected four times as many cancers as standard of care alone.1 Most of what it finds, existing screening was never going to find, because existing screening covers four or five cancers and ignores the rest.

The number the marketing prefers

For the 12 cancers responsible for roughly two-thirds of US cancer deaths, episode sensitivity was 73.7%.1

That is a real and important figure, and the one that will appear on clinic websites. It is not interchangeable with the all-cancer number, and conflating them is the first misrepresentation to watch for. “Detects 74% of cancers” is false. “Detects 74% of the twelve cancers that cause two-thirds of cancer deaths” is true and considerably less catchy.

Randomised evidence adds what PATHFINDER 2 cannot supply. NHS-Galleri, the first and only randomised controlled trial of an MCED, observed a reduction in stage IV cancers of more than 20%.1 Stage shift is a surrogate rather than a mortality endpoint. It is the right surrogate, and a 20% shift is not noise.

Specificity makes this test defensible. Sensitivity makes it insufficient. Both belong in the consent conversation.

Why grade B-, and what is still missing

B- means promising trial evidence that does not reach a hard outcome, or a single registrational study without independent replication. MCED meets both conditions.

The missing endpoint is mortality. Stage shift is a strong proxy and the figure above is encouraging, but the history of cancer screening contains technologies that produced clean stage shifts and no survival benefit because they were detecting indolent disease, lengthening the period of knowing rather than the period of living. Nobody should assume that is happening here. Nobody can rule it out yet either.

The missing replication is independent. The evidence programme spans more than 380,000 participants across multiple studies,1 which is genuinely large and also almost entirely the manufacturer’s. That is how registrational programmes work, and it is still a reason to hold the grade below B.

What this means for programmes that include it

A liquid biopsy is now a standard line item in premium screening, though where it sits in a price list varies more than the marketing suggests. YEARS includes one in its €7,600 Evolve programme and above, and explicitly not in the €1,900 Core tier.5 Biograph reserves its multi-cancer blood test for the $15,000 Black membership and leaves it out of the $7,500 Core. Fountain Life includes one in its annual membership.

For anyone shopping specifically for an MCED test bundled with imaging, that makes Evolve at €7,600 the cheapest route we have found, since the comparable American tier costs $15,000 once you require the blood test. Buying the test on its own, outside any programme, is cheaper again, and worth pricing before committing to a bundle in order to get it.

On current evidence that inclusion is defensible, which is more than can be said for several other things sold in the same packages. The defensibility is conditional, and the condition is how the result gets communicated. Three tests of whether a provider handles this well: whether the consent material states the all-cancer sensitivity rather than only the deadly-twelve figure, whether it says in writing that a negative result does not substitute for mammography, colonoscopy, cervical or lung screening, and whether a named clinician owns the pathway when a result comes back positive.

A provider clearing all three is selling the test the trial supports. A provider presenting a negative result as a clean bill of health is selling something the trial specifically contradicts.