Why Don’t We Catch Cancer Before It Spreads?

Imagine testing 100,000 people, of whom about 1,000 have cancer. A test that catches 80% of cancers and wrongly flags 10% of healthy people finds 800 cancers and sends nearly 10,000 healthy people for scans and biopsies: fewer than one positive in thirteen is cancer (Derived). Raise specificity to 99.5% and the false alarms fall to about 500. This is why multi-cancer blood tests are designed for very high specificity, accepting that they will miss many early cancers, and why a single-cancer test with 90% specificity works only when the diagnostic pathway has demonstrated benefits that justify its costs and harms; colonoscopy is invasive and is not a cheap or risk-free follow-up.

Earlier diagnosis can lengthen measured survival without extending life

Imagine a cancer that would be diagnosed at 68 and cause death at 70. A blood test finds it at 63, but treatment changes nothing. Survival after diagnosis rises from two years to seven, even though the person dies at the same age. This is lead-time bias. Screening also preferentially finds slower-growing cancers, which remain detectable longer, and can identify cancers that would never cause illness. These effects explain why the dramatic survival gap between localized and metastatic cancer is a reason to investigate screening, rather than a measurement of its benefit. NCI’s explanation of screening statistics

The right comparison follows comparable groups invited to screening or usual care and counts cancer deaths, harms and resource use. A sustained reduction in advanced disease can be an earlier signal, but it does not automatically establish fewer deaths. The diagnostic pathway also belongs in the intervention: a positive blood signal needs a located tumour, a confirmed diagnosis and timely effective treatment. Without those steps, a technically accurate result can consume clinical capacity without improving outcomes. NHS-Galleri performance paper, September 2026

The age example is hypothetical. Screening benefit should be assessed against mortality and harms; five-year survival alone cannot establish it.

The screens that already work are underused

Proven screening saves lives now. Low-dose CT scanning of current and former heavy smokers cut lung-cancer deaths by 20–24% in two large trials. In 2024, between about 19% and 25% of eligible Americans were up to date with it, depending on the survey. Colorectal screening has also become easier: in July 2024 the FDA approved Guardant’s Shield, the first blood test cleared as a primary colorectal screening option, which detects 83% of colorectal cancers. A blood draw at a routine appointment may reach people who would never book a colonoscopy, provided those who test positive then get one.

Galleri’s mixed result is more informative than either headline

The NHS trial did not meet its primary goal of reducing combined stage III–IV diagnoses. GRAIL’s May presentation reported a 14% reduction in stage IV diagnoses of the 12 prespecified cancers overall, with larger reductions in later rounds. That secondary result is encouraging, but its inferential status and the failed primary endpoint belong in the same sentence. Cancer mortality follow-up remains necessary. Different screening rounds also represent different populations: the first searches an accumulated backlog, while subsequent rounds increasingly search for newly detectable disease. GRAIL’s May 2026 results factsheet

September’s peer-reviewed performance analysis puts numbers on that distinction. Positive predictive value fell from 58.0% in the first round to 50.4% and 45.8% in later rounds; all-cancer episode sensitivity ranged from 26.7% to 37.2%. These describe test performance, not a mortality benefit. A negative result therefore cannot exclude cancer or replace other screening. Meanwhile, the ACS estimates only one in five eligible Americans received lung screening in 2024. Expanding a proven programme offers a separate route to improvement while the multi-cancer evidence matures. Neal et al., Nature Medicine, September 2026; ACS’s 2024 lung-screening analysis

NHS-Galleri’s aggregate positivity denominator is screening episodes, not unique people. Advisory recommendations, regulatory authorization and proof of mortality benefit are separate milestones.

Who is building what

Blood-test developers, screening programmes and the trials that will decide whether new tests save lives. Search the record, or filter by role.

Sensitivity figures depend heavily on cancer stage and type. A test’s value is judged by stage shift first and, ultimately, by fewer cancer deaths in randomised trials.

Four numbers to watch

First, cancer deaths in the long-term follow-up of NHS-Galleri and similar randomised trials. Second, the stage IV reduction across successive screening rounds. Third, uptake of proven screening, especially lung CT among eligible smokers. Fourth, the specificity of blood tests in routine use, outside the controlled conditions of trials.