A $949 Blood Test Cut Stage IV Cancer Diagnoses by 20%. The Trial That Proved It Technically Failed.

142,000 people in England, aged 50 to 77, gave blood three times over two years. Half got a methylation-based multi-cancer early detection (MCED) test called Galleri, made by GRAIL in Menlo Park, California. Half got standard care. On February 19, 2026, GRAIL published topline results from what is now the largest randomized controlled trial of a multi-cancer screening test ever conducted.

Here is the headline: the primary endpoint did not reach statistical significance.

And here is the number buried three paragraphs deeper: Stage IV diagnoses of the 12 deadliest cancer types dropped more than 20% in the screening arm during the second and third rounds. Emergency cancer presentations, the kind where a patient shows up at a hospital already in crisis, fell substantially. Overall cancer detection climbed four-fold compared to standard screening alone.

Both of those things are true at the same time. Understanding why requires a lesson in how cancer screening actually works at the molecular level.

What the Test Measures

Galleri does not sequence tumor DNA. It reads methylation patterns. When a cell becomes cancerous, the distribution of methyl groups across its genome changes in characteristic ways. Fragments of cell-free DNA (cfDNA) shed by tumors into the bloodstream carry these aberrant methylation signatures. GRAIL's platform uses targeted bisulfite sequencing across more than 100,000 methylation sites, then feeds the results into a machine learning classifier trained on data from the Circulating Cell-free Genome Atlas (CCGA) study.

Two outputs emerge: a binary signal (cancer detected or not) and a Cancer Signal Origin (CSO) prediction that identifies the most likely tissue of origin with 93.4% accuracy. In the PATHFINDER 2 trial, the largest US interventional MCED study at 35,000+ participants, the test delivered a specificity of 99.6% (false positive rate: 0.4%) and a positive predictive value of 61.6%, meaning 6 of every 10 people flagged by the test actually had cancer.

For the 12 cancer types responsible for roughly two-thirds of cancer deaths in England and the United States, the sensitivity was 76.3%. For all 50+ cancer types combined, the sensitivity was 51.5%. More aggressive cancers shed more cfDNA and are more reliably detected: liver/bile duct hit 93.5%, pancreas 83.7%, ovary 83.1%, and esophagus 85.0%.

The Stage Shift Paradox

The trial's pre-registered primary endpoint was a statistically significant reduction in Stage III and Stage IV cancers combined within the 12 deadly cancer types. It missed.

But the reason it missed reveals something important about what the test is actually doing. When you catch a cancer that would have presented at Stage IV and find it instead at Stage III, you have shifted its stage. Stage III diagnoses in the intervention arm rose. Stage IV diagnoses fell. Combined, the two categories partially cancel each other out. GRAIL noted a "higher than anticipated incidence of Stage III cancers" across both arms, suggesting something about the trial's population or the NHS's diagnostic pipeline produced more Stage III disease than projected.

Look at Stage IV alone, and the story changes. Second and third screening rounds cut Stage IV diagnoses by more than 20% for the 12 pre-specified deadly cancers. Similar reductions appeared across all cancers. And the absolute number of Stage I and Stage II diagnoses in the deadly cancer group rose in the intervention arm, exactly what stage shift predicts.

Professor Charles Swanton, thoracic medical oncologist at University College Hospital London and one of the trial's chief investigators, framed the distinction plainly: "When cancer is detected before distant metastatic spread, we can often treat with curative intent, combining surgery, radiotherapy, and systemic therapy in an effort to eradicate all disease. Once distant metastases are established, treatment typically shifts toward long-term disease control and symptom management; durable cures become uncommon in most solid tumours."

The Screening Coverage Gap: A Calculation

In the United States, the USPSTF issues A or B screening recommendations for exactly four cancer types: breast (mammography), cervical (Pap/HPV), colorectal (colonoscopy or FIT), and lung (low-dose CT for high-risk smokers aged 50-80). Prostate screening (PSA) carries a C rating. Everything else has no recommended population-level screen.

Those four cancers account for approximately 196,000 of the roughly 610,000 annual US cancer deaths, or about 32%. That leaves approximately 414,000 deaths per year, 68% of the total, from cancers with no recommended screening pathway. Pancreatic. Ovarian. Liver. Esophageal. Stomach. Head and neck. These are cancers where the five-year survival rate for Stage I can exceed 70% but drops below 10% at Stage IV, and most patients are diagnosed at Stage III or IV precisely because nobody screens for them.

Galleri's sensitivity for these specific cancers is high: pancreas 83.7%, ovary 83.1%, liver 93.5%, esophagus 85.0%, stomach 66.7%, head and neck 85.7%. Those numbers come from case-controlled studies, so real-world performance in asymptomatic populations will differ. PATHFINDER 2's interventional data put the 12-cancer sensitivity at 73.7%, somewhat lower. But even at 73.7%, the test is finding nearly three-quarters of the deadliest cancers that currently have zero screening infrastructure.

No other deployed technology covers this gap. Galleri's four-fold improvement in cancer detection rate over standard screening in the NHS trial reflects the sheer volume of cancers currently invisible to the healthcare system until symptoms manifest.

The Economics

Galleri's list price is $949 per test, with many providers offering it at $799 or less. Most US health insurance plans do not cover it. TRICARE and some employer plans do. GRAIL offers financial assistance for lower-income patients.

A cost-effectiveness analysis published in Cancer Epidemiology, Biomarkers & Prevention in August 2025 modeled annual MCED testing for Americans aged 50-79. The incremental cost-effectiveness ratio (ICER) came in at $66,043 per quality-adjusted life-year (QALY) for the general population, below the $100,000-$150,000 threshold typically used by US health economists. For higher-risk subpopulations (hereditary cancer syndromes, obesity, diabetes, smoking history), the ICER was lower still.

Here is a rough calculation the cost-effectiveness models do not typically show. Annual US cancer care spending is approximately $208 billion (NCI projections). Late-stage cancers consume a disproportionate share because of immunotherapy, targeted therapy, and extended palliative care. A 20% reduction in Stage IV diagnoses, if it held at population scale, would shift hundreds of thousands of treatment trajectories from "manage disease" to "attempt cure." Stage I colorectal cancer treatment costs roughly $30,000-$40,000. Stage IV averages $150,000-$250,000, sometimes far more with immunotherapy combinations. Multiply the per-patient savings across even a fraction of the approximately 200,000 late-stage cancers diagnosed annually in the unscreened categories, and the offset against screening costs becomes significant.

But the cost model has a caveat: it was funded by GRAIL.

Strongest Counterargument

The primary endpoint failed. A greater-than-20% Stage IV reduction sounds clinically meaningful, but it emerged from a subset analysis across screening rounds, not from the pre-registered primary outcome. The trial was powered for combined Stage III and IV reduction, and it did not hit significance. Critics will argue, reasonably, that if you can only demonstrate the benefit by slicing the data a particular way, the robustness of the finding is uncertain.

There is also the false positive problem at scale. A 0.4% false positive rate is low by screening standards, but the US has roughly 100 million people between ages 50 and 79. Annual screening at 0.4% would generate approximately 400,000 false positives per year, each requiring diagnostic imaging, biopsies, specialist visits, and patient anxiety. Medicare covers colonoscopies and mammograms. It does not yet cover the $949 Galleri test or the cascade of diagnostic workups triggered by a positive result.

And the overall sensitivity of 51.5% means the test misses approximately half of all cancers. A negative result could create false reassurance. "No cancer signal detected" does not mean no cancer. That distinction matters at population scale, where people may treat a negative MCED result as permission to skip mammograms or colonoscopies.

Limitations

Several caveats apply to these findings. First, the NHS-Galleri population is demographically narrower than the US population; generalizability across different ethnic groups and cancer mixes is unproven. Second, there is no mortality data yet from this trial. Stage shift is a well-established surrogate for survival improvement, but it is not proven in the MCED context specifically. GRAIL plans to extend follow-up by 6-12 months, but mortality endpoints may require years of additional data. Third, the cost-effectiveness models rely on GRAIL-funded assumptions, and independent replication with different cost structures has not been published. Fourth, the insurance coverage gap creates a two-tier system: patients who can afford $949 get screened for 50+ cancers, while everyone else gets the same four-cancer standard of care. Fifth, time-to-diagnostic-resolution improved over the trial's three rounds as physicians gained familiarity with the test, suggesting that early-round performance may understate steady-state real-world effectiveness, but also that the learning curve itself is a deployment barrier.

What You Can Do

If you are 50 or older and your physician offers the Galleri test, it is worth a conversation. Ask about the false positive rate (0.4%), the sensitivity for your specific risk profile, and whether your insurance or employer health plan covers the $949 cost. HSA and FSA accounts can typically be used. Do not treat a negative result as a replacement for recommended single-cancer screening (mammography, colonoscopy, Pap smears, low-dose CT for former smokers). Galleri is designed as an add-on, not a substitute.

If you have hereditary cancer risk (BRCA, Lynch syndrome, Li-Fraumeni), the cost-effectiveness data is even more favorable for you. Discuss MCED testing with your oncologist or genetic counselor.

If you are a clinician, the PATHFINDER 2 data shows that diagnostic resolution improves with experience. The first positive results in a practice may generate anxiety and over-investigation. By the third round, workup efficiency improves substantially. Community oncology practices that adopt MCED testing should expect a learning curve.

If you are a health policy maker, watch the FDA's pending premarket approval decision for Galleri. An approval would trigger the coverage debate that determines whether this test reaches 100 million Americans or stays a $949 out-of-pocket luxury.

The Bottom Line

The NHS-Galleri trial asked whether a blood test could shift the stage at which cancers are found. The answer, by the pre-registered primary endpoint, is: the evidence is suggestive but not yet statistically definitive. By the secondary endpoints and subset analyses, the answer is more direct: Stage IV diagnoses of the deadliest cancers fell by more than 20%, early-stage detection rose, and cancer discovery quadrupled relative to standard screening. No serious safety concerns emerged. The question is no longer whether methylation-based screening can detect cancer in the blood. It can. It is whether the health system is willing to pay $949 per person per year to close the screening gap that currently leaves 68% of cancer deaths invisible until it is too late.