GRAIL's Cancer Test Just Failed Its Big Trial. Here's Why I Still Use It.

Archive note: This February 2026 issue followed The Longevity Letter’s move from Kit to Beehiiv. The NHS-Galleri results received a fuller presentation in May 2026; the study discussion below includes that update.

Hi there,

This week, the most talked-about blood test in cancer screening stumbled publicly.

GRAIL’s NHS-Galleri randomized trial enrolled roughly 142,000 adults and did not meet its primary endpoint. The result matters more than the stock-market response.

The result prompted strong reactions about both Galleri and the wider multi-cancer detection field.

I want to understand the encouraging findings without explaining away the primary result.

Six people in my family have had cancer. Most of them died from it. One of my cousins was a physician in his 30s. He left behind a wife and two kids.

I trained as a hematology/oncology fellow at Columbia before moving into startups and longevity medicine. Cancer is personal for me in a way it is for a lot of families, and that's exactly why I've spent years tracking this field.

I have built an individualized cancer-screening approach for myself and my patients. I have had Galleri, two whole-body MRIs, and whole-genome sequencing. Those are my choices. Whole-genome sequencing surveys far more DNA than a targeted panel, but no test detects or interprets every inherited risk.

The protocol I walk through in this issue isn't a theoretical framework. It's what I actually do.

In this issue:

• The NHS-Galleri results and the wider detection field.
• My approach to cancer screening, with established tests kept first.
• Dated Quick Hits on research, policy, and drug access.

Galleri missed its primary endpoint. What follows?

The company and what it's betting on

GRAIL’s approach rests on a plausible goal: find more cancers before symptoms. Earlier diagnosis can improve treatment for some cancers. For screening, though, the test must ultimately show that benefits outweigh false alarms, overdiagnosis, and missed cancers.

The company was spun out of Illumina in 2015. Jeff Huber, a former Google executive who had just lost his wife to colorectal cancer, became CEO. Bezos Expeditions and Bill Gates both wrote checks in the Series A. Illumina eventually bought the whole company in 2021 for $8 billion, a deal that immediately drew antitrust scrutiny from both U.S. and European regulators because Illumina makes the sequencing machines that power liquid biopsy. (The regulators won and GRAIL was forced to re-spin as an independent public company in June 2024 and now trades on NASDAQ as GRAL.)

This has attracted large investments. A trial still has to show that screening saves lives.

The science: what methylation-based cfDNA actually means

The Galleri test is a liquid biopsy, and that phrase is doing a lot of work.

Cell-free DNA consists of fragments released into blood by cells through several processes, including cell death. Most comes from non-cancer cells. The test looks for cancer-related patterns within that mixture.

GRAIL reads methylation patterns in that cfDNA. Methylation is a chemical modification (a methyl group attached to specific cytosine bases on the DNA strand) that regulates gene expression. Normal cells have characteristic methylation patterns. Cancer disrupts them in ways that are consistent within a cancer type and distinguishable from normal tissue.

The test sequences cfDNA from a blood draw, reads hundreds of thousands of methylation sites, and runs a machine learning model against a trained reference set. If a cancer-like pattern appears, a second algorithm predicts where in the body the signal originated, which tells the clinician where to look first.

High specificity reduces false-positive results among people without cancer. It doesn’t tell a person with a positive result how likely they are to have cancer. In the NHS-Galleri report, specificity was 99.55%, but only 52% of positive tests were followed by a cancer diagnosis. Sensitivity was lower across all cancers than for the twelve prespecified types. A negative result cannot rule cancer out.

The NHS trial: what actually happened

Here's what the 142,000-person study actually showed.

The trial ran three years across England, enrolling participants aged 50-77 in a randomized controlled design. The pre-specified primary endpoint was a statistically significant reduction in combined Stage III and IV cancer diagnoses in the Galleri arm versus standard of care. That endpoint was not met.

The May 2026 report described a fourfold increase in screen-detected cancers, not fourfold more cancers overall. Stage IV diagnoses of twelve prespecified cancers were 22% and 26% lower in the second and third screening rounds. Those secondary findings are encouraging, but the combined Stage III/IV primary endpoint remained unmet.

Longer follow-up may answer other questions, but short follow-up alone does not explain away a null primary result. An increase in detection is not automatically a survival benefit: lead-time bias and overdiagnosis still matter.

GRAIL submitted its PMA application in January 2026. Submission is not approval. FDA review considers the evidence for the intended use; it does not make a missed randomized endpoint irrelevant.

The February 3, 2026 appropriations law includes a Medicare MCED coverage pathway in Section 6221. It begins in 2029 and requires FDA authorization and a Medicare coverage determination, with age and frequency limits. It doesn’t guarantee immediate coverage for the first test FDA authorizes.

The competitors: a crowded, well-funded field

"GRAIL" has become synonymous with multi-cancer detection. That's both a marketing win and a clinical misconception.

Other companies use different signals. Some focus on one cancer; others aim to detect many. Those are different clinical claims.

Guardant Health: SHIELD

FDA authorized Shield in July 2024 for colorectal cancer screening in average-risk adults forty-five and older. It is not an MCED test. A positive result needs colonoscopy, and Shield does not replace diagnostic or high-risk surveillance colonoscopy. Its approval does not validate Galleri or another company’s multicancer test.

Exact Sciences: CancerGuard

Exact Sciences acquired the older CancerSEEK program, which combined proteins and mutations. The marketed Cancerguard test combines DNA methylation and protein biomarkers. It launched as a laboratory-developed test in 2025; its history should not be used to describe today’s assay as mutation-based. More signal types do not automatically mean better clinical outcomes.

Freenome

Freenome studies multiomics, including signals from nucleic acids and proteins. Its Roche agreement supports development and commercialization. Investment does not prove screening efficacy, and regulatory status can change after this February 2026 issue.

Harbinger Health

Harbinger Health is developing a different multicancer approach, including research in higher-risk groups. Such work may improve our understanding of performance by population; it does not establish superiority for a person with obesity or an FDA timeline.

Delfi Diagnostics

Delfi’s program analyzes cell-free DNA fragmentation patterns, rather than relying only on methylation. This is another detection approach to validate for its particular intended use.

The Medicare statute does not name one company or one winning test. Each eligible test still needs the required authorization and coverage assessment.

Where this is heading

Current assays can detect cancer-related signals. Whether using them in people without symptoms produces a net health benefit remains the central question.

The limitation everyone is racing to solve is Stage I sensitivity. Small early tumors shed very little DNA, and that signal-to-noise problem doesn't disappear with incremental refinement. The next approaches are attacking it differently.

Combining DNA, RNA, proteins, or other signals may improve detection, but every assay needs validation. Adding measurements does not guarantee more sensitivity without a loss of specificity.

MIT’s peptide-sensor research takes a different approach: administer sensors, then measure reporters released after enzyme cleavage. Early work used experimental systems; an at-home multicancer screening kit is a development goal, not an established test.

Inherited risk should guide validated surveillance. A pathogenic BRCA2 or PALB2 variant can change breast-risk management; it does not show that a particular MCED platform is calibrated to that person or can substitute for the recommended surveillance.

The field still needs to answer whether screening reduces important harms and deaths, which patients benefit, and how follow-up should work. Regulation and reimbursement do not settle those clinical questions.

My protocol for cancer screening (updated Feb 2026)

Cancer screening is a strategy, not a single test. The GRAIL news reinforces why I never put all the weight on any one tool, and why the right protocol depends on your biology, not on which test got good press this week.

Some tests I’ve chosen have uncertain screening benefits. My own plan should not become everyone’s checklist.

Layer 1: Family history and genetic risk

Family history comes first in my assessment. Six relatives in my family have had cancer. The cancer types, affected relatives, and ages at diagnosis can help determine whether genetic counseling or syndrome-specific screening is useful. A known familial pathogenic variant deserves a different plan from a vague family history.

Layer 2: Germline genomics

When the history suggests an inherited syndrome, clinical genetic counseling can help choose a targeted panel or broader sequencing. I have had whole-genome sequencing. Broader testing can find additional variants, but it also creates uncertain findings and does not replace interpretation or a plan for the result.

Layer 3: Traditional screening, done right and on time

Keep established cancer screening current: colorectal, breast, cervical, and risk-eligible lung screening, plus a shared prostate-screening decision when appropriate. Starting ages, methods, and intervals depend on the guideline and the person’s risk. My annual-mammography preference is one approach; not every guideline uses the same interval.

Layer 4: Whole-body MRI, for the right person

I have had two whole-body MRIs. MRI has defined roles in some high-risk surveillance plans, but a broad family history alone does not prove whole-body screening will help. ACR notes the lack of proven life extension and the risk of nonspecific findings and unnecessary follow-up. This needs an individual discussion.

Layer 5: MCED tests

I still consider Galleri for selected patients after discussing the uncertainty. PATHFINDER 2 reported more screen-detected cancers when the test was added to recommended screening. Detection yield does not establish fewer deaths. Patients need to understand false positives, missed cancers, follow-up, and possible out-of-pocket costs.

Cancerguard is another available approach with a different assay. Prices and access change. Head-to-head or comparable prospective data are needed before concluding which test serves a patient better.

Do not postpone established screening while deciding whether to add an emerging test. Revisit the plan as your risk and the evidence change.

📋 Takeaways:

• The NHS-Galleri primary endpoint was unmet; some secondary stage and detection findings were favorable.
• The Medicare pathway starts in 2029 with authorization, coverage, age, and frequency conditions.
• Single-cancer FDA authorization is different from MCED validation.
• Start with family history and established screening. Add genetic testing, MRI, or MCED only when their expected use and tradeoffs fit the person.

The technology can detect signals. The harder task is showing when finding those signals improves health enough to justify the harms and costs.

If you found this useful, share it with someone who should be thinking about cancer screening. 

⚡ Quick Hits

🏛️ Up to $144M for aging research: ARPA-H’s February 2026 announcement named seven teams receiving milestone-based contracts over five years. A Brown/Rochester-led project was awarded a contract worth up to $22M to test Censavudine, a drug developed for HIV, against retrotransposon activity and age-related decline. The announcement described a planned human trial, not a proven healthspan benefit.

🇯🇵 iPSC therapy news, February 2026: A Japanese panel recommended conditional approval of AMCHEPRY for Parkinson’s motor symptoms inadequately controlled by medication, and RiHEART for severe heart failure from ischemic cardiomyopathy despite standard treatment. Formal approval followed on March 6, with seven-year limits and required follow-up evaluation. Lasting clinical benefit still needs confirmation.

💊 Price announcement, February 2026: Novo Nordisk announced lower U.S. list prices for selected products beginning in 2027. List prices do not tell us a patient’s final cost; insurance, rebates, and self-pay terms can differ.

⚗️ Peptide sensors: MIT researchers used AI to design enzyme-responsive sensors. Early experiments are promising; the ARPA-H-supported thirty-cancer home-kit goal is not a validated clinical capability.

🧬 Klotho clinical development: In February 2026, Klothea Bio announced an early trial of AKL003, an experimental alpha-Klotho mRNA therapy. The trial was designed chiefly to assess safety and protein expression, not to show that it extends human lifespan.

🫀 GLP-1 safety-label update: In January 2026, FDA requested removal of suicidal-behavior and ideation warnings from the affected GLP-1 weight-management drug labels after its review found no increased risk. That decision addresses this particular risk; it does not mean the drugs have no psychiatric or other adverse effects for an individual.

The trial missed its primary endpoint. The favorable secondary findings are worth following, but they don’t cancel that result. I’m keeping both in view.

The 2026 law may affect access later. It doesn’t change the clinical result or make today’s tests Medicare-covered.

If you do one thing after reading this: ask your doctor which standard screenings you're actually due for. That's still where most of the preventable deaths hide, not in the $1,000 blood tests, but in the colonoscopies people keep rescheduling and the mammograms that slip to "next year."

If you want a starting point, use the 5-layer framework above. Start at Layer 1. Know your family history. That's free, and it changes everything.

To your healthspan,

Hillary Lin, MD

Co-Founder & CEO, Curator

Follow me for more longevity insights: YouTube | LinkedIn | Instagram | TikTok

Want to turn your wellness brand into a full-service health destination? Learn about Curator's platform or Get Started Here

Events listed when this issue was published

👱🏻‍♀️👩🏻‍🦰👩🏻👧🏽👧🏾 Livelong Women's Health Summit: April 17-18, 2026, SF, CA. Delighted to join 50 other thought leaders in speaking on women's longevity! (Code HILLARY50 for $50 off)

Offers listed when this issue was published

This issue included an affiliate offer from NeuroAgeTx. Check current services and prices before buying.

This issue included a Timeline offer for urolithin A. Human studies of selected markers do not establish a human lifespan benefit. Check current terms.

GlycanAge reports an age estimate from IgG glycosylation. It doesn’t diagnose or treat inflammation. The March 2026 discount listed in this issue has expired.

I have liked ZBiotics products. That preference is not proof that they extend life or offset alcohol’s health risks. Check current offer terms.

Have a question or topic you want covered? Reply to this email.

Was this forwarded to you? Subscribe here to get The Longevity Letter every week.