A single blood sample can detect more than 50 types of cancer; the US FDA expert committee votes in favor, heralding the arrival of the AI screening era?

On September 23, 2026, Eastern Time, a landmark moment will arrive in the field of early detection of multiple cancers.

The U.S. FDA's Expert Committee on Molecular and Clinical Genetics reviewed GRAIL's Galleri multi-cancer early detection product and made its stance clear in three key votes:

10:0, indicating that Galleri has reasonable security guarantees;6:4, which considers its effectiveness to be reasonably guaranteed;7:2 (1 abstention) - Galleri's clinical benefits outweigh its risks.

While the FDA's final decision on nationwide sales approval still needs to be made in the coming months, the expert committee's endorsement has significantly boosted market confidence. As Guggenheim analyst Subbu Nambi put it, this has virtually "eliminated the risk."

A single vial of blood can screen for more than 50 types of cancer—it sounds like science fiction. But those in the know understand that the most noteworthy aspect of this revolution is the clarion call that AI and data are ushering in for cancer screening.

Abandoning the traditional approach: it sells an AI-powered cancer detection system.

When liquid biopsies are mentioned, many people think of searching for specific "gene mutations" released by tumors in the blood. But Galleri takes a completely different approach.

When tumor cells die, they release cell-free DNA (cfDNA) into the bloodstream. This DNA retains tissue- and tumor-specific methylation patterns. Galleri's core technology reads these signals through targeted methylation sequencing and then feeds them into a machine learning (AI) model.

The model not only needs to determine whether there is a cancer signal in the sample, but also needs to predict which organ the signal is most likely to come from.

Therefore, from a business and technological perspective, GRAIL is no longer a "testing reagent company" in the traditional sense. What it truly sells is an AI-powered cancer identification system comprised of molecular testing, algorithmic classification, and clinical interpretation.

35% Sensitivity and 99.85% Specificity: A Restrained Product Philosophy

If you're looking for a "one-shot blood test to detect all cancers," Galleri's data might surprise you: in the PATHFINDER 2 study in the United States, Galleri's overall 12-month sensitivity was approximately 35.0%.

This means that a significant portion of cancers go undetected during testing. So why did FDA experts still vote in favor? The answer lies in another astonishing statistic: 99.85% specificity.

In screening tens or even hundreds of millions of ordinary healthy people, the biggest disaster is not "missed diagnoses," but "false positives." Even a false positive rate of only a few percentage points, when amplified to the population level, will create massive panic and trigger countless unnecessary CT scans, PET scans, endoscopy, and even puncture biopsies, enough to overwhelm the existing medical system.

Galleri's 99.85% specificity ensures that it rarely produces false alarms. At the same time, its positive predictive value (PPV) reaches approximately 77.0%, meaning that once it issues an alarm, the probability of a cancer diagnosis is already quite high.

"Prioritize controlling false positives and precisely target critical blind spots"—this is Galleri's product philosophy.

Data shows that Galleri has a sensitivity of approximately 61.3% for 12 types of highly lethal cancers that currently lack mature screening methods (such as pancreatic cancer, ovarian cancer, and bile duct cancer) but cause about two-thirds of cancer deaths in the United States.

The FDA committee explicitly emphasized that Galleri is positioned as a "supplement, not a replacement" to existing standard screening methods (such as mammography and colonoscopy). Its target market is the market that fills the most obvious gap in the existing cancer screening system.

A massive test for 140,000 people: though they didn't reach the finish line, they proved their "true value."

The biggest controversy surrounding Galleri comes from the UK NHS-Galleri randomized controlled trial, which involved more than 140,000 participants.

A large-scale trial of the Galleri cancer screening technology in the UK, involving 140,000 people, did not meet its primary endpoint of reducing the number of patients with intermediate and advanced cancers in total, but it reduced the number of patients with the most deadly advanced cancers by 22% and 26% in the second and third rounds of screening, respectively.

The trial successfully detected deadly cancers that would otherwise be exposed at a late stage, bringing them to the early to mid-stages, prompting a 7-2 vote by a panel of FDA experts to determine that the benefits outweighed the risks.

This trial attempted to answer a fundamental public health question: Could incorporating Galleri into the healthcare system reduce the incidence of advanced (stage III/IV) cancer? From a strictly statistical perspective, the trial did not meet its pre-specified primary endpoint.

In the field of early cancer screening, "detecting more" does not equal "screening success" (there may be overdiagnosis). True success is "to bring forward the detection of deadly cancers that would otherwise only be exposed in stage IV to stage I/II".

This is precisely the core reason why the FDA vote, which ended in a 6:4 split on effectiveness, ultimately concluded with a 7:2 vote that the benefits outweighed the risks. Experts believe that while it is not perfect, its potential to reduce advanced cancer has already crossed the threshold for clinical trials.

Disrupting the old rules: Screening models and data flywheels in the AI era

For the past few decades, cancer screening for humans has been "organ-based": low-dose CT scans were performed to check the lungs, and colonoscopies were performed to check the intestines.

Galleri represents a paradigm shift: changing the basic unit from "organ" to "person." Future health checkups might begin with a blood sample for a full-body scan; if AI flags an abnormality, the algorithm will then guide the targeted examination of specific organs.

In this transformation, AI's opportunities are just beginning:

AI tracking over a longitudinal timeline : Cancer is dynamic. If a person has blood tests for five consecutive years, a single signal may never reach the "positive" standard, but in the eyes of AI, the subtle changes over those five years may have already revealed a high-risk trend.Establishing a closed-loop decision-making process after a positive result : When an abnormality is found in the blood, should the patient undergo a CT scan or an MRI first? When should the testing be stopped? In the future, a large language model combining medical records and family history, along with multimodal AI, will become the most powerful external brain guiding doctors' decisions.An Unreplicable Data Flywheel : GRAIL currently possesses real-world data from CCGA, the PATHFINDER series, and over 140,000 NHS-Galleri individuals. When this methylation data is fully integrated with patient outcomes such as survival and treatment, GRAIL will have a continuously evolving "real-world AI training platform"—a formidable competitive advantage that will be difficult for any newcomer to overcome.

Conclusion: AI, which should not be mythologized, is a future that has already arrived.

Of course, AI cannot create something from nothing. If a tiny tumor in its very early stages does not release enough cfDNA into the bloodstream, even the most powerful algorithm will be useless.

To ensure the safety of population-level screening, the Galleri currently under review by the FDA still uses "locked machine-learning classifiers," which means that it cannot "automatically update its learning" every day in clinical applications at this stage.

However, the expert vote on September 23, 2026, is by no means just a permit for a blood testing product.

It proves that simultaneously capturing dozens of cancer signals in a vast sea of blood is technically feasible. If, in the next few years, it can further demonstrate in the real world its ability to reduce cancer mortality and successfully complete cost calculations for commercialization, then it will be a significant breakthrough.

Today, we stand at the starting point of a new era of cancer screening and hear the first clarion call of the AI era.

Risk Warning and DisclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.