Liquid Biopsy — Early Cancer Detection Before Conventional Diagnosis

The detection of circulating tumour DNA (ctDNA) and cell-free DNA (cfDNA) fragments in blood establishes the technological basis for liquid biopsy. Diehl et al. (2008) demonstrate ctDNA detection in colorectal cancer; subsequent work extends detection across cancer types. Key technical advances include digital droplet PCR (ddPCR) for mutation detection and next-generation sequencing for broader genomic profiling. By 2019, several single-cancer liquid biopsy tests have received FDA breakthrough device designation. The technology establishes that tumour-derived nucleic acids are detectable in blood at clinically relevant cancer stages. This is foundational positive evidence: the proxy signal (blood-based ctDNA) does correlate with the presence of cancer, and the biological basis is sound. The question is whether the signal is detectable early enough and reliably enough to satisfy the claim.

GRAIL's Galleri test, using methylation sequencing of cfDNA, reports prospective evidence from two distinct intended-use settings. PATHFINDER, published in 2023 after enrolment and testing beginning earlier, returned results to asymptomatic adults aged 50 years or older. SYMPLIFY, also published in 2023, evaluated symptomatic patients already referred for cancer investigation in England and Wales: sensitivity was 66.3% overall and specificity 98.4%, while sensitivity rose from 24.2% at Stage I to 95.3% at Stage IV. This distinction matters: SYMPLIFY is a diagnostic-triage cohort, not a high-risk screening population. Across the evidence, detection is materially weaker at the early stages where the record's claim has greatest value, leaving the early-stage sensitivity gap as the primary evidential challenge.

The NHS-Galleri trial randomised 142,250 participants in England (71,122 to Galleri plus standard care and 71,128 to standard care alone) in an asymptomatic population aged 50–77. Its pre-specified primary objective was a reduction in late-stage (Stage III–IV) cancer incidence, first assessed across twelve pre-specified cancer types; mortality was not the primary endpoint. Randomisation and longer-term linkage can support later mortality comparison, but stage shift and mortality benefit are distinct outcomes. At this point in the record's chronology, the late-stage-incidence result and mortality follow-up were both pending. The claim therefore remained FRAGMENTING: early detection was demonstrably achievable in some contexts, while whether it produces clinically meaningful outcome benefit rather than lead-time or overdiagnosis effects remained unresolved.

The cancer screening literature contains multiple examples of earlier detection not producing mortality benefit due to lead-time bias (earlier detection without actual life extension) and overdiagnosis (detection of cancers that would never have caused clinical harm). Prostate-specific antigen (PSA) screening is the canonical case: widespread adoption followed by evidence that many screen-detected cancers were low-risk and treated unnecessarily. Lung cancer screening with low-dose CT demonstrates genuine mortality benefit in high-risk populations, establishing that early detection can work. The liquid biopsy literature has not yet accumulated sufficient follow-up data to distinguish genuine benefit from lead-time bias at scale. This is the specific proxy gap: blood-based ctDNA signal (measured object) as proxy for clinically meaningful early cancer (asserted object). A cancer detectable by liquid biopsy may be a fast-growing, treatment-responsive cancer where earlier detection genuinely helps — or a slow-growing cancer that would never have caused symptoms, making detection harmful through unnecessary treatment. The measurement cannot currently discriminate between these.

Illumina completed its acquisition of GRAIL in 2021 in a transaction valued at approximately $7.1 billion. Competition authorities in both the European Union and United States challenged the transaction; the European Commission ordered restorative divestment measures in 2023, and Illumina announced it would divest after the December 2023 US Fifth Circuit decision. The separation was completed through a GRAIL spin-off on 24 June 2024. Meanwhile, GRAIL and competitors continued commercial deployment and clinical-validation programmes. Galleri was commercially available in the United States at approximately $950 per test without FDA approval for population screening. This constitutes anticipatory institutional evidence: substantial capital and commercial deployment preceded definitive randomised clinical-utility and mortality evidence.

Investigators present the NHS-Galleri primary results orally at the 2026 American Society of Clinical Oncology Annual Meeting on 30 May 2026. The results are published as late-breaking abstract LBA100 in the Journal of Clinical Oncology conference supplement, not as a full peer-reviewed results article. The pre-specified primary endpoint — a statistically significant combined reduction in Stage III and Stage IV diagnoses across twelve cancer types — is not met. Stage IV diagnoses fall by more than 20% in later screening rounds, but a higher-than-anticipated number of Stage III diagnoses offsets that result in the combined endpoint. The Galleri arm reports a four-fold higher overall cancer detection rate than standard-care screening alone. No mortality endpoint is reported. Independent commentary via the Science Media Centre emphasises that stage shift and mortality benefit are distinct questions and describes the evidence as conference-presented results rather than a published full study. The awaited trial event has therefore occurred without producing the clean resolution anticipated by the record's original attractor.

Sasieni et al. publish the NHS-Galleri primary randomized results in the New England Journal of Medicine on 22 September 2026. Among 142,250 randomized participants, the prespecified primary endpoint of Stage III/IV cancer incidence across twelve cancer types is not reduced by MCED screening plus usual care versus usual care alone (incidence-rate ratio 1.03; 95% CI 0.92–1.14; P=0.63). The key Stage IV secondary endpoint has an incidence-rate ratio of 0.86 (95% CI 0.74–1.00). Trial-related adverse events occur in fewer than 1% of participants and none are serious. The peer-reviewed article therefore confirms the mixed evidential state previously represented from ASCO 2026: the primary late-stage endpoint is negative while a Stage IV signal remains compatible with benefit but requires further follow-up. Mortality is not resolved.

Neal et al. publish prespecified secondary test-performance endpoints from the NHS-Galleri intervention arm in Nature Medicine on 22 September 2026. Across the three annual screening rounds, positive predictive value is 58.0%, 50.4% and 45.8%; specificity remains 99.50–99.60%; and episode sensitivity for all cancers is 37.2%, 27.1% and 26.7%, respectively. Cancer-signal-origin accuracy remains above 91% by round. These results strengthen evidence that the assay can identify cancer signals with very high specificity in population screening, while also quantifying a substantial sensitivity limitation that persists across repeated annual rounds. The analyses are descriptive and do not establish mortality benefit or resolve overdiagnosis; those outcomes are reserved for future reporting.