Participation in the NHS-Galleri trial was associated with a modest increase in diagnostic delay rates for suspected head and neck, lung, and upper gastrointestinal cancers during the first year of the study, according to findings published in JAMA.
The analysis was published alongside two editorials that placed the findings in broader context. One editorial emphasized that the regional delay analysis does not show that spillover effects biased the internal randomized comparison in NHS-Galleri. Another noted that the trial, according to previously reported topline results, did not meet its primary endpoint of a statistically significant reduction in combined stage III-IV cancer.
The NHS-Galleri trial evaluated a blood-based multicancer early detection (MCED) test that uses cell-free DNA and artificial intelligence to screen for more than 50 cancer types and predict the cancer signal origin. The spillover analysis examined whether regional participation in the trial affected routine cancer diagnostic services in England.
Researchers analyzed publicly available National Health Service data from all 21 cancer alliance regions in England from April 2021 to September 2024. Eight regions participated in the NHS-Galleri trial, and 13 regions did not. The primary analysis included nearly 1.9 million referrals for suspected head and neck, lung, or upper gastrointestinal cancers, a group selected because these cancers were identified as having higher signal detection with the test and were not subject to routine screening.
The primary outcome was diagnostic delay, defined as the percentage of referrals that took longer than 28 days to reach diagnostic resolution.
During the first 6 months following trial initiation, diagnostic delay rates increased from 29% to 30% in participating regions and decreased from 29% to 26% in nonparticipating regions. In adjusted analyses, participation in the trial was associated with an approximately 3-percentage-point increase in diagnostic delay rates. The difference persisted during the second 6-month period, when delay rates were approximately 5 percentage points higher in participating regions, but was no longer statistically significant following the first year.
The researchers estimated that the differences corresponded to 9,591 additional referrals experiencing diagnostic delay during the first year of the trial. Average time to diagnostic resolution increased by about 2 days in participating regions, with most of the increase occurring following initial specialist consultation.
Referral rates also increased in some early analyses. During the first 6 months, participating regions had about 24 additional referrals per 100,000 population for suspected head and neck, lung, and upper gastrointestinal cancers compared with nonparticipating regions. A broader group of cancers with higher MCED signal detection showed a similar delay pattern during the first year, whereas cancers categorized as lower detection did not show statistically meaningful delay differences.
The mechanism remains uncertain. The researchers noted several possible explanations, including diagnostic workups following positive MCED test results, increased referral-seeking behavior prompted by trial invitations or publicity, and additional demands on local diagnostic services during trial implementation.
However, the number of positive-test-related referrals alone appeared too small to explain the estimated increase in delayed referrals. The researchers estimated that positive MCED test results would have generated about 100 to 400 new referrals during the first year, compared with 9,591 additional referrals experiencing delay. That mismatch suggests broader information spillover, local service disruption, or other system-level effects may have contributed.
Several limitations should be considered. The participating regions were not randomly selected and had greater socioeconomic deprivation and fewer health care staff than nonparticipating regions, raising the possibility of residual confounding. The study used a difference-in-differences design, which depends on the assumption that participating and nonparticipating regions would have followed similar trends without the trial; however, only 6 months of pretrial data were available to assess baseline trends.
The available data also could not distinguish between referrals that resulted in cancer diagnosis and referrals in which cancer was ruled out. Investigators could not determine whether delays occurred among intervention participants, control participants, or other patients in the same regions.
In an accompanying editorial, Issa J. Dahabreh, MD, of the Harvard T.H. Chan School of Public Health, and Roger J. Lewis, MD, of the David Geffen School of Medicine at UCLA, wrote that the findings are most relevant to a policy comparison: cell-free DNA screening within the trial vs usual care in the absence of both the test and the trial. They emphasized that outcomes in the intervention group may reflect nonspecific trial effects as well as the screening strategy itself, and that the regional data cannot determine whether spillover mechanisms affected the randomized study groups differently. As a result, the findings do not, by themselves, demonstrate bias in the internal validity of the NHS-Galleri randomized comparison.
A second editorial by Richard Sullivan, MD, of King’s College London, and Christopher M. Booth, MD, of Queen’s University, placed the findings in the broader debate over multicancer detection testing. They noted that full peer-reviewed NHS-Galleri results remain unpublished and that topline results reported in February 2026 indicated the trial did not meet its primary endpoint of reducing combined stage III -IV cancer.
Sullivan and Booth argued that, based on currently available evidence, the GRAIL multicancer detection test has no demonstrated utility for population-level cancer screening. They also cautioned about potential harms, including false-positive results, false reassurance from negative results, diagnostic cascades, system strain, and opportunity costs. At $949 per test, they estimated that nationwide rollout in England among adults aged 50 to 77 years would cost about $19.2 billion annually for testing alone, excluding infrastructure, downstream diagnostics, and overtreatment.
The researchers concluded that participation in the population-based MCED screening trial was associated with a modest and temporary increase in diagnostic delays that was unlikely to have affected the NHS-Galleri trial’s primary findings.
“Decisions to implement MCED testing should account for downstream capacity needs and potential effects on all patients in the implementing health system, not just those receiving testing,” wrote lead study author Sean Mann, MSc, of RAND in Arlington, Virginia, and colleagues.
Disclosures: Disclosures can be found in the published investigation. The authors of both editorials reported no conflicts of interest.
Source: JAMA