Abstract
Background
NHS-Galleri, a randomized controlled trial, evaluates the clinical utility of a blood-based multi-cancer early detection (MCED) test added to usual care among individuals aged 50-77 years in England.
Methods
Peripheral blood was collected at up to 3 annual visits. Following baseline blood collection, participants were randomized 1:1 to intervention (blood tested) or control (blood stored) arms. Participants and study personnel were initially blinded. Reductions of stage III/IV (primary endpoint) and IV (key secondary) cancers of 12 prespecified types in intervention vs control were measured ~3 years after the last participant’s randomization.
Results
Overall, 71,122 and 71,128 participants were randomized to intervention and control arm, respectively. The primary endpoint was not met (incidence rate ratio [IRR; intervention vs control]:1.03 [0.92-1.14], p=0.63). For secondary endpoints, the IRR for stage IV cancer after 3 screening rounds was 0.86 (0.74-1.00). <1% of participants experienced related adverse events, none were serious.
Conclusions
In NHS-Galleri, we did not observe a reduction in stage III/IV cancers across 12 prespecified cancers after 3 screening rounds when MCED testing was added to standard-of-care. Further follow-up is planned to assess how findings for stage III/IV cancers and the difference in stage IV cancers evolve over time. (Funded by GRAIL, Inc.; ClinicalTrials.gov #NCT05611632)
Background
The survival advantage of earlier-stage cancer detection reflects broader eligibility for curative treatment and lower disease burden.1 Standard-of-care screening contributes to earlier diagnosis and improved survival,2 yet recommended single-cancer screening accounts for <7% of all cancer diagnoses in England (US: <15%).2,3
Multi-cancer early detection (MCED) tests offer a different approach by simultaneously screening for multiple cancer types using a biological specimen (e.g., blood).4–6 One MCED test analyses methylation patterns in cell-free DNA (cfDNA)6 to detect a cancer signal and predict the associated tissue type or organ (cancer signal origin) has demonstrated detection across a broad range of cancer types. The test has shown strong performance in case-control7 and prospective clinical studies,8,9 as well as in real-world experience.10–12
Here, we report the primary results from the NHS-Galleri trial (NCT05611632), a randomized controlled trial evaluating the clinical utility of a clinically validated MCED test when added to usual care in a national healthcare system among participants aged 50-77 years at invitation, over 3 annual screening rounds.
Methods
Trial oversight and design
The NHS-Galleri trial (ISRCTN91431511, registration date: 16 September 2021; NCT05611632, registration date: 28 October 2022), sponsored and funded by GRAIL, Inc. (Menlo Park, CA, USA), is a prospective randomized controlled trial to evaluate the clinical utility and effectiveness of an MCED blood test (Galleri® [GRAIL, Inc.]) when added to usual care.6 Briefly, eligible individuals (those aged 50-77 years who did not have a cancer diagnosis nor treatment within the 3 years before enrolment, and were not being investigated or under clinical suspicion for cancer; Supplementary Methods section 1) in 8 participating English Cancer Alliance regions were invited. Invitations, information sheets, and consent forms were developed with input from diverse members of the public.13 Further information on trial oversight and author responsibilities are provided in Supplementary Methods section 2. Study design details (e.g., recruitment methods, trial objectives) are provided in publications6,13 and Supplementary Methods sections 1-11.
GRAIL, collaborating with The Cancer Research UK Cancer Prevention Trials Unit, was responsible for study design, conduct, reporting, and data analysis. The trial protocol, available at nejm.org, (Supplementary Methods section 10) received national ethical approval from Wales Research Ethics Committee 1 (Ref: 21/WA/0141). An Independent Data Monitoring Committee (IDMC) oversaw and assessed clinical study progress, reviewed clinical/safety data, and provided design and conduct recommendations to GRAIL.
Procedures
At initial appointments conducted in mobile phlebotomy clinics, after informed consent, a baseline peripheral blood sample was collected by routine venipuncture, and participants were then randomized 1:1 into intervention or control arm (Figure S1).6 All participants, except those for whom the study team had received death notification, were invited to 2 further annual (12 months±6 weeks) blood draws; participants missing the second visit could return for the third. All participants and trial staff were initially blinded; intervention-arm participants with a positive (cancer signal detected) MCED test result had results communicated by unblinded trial nurses.13 Unblinded statisticians reviewed safety data and presented results to the IDMC in closed biannual meetings.
Blood samples from intervention-arm participants were analyzed using the MCED test. The study team facilitated referral of participants with a positive result for diagnostic testing via established National Health Service (NHS) urgent cancer referral pathways, following existing anatomical site-specific suspected cancer guidelines based on the MCED test cancer signal origin prediction.6,14 All control-arm and remaining intervention-arm samples were stored for consented future testing.
Cancer outcomes were collected via linked national healthcare datasets. Participants were passively followed for cancer diagnoses and death, until emigration or consent withdrawal.
Important protocol deviations are summarized in Tables S1-S2.
Statistical analysis
The statistical analysis plan is available in Supplementary Methods section 11. Microsimulation indicated ~140,000 participants were needed for ≥90% power for the between-arm primary endpoint after three screening rounds, with follow-up through 12 months after the last participant’s third visit, across varied natural history assumptions;15 an approximately 20% relative reduction in stage III/IV cancers at the 12 prespecified sites was projected under the then most likely scenario.6 All analyses comparing the two arms were intention-to-treat analyses (conducted among randomized participants [Randomized Analysis Set]).
Cancer diagnoses from randomization to 12 July 2025 registered by the National Cancer Registration and Analysis Service (NCRAS) were included, for a follow-up period of 3 screening rounds plus 12 months after the last participant’s third visit (Primary Analysis Window). Person-years of follow-up were measured from randomization to the first event of interest (i.e., stage III or IV cancer diagnosis, death, consent withdrawal, or emigration).
Incidence rates were calculated as the number of first eligible events divided by person-years at risk, with 95% confidence intervals (CI) for incidence rates calculated assuming a Poisson distribution. Incidence rate ratios (IRR), 95% CI, and two-sided p-values were calculated using exact Poisson methods. Unstageable cancers and cancers with missing stage (Supplemental Methods section 7) were excluded from stage-specific analysis. The primary endpoint was the between-arm comparison of stage III/IV cancer incidence rates after 3 screening rounds and ≥12 months follow-up after the last participant’s third visit. Statistical testing followed a prespecified fixed sequence:
12 prespecified cancer types responsible for >60% of cancer deaths6,7: lung, head and neck, colon/rectum, pancreas, myeloma/plasma cell neoplasm, liver/bile duct, stomach, esophagus, anus, lymphoma, ovary, and bladder.
All routinely staged cancer types excluding prostate cancer.
All routinely staged cancer types.
Testing stopped at the first non-significant step to control type I error. Secondary endpoints (Supplemental Methods section 6), including stage IV cancer incidence reduction, were interpreted descriptively with 95% CIs. The width of the CIs are not adjusted for multiple testing and should not be used for hypothesis testing.
Safety
Adverse events were reported by participants or recorded by qualified trial members. Study-related adverse events were those related to study procedures until NHS referral; “device-related” adverse events occurred after NHS referral with a reasonable temporal possibility of a causal relationship with the onward referral, investigational device, or its use, despite these activities occurring outside of the trial. Adverse events were assessed for relatedness, expectedness, and severity, aligned with the medical safety and monitoring plan (Supplementary Methods section 8). Further information on anxiety among those receiving a positive MCED test result,16 and radiation exposure and complications associated with diagnostic procedures outside the study, will be published separately.
Data sharing
Individual-level participant cancer outcome data are not publicly available and cannot be shared by the authors due to confidentiality requirements and restrictions under the applicable data-sharing agreements with National Disease Registry Service (NDRS), an organization within NHS England (formally called the National Cancer Registry and Analysis Service [NCRAS]). Aggregated, anonymized data supporting the findings are provided in the manuscript and supplementary materials.
Results
Population
1,496,311 individuals were invited, and 142,826 participants consented to the trial over 10.5 months13; 142,250 (99.6%) had their first blood draw and were randomized to intervention (71,122) or control (71,128) arms (Figure 1). Of the randomized intervention- and control-arm participants, 64,888 (91.2%) and 65,066 (91.5%) returned for their second and 62,867 (88.4%) and 63,101 (88.7%) for their third annual blood draw, respectively. Participants had 12 months follow-up after each of the first two MCED screening rounds, and a median of 17 (range: 12-22) months follow-up after the third round.
Figure 1. CONSORT diagram for the NHS-Galleri study cohort.
Clinically eligible refers to randomized participants who satisfied the inclusion and exclusion criteria. The clinically evaluable analysis set comprised participants who were successfully linked in the routine datasets used in this study.
By design, participants from the most socioeconomically deprived areas were over-represented (Table 1). Ethnic distribution broadly represented the equivalent-aged English population (Table S3).13 Most participants never smoked and reported being up to date on standard-of-care screening. Demographics and characteristics were well balanced between arms at baseline and across screening rounds (Table 1, S4).
Table 1. Demographics and baseline characteristics of the Randomized Analysis set in the NHS-Galleri trial.
| Intervention arm (N = 71,122) | Control arm (N = 71,128) | Total (N = 142,250) | ||
|---|---|---|---|---|
| Median age at randomization, years (IQR) | 66 (59-71) | 66 (59-71) | 66 (59-71) | |
| Age groups at randomiz ation, n (%) | 50-59 years | 18,791 (26.4) | 18,889 (26.6) | 37,680 (26.5) |
| 60-69 years | 28,596 (40.2) | 28,792 (40.5) | 57,388 (40.3) | |
| 70-79 yearsa | 23,734 (33.4) | 23,447 (33.0) | 47,181 (33.2) | |
| Sex, n (%) | Female | 35,723 (50.2) | 35,739 (50.2) | 71,462 (50.2) |
| Male | 35,399 (49.8) | 35,389 (49.8) | 70,788 (49.8) | |
| Race/ethni city, n (%) | Asian or Asian British | 2360 (3.3) | 2365 (3.3) | 4725 (3.3) |
| Black, Black British, Caribbean, or African | 1016 (1.4) | 1038 (1.5) | 2054 (1.4) | |
| White | 66,552 (93.6) | 66,566 (93.6) | 133,118 (93.6) | |
| Mixedb | 776 (1.1) | 768 (1.1) | 1544 (1.1) | |
| Otherc | 318 (0.4) | 309 (0.4) | 627 (0.4) | |
| Missing | 100 (0.1%) | 82 (0.1%) | 182 (0.1%) | |
|
IMD quintile, n
(%) |
1 - Most deprived | 16,071 (22.6) | 16,062 (22.6) | 32,133 (22.6) |
| 2 | 13,942 (19.6) | 13,980 (19.7) | 27,922 (19.6) | |
| 3 | 15,034 (21.1) | 14,956 (21.0) | 29,990 (21.1) | |
| 4 | 14,372 (20.2) | 14,476 (20.4) | 28,848 (20.3) | |
| 5 - Least deprived | 11,469 (16.1) | 11,414 (16.0) | 22,883 (16.1) | |
| Missing | 234 (0.3%) | 240 (0.3%) | 474 (0.3%) | |
| Prior history of cancer | n (%) | 5328 (7.5) | 5355 (7.5) | 10,683 (7.5) |
| Median (IQR) time since prior cancer diagnosis-yrsd | 10.2 (6.7–15.0) | 10.2 (6.7–14.8) | 10.2 (6.7–14.9) | |
| Self-reported screening participati on, n (%)e | Never heard of cancer screening | 2056 (2.9) | 2108 (3.0) | 4164 (2.9) |
| Heard of cancer screening but never been invited | 13,955 (19.6) | 13,871 (19.5) | 27,826 (19.6) | |
| Invited for cancer screening but never done it | 861 (1.2) | 883 (1.2) | 1744 (1.2) | |
| Invited for cancer screening, but did not participate every time | 2428 (3.4) | 2361 (3.3) | 4789 (3.4) | |
| Invited for cancer screening and participated every time | 51,013 (71.8) | 51,150 (72.0) | 102,163 (71.9) | |
| Don’t know | 655 (0.9) | 608 (0.9) | 1263 (0.9) | |
| Decline to answer | 51 (0.1) | 58 (0.1) | 109 (0.1) | |
| Missing | 103 (0.1) | 89 (0.1) | 192 (0.1) | |
|
Smoking status, n
(%) |
Currently smokes | 4744 (6.7) | 4708 (6.6) | 9452 (6.6) |
| Formerly smoked | 27,272 (38.3) | 27,148 (38.2) | 54,420 (38.3) | |
| Never smoked | 39,004 (54.8) | 39,187 (55.1) | 78,191 (55.0) | |
| Missing | 102 (0.1%) | 85 (0.1%) | 187 (0.1%) | |
|
BMI (kg/m2), n
(%) |
<25 | 23,058 (32.4) | 23,203 (32.6) | 46,261 (32.5) |
| 25–30 | 28,009 (39.4) | 27,925 (39.3) | 55,934 (39.3) | |
| ≥30.0 | 19,739 (27.8) | 19,706 (27.7) | 39,445 (27.7) | |
| Missing | 316 (0.4) | 294 (0.4) | 610 (0.4) | |
| History of selected medical diagnoses, n (%) | Barrett’s esophagus | 811 (1.1) | 810 (1.1) | 1621 (1.1) |
| Chronic bowel disease | 1660 (2.3) | 1632 (2.3) | 3292 (2.3) | |
| Chronic liver disease | 1167 (1.6) | 1056 (1.5) | 2223 (1.6) | |
| Diabetes | 7091 (10.0) | 7129 (10.0) | 14,220 (10.0) | |
| Chronic pancreas disease | 304 (0.4) | 298 (0.4) | 602 (0.4) | |
| Systemic lupus erythematosus | 129 (0.2) | 140 (0.2) | 269 (0.2) | |
| Education, n (%) | School at or before the age of 15 years | 12,672 (17.8) | 12,887 (18.1) | 25,559 (18.0) |
| CSEs, O-levels, or equivalent | 19,262 (27.1) | 19,260 (27.1) | 38,522 (27.1) | |
| A-levels or equivalent | 6517 (9.2) | 6509 (9.2) | 13026 (9.2) | |
| Further education without a degree | 14,776 (20.8) | 14,570 (20.5) | 29,346 (20.6) | |
| Bachelor’s degree or equivalent | 11,953 (16.8) | 11,970 (16.8) | 23,923 (16.8) | |
| Further degree (e.g., Masters, PhD | 5842 (8.2) | 5850 (8.2) | 11,692 (8.2) | |
| Missing | 100 (0.1) | 82 (0.1) | 182 (0.1) | |
BMI, body mass index; IMD, index of multiple deprivation (an area-based measure of relative deprivation); IQR, interquartile range.
Individuals had to be aged 50-77 years at invitation; however, they could be aged 78 years by enrolment.
Includes White and Black Caribbean; White and Black African; White and Asian; and any other Mixed or multiple ethnic background.
Includes Arab and any other ethnic background not listed.
If a participant had multiple prior cancers, the number of years since the most recent prior cancer to randomisation date is used.
Percentage is based on the count out of all non-missing responses. Percentages may not sum to 100% across groups due to rounding.
Late-stage cancer outcomes with MCED screening
Overall, 3637 intervention- and 3400 control-arm participants were diagnosed with cancer, of which 3431 and 3177 were routinely staged cancer types; 1421 and 1298 were of the 12 prespecified cancer types (Table S5-S6). There were 706 and 688 participants diagnosed with stage III/IV cancer of the 12 prespecified types in the intervention vs control arm after 3 screening rounds, yielding an incidence rate of 300.5 (95% CI: 278.7-323.5) and 292.5 (271.0-315.1) per 100,000 person-years, respectively (Figure 2, Table 2). Stage III/IV cancer incidence rate did not differ significantly between intervention and control arms (IRR: 1.03 [0.92-1.14], p=0.63). Participant-level descriptive counts (Table S7) revealed an increase in stage III cancers of the 12 prespecified types diagnosed in the intervention (360) vs control (287) arm, outweighing the decrease in stage IV cancers (341 vs 394). The overall result was driven by the higher incidence rate of stage III/IV cancers in the intervention vs control arm in the prevalent round (IRR: 1.19 [0.98-1.43]). In contrast, IRR for stage III/IV cancers in incident rounds comparing intervention vs control were 0.95 (0.77-1.17) for second round (post-hoc); 0.88 (0.73-1.07) for third round (Table S8).
Figure 2. Cumulative probability of stage III/IV cancers and stage IV cancers by trial arm after 3 screening rounds.
Cumulative probability of A) stage III/IV, 12 prespecified cancer types; B) stage IV, 12 prespecified cancer types; C) stage III/IV, all routinely staged cancer types; and D) stage IV, all routinely staged cancer types (all post-hoc analyses). Distribution of cancers by stage and trial arm shown for E) 12 prespecified cancer types (post-hoc analysis) and F) all routinely staged cancer types. There were two participants with stage 0 cancer in each of the intervention and control arms. All of these were Paget’s disease, which is invasive at stage 0.
IRR, incidence rate ratio.
Table 2. Stage III/IV and stage IV cancer outcomes after 3 annual screening rounds in the NHS-Galleri trial.
| Prespecified group of 12 cancer typesa | All routinely staged cancer types excluding prostate cancer | All routinely staged cancer types | ||||
|---|---|---|---|---|---|---|
| Intervention arm (N = 71,122) | Control arm (N = 71,128) | Intervention arm (N = 71,122) | Control arm (N = 71,128) | Intervention arm (N = 71,122) | Control arm (N = 71,128) | |
| Stage III/IV | ||||||
| Participants with stage III/IV cancer, n (%) | 706 (1.0) | 688 (1.0) | 903 (1.3) | 848 (1.2) | 1266 (1.8) | 1208 (1.7) |
| Follow up, person-years | 234,959 | 235,249 | 234,667 | 235,044 | 234,062 | 234,495 |
| Incidence rate per 100,000 person-years (95% CI) | 300.5 (278.7, 323.5) |
292.5 (271.0, 315.1) |
384.8 (360.1, 410.7) |
360.8 (336.9, 385.9) | 540.9 (511.5, 571.5) |
515.1 (486.5, 545.0) |
|
Incidence rate difference
(95% CI) |
8.0 (−23.1, 39.2) | 24.0 (−10.9, 59.0) | 25.7 (−15.9, 67.4) | |||
| Incidence rate ratio (95% CI) | 1.03 (0.92, 1.14) p=0.63 | 1.07 (0.97, 1.17) | 1.05 (0.97, 1.14) | |||
| Stage IV | ||||||
| Participants with stage IV cancer, n (%) | 342 (0.5) | 397 (0.6) | 412 (0.6) | 467 (0.7) | 503 (0.7) | 563 (0.8) |
| Follow up, person-years | 235,584 | 235,662 | 235,501 | 235,585 | 235,347 | 235,434 |
| Incidence rate per 100,000 person-years (95% CI) | 145.2 (130.2, 161.4) |
168.5 (152.3, 185.9) |
174.9 (158.5, 192.7) |
198.2 (180.7, 217.1) | 213.7 (195.5, 233.2) |
239.1 (219.8, 259.7) |
|
Incidence rate difference
(95% CI) |
−23.3 (−46.0, −0.7) | −23.3 (−48.0, 1.4) | −25.4 (−52.6, 1.8) | |||
| Incidence rate ratio (95% CI) | 0.86 (0.74, 1.00) | 0.88 (0.77, 1.01) | 0.89 (0.79, 1.01) | |||
Lung, head and neck, colon/rectum, pancreas, myeloma/plasma cell neoplasm, liver/bile duct, stomach, esophagus, anus, lymphoma, ovary, and bladder.
Note: In this table for event specific IRR analysis, a participant is counted based on the first qualifying event meeting the corresponding endpoint definition during follow-up. Counts may differ from descriptive analyses tables in which a single cancer/stage is assigned using the highest stage and earliest diagnosis if tied.
Among the 12 prespecified cancer types, after 3 screening rounds stage IV cancer incidence rate was 145.2 (130.2-161.4) per 100,000 people-years in the intervention arm vs 168.5 (152.3-185.9) per 100,000 people-years in the control arm; IRR: 0.86 (0.74-1.00) (Figure 2, Table 2). In the intervention vs control arm, IRR for stage IV cancers were 0.91 (0.71-1.18) in the prevalent (first) round, 0.78 (0.57-1.06) in the second round, and 0.74 (0.57-0.95) in the third round (Table S9).
The observed IRRs for stage III/IV and IV cancers appeared similar between the 12 prespecified types and all routinely staged types (with and without prostate; Figures 2, S2; Tables 2, S8-S9). Notably, across all routinely staged cancers, IRR for stage IV cancer was 0.78 (0.61-1.00) in the second round and 0.80 (0.65-0.99) in the third round (Table S9).
As a prespecified secondary endpoint, we combined fatal cancers (regardless of stage) with stage III/IV cancers. With this definition, there were 1415 intervention-arm and 1359 control-arm events from all routinely staged cancer types after 3 screening rounds (IRR: 1.04 [0.97-1.12]; Table S10). Excluding cancers with standard-of-care screening (i.e., breast, cervix, colon/rectum), 1077 intervention-arm and 1030 control-arm participants had stage III/IV cancer diagnosed after 3 rounds (IRR: 1.05 [0.96-1.14]; Table S11).
In prespecified exploratory subgroup analyses, the greatest (relative and absolute) differences in stage III/IV and IV cancer rates in the intervention vs control arm were observed among men, those with prior cancer history, and with lower education status; the greatest stage IV decrease among IMD quintiles was observed in the most deprived (Figure S3).
Increase in early-stage cancers with MCED screening
The observed cancer detection rate in the first round was higher in the intervention arm than the control arm: ratio 1.21 (1.11-1.31; Table S12).
In the third round, the proportion of routinely staged cancers that were stage I/II was 62.8% in the intervention vs 59.9% in the control arm (Table S13). Relatedly, in a post-hoc analysis, after 3 screening rounds, there were 647 stage I/II cancers from the 12 prespecified types in the intervention arm and 559 in the control arm (relative risk: 1.16 [1.03-1.30]) and 1007 vs 846 for stage I/II/III cancers (relative risk 1.19 [1.00-1.30]; Table S14). Similar results were observed for participants with all routinely staged cancer types (with and without prostate; Table S14).
Route to cancer diagnosis
Based on the ‘route to diagnosis’ recorded by the cancer registry, in the intervention vs control arm, there was an 18.9% (1775 vs 1493) increase in diagnoses following urgent cancer referral (the referral route for positive MCED test results)14 and a 21.3% (225 vs 286) decrease in emergency presentations after 3 screening rounds (Table S5).
In the intervention arm, 1173/3637 (32.3%) participants had screen-detected cancers: 937/3637 (25.8%) following a positive MCED test result and 236/3637 (6.5%) by NHS screening programs (Figure 3). In the control arm, 290/3400 (8.5%) had cancer detected by NHS screening programs. Thus, adding MCED to standard-of-care screening was associated with an apparent increase in screen-detected cancers (1173 vs 290) and in screening program sensitivity (32.3% vs 8.5%; Table S15).
Figure 3. Distribution of the route to cancer diagnosis by trial arm after 3 annual screening rounds.
For participants with multiple cancers diagnosed, only the route to diagnosis of the first cancer is included.
There were 20.8% fewer clinically detected cancers (i.e., cancers neither screen-detected nor MCED-detected): 2464/3637 (67.7%) vs 3110/3400 (91.5%) in the intervention vs control arm (Figure 3). Similar results were observed across all 3 screening rounds (Figure S4).
Safety
Across 3 screening rounds, 371 (0.52%) randomized intervention-arm and 321 (0.45%) randomized control-arm participants had related adverse events; none were serious (Tables S16-S17). The excess was primarily from anxiety following receipt of a positive MCED result. Among intervention-arm participants, 337 (0.47%) had study-related adverse events, 27 (0.04%) had device-related adverse events, and 11 (0.02%) had study- and device-related adverse events (Table S16). Among control-arm participants, 321 (0.45%) had study-related adverse events. The most common study-related adverse events were vessel puncture site bruises, syncope, dizziness, vessel puncture site hematoma, and anxiety.
Participants reported high adherence to standard-of-care cancer screenings (Table S18).
Discussion
NHS-Galleri is a randomized controlled trial evaluating the clinical utility of an MCED test for population screening. After three annual MCED screening rounds with 12 months follow-up after the last participant’s third visit, we did not observe a reduction in the primary outcome of stage III/IV cancers among the 12 prespecified types. Although the prespecified hierarchical plan to adjust for multiple testing precluded any formal testing of secondary outcomes, the results suggest the possibility that the intervention was associated with fewer stage IV cancer diagnosed in incident rounds, in which observed trends would be expected to reflect those of a stable screening program with a routinely screened population. Stage IV cancer reduction is clinically important since stage IV disease confers worse survival than earlier-stage.1
The observed lower rate of stage IV cancer was offset by an increase in stage III cancers, resulting in no overall reduction in stage III/IV combined. Prevalent-round stage III cancers appeared to be the predominant driver of the missed primary endpoint for the 12 prespecified types, with the higher number in the intervention arm potentially reflecting several non-mutually exclusive mechanisms, including earlier detection of cancers that might otherwise have presented at stage IV, and longer than anticipated lead-time effects. Additional follow-up will provide more evidence about the intervention’s effects on long-term III/IV incidence. It is possible that the full clinical utility of MCED screening will emerge over time, after depletion of the prevalent pool, consistent with published screening studies.17–20 Longer follow-up is needed to determine whether the excess of stage III cancers is transient and the observed stage distribution translates into lower cancer-specific mortality.
Late-stage cancer incidence was chosen as the primary endpoint because it is clinically relevant and can be assessed earlier than cancer-specific mortality, although its value as a surrogate may vary by cancer type and screening context.21–23 A favorable result requires a lower late-stage cancer incidence rate, not merely more early-stage diagnoses or a lower late-stage proportion, which could reflect overdiagnosis without reducing advanced disease. At trial design, microsimulation enabled sample-size planning despite limited evidence on cancer dwell times and preclinical ctDNA detectability. These simplified assumptions projected a larger and quicker stage shift than observed. Subsequent model development has considered broader sojourn-time ranges and lower screening sensitivity.24 Further analyses incorporating trial data should better characterize these parameters and inform the design of future studies. Cancer-specific and nested mortality will be analyzed at 3 and 6 years after final trial visit as secondary endpoints; the trial was not powered for mortality endpoints.
The intervention arm had more cancers diagnosed at earlier stages than the control arm. The current analysis cannot distinguish between more cancers being found before the equivalent cancers are diagnosed in the control arm versus overdiagnosis. Given that cfDNA shedding is a hallmark of more aggressive cancers, these excess cancers may be more indolent. Recent evidence suggests that cancers detected by circulating tumor DNA-based MCED tests are unlikely to be over diagnosed, as survival outcomes in case control studies in participants with positive MCED results were consistent with population averages, even in early stages.25 Rather, the excess stage I/II cancers diagnosed in the intervention arm are likely due to earlier detection of clinically significant disease. Longer follow-up will help address these questions. Future analysis to assess overdiagnosis is planned and will be published separately.
The addition of MCED testing increased the number of screen-detected cancers and decreased the number of clinically detected cancers. Diagnoses via emergency presentations were reduced by 21.3%; such diagnoses are associated with poorer survival, even for the same cancer type and stage.2,26
The performance characteristics of the MCED test are the subject of another publication.27 Briefly summarized, positive MCED tests were obtained in 722 (1.03%) in the first round, 518 (0.80%) in the second round, and 561 (0.90%) in the third round. Of these 1801 positive MCED tests, 937 (52%) were found to have cancer; of these, 815 (87.0%) were located in the first-ranked predicted site of origin predicted by the positive MCED.
The NHS-Galleri trial is a randomized controlled trial designed to evaluate an MCED test at a population level, providing robust evidence on the implementation of this approach in routine screening settings. The findings from this trial are consistent with those from the prospective PATHFINDER 2 study.9 The fact that these two independent prospective datasets demonstrate reproducibility across different healthcare settings (UK and US) supports the feasibility of implementing this MCED screening at scale. Future analyses of health economics and outcomes research endpoints that provide further insights will be published separately.
Strengths
Trial strengths included the randomized, double-blind design, ascertainment of outcome data from a quality-assured national cancer registry, socioeconomically diverse representation of the target population, high retention throughout the trial, and the use of existing standardized diagnostic pathways within the NHS,14 ensuring clinical management as close to usual care as possible. These features reduced bias and supported generalizability of the trial results.
Limitations
The follow-up was relatively short for a screening trial and cancer incidence between the two arms has not yet converged, which suggests the full effects of the MCED intervention may not yet be mature. Continued follow-up is needed to assess how these effects evolve; for example, it is plausible that the excess of stage III cancers in the intervention arm reflects earlier diagnosis of cancers that remain asymptomatic in the control arm and may later present at stage III or IV. The combined stage III/IV primary endpoint may not capture all potentially clinically meaningful effects of screening, such as a shift from stage IV to stage III, diagnosis at an earlier substage within the same stage, and diagnosis before symptom onset. These findings suggest that future screening trials should carefully consider the primary and secondary endpoints including the duration of follow-up after the last screen. Additionally, traditional stage categories should be complemented by outcomes that capture clinically meaningful effects within and beyond stage, such as circulating tumor DNA (ctDNA) fraction. The study was not powered to evaluate a reduction in cancer mortality. The study was not designed to assess different MCED screening intervals or likely participation in a routine MCED screening program. Negative MCED test results were not returned and individual medical records were unavailable, limiting assessment of effects on screening behavior and diagnostic work-up. The study was not powered to assess clinical utility of MCED screening within demographic subgroups or individual cancer types. Adverse events in the trial were systematically collected up until referral into the NHS system for diagnostic workup; thereafter, adverse events were captured only if reported directly by the participant or a clinician. As a result, this trial does not fully characterize the safety profile during diagnostic evaluation following a positive MCED test. This limitation is addressed with the PATHFINDER2 study.9
Conclusion
In this large randomized controlled trial, annual MCED testing added to usual care did not significantly reduce stage III/IV cancers across 12 prespecified types after 3 screening rounds and ≥12 months follow-up. Further follow-up is planned to assess how findings for stage III/IV cancers and the observed difference in stage IV cancers translate to clinical benefit over time.
Disclosure forms provided by the authors are available with the full text of this article at NEJM.org.
Supplementary Material
Acknowledgements
This work uses data provided by patients and collected by the NHS as part of their care and support. The data are collated, maintained, and quality-assured by the National Cancer Registration and Analysis Service (NCRAS), which is part of NHS England. Eligible individuals were identified and invited from participating Cancer Alliances. The NHS DigiTrials population database was used to recruit the majority of participants in this trial. The Cancer Research UK Cancer Prevention Trials Unit collaborated with GRAIL, Inc. on study design, conduct, reporting, and data analysis. We acknowledge and thank all participants in the NHS-Galleri trial and their families, as well as the clinical site staff, investigators, the Trial Steering Committee, the IDMC, research coordinators, data managers, and operational teams who supported the study. We acknowledge members of the NHS England Cancer Alliances in Cheshire and Merseyside, East of England, East Midlands, Greater Manchester, Kent and Medway, Northern, South East London, West Midlands, overseen by the NHS England Senior Operations Team. We acknowledge the contributions of Nan Zhang for biostatistics; Liz Holmes and Poorwa Roy for clinical monitoring; Graham Cheetham and Stephanie Hamilton for clinical operations; Hemanshi Chawla and Landi Luo for data analysis; Ting Ma and Hannah Persampieri for data engineering; Sophia Jung and Harit Nandani for data management; Sam Thulburn for laboratory and sample management; Jackie Brooks for Tidy Data; and Sara Hiom for External Affairs (all formerly at GRAIL, Inc.). We acknowledge the contributions of James Dai, Rita Lopatin, and Tong Meng for biostatistics; Velicia Bachtiar, Sarah Day, Rachael Neale, Kaveh Riahi, and Susan Wan for clinical operations; Cavan Donohoe, Dongjing Guo, Mandy Ho, Cindy (Xinyi) Hou, Victoria Le, Evgueni Neiman, Sarah Yam, and Amber Yeh for data analysis; Zhao Dong, Terri Shen, and Katie Truong for data engineering; Shelby Dixon for data management; Laura Farrell for laboratory and sample management; Laura King for media relations; Jeffrey Diament for sample selection; and Tuan Ahn Le, Jacob Metcalf, Paddy Steed, and Corina Tamas for software support (all from GRAIL, Inc.). We acknowledge the contributions of Martine Bomb, Charlotte Eversfield, Sean McPhail, and Brian Rous for NCRAS data (all from NDRS); and Seema Chauhan, Gemma Edney, Fiona Jones, Amy Johnston, Louise Southan, and the EMS clinical unit leads and clinical staff from EMS Healthcare. From Queen Mary University of London, we acknowledge Sandra Beare, Harriet Gilbert-Jones, Sheetal Karavadra, Ian Lowenhoff, Simranjit Mehta, Cherry Paice, Michelle Sleeth, and Benjamin Wan for trial management; Jennifer Ching, Zoi Katsirma, and Roger Newson for statistics and data management; Goldy Bampoe-Parry, Chris Deeley, Emily Jobson, and Anjeli Ketkar for participant support; Rima Begum, Anne Marie Borges Da Silva, Kate Dear, and Genee Tolentino for monitoring; and James Aroloye for clinical team support. We also acknowledge members of the clinical team at Guy’s and St Thomas’ NHS Foundation Trust, including Serena Dervishi, Jisha Mathew, Rebecca Smith, Jennifer Squires, and Simone Tal. Libby Ellis, Gabriel Gasque and Rachel McMullan (all from GRAIL, Inc.) provided manuscript development support. Medical writing support was provided by Emma B. Saxon (Royston, UK), Alexandra L. Thomas, PhD, from Citrus Health Group, Inc. (Chicago, IL), and Ichha Khanal, PhD (West Sussex, UK), and funded by GRAIL, Inc. Graphic design support was provided by Kristi Whitefield from PosterDocs (Oakland, CA) and copy editing support was provided by Erin Sophr from Engage Labs, LLC (Oak Ridge, NJ), both funded by GRAIL, Inc.
Funding Support
The study was funded by GRAIL, Inc. The funder participated in the study design, data curation, data interpretation, and drafting and approval of the publication. Data analysis was conducted independently by QMUL and verified by GRAIL.
References
- 1.Cancer survival in England, cancers diagnosed 2018 to 2022, followed up to 2023. NHS England Digital; 2026. ( https://digital.nhs.uk/data-and-information/publications/statistical/cancer-survival-in-england/cancers-diagnosed-2018-to-2022-followed-up-to-2023) [Google Scholar]
- 2.Routes to diagnosis. NHS England Digital - National Disease Registration Service; 2024. https://nhsd-ndrs.shinyapps.io/routes_to_diagnosis/ [Google Scholar]
- 3.Only 14% of cancers are detected through a preventive screening test. National Opinion Research Center; 2022. https://www.norc.org/content/dam/norc-org/pdfs/State-Specific%20PCDSs%20chart%201213.pdf . [Google Scholar]
- 4.Clarke CA, Mitchell BL, Putcha G, et al. Lexicon for blood-based early detection and screening: BLOODPAC consensus document. Clin Transl Sci. 2024;17(9):e70016. doi: 10.1111/cts.70016. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Brito-Rocha T, Constâncio V, Henrique R, Jerónimo C. Shifting the cancer screening paradigm: the rising potential of blood-based multi-cancer early detection tests. Cells. 2023;12(6):935. doi: 10.3390/cells12060935. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Neal RD, Johnson P, Clarke CA, et al. Cell-free DNA–based multi-cancer early detection test in an asymptomatic screening population (NHS-Galleri): design of a pragmatic, prospective randomised controlled trial. Cancers. 2022;14(19):4818. doi: 10.3390/cancers14194818. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Klein EA, Richards D, Cohn A, et al. Clinical validation of a targeted methylation-based multi-cancer early detection test using an independent validation set. Ann Oncol. 2021;32(9):1167–77. doi: 10.1016/j.annonc.2021.05.806. [DOI] [PubMed] [Google Scholar]
- 8.Schrag D, Beer TM, McDonnell CH, et al. Blood-based tests for multicancer early detection (PATHFINDER): a prospective cohort study. The Lancet. 2023;402(10409):1251–60. doi: 10.1016/S0140-6736(23)01700-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Nabavizadeh N, McDonnell C, Kurbegov D, et al. Safety and performance of a multi-cancer early detection (MCED) test in an intended-use population: initial results from the registrational PATHFINDER II study. Ann Oncol. 2025;36:S1605 [Google Scholar]
- 10.Matrana M, Shukla V, Kingsbury D, et al. Real-world data and clinical experience from over 100,000 multi-cancer early detection tests. Nat Commun. 2025;16(1):9625. doi: 10.1038/s41467-025-64094-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Hurt RT, Ghosh AK, Dougan BM, et al. Implementation of a multicancer detection (MCD) test in a tertiary referral center in asymptomatic patients: an 18-month prospective cohort study. J Prim Care Community Health. 2025;16:21501319251329290. doi: 10.1177/21501319251329290. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.O’Donnell E, Kauffman T, Beckwith J, et al. Implementation and evaluation of multi-cancer early detection testing at the Dana-Farber Cancer Institute: A retrospective analysis of clinical outcomes and diagnostic pathways. J Clin Oncol. 2025;43(16_suppl):11159–11159. [Google Scholar]
- 13.Swanton C, Bachtiar V, Mathews C, et al. NHS-Galleri trial: enriched enrolment approaches and sociodemographic characteristics of enrolled participants. Clin Trials. 2025;22(2):227–38. doi: 10.1177/17407745241302477. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Lowenhoff I, Dolly S, Dowinton Smith R, et al. Clinical referral to the NHS following multi-cancer early detection test results from the NHS-Galleri trial. Front Oncol. 2025;15:1511816. doi: 10.3389/fonc.2025.1511816. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Zhang J, Braun J, Simon N, Hubbell E, Zhang N. Stage shift and mortality reduction by screening with a multi-cancer early-detection test: microsimulation modeling of a randomized controlled trial. JSM 2021 Online Program. 2021 https://ww2.amstat.org/meetings/jsm/2021/onlineprogram/AbstractDetails.cfm?abstractid=319050 . [Google Scholar]
- 16.Marlow LAV, Schmeising-Barnes N, Warwick J, Waller J. Psychological Impact of the Galleri test (sIG(n)al): protocol for a longitudinal evaluation of the psychological impact of receiving a cancer signal in the NHS-Galleri trial. BMJ Open. 2023;13(7):e072657. doi: 10.1136/bmjopen-2023-072657. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.de Koning HJ, van der Aalst CM, de Jong PA, et al. Reduced lung-cancer mortality with volume CT screening in a randomized trial. N Engl J Med. 2020;382(6):503–13. doi: 10.1056/NEJMoa1911793. [DOI] [PubMed] [Google Scholar]
- 18.Bretthauer M, Løberg M, Wieszczy P, et al. Effect of colonoscopy screening on risks of colorectal cancer and related death. N Engl J Med. 2022;387(17):1547–56. doi: 10.1056/NEJMoa2208375. [DOI] [PubMed] [Google Scholar]
- 19.Schoen RE, Pinsky PF, Weissfeld JL, et al. Colorectal-cancer incidence and mortality with screening flexible sigmoidoscopy. N Engl J Med. 2012;366(25):2345–57. doi: 10.1056/NEJMoa1114635. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Wooldrage K, Robbins E, Duffy S, Cross A. Long-term effects of once-only flexible sigmoidoscopy screening on colorectal cancer incidence and mortality: 21-year follow-up of the UK Flexible Sigmoidoscopy Screening randomised controlled trial. Lancet Gastroenterol Hepatol. 2024;9(9):811–24. doi: 10.1016/S2468-1253(24)00190-0. [DOI] [PubMed] [Google Scholar]
- 21.Feng X, Zahed H, Onwuka J, et al. Cancer stage compared with mortality as end points in randomized clinical trials of cancer screening: a systematic review and meta-analysis. JAMA. 2024;331(22):1910–7. doi: 10.1001/jama.2024.5814. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Webb AB, Berg CD, Castle PE, et al. Considerations for using potential surrogate endpoints in cancer screening trials. Lancet Oncol. 2024;25(5):e183–92. doi: 10.1016/S1470-2045(24)00015-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Rebolj M, Brentnall AR, Geppert J, et al. Late-stage outcomes as surrogates for mortality in cancer screening trials: a systematic review and meta-analysis. Cancer Epidemiol Biomarkers Prev. 2025;34(10):1694–709. doi: 10.1158/1055-9965.EPI-25-0201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Dai JY, Zhang J, Braun JV, Simon N, Hubbell E, Zhang N. Clinical performance and utility: a microsimulation model to inform the design of screening trials for a multi-cancer early detection test. J Med Screen. 2024;31(3):140–9. doi: 10.1177/09691413241228041. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Swanton C, Bryce A, Cohn A, et al. Prognostic significance of blood-based multicancer detection in circulating tumor DNA: five-year outcomes analysis. JCO Precis Oncol. 2026;10(3):e2500725. doi: 10.1200/PO-25-00725. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Heitkamp S, Olsen EB, Tafazzoli A, et al. Association between emergency department (ED) involvement in cancer diagnosis and survival across cancer types in the Medicare population. San Diego, CA: 2026. [submitted for publication] [Google Scholar]
- 27.Neal RD, Dolly S, Johnson P, et al. Performance of a multi-cancer early detection test across three annual screening rounds in the randomised controlled NHS-Galleri trial. Nat Med. 2026 doi: 10.1038/s41591-026-04652-8. [under review for publication] [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.