October 17, 2025

expert reaction to conference abstract on the PATHFINDER II trial into the safety and performance of the Galleri test in people without cancer symptoms

A conference abstract presented at the European Society for Medical Oncology (ESMO) 2025 meeting looks at the safety and performance of the Galleri test. 

Professor Clare Turnbull, Professor of Translational Cancer Genetics at The Institute of Cancer Research, London, said:

“Improvement in positive predictive value is positive to note; screen-false positives are not too problematic to deal with for site-specific screening. Positive Multi-Cancer Detection (MCD) results for which no cancer can be detected are a big problem.

“It will be useful to see more detail on these results. Were haematological cancers included? Inclusion of these distort the results unhelpfully as haematological malignancy is readily detectable on a full blood count.

“What was the sensitivity for early-stage cancers (stage 1-2), namely those cancers for which prompt surgical (or radiotherapy) intervention may provide cure? The number of early-stage cancers detected and the overall sensitivity have been highlighted, but it is important to know the sensitivity for early-stage cancers. Detection of late-stage cancers is not the goal for designing new screening programmes. There are little data to indicate that finding a stage 4 cancer earlier will alter its outcome: largely this just confers lead-time meaning that the patient knows about the cancer for a longer time.

“Follow-up of the PATHFINDER 2 data will be essential to examine outcomes for detected cancers that appeared to be early stage. 5-year cancer-specific mortality even for stage 1-2 disease following surgery is high for many of these abdominal cancers (pancreatic, hepatobiliary, oesophageal, gastric, (i.e. a sizeable proportion of cases still die from their cancer), demonstrating that our staging for these cancer types is somewhat imperfect. Previous data have demonstrated that Galleri-positive cancers have higher mortality stage-to-stage than Galleri-negative cancers. Is Galleri just more sensitive than conventional staging at identifying cancers that have already metastasized and will do poorly regardless (even if seemingly early-stage on imaging at detection)?

“Data from randomised studies, with mortality as an endpoint, will therefore be absolutely essential to establish whether seemingly earlier-stage detection by Galleri translates into benefits in mortality. The results of the NHS Galleri study will be invaluable for addressing this question, providing a unique opportunity for randomized comparison with evaluation of all-cancer-specific mortality as the end point.”

Professor Nitzan Rosenfeld, Director of the Barts Cancer Institute, Queen Mary University of London (QMUL), said:

“The updated headline results for the Pathfinder II study, as reported in summary in abstract form, continue to build the evidence base for the potential value of Multi-Cancer Early Detection (MCED) liquid biopsy testing, for blood-based screening for multiple cancer types. The impressive results reported today, for a population of 35,878 individuals ages 50+, are consistent with results previously reported for the Pathfinder study for 6,621 participants [link from Grail.com website: https://view-su2.highspot.com/viewer/64540ec70e89bee795cab05e]. The specificity in the Pathfinder II study continues to be very high at >99.5% and the positive predictive value (PPV), the likelihood that a person has cancer when a positive test result is returned, has now been updated to >60% which is very encouraging and provides strong evidence that this test could be safe and informative. Importantly, >50% of the cancers detected by the Galleri test in this study were early stage (stages I-II), and >75% of them do not have common screening options. These results together, provide important evidence in support of the utility of the test in screening of certain populations, bearing in mind limitations of the study notably the characteristics of the population that has been recruited and has taken part and the length of follow up and testing intervals. Outcome data which will take longer to accumulate, including cancer mortality and the analysis of false-positive and false-negative results, will help confirm the findings and establish other important performance parameters of the test, the populations and the frequency where this test could be effectively deployed.”

Prof Anna Schuh, Professor of Molecular Diagnostics, and Honorary Consultant Haematologist, University of Oxford, said:

“Pathfinder 2 is an important study and follows an appropriate design in a challenging indication, i.e.: testing for cancer in asymptomatic people. The sample size calculation was probably based on either better expected performance of the assay or an expected higher cancer incidence in this cohort although it is usually estimated at about 1% in asymptomatic people as is the case here. I therefore suspect that the investigators expected their test to perform to a higher level.

“The results for about a third of enrolled participants were not analysable at this time. This requires an explanation. What is the turnaround time for this test? The follow-up of 12 months in the trial is short, but it should be possible to analyse samples as they come through the door. Longer follow-up data will increase the number of analysable samples and might somewhat improve the overall performance of the assay; however, we must wait for another 2 years to know for sure.

“As results are at present: in this (unenriched) cohort, the positive predictive value (PPV) is 55% to 68%. This means that when used in this specific cohort, the likelihood that an individual with a positive test result actually has cancer is only about 60%. Or in other words: almost half of the time, the test gets it wrong when it calls a positive result. This is disappointing as it is only fractionally better compared to tossing a coin, although better compared to current screening tests where still most positive results turn out to be nothing.

“133 cancers were detected by the test, so only roughly about a third of what the test should be expected to detect. To achieve this, over 23,000 people had to be screened. This would hardly be acceptable in a cost-utility analysis for the NHS for example, where we would be looking to detect 300 cancers from 10,000 screens, based on the current test price of USD 996 and the rule of thumb that introducing a multi-cancer early detection (MCED) intervention should not cost the NHS more than about £30,000 per additional cancer case detected/quality of life adjusted (Qaly) years gained; i.e. roughly the threshold applied by the NHS to new cancer drugs. Currently, assuming the above price tag of USD 996 per test, this means an estimated cost of USD 174,000 per additional cancer diagnosed.

“It is encouraging to see that almost 50% of the cases detected were early stage (1 or 2). In line with what is expected from a test that uses only a single modality (i.e. cancer specific methylation changes), this implies that sensitivity is good for some of the common cancers (where it was 74%), but not so much for the others (that together make up more than half of all cancers) as for these the clinical sensitivity is poor (40%).

“Interestingly, the abstract does not mention the absolute number of false positives directly. The specificity of 99.6% is better compared to what was reported previously and testifies that the investigators trained the model towards avoiding false positives, but I calculate that there are still about 100 odd false positives. I think this is acceptable, but we must ensure that follow-up for these people (not patients) is put in place.

“The conclusions are surprising. It is not the size of the study I am worried about (although confidence intervals are big and hence, I quote the lower numbers); it is the low pick-up rate in this specific cohort combined with the cost of the assay that in my view makes this approach currently unsuitable for population screening based on this current data, though as I mention earlier, we must wait for another 2 years for follow up data to know for sure. In other words: based on this data, does this test in this specific unselected cohort (i.e. not enriched for cancer risk) represent good value for tax payers’ money if it picks up only a third of the people with cancers in this cohort and of these a quarter are cancers that could be picked up with conventional screening?”

The conference abstract ‘Safety and performance of a multi-cancer early detection (MCED) test in an intended-use population: Initial results from the registrational PATHFINDER II study’ by K Giridhar et al. was presented at European Society for Medical Oncology (ESMO) 2025 meeting. The embargo on the abstract lifted at 23:05 UK time on Friday 17 October 2025. 

Declared interests

Prof Anna Schuh: I have a few shares worth nothing in GRAIL from a previous start-up sale. I am a co-inventor of TriOx, an algorithm to detect cancer signals from whole genome sequencing. We have not raised any funding to support commercial exploitation and are seeking research funding to test its diagnostic accuracy in patients with non-specific symptoms of cancer (so different from pathfinder).

For all other experts, no reply to our request for DOIs was received.