GRAIL Multi-Cancer Early Detection (MCED) — Phiên báo cáo ASCO 2026ASCO 2026 Session

"Let me start with the status quo of cancer screening in the United States. Today, the USPSTF recommends screening for only five cancers — cervical, colorectal, breast, lung, and prostate. And yet more than 600,000 Americans die of cancer every single year. The sobering reality, shown here, is that 87% of those cancer deaths come from cancers for which we have no recommended screening at all. So no matter how well we perform with the tools we have, we are only addressing a small fraction of the problem. That is precisely the gap we set out to close."

"This is where Galleri comes in. Galleri is a single blood test, built around very high specificity — that is, a very low false-positive rate. And it does two things that matter a great deal clinically. First, it predicts the cancer signal origin, the CSO — it tells you where in the body to go look. Second, it tends to preferentially detect the more aggressive cancers, which are exactly the ones we most want to catch early. It has been validated in our CCGA study and tested in the real-world PATHFINDER implementation study."

"I want to give you a sense of the scale of evidence behind this. GRAIL has assembled one of the largest evidence programs in all of genomic medicine — nine studies, more than 385,000 participants. You can see them here: CCGA, PATHFINDER, NHS-Galleri, PATHFINDER 2, REACH, REFLECTION, STRIVE, SUMMIT, and SYMPLIFY. Together they span the full arc — from developing the test, to clinical implementation, to population-level screening, and ultimately to clinical utility. This is not a single study; it is a coherent, deliberate program."

"If you map those studies onto the classic phases of biomarker development, you can see just how comprehensive the program is. The preclinical discovery work was CCGA1. Clinical assay development was CCGA2. The retrospective case-control validation was CCGA3. The prospective screening studies are PATHFINDER 1 and 2. And at the very top of that pyramid sits NHS-Galleri — the first and only randomized controlled trial of clinical utility for an MCED test anywhere in the world. Harpal will tell you a great deal more about that one shortly."

"Let me now turn to PATHFINDER 2 itself, which is the largest interventional MCED study conducted in North America to date. We enrolled 35,878 participants aged 50 and older, across 32 sites in North America, with up to three years of follow-up. Anyone with a cancer signal detected goes on to a CSO-guided diagnostic work-up. The data I'm showing you today reflect about 12 months of follow-up; we have planned interim analyses at 24 months and a final analysis at 36 months."

"Before I show you the results, a word of caution about our population — because it matters for how you interpret everything that follows. When we compare our PATHFINDER 2 participants with the NHANES intended-use population in the US, they are similar in age, around 64. But our participants were more highly educated — nearly 59% had a bachelor's degree or more, versus about 29% — and they smoked far less, under 4% versus nearly 14%. In other words, this was a relatively healthier cohort than the general population, and we should keep that firmly in mind."

"Now to the impact. Of the 440 participants diagnosed with cancer in the study, 264 — about 60% — were detected through screening. And here is the key number: when you add the MCED test on top of the recommended USPSTF A/B screening, you find roughly six and a half times more screen-detected cancers. Even if you include prostate screening — USPSTF category C — you still find about three times more. So the test is not replacing existing screening; it is dramatically expanding what we are able to catch."

"Performance was robust, and consistently so. In PATHFINDER 2 we saw a positive predictive value of 60.3%, a negative predictive value of 99.2%, episode sensitivity of 39.3%, and specificity of 99.6%. To put that PPV in context, it is roughly three- to twenty-fold higher than the single-cancer USPSTF A/B screening tests we rely on today. The message I want you to take away is simple: Galleri is meant to be used in addition to recommended screening — not instead of it."

"There's one subtlety I want to make sure is clear, because it often causes confusion. There is a difference between 'episode sensitivity' and 'test sensitivity.' Episode sensitivity — which is what we report in PATHFINDER 2, an interventional study where cancer status is unknown at the time of the test — counts the cancers we detect at the moment of screening out of all the cancers diagnosed over a pre-defined follow-up period. That is different from the 'test sensitivity' you get from a case-control study like CCGA3, where you already know who has cancer. Episode sensitivity is the more demanding, real-world number, and it can only be measured in an interventional intended-use screening study like this one."

"For the 12 cancers responsible for two-thirds of US cancer deaths, sensitivity is 69%. We can also look at the 6 aggressive cancers with the lowest 5-year survival. We won't go through all of these, but you can see that, for example, for cancers exclusive of breast and prostate cancer, we have 54.9% sensitivity. One reason we think about it this way is that, as mentioned earlier, many cancers are detected through conventional screening for breast and prostate. So we don't think of this test as a full screening solution, but really as part of a program. And when we think of these together, the aggregate sensitivity across all these cancer types continues to increase."

"One of the most important questions asked over time is: what is the safety profile of MCED testing? Are you going to induce a lot of unnecessary work-ups? Introduce harm? Induce anxiety? What we can say is that over the 35,000 individuals who participated in PATHFINDER 2, less than 1% had an invasive procedure to evaluate a positive MCED test. The majority of diagnostic evaluations are non-invasive — blood tests, imaging, and so forth. And as you might suspect, invasive procedures are much more common in individuals with cancer. For many work-ups, the first test is imaging or a blood test, followed — if there is a suspicious finding — by biopsy."

"The cancer signal origin (CSO) can guide an efficient diagnostic work-up following a positive MCED test. For all individuals with a positive result, the median time to diagnostic resolution — when the physician felt the work-up was done — was 48 days. We can parse that by who had cancer: time to resolution was faster for those who did have cancer than for those who did not. Importantly, at the time of this study we did not yet have a 'retest' program. A colleague's presentation last year showed that for individuals with a negative work-up who get a second Galleri test: if it converts to no-signal-detected, there is almost a 0% chance of cancer within that year; if the signal persists, there is about a 30% chance of cancer. The CSO helped guide diagnostic evaluation in about 88% of cases. People ask: why didn't everyone simply follow the CSO-based pathway? Because medicine is messy — for example, a woman with prior breast cancer and mastectomy would not get a mammogram for a breast signal; it might be time for whole-body imaging to look for distant metastasis. Our clinical collaborators authored a white paper to help physicians with CSO-guided recommendations. And of course, this is only valuable with high CSO accuracy — which is over 90%."

"Another concern: will screening induce a lot of anxiety? We have experience from other screening tests — anxiety has been measured in, for example, mammography and low-dose CT screening for lung cancer. Those have shown a transient increase in anxiety that returns to baseline — which is exactly what you see here with MCED testing. So MCED testing is really no different from other cancer-screening approaches."

"In conclusion, we had a very high positive predictive value — driven by our ability to detect a shared cancer signal — at average aggregate prevalence, a very low false-positive rate of ~0.36%, sensitivity of 39%, and sensitivity for the aggressive cancers of over 69%. Most of the cancers — 71% — were stages I through III. On the safety side, a very low proportion of individuals underwent an invasive procedure, and anxiety was quite low. With that, I'll close on PATHFINDER 2 and turn it over to Sir Harpal Kumar to describe NHS-Galleri. Thank you."

"Thank you very much, Eric. I'm going to talk about NHS-Galleri. This is the only randomized trial of an MCED carried out anywhere. As you'll see, it's a very large trial, carried out in England, and I'll show you the results we presented yesterday."

"The study schema is shown here. This is a randomized trial — an enormous one. We enrolled over 142,000 people extremely rapidly, in just over 10 months. People were invited to provide a blood sample; once they had, they were randomized 1:1 to the control arm or the intervention arm. It was double-blinded — neither participants nor their physicians knew which arm they were in — and people only became unblinded if they had a positive Galleri test. At that point they were referred into the standard NHS referral pathway for investigation of the signal and, if a diagnosis was confirmed, treatment — all within the NHS. Those in the control arm, or with a negative test, took part in all three screening rounds, and data was collected through standard, quality-controlled NHS datasets. Participants were 50 to 77 at enrollment — so by their third screening round they'd be at most 80 — with no cancer diagnosis or treatment in at least the last 3 years, no current clinical suspicion of cancer, and living in one of the enrolling NHS Cancer Alliance regions."

"We took three blood draws spaced one year apart. It's important to note this is a trial with a prevalent screening round — the first round — where you sweep up the cancers already prevalent in the population (which turns out to be a large number of undiagnosed cancers). Once past that prevalent round, you hope to see a reduction in late-stage cancers as the screening effect takes hold — these are the incident screening rounds, showing what a steady-state annual program might look like. I'll show results with data collected 12 months after the third-round blood draw; we are continuing follow-up. When this trial was designed, there was an NHS policy objective to increase cancers diagnosed at stages I and II — an ambitious target of 75% diagnosed early by 2028. The corollary is reducing those diagnosed at stages III and IV. So the primary objective was set as reducing stage III and IV diagnoses, in aggregate across the three rounds. But we were clear this is only one of many ways to look at clinical utility — so we also pre-specified a number of secondary objectives."

"We enrolled this population very rapidly — over 142,000 people. Randomization worked very well and the two arms were very well balanced. What I'll point out first is that we had extremely high retention: around 90% came back for their second and third blood draws, despite being people with no symptoms and no reason for clinical suspicion. On demographics: a very good balance of men and women, about 50:50. The racial and ethnic mix is representative of the England population (different from the US). We were determined to enroll people from the most socioeconomically deprived communities — those who unfortunately have the largest cancer burden and worst outcomes — and we actually over-enrolled from the most deprived communities, and enrolled a large number of people with very low education levels, who again carry the highest burden."

"This is the summary of the primary objective — the difference in stage III/IV cancers diagnosed in the intervention versus control arm. Focus first on the left-hand side: there was no difference between the two arms (p-value 0.6); if anything, a marginal increase in stage III/IV cancers in the intervention arm. But to focus only on that is to miss an enormous amount of important information. The first part is to look at what happened round by round. As expected, in the first round you sweep up a lot of already-prevalent cancer — and that's exactly what happened: a large increase in stage III/IV cancers (already prevalent but undiagnosed). As we move to the incident rounds, we do start to see a reduction in stage III/IV, reaching 12% by the third round. Cumulatively, those reductions weren't enough to overcome the big first-round increase — which is why, in aggregate across three rounds, there was no net decrease. But you can see the decrease emerging in the incident rounds."

"Another way to look at this: that combined endpoint pooled stage III and IV together. What if we separate them? You see two distinct, opposing patterns — a large increase in stage III cancers, but a large decrease in stage IV cancers. These two effects counterbalanced each other in the combined endpoint. So why the large excess of stage III cancers in the intervention arm? Three effects are at play: (1) as described, a large amount of undiagnosed late-stage cancer — including stage III — was already prevalent and got swept up in the first round; (2) we are indeed stage-shifting — moving some cancers that would have progressed to stage IV to be diagnosed at stage III or earlier, increasing the stage III count; and (3) one you can't see on this chart — overall, across the three rounds, there was simply more cancer found in the intervention arm. We'd expect the impact of that to attenuate, if not reverse, with further follow-up. But as I say, we do see the decrease in stage IV."

"Looking in detail now at the stage IV cancers, starting with this pre-specified group of 12 cancers: overall we did see a significant reduction in stage IV diagnoses, with the confidence interval not straddling one. And you can see the impact of the incident rounds — the reduction exceeded 22% in both incident rounds and reached over 26% by the third screening round, again with the confidence interval not straddling one."

"This shows the picture of stage III and IV cancers graphically. The chart on the left shows the cumulative incidence of stage III/IV cancers from the point of randomization for each individual. The yellow line — the control arm — shows a steady, linear rise, as you'd expect. The purple line — the intervention arm — shows that within the first couple of months after the blood draw there is a large increase in stage III/IV cancers, representing those detected by the MCED test. The lines then start to converge — until the next screening round produces another increase. It's really as we get toward the end of the trial, with further follow-up, that the lines converge. The importance of that is: overall there was an excess of cancer in the intervention arm; as we continue follow-up, we expect these lines to cross, and a net reduction in stage III/IV cancers overall."

"Now the picture for stage IV diagnoses is somewhat different. Here we start to see a net reduction in stage IV cancers from within the first screening round — because we're finding cancers that would otherwise have progressed to stage IV at earlier stages, very early in the study. As you can see, the curves are still separating by the end of the trial. And on the right are the numbers I showed you earlier — the reductions in stage IV in the subsequent screening rounds."

"This was the picture for the 12 pre-specified cancers. What was also very exciting is that we saw an almost identical pattern across all routinely staged cancers: in the incident screening rounds, reductions in stage IV cancers in excess of 20%. When you think about the impact at population scale, we are talking about all the people for whom we're finding cancers before they have progressed to a metastatic stage — and again, the confidence intervals by the third screening round do not straddle one."

"This shows examples of the variation in stage IV diagnoses across those 12 cancer types. The only two of the 12 not shown are lymphoma and plasma cell neoplasm, because the counts were just 0 or 1 — not enough to show. This is a descriptive analysis and some counts are small. With some cancer types we're seeing really substantial reductions in stage IV diagnoses. The importance is that, with the advances being made in treatment, we now have many effective treatments for cancers diagnosed at stage III, such that 5-year survival can be quite promising. For example, for colorectal cancer we saw a 34% reduction in stage IV diagnoses; in England, 5-year survival for stage IV colorectal cancer is ~11%, versus ~64% for stage III — so the potential impact on survival is dramatic (and not very different in the US). Not shown on this chart, but a very interesting example: we also saw a 25% reduction in stage IV prostate cancer — and the difference in 5-year survival between stage III and stage IV prostate cancer is almost double, so reducing stage IV prostate diagnoses is again very meaningful."

"So that's late-stage reduction. What about stages I and II diagnoses? Shown here for the 12 cancer types: overall we saw a 16% increase in cancers diagnosed at stages I and II — and these are often cancers diagnosed very late. That 16% increase is predominantly found in colorectal, upper GI, hematologic, and ovarian cancers. For example, we saw a 100% increase in ovarian cancers found at stages I and II, and a 22% increase in pancreatic cancers caught at stages I and II, and so on."

"Test performance is very consistent with what we saw in PATHFINDER 2. Across the three screening rounds we saw a PPV of 52%, but in the first screening round it was 58% — and in the report Eric discussed it was 60% in PATHFINDER 2. PATHFINDER 2 also includes cancer recurrences, which aren't included here; if we include recurrences, this is also 60% — almost identical, despite these being two very differently conducted studies in different populations. So test performance is very consistent across our studies: NPV very close to 99%, specificity again extremely high at 99.55% (a false-positive rate of 0.45%), and a very high CSO accuracy of 92.5%. Focusing on the 12 pre-specified cancer types, episode sensitivity in the first round was again ~60%, very similar to what we saw before."

"Now, what about the false positives? I'm asked this very frequently: okay, it's a very low false-positive rate — but what actually happens to those false positives? This chart shows an analysis we were able to do on the false positives declared in the first screen. I'm showing these because we have two further years of follow-up. Of the ~300 'false positives' in that first round, 54 — close to 18% — were subsequently diagnosed with cancer within the next two years. And of those, over 80% had a cancer consistent with their CSO prediction in that first screening round — suggesting the cancer was there at the time of that first screen but wasn't diagnosed, for a multitude of reasons. This mirrors what we saw in another study, SYMPLIFY, where with two years of further follow-up, over one-third of people initially called false-positive did have a cancer diagnosis within the next two years — again, the vast majority consistent with the original CSO prediction."

"With respect to safety — as expected, and as just shown with PATHFINDER 2 — the test was very safe. The vast majority of adverse events were phlebotomy-related, things like bruising. In the intervention arm, as you'd expect, there was some increase in anxiety; though not yet published, the anxiety experience is similar to PATHFINDER 2 — a temporary increase that later returns to baseline, as seen in all other screening programs. No serious study-related adverse events occurred from blood sample collection or return of test results."

"What about the overall cancer-detection rate? This is analogous to the slides Eric showed. Overall, we detected four times as many cancers when adding the MCED test to existing standard-of-care screening. In the control arm, there were 290 cancers detected through all existing NHS screening programs; in the intervention arm, over 1,100 cancers detected through screening — and you can see the breakdown by stage. People like to tell me 'ctDNA only detects late-stage cancers' — but you can see that's not the case: 40% of cancers were detected at stages I–II, and just under 70% at stages I–III. Indeed, at stages I and II alone we detected 366 cancers — more than ALL the cancers detected across every NHS screening program regardless of stage. We really are finding a substantial number of cancers at those very early stages."

"The fact that so many more cancers were found through screening means we then had 21% fewer cancers that needed to be detected clinically. And another exciting finding: a 25% reduction in cancers diagnosed as a result of an emergency presentation. As you all know, individuals who present as an emergency do, stage for stage, a lot worse than those diagnosed through other routes."

"This is an exploratory analysis, but I think it's provocative and intriguing. Because we saw such a large increase in stage III cancers, we looked in more detail — we had large numbers of stage III colorectal and lung cancers. We noticed a shift WITHIN stage as we went from the first screening round to the second and third: a shift from stage IIIC toward IIIB and then IIIA. This is profoundly important for these cancer types — look at the survival rates for those substages at the bottom. For colorectal, a stage IIIC cancer has a 5-year survival of 55%, but a stage IIIA cancer has 89%. So the fact that we're seeing this within-stage shift is also very intriguing. This is an exploratory endpoint and the counts are not large; the reason you don't see the exact n's is an agreement with the NHS about minimum numbers below which we can't show specifics — but you can see the performance."

"I've touched on this, but as we think about what we've seen — this dramatic reduction in stage IV cancers — it turns out to be very clinically important. On the left you can see survival rates across all cancer types; I'll focus you on the 5-year survival, where there's a significant 'cliff' in survival between stage III and stage IV cancers. Of course it varies by cancer type, as I've shown. But overall, if we can shift from metastatic disease to cancers diagnosed at an earlier stage, in many cases we have the opportunity for curative-intent treatment — opportunities we unfortunately no longer have for those diagnosed at stage IV."

"In summary, from NHS-Galleri: we did not meet the primary endpoint of a combined stage III/IV reduction. But across many pre-specified secondary endpoints, we observed a greater than 20% reduction in stage IV cancers in the incident rounds — both for the 12 pre-specified cancers and for all routinely staged cancers — and a substantial increase in cancers detected at stages I and II. The test was safe, with a very high PPV and very high specificity, consistent with our previous studies. Overall we found four times as many cancers through screening as the NHS screening programs, and a 25% reduction in emergency-presentation diagnoses."

"One thing I'll conclude on: when we think about these PPVs and the importance of a high PPV — as Eric pointed out, they're an order of magnitude higher than what we see with single-cancer screening programs today. And encouragingly, we're seeing these results replicated in real-world implementation. On the left is now-published data from some of our partners — the Mayo Clinic, Dana-Farber, and Century City. The levels we're seeing in the trial are indeed reflected there. The reason some are even higher is that, with these providers' confidence in the test performance, they work patients up more thoroughly when they present with a positive test."

"To summarize across the two studies we've talked through today: a clinical program of almost 180,000 people across two interventional studies on two different continents, with results that are consistent across both. We see clinical utility — greater than 20% reductions in stage IV cancers, increases in stage I and II cancers, fewer cancers diagnosed through emergency presentation, and substantially more cancers detected through screening overall — between 4 and 6.5 times as many, including many cancers we don't currently screen for. High PPV levels, low false-positive rates, and a strong overall safety profile for this kind of technology. With that, I'll stop there."

"I'm a surgical oncologist. I spent 35 years at the Cleveland Clinic treating patients with urologic cancers before I joined GRAIL — so this is my second career. I want to emphasize some important points about screening from a clinical perspective. The value of screening depends on our ability to treat and cure cancers when they're detected — and the treatment landscape is not static. When I started my career in 1989, 5-year overall survival for all cancers combined was about 50%. In the most recent Annual Report to the Nation on cancer statistics, for the first time, 5-year survival in aggregate for all cancers is 70% — a 40% increase in 35 years. We saw today at the plenary many randomized trials that will change the standard of care — one you're probably familiar with is the new RAS inhibitor that improves survival in patients with metastatic pancreatic cancer in the first-line setting, and that's now available. [There is also] a big increase in early-onset cancer — something I think we're going to have to look at more closely in the future."

"The most important point is that there's no doubt test sensitivity improves with stage — sensitivity is higher for late-stage than early-stage cancers. What we're starting to see is that for some of the more aggressive cancers, they really are progressing very rapidly — going from undetectable to late-stage cancer within a 12-month period. So there will be some cancer types that are completely undetectable and then become a stage III or even stage IV cancer within 12 months. That's some of the biology we're starting to uncover. And from a population perspective: in the NHS-Galleri data, there were more stage I and II cancers detected by MCED than the total number of cancers across all stages detected by standard-of-care screening."

"It's a really good question. To my knowledge, this is the first time this endpoint (combined stage III/IV reduction) has been used in a screening study; I explained the reasons for that choice earlier. We designed this study seven or eight years ago. As Dr. Klein just said, the treatment landscape is evolving all the time — that's the biggest single change of the last 10 years. There's much more we can do now. So I do think endpoints will evolve — and I think they should evolve. That would be my perspective."