Milestone Monday: A Win in China. A Shift in England. A Gap Nobody Named.
Three ASCO readouts, three announcements that claimed more than the data supported — and one question the conference floor has been too busy to ask.
The data from the first few days at ASCO is real. The conclusions being drawn from it are doing considerably more work than the evidence supports. Ivonescimab beat a PD-1 inhibitor on OS in China. Grail showed stage shift across 140,000 NHS patients. MRD posted its strongest conference presence to date. Each of those results is worth taking seriously.
Each of the narratives built on top of them, about global commercial destiny, proven screening utility, and a solved sensitivity problem, is running well ahead of what was actually demonstrated in Chicago. That distance is the story this week.
China Proved It. The World Is Next, If Summit Can Get There.
Akeso's ivonescimab beat tislelizumab plus chemotherapy on overall survival in the Phase III HARMONi-6 trial in first-line squamous NSCLC, cutting the risk of death by 34% and extending median OS by 4.2 months to 27.9 months. It is the first OS superiority result for a PD-(L)1 x VEGF bispecific over a PD-(L)1 inhibitor in a head-to-head trial.
WHAT HAPPENED
HARMONi-6 enrolled 532 patients in China and compared ivonescimab plus platinum-doublet chemotherapy against tislelizumab plus the same regimen. The OS hazard ratio of 0.66 cleared the prespecified boundary and, by Bernstein's framing, meets the threshold for what analysts consider a home-run result. The two-year OS rate for ivonescimab reached 64.7%, against 56.9% for Keytruda plus chemotherapy in the KEYNOTE-407 China studies, a cross-trial comparison that is imperfect but not meaningless. The PFS win in HARMONi-6 was first reported at ESMO in October 2025. OS was a key secondary endpoint. Full results published simultaneously in The Lancet on May 31.
THE CHATTER
The comparison that matters most is the one the data cannot cleanly support. The KEYNOTE-407 China cohort had a meaningfully more favorable patient population: 30.8% of Keytruda patients had an ECOG performance status of 0, against 15.8% in the ivonescimab arm, and follow-up in the Keytruda studies ran longer. Ivonescimab's two-year OS rate still looks better on the raw numbers. Whether that holds in a global trial with a more diverse and older patient population is the question HARMONi-3 was supposed to begin answering. In April, it did not reach statistical significance on PFS at interim analysis. Summit Therapeutics, which holds ex-China rights, offered limited explanation. The OS readout in HARMONi-3 is not expected before late 2027 at the earliest.
WHY YOU SHOULD CARE
If you are evaluating a bispecific platform or an IO partnership in lung cancer, ivonescimab has now demonstrated OS benefit in a head-to-head setting, which changes the class conversation even with HARMONi-3's interim failure sitting underneath it. The drug works in Chinese patients against a PD-1 follower. Whether it outperforms Keytruda in a Western population, in older patients, at commercial scale, is still unanswered. Any BD conversation touching PD-(L)1 x VEGF bispecifics in NSCLC is now being priced against a data package that is strong in one geography and incomplete in the one that matters most commercially.
LOOK OUT FOR
Summit has guided to a final PFS analysis in HARMONi-3 in the second half of 2026. A positive result restores momentum. A second miss narrows ivonescimab's global commercial case before the OS data even arrives, and changes the calculus for every licensing conversation currently in motion around the class.
Galleri Showed Stage Shift. It Did Not Show It Where It Matters.
Grail presented full results from the NHS-Galleri trial and PATHFINDER 2 at ASCO this weekend. The NHS trial, run across roughly 140,000 participants in England, is the largest prospective multi-cancer early detection study conducted to date. The data landed on Saturday and is being absorbed in real time.
WHAT HAPPENED
The headline numbers are real: a 25% reduction in emergency cancer presentations, a more than 20% reduction in stage IV diagnoses, and a 6.5x increase in detected cancers on top of standard-of-care screening in PATHFINDER 2. Specificity across three rounds of the NHS trial held at 99.55%, with a positive predictive value of roughly 52 to 53%. The problem is where the benefit was concentrated. Colorectal cancer drove a disproportionate share of the stage-shift result, and CRC already has an established, functioning screening pathway in the NHS. Whether Galleri adds meaningful detection above standard screening for cancers that currently have no early detection option remains unanswered. Panelists in the Q&A flagged the trial design, noting that a study focused on higher-risk cohorts or single-cancer survival may have produced a cleaner regulatory answer.
THE CHATTER
The number not getting enough attention on the conference floor: 449 patients in the NHS trial underwent unnecessary invasive follow-up procedures as a result of false positive results. At 99.55% specificity, that sounds reassuring. Across a national screening program at scale, it is a harm number that regulators will have to price into any coverage decision. Grail's management argued on the analyst call that metastatic disease reduction may ultimately prove a more relevant regulatory endpoint than mortality, which is a creative reframe of what the data actually demonstrated. Whether the FDA reads it the same way is a different question entirely.
WHY YOU SHOULD CARE
If you are in a cancer indication with no current screening standard, the MCED story coming out of this ASCO is more complicated than the press releases suggest. The NHS data gives regulators a reason to ask for cancer-specific evidence before approving a pan-cancer label, and Grail's own management has signaled the label may end up being cancer-specific. That narrows the addressable market in ways that affect every early detection partnership conversation happening right now, including ones that have nothing to do with Grail directly. A cancer-by-cancer regulatory path means longer timelines, higher development costs, and a different business model than the field has been building toward.
LOOK OUT FOR
Grail has flagged analysis of MCED-positive patients in the NHS control arm as forthcoming. That dataset will be the first real test of whether a positive Galleri result in an unscreened population changes clinical outcomes. Until it publishes, the mortality question stays open and so does the regulatory conversation about what kind of label this data actually supports.
MRD Is in the Clinic. The Confidence Isn't.
The MRD space is having its most substantive ASCO to date, with competing approaches from NeoGenomics, Myriad Genetics, and Illumina each presenting data or announcing commercial timelines across the first four days of the conference. The field is consolidating around a technical question that is also a commercial one: how sensitive is sensitive enough, and who gets to define it?
WHAT HAPPENED
The debate centers on the limit of detection threshold below which MRD results carry clinical meaning. Myriad's PreciseMRD session made the case that true ultrasensitivity requires tumor-informed whole genome sequencing, phased variants, and approximately 1,000 variant sites to reach a limit of detection of 5 parts per million with 95% confidence. MONSTAR-SCREEN-3 data showed 99.5% baseline sensitivity across tumor types and a hazard ratio of 24.5 for patients above 100 ppm, with commercial launch in breast cancer, CRC, and renal cancer targeted for this summer. Illumina is moving toward a decentralized hospital-level MRD kit in 2027, targeting single-digit ppm sensitivity with five-day turnaround, a model not unlike what Roche is pursuing with SAGA Diagnostics. NeoGenomics is making a different argument: two SURVIVE study abstracts show that treating at molecular relapse, ctDNA positivity before imaging confirmation, improves long-term outcomes, making the case for moving the intervention point earlier rather than simply measuring more precisely.
THE CHATTER
Physicians are using MRD now, but not uniformly and not with full confidence. At a KOL dinner on the sidelines of ASCO, one breast oncologist using Signatera reported multiple false negatives preceding clinical recurrence in a triple-negative patient. He has not stopped using the assay, because the alternative is repeated imaging and radiation exposure. That is an endorsement of MRD as the lesser of two imperfect options, not as a validated standard of care. The gap between clinical adoption and clinical confidence is wider than the commercial narrative in this space currently acknowledges.
WHY YOU SHOULD CARE
If you are building a therapeutic program where MRD is becoming standard of care, the choice of CDx partner is a live decision, not a future one. The sensitivity bar is being set through commercial competition, not regulatory mandate. Illumina's distributed model, if it executes, changes the cost and access structure for the whole category by 2027, affecting reimbursement negotiations for every assay currently running through a centralized lab. The IMvigor011 readout and the Tecentriq/Signatera CDx label in bladder cancer are the nearest post-ASCO catalysts that will test whether MRD actually changes treatment decisions at the prescriber level, rather than just informing them.
LOOK OUT FOR
Myriad's PreciseMRD launch is targeted for summer 2026. The MONSTAR-SCREEN-3 data is the strongest ultrasensitive MRD dataset at this ASCO. Whether Myriad can convert it into reimbursed volume before NeoGenomics and Guardant close the sensitivity gap is the question the next two quarters will answer. The data, for once, is not the variable.
KEEP AN EYE OUT FOR
The last sessions land tomorrow. We'll be watching.
