A China-originated CLDN18.2 ADC licensed to AstraZeneca: gastric cancer Phase 3 "wins OS across the board, stumbles on PFS" — how should the China battle be fought?

To understand the molecule's value, first understand the battlefield it targets. China is the absolute home ground of global gastric cancer: according to GLOBOCAN 2022, China has about 359,000 new gastric cancer cases a year, about 37% of the global total. This means that for any gastric cancer drug, China is not an "optional market" but a core market — especially so for a China-originated molecule.

Gastric cancer treatment is highly stratified. In first line, HER2-positive disease gets trastuzumab-based combinations, HER2-negative gets PD-1 plus chemotherapy, and CLDN18.2-high adds zolbetuximab; in second line, ramucirumab plus paclitaxel is the mainstay; after third line, options dwindle, and patients' median survival is often under a year. This "the further back, the more barren" space has become the battlefield where new CLDN18.2 weapons are converging — because unmet need here is most real, and regulators' recognition of survival benefit most direct.

How big the CLDN18.2 patient pool is depends on the criterion used to count. At zolbetuximab's ≥75% moderate-to-strong staining, positivity is about 30–40%; Sone-Ve uses a broader criterion of ≥25% at any intensity, with positivity reaching about 60% by AstraZeneca's estimate, using the Ventana SP455 assay to identify patients. In other words, from the outset it chose a wider door than its "predecessor", trying to enlarge the accessible population — both an opportunity and the seed of a question: "is efficacy in low expressers good enough?"

Sone-Ve's technical composition is not aggressive; one could even call it "textbook": a humanized IgG1 antibody targeting CLDN18.2, carrying MMAE (monomethyl auristatin E, a microtubule inhibitor) via a protease-cleavable Val-Cit linker, with an average drug-to-antibody ratio (DAR) of about 4.

CLDN18.2 is a protein normally buried in the tight junctions of the gastric mucosa, out of reach of the immune system; once cells become malignant and lose polarity, it becomes exposed on the tumor cell surface. This target has already been validated as druggable by predecessors with real money: the monoclonal antibody zolbetuximab and the CAR-T satri-cel have both been approved. That is, for Sone-Ve, target risk has long been dissolved by the industry; the decisive factor is not the target but delivery, safety profile and differentiation. The cleavable linker also gives it some "bystander effect", killing adjacent low-expressing cells — neatly matching its broad ≥25% enrollment criterion.

The most typical toxicities of microtubule payloads like MMAE are peripheral sensory neuropathy and hematologic toxicity (neutropenia, anemia), plus some gastrointestinal reactions. It should be noted that MMAE and topoisomerase I inhibitor payloads (such as that used by competitor IBI343) have different toxicity monitoring priorities — the former mainly neuro/hematologic, the latter requiring additional attention to interstitial lung disease (ILD) — but this is only a difference in monitoring emphasis; whether Sone-Ve has an advantage over IBI343 in ILD risk is currently unsupported by comparative clinical evidence and should not be concluded prematurely.

Early data: the bases need to be separated

In the Phase 1 study KYM901 (formally published in Lancet Oncology, 2025; single-arm), in the gastric/GEJ cancer full analysis set at 2.2–3.0 mg/kg, investigator-assessed confirmed objective response rate (ORR) was 28% (95% CI 20–38); data for the CLDN18.2-high subgroup and the specific 2.2 mg/kg dose were higher, but these should be presented separately and not conflated with the overall figure (the "48%" that circulated earlier came from a specific high-expression subgroup, not the overall population). Global Phase 2 data released in 2026 gave a consistent order of magnitude: in the 2.2 mg/kg group, ORR 28.4%, median PFS 4.2 months, grade ≥3 adverse events 34.3%. Overall, this is a "solid but not stunning" efficacy signal in previously treated advanced gastric cancer, with safety described as manageable.

CLARITY-Gastric01: OS wins in two populations, PFS primary endpoint missed

The pivotal Phase 3 CLARITY-Gastric01 (NCT06346392) is a global multicenter, open-label, randomized controlled study that enrolled 594 patients with CLDN18.2-positive, HER2-negative advanced/metastatic gastric, gastroesophageal junction or esophageal adenocarcinoma, second line and later (2L+), comparing Sone-Ve monotherapy with investigator's choice of chemotherapy (ramucirumab + paclitaxel, paclitaxel, docetaxel, irinotecan, TAS-102, and, specific to China, apatinib). It had two primary endpoints, with the following readout:

There is one point that must be clarified and is most easily misread: "OS won, PFS lost" can only describe the contrast between the two primary endpoints, and must never be used to infer "no OS benefit in second line". Quite the opposite — in the overall 2L+ population, OS as a key secondary endpoint was significantly improved. That is, this drug genuinely prolonged life in the overall second-line-and-later population; it just failed to clear the significance threshold on the "radiographic progression" yardstick of PFS.

The combination "OS met but PFS not met" is not itself rare, and may relate to cross-line treatment, subsequent therapies and delayed effects of ADCs, but should not be over-interpreted before the full data are out. Equally worth stressing: the company has not yet released hazard ratios (HR), median OS / PFS or subgroup data; "highly clinically meaningful" is still a qualitative description, and the real magnitude of efficacy must wait for a scientific meeting. On safety, the readout says "well tolerated, no new safety signals", consistent with the known MMAE profile.

Breaking whether Sone-Ve can succeed in China into several thresholds it must clear — red for hard flaws, orange for uncertainties, blue for the positive cards in its hand:

Sone-Ve's registration logic is relatively clean — it is itself a randomized controlled Phase 3 including Chinese patients, not relying on single-arm or real-world evidence. But note: mainland China had fairly broad site participation (about 35 sites publicly registered), providing a basis for including Chinese patient data, but whether it can directly support Chinese registration still depends on actual enrollment size, regional consistency analysis and CDE's requirements; "no bridging needed" cannot be asserted from site count alone.

Every drug that defines its population by a biomarker cannot escape one commercial loop: no testing, no prescriptions; no prescriptions, no data; no data, no good payment negotiation; and inadequate payment in turn suppresses the motivation to order tests. Sone-Ve is no exception, and its testing step carries an extra layer of cutoff complexity:

To understand Sone-Ve's situation, one must separate "global" from "China" and "third line" from "earlier lines". In China's third-line gastric cancer, it faces a battlefield already under fire:

The most glaring row in this table is the last one. In China's third-line gastric cancer, Sone-Ve started latest: satri-cel is already marketed, the marketing applications of IBI343 and LM-302 are both in priority review, and its filing hasn't even been accepted. LM-302 in particular — the same class of CLDN18.2-MMAE ADC as Sone-Ve, almost a "mirror opponent" — is a step ahead in China. This is a negative footnote to "China-originated, going global": when a molecule sells its global rights to an MNC and the latter advances it on a global schedule, it may actually be overtaken in its home market by peers who "stayed home, do only China, and run faster".

But widen the lens, and Sone-Ve's situation is far less passive than that one table row suggests. Its ability to fight back lies precisely in not having to be trapped in third line: First, hard dual-population OS data — satri-cel won on PFS, IBI343 and LM-302 announced "met primary endpoint", while Sone-Ve directly produced positive OS in both 3L+ and 2L+, which carries the most weight at the regulatory and guideline level (cross-study comparisons are not direct, but the difference in level of evidence is real). Second, off-the-shelf — for the many primary-level and prefecture-city hospitals without cell therapy capability, an IV-infused ADC is an order of magnitude more accessible than a CAR-T requiring apheresis + reinfusion + CRS management. Third, depth across lines — the three front-runners are all crowded in third line, while Sone-Ve has 2L+ OS data and is already running first line (detailed in the next section). Its real differentiation is not "catching up with someone in third line" but pushing the battlefield forward to second or even first line.

Pull the lens back further for the structural lesson of this game: why do "those who stayed home run faster"? Because companies like CARsgen, Innovent and Sinopharm keep the entire chain of R&D, registration and commercialization in their own hands, and can use a "China first" pace to make the most of local acceleration tools like Breakthrough Therapy and priority review; whereas a molecule licensed to an MNC must follow the partner's global development schedule and portfolio priorities. Agility comes at a price — going global bought global resources and certainty, but gave up the ability to sprint at home. This is not a problem unique to Sone-Ve, but a shared proposition for all "invented in China, licensed abroad" assets.

In deal economics, this 2023 license was favorable to the originators: the US$63 million upfront was just an entry ticket, with the real value in up to US$1.1 billion of development/sales milestones and tiered sales royalties up to low double digits. It should be noted that there is no public evidence that this OS result has directly triggered any milestone payment, and it should not be speculated. But the direction is clear: Keymed and Lepu Biopharma neither bear the huge cost of a global Phase 3 nor lose a continuing share when AstraZeneca sells the drug — an "asset-light participation in the global market" paradigm. Only, this participation is premised on giving up local control: the math works, but they don't get to make the moves.

This timeline hides a strategic contrast: the agility of R&D came from a Chinese biotech, the weight of commercialization from an MNC. After AstraZeneca took over, Sone-Ve gained global Phase 3 resources, global registration capability and a mature China commercial network, and has already been pushed into first-line combination therapy; but at the same time it has been slotted into the ranking of AstraZeneca's own gastric cancer portfolio — the company already has T-DXd for HER2-positive gastric cancer, and CLDN18.2 is only one piece of the puzzle. The real open variable is how much acceleration AstraZeneca is willing to invest in China for this China-originated molecule.

There is only one core variable behind the three scenarios: speed. Not R&D speed — the 2L+ Phase 3 has already read out; but the speed of converting OS data into a China marketing application and then into a prescription entry point. In a third-line market where a marketed CAR-T and two ADCs in priority review are already standing in front, every quarter's delay in filing diverts another batch of accessible third-line patients — which is exactly why it should put its weight on second and first line.

Sone-Ve is a mirror. It reflects the strength of Chinese innovative drugs — a molecule made in a local lab can become the world's first CLDN18.2 ADC with positive OS, winning in both 3L+ and 2L+; it also reflects the other side of the "out-licensing" road — handing the steering wheel to an MNC in exchange for global resources and certainty may also mean losing the third-line lead in one's most familiar home market to faster-running peers.

It holds the hardest card in the world (dual-population positive OS), yet stands at the very back in China's most crowded third line. But the real way out of this game may not be in third line at all — it has 2L+ OS data and a first-line Phase 3 already under way, and can entirely skip the overcrowded third line and build its differentiation in earlier lines. The PFS miss limits its voice on radiographic progression, but does not take away its most essential value: prolonging life.

So the answer is not in the molecule — the molecule has already proven itself; the answer lies in AstraZeneca's moves over the next few quarters: whether it can quickly turn a fine set of survival data into as broad a label as possible, a testing standard and a prescription pathway, and play the first-line Phase 3 trump card well. Speed and choice of line are the variables it can still control in China.