cartanis07/30/2026

Not another PD(L)1xVEGF POV

This class of bispecifics has been the talk of the immuno-oncology town for some time. No, the world probably doesn't need another opinion on it. Yes, it makes for a great case study around which I can build the analytics around a decision support tool like cartanis.

I started with the question "Which of the yet-to-be-partnered PD(L)1xVEGF programs are most promising, and what should an interested search & evaluation team first do?" and I wanted to see if I could come up with a better answer than "naked" Claude.

So, if you'd like, go spin up your favorite agent and ask it in whatever manner you'd like, turn on high thinking if tokenmaxxing is still in vogue when you read this. I'd love feedback on any interesting angles my analysis misses.

My approach:

First of all, I don't just want an agent to say "go buy asset XYZ." I also don't want to just show a red/yellow/green matrix of assets vs criteria from which the user has to infer the next steps. I am building cartanis to recommend concrete actions the hypothetical BD team can take based on the information they have. For example:

  • Interview KOLs
  • Buy a proprietary dataset
  • Wait for the target or competitor clinical readout
  • Reach out to target company to solicit information

In this way, we're not just enabling decisions around which assets look the most "green" ("what should buy"?), we're guiding decisions around which actions would give us more confidence in that scoring. Identifying reds which are most prone to flipping green with a cheap experiment ("where can we find hidden pockets of value?") and vice versa, green to red ("where should our diligence focus?"). How do we automate this?

To answer these kinds of questions, cartanis uses:

  • A "pareto frontier" representation of the standard of care (read more from Formation Bio on this) across relevant tumor types
  • Estimates of future frontiers based on the known pipeline and existing data, and probability of success estimates for each program
  • Estimates of uncertainty around both the future SOC and the asset in question
  • A catalog of activities which can resolve those uncertainties

I look forward to talking about those modules in future posts, but today's focus is the answer to the PD-(L)1×VEGF question: Claude Opus 4.8 drafted the report below using a human-guided workflow built on top of cartanis. The animations were human-conceived, and Claude built. Em-dashes are all Claude. And frankly… I cheated a decent bit. The blog post took a few iterations to get to this point: the next analysis will materialize with less guidance as each workflow improves.

Finally, I was at first tempted to replace the exact numbers produced from cartanis with directional statements, but a good friend once told me that the best way to get the right answer is to say the wrong one loudly in public.

The setup

PD-(L)1×VEGF bispecifics do two jobs with one molecule: block the PD-1/PD-L1 checkpoint that most modern immunotherapy targets, and block VEGF, the blood-vessel-growth pathway that drugs like bevacizumab hit. The premise is that shutting down VEGF remodels the tumor's blood supply and immune environment so the checkpoint half lands harder.

One caveat before the count: the class is that PD-(L)1×VEGF pairing, but four of the assets — CS2009, GB268, DR30206, and SCTB41 — are trispecifics that bolt on a third target (CTLA-4 or TGF-β). That's a different mechanism and risk profile, so they aren't a like-for-like comparison with the bispecifics.

That was theory until ivonescimab tested it head-to-head. Akeso's bispecific — licensed outside China to Summit Therapeutics1 — cut the risk of progression roughly in half versus pembrolizumab in a randomized 1L NSCLC trial (HARMONi-2, PFS HR 0.512), the first time a PD-(L)1×VEGF beat an approved checkpoint inhibitor in a controlled trial. That single result set off a gold rush: Summit's ~$5B deal, then Merck3, Pfizer4, BMS with BioNTech5, and AbbVie6 all paying up for their own molecules across 2024–2026.

The open questions are what this piece is about: whether the mostly-China data will translate to a US approval, whether the added toxicity is manageable, and — with nineteen molecules chasing one mechanism — whether anyone is genuinely differentiated. Which sets up the search-and-evaluation problem: of the assets not yet spoken for, which is the best bet, and what should a buyer do first?

The short version

Three readouts decide this class: SCTB14's blinded pivotal, CS2009's first randomized trial, and pumitamig's first 1L NSCLC disclosure.

01Nineteen assets in the class — and five are already gone.

Nineteen molecules — who owns them, and where they're running

19 disclosed PD-(L)1×VEGF programs across 13 tumor rows (95 placements). Each cell is the most advanced trial an asset runs in that tumor, plus any distinct non-chemo combination; columns are colored by what a BD team could do about the asset, dots drawn by phase.

assetlicensablealready partnered
trialphase 3phase 2phase 1
diseaseIVONESCIMABPUMITAMIGJS207RC148HB0025IMM2510SSGJ707SCTB14SCTB41△AI081AP505CR001CS2009△DR30206△GB268△HLX37LM299MHB039ABCG036
Solid tumors (basket)
NSCLC
Colorectal
Breast
Liver (HCC)
Pancreatic
Gynecologic
Gastric / GEJ
Biliary
Genitourinary
Head & neck
Small-cell lung
Melanoma
Asset × tumor type · ClinicalTrials.gov / AACT (2026-07-05) ↗
how each cell is counted

Each cell shows the MOST ADVANCED registered trial for that asset-tumor (ties broken toward an ongoing study), plus any additional ongoing trial combining the asset with another named agent — distinct non-chemo combinations are kept; chemotherapy-only and earlier-phase duplicates are folded away. Combined-phase studies (e.g. Ph1/2) are drawn at their higher phase. So a column reads as commitment per tumor. BCG036 has no registered phase study at all — the column is kept blank rather than dropped, because that absence is itself the thing worth knowing.

△ trispecific — bolts a third target (CTLA-4 or TGF-β) onto the PD-(L)1×VEGF backbone.

The class is one asset and a long tail. Ivonescimab alone is 38 of those 95 placements (155 registered trials in all) and appears in every tumor row; most of the rest live in the basket row, where a first-in-human study in "advanced solid tumors" is a molecule without a stated program. 12 of the 19 assets have two or fewer trials on the grid.

What you could do a deal on

5 of the 19 are partnered; the other 14 are reachable — mostly straight licensing, plus CR-001, where US rights come only by acquiring the company. Each line below is the rights position as its source states it, disclosed terms included.

Partnered — US rights are taken

  • IVONESCIMABex-China rights held by Summit Therapeutics ($500M upfront / up to ~$5B, Dec 2022); Summit submitted a US NSCLC BLA to FDA in Jan 2026 — US rights PARTNERED (taken). AstraZeneca ~$15B licensing interest reported Jul 2025 remains UNCONFIRMED.
    Ivonescimab (AK112/SMT112, Akeso) ex-China rights held by Summit Therapeutics ($500M upfront / up to ~$5B, Dec 2022); Summit submitted a US NSCLC BLA to FDA in Jan 2026 — US rights PARTNERED (taken). AstraZeneca ~$15B licensing interest reported Jul 2025 remains UNCONFIRMED.1
  • LM299ex-China rights held by Merck; $588M upfront / up to ~$3.3B total (Nov 2024) — US rights PARTNERED (taken).
    LM-299 (MK-2010, LaNova) ex-China rights held by Merck; $588M upfront / up to ~$3.3B total (Nov 2024) — US rights PARTNERED (taken).3
  • PUMITAMIGUS rights held by BioNTech + BMS; a 50/50 co-develop / co-commercialize GLOBAL PARTNERSHIP (not a clean ex-US license): $1.5B upfront + $2.0B non-contingent through 2028 (~$3.5B guaranteed) + milestones to ~$11.1B total (Jun 2 2025); BioNTech acquired originator Biotheus first (buy-then-partner) — US rights PARTNERED (taken).
    Pumitamig (BNT327/PM8002, Biotheus) US rights held by BioNTech + BMS; a 50/50 co-develop / co-commercialize GLOBAL PARTNERSHIP (not a clean ex-US license): $1.5B upfront + $2.0B non-contingent through 2028 (~$3.5B guaranteed) + milestones to ~$11.1B total (Jun 2 2025); BioNTech acquired originator Biotheus first (buy-then-partner) — US rights PARTNERED (taken).5
  • RC148ex-Greater-China rights licensed to AbbVie; $650M upfront + up to ~$4.95B milestones (~$5.6B total), announced Jan 12 2026 — the most recent partnering of the validated set; US rights PARTNERED (taken).
    RC148 (RemeGen) ex-Greater-China rights licensed to AbbVie; $650M upfront + up to ~$4.95B milestones (~$5.6B total), announced Jan 12 2026 — the most recent partnering of the validated set; US rights PARTNERED (taken).6
  • SSGJ707ex-China global rights held by Pfizer; $1.25B upfront + up to $4.8B milestones PLUS $100M equity + $100M China-option fee (+$50M exercise) — true cost above ~$6B (announced May 20 2025) — US rights PARTNERED (taken).
    SSGJ-707 (3SBio) ex-China global rights held by Pfizer; $1.25B upfront + up to $4.8B milestones PLUS $100M equity + $100M China-option fee (+$50M exercise) — true cost above ~$6B (announced May 20 2025) — US rights PARTNERED (taken).4

Licensable

  • AI081FULLY OWNED by OncoC4, no external licensee; Ph2 portion of BIPAVE-001 dosing in the US (11 US sites, first Ph2 patient Jul 2025) — US originator with CLEAN, fully-available US rights, but early-stage.
    AI-081 (OncoC4, US; PD-1×VEGF, from proprietary anti-PD1 AI-025 + anti-VEGF AI-011) — FULLY OWNED by OncoC4, no external licensee; Ph2 portion of BIPAVE-001 dosing in the US (11 US sites, first Ph2 patient Jul 2025) — US originator with CLEAN, fully-available US rights, but early-stage.7
  • AP505Ph1 dose-escalation in advanced solid tumors; global out-licensing in progress; UNPARTNERED for US.
    B1962/AP505 (AP Biosciences/Tasly; PD-L1×VEGF fusion protein) — Ph1 dose-escalation in advanced solid tumors; global out-licensing in progress; UNPARTNERED for US.8
  • BCG036PRECLINICAL next-gen candidate on Biocytogen's out-licensing platform; available for partnering but not yet clinical (optionality only, not rankable on clinical data).
    BCG036 (Biocytogen; PD-1xVEGF-A bispecific, VEGF VHH/nanobody 'RenNano' design) — PRECLINICAL next-gen candidate on Biocytogen's out-licensing platform; available for partnering but not yet clinical (optionality only, not rankable on clinical data).9
  • CR001Crescent granted Kelun-Biotech exclusive GREATER-CHINA rights (Dec 2025) but RETAINS ex-China/US rights and runs the global ASCEND trial (US FDA IND cleared Jan 2026); US access is only via ACQUIRING Crescent, not a clean out-license.
    CR-001/SKB118 (Crescent Biopharma, US; PD-1×VEGF) — Crescent granted Kelun-Biotech exclusive GREATER-CHINA rights (Dec 2025) but RETAINS ex-China/US rights and runs the global ASCEND trial (US FDA IND cleared Jan 2026); US access is only via ACQUIRING Crescent, not a clean out-license.10
  • CS20091L NSCLC PD-L1-high (TPS≥50%) monotherapy ORR 81.3% / DCR 100% (updated ASCO 2026); US FDA IND cleared; CStone preparing first global Ph3 MRCT by end-2026 — highest efficacy ceiling of the field, added CTLA-4 safety/complexity; UNPARTNERED.
    CS2009 (CStone; PD-1×VEGF×CTLA-4 TRISPECIFIC) — 1L NSCLC PD-L1-high (TPS≥50%) monotherapy ORR 81.3% / DCR 100% (updated ASCO 2026); US FDA IND cleared; CStone preparing first global Ph3 MRCT by end-2026 — highest efficacy ceiling of the field, added CTLA-4 safety/complexity; UNPARTNERED.11
  • DR30206anti-PD-L1 × VEGF × TGF-β TRISPECIFIC trifunctional fusion protein (anti-PD-L1 VHH + bevacizumab-derived anti-VEGF + TGF-βRII ECD trap); non-CTLA-4 third arm is TGF-β. Ph1 FIH in advanced solid tumors (NCT06132828, recruiting from Nov 2023) + Ph1b/2a GI-cancer combo (NCT07056777); China-only development so far, ex-China/US rights UNPARTNERED (no disclosed licensee).
    DR30206 (Zhejiang Doer Biologics, Hangzhou China) — anti-PD-L1 × VEGF × TGF-β TRISPECIFIC trifunctional fusion protein (anti-PD-L1 VHH + bevacizumab-derived anti-VEGF + TGF-βRII ECD trap); non-CTLA-4 third arm is TGF-β. Ph1 FIH in advanced solid tumors (NCT06132828, recruiting from Nov 2023) + Ph1b/2a GI-cancer combo (NCT07056777); China-only development so far, ex-China/US rights UNPARTNERED (no disclosed licensee).12
  • GB268Phase 1 first-in-human in advanced solid tumors (study start 2025-04-24); a second in-class trispecific alongside CS2009; UNPARTNERED, very early.
    GB268 (Genor Biopharma; PD-1/CTLA-4/VEGF TRISPECIFIC antibody) — Phase 1 first-in-human in advanced solid tumors (study start 2025-04-24); a second in-class trispecific alongside CS2009; UNPARTNERED, very early.13
  • HB0025Ph2 combo with chemo in 1L advanced/recurrent ENDOMETRIAL cancer (differentiated indication focus); UNPARTNERED, limited disclosed data.
    HB0025 (Huabo/Harbour; PD-L1×VEGF-trap) — Ph2 combo with chemo in 1L advanced/recurrent ENDOMETRIAL cancer (differentiated indication focus); UNPARTNERED, limited disclosed data.14
  • HLX37Ph1, first patient dosed; novel VEGF-induced-PD-L1-internalization design; rights status likely open/UNPARTNERED but not formally confirmed.
    HLX37 (Henlius; PD-L1×VEGF) — Ph1, first patient dosed; novel VEGF-induced-PD-L1-internalization design; rights status likely open/UNPARTNERED but not formally confirmed.15
  • IMM2510ex-China rights REVERTED from Instil Bio (Jan 6 2026) because Instil wound down its ENTIRE clinical pipeline — a licensee STRATEGIC/FINANCIAL decision, NOT a data verdict on IMM2510 (~$35M paid); IMM2510 + IMM27M handed back — ACTIVELY AVAILABLE ex-China; carries early 1L NSCLC efficacy (Ph1/2, China).
    IMM2510/AXN-2510 (ImmuneOnco; PD-L1×VEGF-trap, VEGFR1-D2 trap binding VEGF-A/B + PlGF, ~169 kDa, Fc-engineered for enhanced ADCC — the only class molecule with ADCC) ex-China rights REVERTED from Instil Bio (Jan 6 2026) because Instil wound down its ENTIRE clinical pipeline — a licensee STRATEGIC/FINANCIAL decision, NOT a data verdict on IMM2510 (~$35M paid); IMM2510 + IMM27M handed back — ACTIVELY AVAILABLE ex-China; carries early 1L NSCLC efficacy (Ph1/2, China).16
  • JS207FDA IND cleared Oct 2025 for a Ph2/3 confirmatory study vs. nivolumab in neoadjuvant resectable NSCLC — the first PD-1×VEGF to enter a confirmatory/registrational study; 11 ongoing Ph2 combos; UNPARTNERED ex-China (most advanced regulatory path of the unpartnered set).
    JS207 (Junshi; PD-1×VEGF-A) FDA IND cleared Oct 2025 for a Ph2/3 confirmatory study vs. nivolumab in neoadjuvant resectable NSCLC — the first PD-1×VEGF to enter a confirmatory/registrational study; 11 ongoing Ph2 combos; UNPARTNERED ex-China (most advanced regulatory path of the unpartnered set).17
  • MHB039APh1 in relapsed/refractory solid tumors, well tolerated to 20 mg/kg (MTD not reached, no DLTs), tumor reduction in prior-PD-1/chemo NSCLC incl. post-3G-TKI EGFR-mutant; Qilu took GREATER-CHINA rights (~$38M, May 2026) but Minghui RETAINS ex-China (incl. US) and is seeking ex-China partners — UNPARTNERED/AVAILABLE for US.
    MHB039A (Minghui Pharmaceutical; PD-1xVEGF bispecific) — Ph1 in relapsed/refractory solid tumors, well tolerated to 20 mg/kg (MTD not reached, no DLTs), tumor reduction in prior-PD-1/chemo NSCLC incl. post-3G-TKI EGFR-mutant; Qilu took GREATER-CHINA rights (~$38M, May 2026) but Minghui RETAINS ex-China (incl. US) and is seeking ex-China partners — UNPARTNERED/AVAILABLE for US.18
  • SCTB14Ph2/3 pivotal 1L NSCLC vs. pembrolizumab; UNPARTNERED, limited disclosed data.
    SCTB14 (Sinocelltech; PD-1×VEGF) — Ph2/3 pivotal 1L NSCLC vs. pembrolizumab; UNPARTNERED, limited disclosed data.19
  • SCTB41is a clinical-stage anti-tumor antibody: Ph1/2 monotherapy dose-escalation/expansion in advanced solid tumors (NCT06600022, recruiting from 2024-10-10) advancing to a Ph2/3 study in 1L squamous NSCLC vs tislelizumab+chemo (NCT07662395). Sponsor Sinocelltech; UNPARTNERED. Its exact targets are NOT stated in the trial registry; secondary sources report it as a PD-(L)1×VEGF×TGF-β trispecific (third arm TGF-β, non-CTLA-4) but the PD arm (PD-1 vs PD-L1) is UNCONFIRMED against any primary source that names SCTB41.
    SCTB41 (Sinocelltech Ltd., China) is a clinical-stage anti-tumor antibody: Ph1/2 monotherapy dose-escalation/expansion in advanced solid tumors (NCT06600022, recruiting from 2024-10-10) advancing to a Ph2/3 study in 1L squamous NSCLC vs tislelizumab+chemo (NCT07662395). Sponsor Sinocelltech; UNPARTNERED. Its exact targets are NOT stated in the trial registry; secondary sources report it as a PD-(L)1×VEGF×TGF-β trispecific (third arm TGF-β, non-CTLA-4) but the PD arm (PD-1 vs PD-L1) is UNCONFIRMED against any primary source that names SCTB41.20

The leader's data — and what it would actually take to beat it

Ivonescimab is the class's only randomized readout, fields the most trials (38 on the grid, more than the next three combined), and appears in all 13 tumor rows. Any follower's pitch is implicitly a claim to beat it on one of four axes — efficacy, safety, format/dosing, or tumor choice — and three of the four are largely closed, for mechanical reasons.

IVONESCIMABHR 0.51median PFS 11.1 mo vs 5.8 moPD-L1-positive, monotherapy · randomized vs pembrolizumab (Ph3, China) — the only controlled result in the class

Efficacy — the ORRs aren't measuring the same thing

Ivonescimab's HARMONi-2 (PFS HR 0.51 vs pembrolizumab) is the only PD-(L)1×VEGF result with a control arm. Five other assets report 1L NSCLC ORRs between 55% and 81% — in five different cohorts (PD-L1-high, PD-L1-positive, unselected, monotherapy, plus-chemotherapy), so they aren't measuring the same thing. And ORR is a weak surrogate for survival in checkpoint therapy21, so even a clean cross-cohort ranking wouldn't tell you which asset extends life.

  1. IVONESCIMABHR 0.5111.1 mo vs 5.8 morandomizedChina
    PD-L1-positive, monotherapy · randomized vs pembrolizumab (Ph3, China)
    HARMONi-2 stratified PFS HR 0.51 (95% CI 0.38-0.69), one-sided p<0.0001, vs pembrolizumab2
  2. CS2009ORR 81.3%single-armChina
    PD-L1-high (TPS≥50), monotherapy · single-arm · China
    1L NSCLC PD-L1-high (TPS>=50) monotherapy ORR 81.3% / DCR 100% (updated ASCO 2026)11
  3. IMM2510ORR 62.0%single-armChina
    unselected, + chemotherapy · single-arm · China (n=21)
    PR in 62% of efficacy-evaluable patients (8/10 squamous, 5/11 non-squamous), 1L NSCLC + chemotherapy (Ph2, China)22
  4. SSGJ707ORR 61.8%single-armChina
    PD-L1 ≥1%, monotherapy at 10 mg/kg · single-arm · China (n=34)
    1L advanced NSCLC (PD-L1 >=1%) monotherapy, 10 mg/kg Q3W: ORR 61.8% (21/34); PD-L1 TPS 1-49% 57% (12/21), TPS >=50% 69% (9/13); non-squamous 54.5%, squamous 75% (ASCO 2025)23
  5. JS207ORR 58.1%single-armChina
    PD-L1-positive, monotherapy · single-arm · China (n=62)
    1L PD-L1-positive NSCLC monotherapy ORR 58.1%, DCR 87.1% (n=62), ESMO Asia 202524
  6. LM299ORR 55.0%single-armChina
    treatment-naive, monotherapy at 20 mg/kg · single-arm · China, responses UNCONFIRMED
    NSCLC expansion, treatment-naive: unconfirmed ORR 55% (20 mg/kg Q3W) and 44% (30 mg/kg Q3W); Ph1/2 n=112 (AACR 2026)25

So a higher ORR than ivonescimab's is a cohort claim, and the only thing that settles it is another controlled readout — of which exactly one is running. There is no fair side-by-side efficacy table to build yet: every follower's number is single-arm and mostly China-only, with its global randomized trial still recruiting.

Safety — the excess toxicity is the mechanism, not the molecule

There is one controlled safety comparison in the class, and it's the same trial that produced the only controlled efficacy result.

HARMONi-2 · ivonescimab monotherapy vs pembrolizumab monotherapy, 1L PD-L1-positive advanced NSCLC · comparator bevacizumab, ECOG E459926
about this comparison

randomized, double-blind, phase 3, single-country (China). The only randomized safety comparison in the class. Anti-VEGF class reference, from a DIFFERENT trial: ECOG E4599 (bevacizumab + carboplatin/paclitaxel vs the same chemo alone, 1L advanced non-squamous NSCLC; Sandler et al., NEJM 2006). The grade 3+ VEGF-pathway events shown are the bevacizumab-minus-chemo delta — i.e. what anti-VEGF alone adds — so they read as the on-target class effect, not a head-to-head with ivonescimab's monotherapy arm. The composite grade-3+ row is reported per-event rather than pooled (E4599 lists events individually, over a chemo backbone); the immune-related row is not applicable to an anti-VEGF agent.

Endpointivonescimab monotherapypembrolizumab monotherapybevacizumab + chemo1L non-sq NSCLC · ECOG E4599 · different study
grade 3+ treatment-related AEs29.4%15.6%per event ↓
treatment-related serious AEs20.8%16.1%—
grade 3+ immune-related AEsnot disclosednot disclosedn/a
grade 3+ hypertensionhigherreference7–8%
grade 3+ proteinuriahigherreference3%
grade 3+ bleeding / hemorrhagehigherreference~4%

The excess over a checkpoint-monotherapy control is VEGF-pathway shaped — proteinuria and hypertension, the canonical systemic VEGF-A blockade effects, whose magnitude varies by molecule but whose kind is a class effect. That is on-target-pathway toxicity: the drug working where you did not want it to. A cleaner antibody does not remove it. Only a therapeutic-index play does — restricting WHERE or WHEN the VEGF arm acts.

Format and dosing — the two obvious levers are already taken

If a cleaner antibody doesn't fix a pathway-shaped burden, differentiation has to come from the format. There are only a few such levers, and the leader holds most of them.

  • effector-silencingtaken by the leaderread more ↗
    Ivonescimab discloses Fc-silencing mutations (L235A/L236A).27
  • avidity-gatingtaken by the leaderread more ↗
    Ivonescimab discloses cooperative, VEGF-dimer-mediated binding (a reported ~18-fold PD-1 avidity increase in the presence of VEGF).27
  • conditional-activationopenread more ↗
    The field's one genuinely unclaimed index lever: no asset in the class has disclosed a masked or tumour-conditional arm. A full-class check (July 2026) found only avidity/valency and tumour-anchoring designs, none of which is conditional activation.
    unclaimed in the DISCLOSURE, which is not the same as unclaimed in the molecule: ten of nineteen assets disclose no format at all
  • exposure-shapingopenread more ↗
    Only ivonescimab discloses a half-life (≈5–7 days), and BCG036 is the sole asset to disclose an explicit Fc half-life-extension design (preclinical — more sustained exposure than ivonescimab in FcRn mice, no absolute figure). Every other asset's PK is undisclosed, so the lever is almost entirely unclaimed in public.28
    claimed only preclinically and by a single preclinical asset; the leader's disclosed half-life already sets a bar a follower would have to beat

There's a cheap move on the ten assets that disclose no format: the one open index lever — a masked or tumour-conditional VEGF arm — and the real dosing and half-life data live in a management conversation, not a press release. Calls with the earlier-stage originators, or a China-based competitive-intelligence agency, are the low-cost way to learn whether any of the silent ten holds the lever the leader hasn't taken.

And four of the nineteen skip the format question by adding a target instead. CS2009 and GB268 add CTLA-4, the checkpoint ipilimumab hits: pairing it with PD-1 is the nivolumab-plus-ipilimumab playbook — deeper, more durable responses, bought with the immune-related toxicity that combination is known for, grade 3/4 events in roughly a third of patients on the NSCLC-approved doublet.29 DR30206 and SCTB41 add a TGF-βtrap. TGF-β is one of the tumour's dominant immunosuppressive signals — it excludes T cells and feeds the stroma and angiogenesis VEGF blockade already targets, so trapping it alongside PD-(L)1 and VEGF is mechanistically appealing.30 But it carries a hard clinical history: bintrafusp alfa, the lead PD-L1×TGF-β fusion, missed a string of phase 2/3 trials and was largely shelved.31 A third target adds a differentiation story and a second failure mode.

Tumor choice — the one axis that's genuinely open

Here the grid says something the ORR tables can't. Ivonescimab appears in every tumor row, but its phase 3 programs sit in only 8 of the 13 it occupies — none in Gynecologic, Gastric / GEJ, Genitourinary, Melanoma. A follower who lands a randomized readout in one of those rows arrives before a registrational program exists.

tumorivonescimabPUMITAMIGJS207SSGJ707HB0025IMM2510
BreastPh3*
ColorectalPh3*
GynecologicPh2
Liver (HCC)Ph3*
Gastric / GEJPh2
GenitourinaryPh2
Small-cell lungPh3*
Where a follower could reach the US market first — tumors where ivonescimab has no phase-3 lead. The ivonescimab column shows its own most-advanced stage there:Ph2 / Ph1 = no registrational program; Ph3* = it has a phase 3 but started itafter the peer's most-advanced trial. A dot marks a peer that could reach the US ahead of ivonescimab; a haloed dot is a front-runner — level with, or ahead of, every other program in that tumor, so it could be first to market. A plain dot clears ivonescimab but sits a phase behind another program. Read off registered stage and timing, not disclosed efficacy.
has an early trial here but sits behind ivonescimab's phase-3 lead — not a route to first-to-market.

Read outside 1L NSCLC and several of these assets already have data — in tumors where the leader has trials running but nothing registrational:

  1. HB0025ORR 84.0%EndometrialChina
    1L advanced/recurrent, + carboplatin/paclitaxel · single-arm · China (n=39, early cutoff)
    ~84% ORR across MMR status (39 patients, early cutoff), 1L advanced/recurrent endometrial cancer + carboplatin/paclitaxel (Ph2, ASCO 2025); improving on historical 40-68%14
  2. JS207ORR 45.5%Liver (HCC)China
    1L, + JS007 (anti-CTLA-4) · single-arm · China (n=22 evaluable)
    1L HCC with JS007 (anti-CTLA-4): ORR 45.5%, DCR 86.4% (22 evaluable), no DLTs, AACR 202632
  3. PUMITAMIGORR 76.3%Small-cell (ES-SCLC)global
    1L extensive-stage, + chemotherapy · global interim Ph2, single-arm
    Global interim Ph2: confirmed ORR 76.3%, DCR 100%, mPFS 6.8 months, pumitamig + chemotherapy in 1L extensive-stage SCLC33
  4. HB0025ORR 9.1%Solid tumors (first-in-human)China
    heavily pretreated, monotherapy ≥ 3 mg/kg · single-arm · China (n=22)
    First-in-human monotherapy ORR 9.1% (2/22 at >=3 mg/kg), heavily pretreated, 202334
  5. IMM2510trial ongoing · no data disclosedBreast (TNBC)
    1L triple-negative breast cancer — Ph2 combination-with-chemotherapy cohort

    Ph2 first-line combination with chemotherapy (NCT06746870); triple-negative breast cohort enrolling — no efficacy disclosed yet.35

None of these are controlled and none are large. But an 84% ORR in first-line endometrial and a 76.3% in first-line ES-SCLC are pointed at rows where the leader has committed nothing — a materially different proposition from being the seventh 1L NSCLC ORR.

What we'd want to know

  1. The format, not the AE table, is where a real difference between two assets lives — and ten of nineteen assets disclose nothing about how the molecule is built.
  2. No asset discloses whether either arm is masked or conditionally activated, the missing input to any therapeutic-index claim; half-life is disclosed only by the leader (≈5–7 days) and one preclinical asset, so every follower's PK is dark.
  3. The per-organ grade 3+ split matters more than the headline rate: only it says whether a follower's burden is the VEGF pathway (unfixable by a cleaner molecule) or something engineerable.

Where the partnered assets are playing — and whether they're actually differentiated.

Ivonescimab set the bar; then the class went on sale — pumitamig to BioNTech and BMS, SSGJ-707 to Pfizer, LM-299 to Merck, RC148 to AbbVie. On the disclosed data, how little separates them is striking.

Pumitamig (BNT327, BioNTech/BMS) led with 1L ES-SCLC; when BMS signed (Jun 2025) it had disclosed nothing in 1L NSCLC. Its first 1L NSCLC readout came only at ASCO 2026 — ROSETTA Lung-02, confirmed ORR ~57% non-squamous / ~68% squamous, still early single-arm Phase 2, not the randomized bar ivonescimab set. The biggest bet in the class was priced on adjacent tumors, not on the shared battleground.

Of the five, two (SSGJ-707, LM-299) have a single-arm 1L NSCLC ORR and nothing randomized; one (RC148) has no disclosed efficacy at all and a handful of registered trials; and the largest deal in the class rests on a tumor the leader hasn't committed a phase 3 to. On the stand-alone data, pharma is buying position more than a distinguished molecule — an option premium paid before China-to-US translation and the pending global phase 3s resolve, which is why every one of these deals is back-loaded into milestones.

Intersect what's still reachable with what has disclosed 1L NSCLC efficacy and you get a three-name list: CS2009, IMM2510, JS207 — the entire licensable-with-data set.

Which leaves the unpartnered ones — positioned how, and known how well?

The reachable set splits cleanly in two. Three assets have data and a position you can argue about. The rest have a molecule and an intention.

  • CS2009 — highest disclosed ceiling — and the biggest open question
  • SCTB14 — option on a blinded pivotal readout
  • IMM2510 — available today, data in hand, rights just reverted
  • JS207 — furthest down a US regulatory path

For that bucket, no amount of waiting produces a number. They are unscoreable from public data by construction, which is why our engine reports value-of-next-move for them instead of a score.

For this bucket the information isn't coming from press releases; it's in a data room. SCTB14 is where we'd start — highest value-of-next-move in the class, and a blinded pivotal already running.

One lens reshuffles the ranking: whether an asset can actually reach the US market. The class is almost entirely China-originated, and a China-only pivotal — single-country, usually on PFS, against a non-US comparator — is close to the fact pattern the FDA's advisory committee rejected for sintilimab36; it doesn't support a US approval on its own. So the axis that deserves the most weight is registrational-path quality, and on it the field reorders: JS207 (first PD-1×VEGF with a US IND, in neoadjuvant NSCLC) and CS2009 (a global phase-3 MRCT planned for later this year) are on a credible US route, while SCTB14's blinded pivotal — its entire appeal — is a China trial vs pembrolizumab that, as designed, doesn't clear the FDA by itself.

Are the reachable assets the picked-over remainder, or available for reasons that survive scrutiny? Mostly the latter. IMM2510's rights reverted because its licensee Instil wound down its entire pipeline — a corporate implosion, not a verdict on the drug (which carries the most clinical data of the unpartnered set, and is the class's only ADCC molecule). JS207 is available because Junshi, which already took a China PD-1 to US approval, can hold for better terms after its readout. CS2009 is differentiated and simply early; its CTLA-4 arm is a real complexity a bispecific-only buyer would pass on. SCTB14 is a black box, unpriceable from public data. Only the long tail of phase-1 and preclinical trispecifics is genuinely picked-over.

What's still blind — and the risk nobody's pricing

The class turns on a short list of unread readouts: to the left of the graph is what each watched asset has disclosed; to the right, the trial that would decide it — still blind.

And the risk almost nobody prices. Partnering rates fall from about 48% at second-or-third into a mechanism class to about 30% at ninth-or-later — a headwind, not a closed window. The bigger issue is that these 19 assets are not 19 independent shots: they share one mechanism, so they tend to fail together, not separately. The shared PD-(L)1×VEGF pathway alone puts whole-class failure in 1L NSCLC at about 8.1%; layer on the China-to-US translation risk that runs under every name — most of the class has only China data — and the model lands at 9.0% (8.1%–10.0%). That correlated tail is the one risk a multi-asset position doesn't diversify away.

The crowding is not reassurance: because every asset shares one mechanism, "the class is so busy that someone will make it work" is nearly the same statement as "this bet works." The field's size does not de-risk the single name you back; price that shared-failure risk into whichever asset you choose.

The shortlist, and what would decide each one

Everything above collapses to four names and one question apiece — each a reachable asset with disclosed 1L NSCLC data or the highest value-of-next-move.

  1. CS2009highest disclosed ceiling — and the biggest open question
    known todayPD-1×VEGF×CTLA-4 trispecific · 1L NSCLC ORR 81.3% — PD-L1-high (TPS≥50), monotherapy, single-arm
    what would decide itdoes 81.3% survive randomization and hold outside PD-L1-high — the first global Ph3 MRCT (planned end-2026)
  2. SCTB14option on a blinded pivotal readout
    known todayno disclosed efficacy — carries the highest value-of-next-move in the engine
    what would decide itits blinded pivotal Ph3 vs pembrolizumab (NCT07362459) — the single highest-information unknown in the class
  3. IMM2510available today, data in hand, rights just reverted
    known today1L NSCLC ORR 62.0% (+ chemo, single-arm); later-line post-checkpoint NSCLC data is resilience, not a new segment
    what would decide itwhether the 1L NSCLC + chemo signal holds in a controlled trial — the Jan-2026 hand-back was a strategic wind-down by Instil, not a disclosed verdict
  4. JS207furthest down a US regulatory path
    known today1L NSCLC ORR 58.1% — PD-L1-positive, monotherapy, single-arm
    what would decide itits confirmatory Ph2/3 vs nivolumab in neoadjuvant NSCLC — the first PD-1×VEGF in a registrational study

And two things that move the whole thesis rather than any single name:

  • China-to-US translatability — most of the class is China-originated and China-run, and the one randomized dataset (HARMONi-2) is a China trial — every China-only ORR inherits this discount until global randomized data lands
  • the correlated whole-class failure tail — 9.0% (8.1–10.0%) that the shared mechanism fails in 1L NSCLC — these assets share one PD-(L)1×VEGF pathway, and most of the data is China-only and may not translate to a US approval. The buyer's risk to hold, and it does not diversify across the names

References

  1. Summit Therapeutics — SEC filing (Summit BLA submission) ↩
  2. HARMONi-2, PFS HR 0.51 — The Lancet ↩
  3. Merck–LaNova exclusive global license for LM-299 (Merck newsroom, Nov 14 2024) ↩
  4. Pfizer / 3SBio exclusive licensing agreement ↩
  5. BMS & BioNTech global strategic partnership (BNT327) ↩
  6. AbbVie & RemeGen exclusive licensing agreement ↩
  7. OncoC4 AI-081 Ph2 dosing PR (primary; fetched & confirmed 'fully owned') ↩
  8. B1962 Ph1 trial registry ↩
  9. Biocytogen BCG036 program page (developer/design/stage); platform out-licensor ↩
  10. Kelun-Biotech SKB118/CR-001 China IND + Crescent collaboration terms ↩
  11. CStone ASCO 2026 CS2009 PR (primary; fetched & confirmed) ↩
  12. peer-reviewed: DR30206 = PD-L1×VEGF×TGF-β trifunctional fusion protein; Zhejiang Doer Biologics (fetched & confirmed targets + developer) ↩
  13. ClinicalTrials.gov API NCT06934616 (primary registry; fetched & confirmed Genor, Ph1 FIH, trispecific) ↩
  14. JCO 2025 abstr 5602 HB0025 endometrial Ph2 ↩
  15. HLX37 Ph1 trial — Henlius, PD-L1×VEGF (ClinicalTrials.gov NCT07274813) ↩
  16. Instil/Axion discontinues AXN-2510; ex-China rights revert to ImmuneOnco ($35M paid), Jan 6 2026 — FierceBiotech ↩
  17. Junshi JS207 FDA IND PR (primary) ↩
  18. Minghui MHB039A Ph1 PR (primary; fetched & confirmed); Qilu China deal per FierceBiotech ↩
  19. SCTB14 Ph2/3 pivotal 1L NSCLC vs pembrolizumab — Sinocelltech (ClinicalTrials.gov NCT07362459) ↩
  20. ClinicalTrials.gov NCT06600022 — SCTB41 Ph1/2 monotherapy, sponsor Sinocelltech (fetched & confirmed drug name + stage + sponsor; registry silent on molecular targets) ↩
  21. In NSCLC immunotherapy trials, response rate and PFS correlate only moderately with overall survival — FDA pooled analysis (Lancet Oncology, 2024) ↩
  22. FierceBiotech on ImmuneOnco IMM2510 1L NSCLC+chemo Ph2 (baseline ref [20]) ↩
  23. JCO ASCO 2025 abstr 8543 SSGJ-707 monotherapy advanced NSCLC (fetched 2026-07-05) ↩
  24. PR Newswire / Antengene x Junshi, ESMO Asia 2025 JS207 1L PD-L1+ NSCLC mono (baseline ref [26]) ↩
  25. Sino Biopharm / LaNova AACR 2026 preliminary MK-2010/LM-299 (fetched 2026-07-05; UNCONFIRMED ORR — flagged) ↩
  26. bevacizumab, ECOG E4599 (comparator reference) ↩
  27. Ivonescimab (AK112) design paper — Fc-silent (L235A/L236A) tetravalent bispecific (Zhong et al., iScience 2024) ↩
  28. Ivonescimab PK — terminal half-life 5.0–7.3 days (Chinese Ph1, PMC11925807) ↩
  29. Combination checkpoint inhibitors for NSCLC — dual anti-CTLA-4 and anti-PD-1/PD-L1 (review, PMC7048109) ↩
  30. Therapeutic targeting of VEGF and/or TGF-β to enhance anti-PD-(L)1 therapy (Frontiers in Oncology, 2022) ↩
  31. The difficulty in translating combined TGF-β and immune-checkpoint inhibition to the clinic (eBioMedicine, 2022) ↩
  32. Junshi / GlobeNewswire AACR 2026 JS207+JS007 1L HCC (baseline ref [24]) ↩
  33. BMS press release, global interim Ph2 pumitamig (BNT327) 1L ES-SCLC (fetched 2026-07-05) ↩
  34. JCO 2023 abstr 2589 HB0025 FIH monotherapy (baseline ref [38]) ↩
  35. IMM2510 Ph2 1L NSCLC/TNBC combination trial (NCT06746870) — ClinicalTrials.gov ↩
  36. Sintilimab ORIENT-11 FDA ODAC (Feb 10 2022) — voted 14-1 that China-only single-country data required additional US/multiregional data ↩