EA2251: Summit taps ECOG-ACRIN to rerun a first-line biliary tract cancer OS trial in the US, supplying the US data the FDA wants for HARMONi-GI1

Data as of 2026-10-26 (Taipei time). [Fact] statements are sourced; [Inference] statements are this article's judgment.

1. One-line conclusion

[Fact] On 2026-10-21 Summit announced a collaboration with ECOG-ACRIN to run EA2251, a registration-enabling Phase 2/3 trial in the US comparing ivonescimab plus GemCis with pembrolizumab or durvalumab plus GemCis, with OS as the primary endpoint (Summit). [Inference] Summit is effectively conceding that China's HARMONi-GI1 alone is not enough for a US filing. For sponsors, this models a "read out in China first, add evidence through a US cooperative group" path, at the cost of waiting at least two more years (ClinicalTrials.gov).

For sponsors: when planning a US filing on Chinese data, budget the time and cost of a US OS confirmatory trial from day one (Summit).

2. What happened

[Fact] EA2251 is an open-label Phase 2/3 trial that plans to enroll in the US, randomizing patients to ivonescimab plus gemcitabine/cisplatin or pembrolizumab/durvalumab plus gemcitabine/cisplatin; the primary endpoint is OS; site activation is expected to begin in Q4 2026; the lead investigator is Deirdre J. Cohen of Mount Sinai (Summit). ClinicalTrials.gov lists the status as "not yet recruiting", with an estimated 336 patients, OS followed for up to 3 years, and estimated primary completion on 2028-12-31 (ClinicalTrials.gov). HR, CI and p-values: the trial has not started, so there are no efficacy data.

[Fact] Summit calls EA2251 the second Phase 3 study of ivonescimab in BTC, the first being HARMONi-GI1, and states that HARMONi-GI1 is a single-region study conducted in China, with all data generated, managed and analyzed by Akeso (Summit).

For sponsors: a registration trial run by a US cooperative group is one workable way to fill the single-country data gap, but its design (open-label, a choice of two comparators) will directly affect FDA acceptance.

3. Why it is designed this way (and the risks)

[Inference] Allowing pembrolizumab or durvalumab as the comparator matches current US practice and answers ODAC's criticism that ORIENT-11's control arm "did not reflect the US standard of care" (OncLive). Choosing OS as the primary endpoint also echoes the FDA briefing statement that "all approvals of first-line immunotherapy-based regimens were based on OS"; but that statement concerned metastatic NSCLC (OncLive), and BTC has no equally explicit precedent, so this is only an analogy.

[Inference] Risk one: sample size. The estimated 336 patients (ClinicalTrials.gov) is about half of HARMONi-GI1's 682 (CancerNetwork); if the true HR is near 0.75, a 336-patient Phase 3 will struggle to reach significance, and it may need to lean on a seamless Phase 2 portion or a combined argument with HARMONi-GI1, but we could not find the specific statistical design. Risk two: the open-label design matters little for OS but does affect PFS and patient retention. Risk three: the etiologic mix of BTC in the US differs from China; if the effect shrinks, that itself shows the limits of extrapolation.

[Inference] Why a cooperative group: the ECOG-ACRIN network includes more than 1,400 hospitals and cancer centers (Summit). BTC is a rare cancer, a sponsor opening its own sites enrolls slowly, and a cooperative group can reuse an existing network; the cost is less sponsor control over timelines and data.

For sponsors: a confirmatory trial that supplies US data should use OS as the primary endpoint and the current US standard as the comparator, and before launch it should check whether the sample size can detect a smaller effect.

4. Regulatory and commercial implications

[Fact] The FDA issued a CRL for ORIENT-11 because it was conducted only in China, and recommended a multiregional trial with an OS endpoint (Lilly); the ODAC vote was 14 to 1 (OncLive). [Inference] If EA2251 succeeds, the FDA can weigh HARMONi-GI1 (Akeso) together with EA2251, a pairing of "one large foreign trial plus one US confirmatory trial". Both use OS as the primary endpoint, which is far stronger than ORIENT-11 with PFS alone. In China, HARMONi-GI1 has been proposed for NMPA priority review (Akeso), so a China launch will come well before the US. Commercially, Summit's US BTC timeline therefore lags China until at least after 2028 (ClinicalTrials.gov).

For sponsors: the US competitive window in first-line BTC lasts at least until the end of 2028 (ClinicalTrials.gov); sponsors running BTC trials in the US can keep PD-(L)1 plus GemCis as the comparator during that period.

5. Recommendations for the next trial and OncoLattice

[Inference] First, for assets that Chinese biotechs license to the West, the licensor should add US or European sites when designing the Chinese Phase 3; otherwise the buyer (such as Summit) has to run another OS trial, as with EA2251 (Summit). Second, if a cooperative group is used, the sponsor should agree in the contract on data access, interim analysis timing and the data format for filing. Third, OncoLattice can help clients estimate US BTC enrollment speed: 336 patients (ClinicalTrials.gov) is mid-sized for a rare cancer, and the geographic spread of activated sites and referral networks is the key.

For OncoLattice: EA2251 is the template for "single-country data plus a US confirmatory trial"; clients planning global development should budget the cost and time of that confirmatory trial up front.

6. Open questions and what to watch

• Statistical design (Phase 2 to Phase 3 transition criteria, target HR, interim analyses): not found; watch updates to NCT07765433 and ECOG-ACRIN's trial description.
• First patient enrolled: Summit says site activation begins in Q4 2026 (Summit); watch for NCT07765433 to change to "recruiting".
• Primary completion: estimated 2028-12-31 (ClinicalTrials.gov).
• The etiologic distribution in the full HARMONi-GI1 paper will determine whether EA2251's effect-size assumptions are reasonable (Akeso).

The next thing to watch is when NCT07765433 moves from "not yet recruiting" to "recruiting", which sets the earliest date for Summit's US BTC filing.

Evidence limitations

EA2251 has not started enrolling, so every judgment here about its chance of success is inference; the registered estimate of 336 patients may change (ClinicalTrials.gov).

Advice for readers: until EA2251's statistical design is public, do not assume it will read out before 2028; keep the US BTC time window wide when planning competing programs.

Note: differences from the original relay

The relayed content (US Phase 2/3, ivonescimab plus GemCis vs PD-(L)1 plus GemCis, OS primary endpoint, NCT07765433) matches the sources. Additions: the estimated 336 patients and primary completion of 2028-12-31 come from ClinicalTrials.gov; Summit's press release does not give a sample size.