Clinical Development Critiques
Summit Q3: before the 11/14 PDUFA for ivonescimab, the company itself restates that “FDA wants statistically significant OS” — and HARMONi’s primary OS analysis missed
[Fact] HARMONi (EGFR-mutant nonsquamous NSCLC after EGFR-TKI) met PFS; the primary OS analysis HR 0.79 (p=0.057) was not statistically significant; the June 2026 update was OS HR 0.76 (95% CI 0.61–0.95; nominal p=0.0151). FDA has said this indication needs statistically significant OS; the PDUFA date is 2026-11-14. [Inference] A post hoc nominal p-value cannot replace the prespecified test; under standard review logic the chance of approval is low. The real test for ivonescimab is HARMONi-3’s squamous cohort PFS and early OS in Q4 2026.
Apex AI Institute / OncoLattice team42 min read
- NSCLC
- EGFR
- PD-1xVEGF
- Bispecific
- FDA
- PDUFA
- OS
Data as of 2026-11-05 (Taipei time). Direct fetches of the Business Wire original timed out twice in this round; content was cross-checked against full-text excerpts of the press release returned by search engines and the WCLC (World Conference on Lung Cancer) release on Summit’s website. [Fact] statements are sourced; [Inference] statements are this article’s judgment.
1. One-line conclusion
[Fact] Summit Therapeutics’ third-quarter 2026 results (released 2026-11-04 Eastern Time) stated that the BLA (Biologics License Application) PDUFA (Prescription Drug User Fee Act) target date for ivonescimab (a PD-1 × VEGF bispecific antibody; PD-1 is programmed cell death protein 1; VEGF is vascular endothelial growth factor) plus chemotherapy remains 2026-11-14; the primary OS (overall survival) analysis of HARMONi was not statistically significant, and “FDA (Food and Drug Administration) has noted that statistically significant OS benefit would be needed to support approval in this indication” (Summit Q3 press release).
[Inference] By restating that FDA sentence in its own earnings release, Summit is managing market expectations in advance. Under standard review logic, the chance of a CRL (complete response letter) on 11/14 is higher than the chance of approval; even if approved, the label is likely to be tightly restricted.
For sponsors: once FDA has said “OS is required,” there is almost no precedent for overturning a miss with a nominal p-value from longer follow-up; alpha for dual primary endpoints should be allocated so that OS has enough events from the start.
2. What happened
[Fact] HARMONi: Phase 3, multiregional, ivonescimab plus chemotherapy vs placebo plus chemotherapy, in EGFR (epidermal growth factor receptor)-mutant, locally advanced or metastatic nonsquamous NSCLC (non-small cell lung cancer) after an EGFR TKI (tyrosine kinase inhibitor); dual primary endpoints were PFS (progression-free survival) and OS; PFS was statistically significant (Summit Q3 press release).
[Fact] OS: the April 2025 primary analysis HR (hazard ratio) was 0.79 (95% CI [confidence interval] 0.62–1.01; p=0.057), median OS 16.8 vs 14.0 months; median follow-up in Western patients was then only 9.2 months (Summit WCLC 2026 press release). Updated analysis with a June 2026 cutoff: ITT (intention-to-treat) OS HR 0.76 (95% CI 0.61–0.95; nominal p=0.0151), Western subgroup HR 0.76 (95% CI 0.52–1.10); median follow-up was 23.2 months in Western patients, while Asian-patient data had been locked in April 2025 (median follow-up 32.7 months) (Summit WCLC press release, Summit Q3 press release).
[Fact] Other pipeline: HARMONi-3 (first-line metastatic NSCLC, ivonescimab plus chemotherapy vs pembrolizumab [Keytruda] plus chemotherapy, BioSpace) has completed enrollment in both the squamous and nonsquamous cohorts; the event count needed for the primary PFS analysis in the squamous cohort is expected in Q4 2026, with a prespecified early OS interim at the same time, and a separate OS interim independent of PFS expected in the first half of 2027; PFS events in the nonsquamous cohort are expected in the first half of 2027; HARMONi-7 (first-line NSCLC with high PD-L1 [programmed death-ligand 1] expression, monotherapy) is expected to complete enrollment in 2027 (Summit Q3 press release). AstraZeneca has made a $2 billion equity investment in Summit (BioSpace).
For sponsors: an update after a nonsignificant primary analysis on the same dataset is statistically descriptive only; plan “follow-up maturity at the primary OS analysis” as a design parameter rather than something to patch afterwards.
3. Why OS missed
[Fact] At the primary analysis, median follow-up in Western patients was only 9.2 months, shorter than the chemotherapy arm’s median OS of 14.0 months; by the updated analysis, median follow-up was 23.2 months in Western patients and 32.7 months in Asian patients (Summit WCLC press release). [Inference] Asian patients were enrolled and matured first, Western patients later, so the regional mix of OS events was imbalanced at the primary analysis; that is the typical cost of a “China first, then global” enrollment pattern in a multiregional trial.
[Inference] Subsequent therapy in the control arm may have diluted the OS gap. Later-line treatments after progression in the chemotherapy arm (including other immunotherapy or an ADC [antibody-drug conjugate]) can shrink the survival difference; Summit’s public materials reviewed for this article do not list subsequent-therapy rates, so this article cannot quantify that.
[Fact] Primary-analysis HR 0.79 (p=0.057); updated-analysis HR 0.76 (nominal p=0.0151) (Summit WCLC press release). [Inference] HR 0.76–0.79 is a moderate effect size; if the sample was sized for a larger effect, power was inadequate from the start. The updated analysis had no prespecified alpha; a nominal p-value is not a re-test, and FDA usually will not accept it as confirmatory evidence.
For sponsors: if a multiregional trial finishes Asia first, the protocol should prespecify a minimum median follow-up in each region at the primary OS analysis, or “regional consistency” will become a review dispute.
4. Regulatory and commercial implications
[Fact] In the Q3 release Summit argued benefit–risk on the grounds that “no currently FDA-approved option has demonstrated statistically significant OS,” while also restating FDA’s requirement of significant OS for this indication (Summit Q3 press release). [Inference] FDA has already set the bar, so that argument has limited force; the most likely outcomes are a CRL or a delayed decision.
[Fact] AstraZeneca has invested $2 billion in equity, and several companies are racing to pair with ivonescimab ahead of the FDA decision (BioSpace). On the competitive side, BioNTech’s third-quarter 2026 release states that pumitamig (BNT327/BMS986545, a PD-L1 × VEGF-A bispecific co-developed with Bristol Myers Squibb) currently has 7 registration-relevant trials under the ROSETTA clinical development program (original wording: sieben zulassungsrelevante Studien), with 5 new global trials started in 2026 covering triple-negative breast cancer, colorectal cancer, gastric cancer, and two NSCLC trials (unresectable stage III, and advanced patients with PD-L1 TPS [tumor proportion score] ≥50%) (BioNTech Q3 press release). [Inference] The industry already treats PD-(L)1 × VEGF as a potential immuno-oncology backbone; even if this HARMONi indication is set back, attention has shifted to first line.
[Inference] The post–EGFR-TKI standard of care will not change in the near term; whether first line moves from pembrolizumab toward PD-(L)1 × VEGF depends on global data from HARMONi-3 and ROSETTA, not HARMONi.
For sponsors: HARMONi has anchored the regulatory bar in the PD-(L)1 × VEGF class at “OS”; new entrants planning later-line or EGFR-mutant studies should make OS a primary or key secondary endpoint and size events accordingly.
5. Implications for the next trial design (OncoLattice)
[Fact] FDA’s September 2024 draft guidance on multiregional clinical development advises that oncology multiregional trials supporting a U.S. filing should ensure results apply to U.S. patients and U.S. medical practice, and should consider U.S. population representativeness, site selection, and regional differences in disease and treatment when planning (FDA MRCT draft guidance). At HARMONi’s primary analysis, median follow-up in Western patients was only 9.2 months (Summit WCLC press release). [Inference] When designing a multiregional Phase 3, simulate “regional enrollment curves + OS event accrual” so that, at the primary analysis, the Western subgroup has enough information to support a consistency judgment.
[Fact] HARMONi-3’s squamous cohort schedules an early OS interim at the same time as the primary PFS analysis, plus a separate OS interim independent of PFS in the first half of 2027 (Summit Q3 press release). [Inference] That shows Summit has already separated the OS and PFS time points; when PFS and OS are dual primaries, OS alpha and interim timing should be triggered by events, not by calendar time.
[Inference] EGFR-mutant disease is less common in the West; to support an FDA review, North American and European sites (or U.S. sites with a high Asian-ancestry share) with larger EGFR caseloads should be opened early, so that Western patients are not concentrated at the back of enrollment; subsequent-therapy data should also be prespecified and collected for sensitivity analyses.
For sponsors: a multiregional Phase 3 protocol should include a trigger of “minimum Western events or minimum follow-up before the primary OS analysis.”
For OncoLattice: HARMONi is a textbook case of “regional enrollment order determining OS maturity” and can be used directly as a demo scenario for an enrollment-simulation tool.
6. Open questions and what to watch
[Fact] FDA’s decision date is 2026-11-14; HARMONi-3’s squamous primary PFS result and early OS interim are expected in Q4 2026 (Summit Q3 press release).
[Inference] Public materials reviewed for this article do not give HARMONi-3’s regional mix or Western-patient share, which is the key to whether HARMONi-3 will repeat HARMONi’s problem. If 11/14 is a CRL, whether FDA requires a new trial and whether it will accept HARMONi-3 as additional evidence will also have to wait on company disclosure.
[Inference] The table below maps results that may appear after 2026-11-14 to concrete decisions for the next trial. The scenarios are this article’s inferences, not statements by the company or FDA; dates in the timing column all come from Summit’s Q3 release (Summit Q3 press release), and OS figures come from Summit’s WCLC release (Summit WCLC press release).
| Event | Expected timing | Source | Decision for the next trial |
|---|---|---|---|
| FDA decision on the HARMONi BLA | PDUFA target date 2026-11-14 | Summit Q3 press release | If it is a CRL because OS missed the prespecified significance bar, the next post–EGFR-TKI trial should make OS the primary endpoint (or the first key secondary in a hierarchical test) and size OS events on a moderate HR of about 0.76–0.79, not a larger effect; if it is an approval (whether or not the label is restricted), the next same-line trial’s control arm should prespecify a switch or stratification clause “if ivonescimab plus chemotherapy becomes an available option” |
| Regional-consistency comments in a CRL or approval document | Released with the FDA decision; public disclosure depends on Summit | Summit Q3 press release | If Summit discloses that FDA objected to insufficient Western follow-up (median 9.2 months at the primary analysis, Summit WCLC press release), the next multiregional Phase 3 should write into the protocol a trigger of “minimum Western events or minimum median follow-up before the primary OS analysis”; if no such comment is disclosed, keeping the condition is still advisable, but the minimum follow-up can be a sensitivity analysis rather than a trigger |
| HARMONi-3 squamous primary PFS analysis + early OS interim | Q4 2026 | Summit Q3 press release | If PFS is significant and early OS points the same way, the next first-line squamous NSCLC trial should expect that the control arm may move off pembrolizumab plus chemotherapy and should reserve a control-arm switch clause; if PFS is significant but early OS is unclear, the next trial should trigger the OS interim by events, to avoid repeating HARMONi’s immature follow-up |
| HARMONi-3 independent OS interim; nonsquamous PFS | Both first half of 2027 | Summit Q3 press release | If OS is significant, PD-(L)1 × VEGF will have first-line OS evidence, and the next first-line nonsquamous trial’s effect-size assumption should use that as the new control benchmark; if OS is not significant, the next trial should not treat PD-(L)1 × VEGF as an established backbone |
For sponsors: do not lock endpoints and sample size for the next post–EGFR-TKI or first-line PD-(L)1 × VEGF trial before 11/14; prepare two protocol versions along the branches above (a CRL version with OS as primary, an approval version that reserves a control-arm switch) and pick one within two weeks of the result.
Evidence limitations
[Fact] The Summit Q3 press release could not be opened directly this round; content comes from search-engine excerpts of the release, cross-checked against the WCLC release on Summit’s website (Summit Q3 press release, Summit WCLC press release). Pumitamig’s 7 trials come from BioNTech’s German-language release (BioNTech Q3 press release). This article does not have original FDA–Summit correspondence, and subsequent-therapy rates in HARMONi and HARMONi-3’s regional mix could not be found.
For sponsors: “significant OS is required” is currently only Summit’s retelling, with no public FDA document; sponsors planning a post–EGFR-TKI or other PD-(L)1 × VEGF trial should not wait for a public FDA statement before deciding. They should obtain a written reply at the next FDA meeting (for example a pre–Phase 3 meeting) on whether OS is an approval condition for that indication, and use the reply to decide whether OS is primary or a key secondary in a hierarchical test, and how many OS events to plan. Update again if 11/14 brings a public FDA explanation or Summit discloses the CRL.
Apex AI Institute / OncoLattice team