Ivonescimab's Western OS hazard ratio falls to 0.76 — Summit hasn't yet disclosed the number the FDA said it requires

What happened

On July 22, 2026, Summit Therapeutics disclosed in an SEC 8-K an updated overall-survival analysis from the global Phase III HARMONi trial of ivonescimab plus platinum-doublet chemotherapy, reporting a Western-patient OS hazard ratio of 0.76 that it called consistent with Asian patients ^[1]. In the June 2026 data cut-off, that same 0.76 hazard ratio appeared in both the full intention-to-treat population and the Western subgroup, with Western median follow-up now at 23.2 months while the Asian arm stayed locked at the 32.7-month primary-analysis mark; the June 2026 analysis reported 219 patients per arm in the ITT population — 438 in total, above the 420 enrollment figure the trial registry lists ^[2][6].

The number matters because of where it started. At the April 2025 primary OS analysis, ivonescimab + chemo missed statistical significance: HR 0.79 (95% CI 0.62–1.01; p=0.057), with median OS of 16.8 vs 14.0 months and Western follow-up of just 9.2 months — shorter than the median OS itself ^[1]. A September 2025 re-read with Western follow-up extended to 13.7 months moved the HR to 0.78 (95% CI 0.62–0.98; nominal p=0.0332), though median OS was unchanged in both arms ^[1]. The July update completes a three-step trajectory — 0.79 → 0.78 → 0.76 — each step bought with more Western maturity ^[1][2].

The number Summit hasn't published

Here is the discipline point. Summit did not disclose median OS, confidence intervals, or a p-value for the June 2026 analysis, deferring the detail to an upcoming medical conference ^[1]. Endpoints News flagged exactly this — the two-year Western survival figure landed without the context needed to judge it ^[8]. Fierce Pharma similarly reported the same day that Summit updated the survival data ahead of its FDA decision date ^[9]. Prognyx therefore cannot confirm from public sources whether the June 2026 HR of 0.76 clears the statistical-significance bar the FDA said it requires.

That bar is explicit and it is not rhetorical. At the May 2025 topline, the FDA told Summit that a statistically significant OS benefit is necessary to support marketing authorization in this setting, and noted that no current FDA-approved regimen carries a statistically significant OS benefit in this population ^[3]. The September 2025 p-value of 0.0332 was the company's own "nominal" figure — a descriptor that ordinarily signals a result sitting outside a trial's pre-specified, alpha-controlled hierarchy — and the June 2026 release carries no p-value at all ^[1]. The direction is right. The formal statistical status of the co-primary OS endpoint remains unconfirmed.

Hold both facts at once: the Western point estimate improved and converged on the ITT estimate as events accrued, and the single metric the FDA said it requires has not been made public.

Why it matters

The question hanging over ivonescimab since May 2025 has been whether Akeso's China-validated asset — engineered as a tetravalent PD-1/VEGF bispecific, with over 60,000 patients treated commercially in China per Akeso — travels to Western regulators ^[4]. HARMONi enrolled roughly 38% Western patients, and the fear was that the OS signal was carried by mature Asian data while immature Western follow-up masked a weaker Western effect ^[3]. The July update is Summit's answer to that specific worry: as Western follow-up more than doubled, the Western HR moved toward the ITT HR rather than away from it, so the falsifiable concern — that ivonescimab helps Asian but not Western patients — did not show up in the point estimate ^[1][2]. Prior single-region evidence pointed the same way, with the HARMONi-A study having shown an OS HR of 0.80 at 52% data maturity ^[3].

HARMONi already met its other primary endpoint decisively — PFS HR 0.52 (95% CI 0.41–0.66; p<0.00001) at the primary analysis ^[3] — and the June 2026 safety read was described as acceptable and manageable with no additional safety signals ^[1], consistent with the topline Grade ≥3 TEAE rate of 56.9% vs 50.0% and fatal TEAEs of 1.8% vs 2.8% ^[3]. Efficacy direction and safety are not the debate. Statistical significance on OS is.

Who is exposed, who gained

Summit is binary. The BLA for 2L+ EGFRm NSCLC carries a PDUFA goal action date of November 14, 2026, the updated OS results have been made available to the FDA, and the addressable US population is roughly 14,000 patients a year; Summit entered 2026 with about $710 million in cash ^[1][4]. One regulatory decision moves the whole equity.

Akeso (HKEX: 9926.HK) gained optionality. It engineered ivonescimab (SMT112 in Summit territories, AK112 elsewhere) and, under the January 2023 license, retains China and the rest of the world while Summit holds the US, Canada, Europe and Japan ^[4][5]. A Western OS win validates the molecule everywhere it is sold.

The PD-1/VEGF bispecific class read-through is the wider prize. Ivonescimab is positioned as potential first-in-class ^[4], and — per the FDA's own statement — an OS-significant approval here would be the first FDA-cleared regimen with a statistically significant OS benefit in this setting ^[3]. That read-through is real only if 0.76 proves significant, not merely favorable.

Competitors in EGFRm 2L+ NSCLC should watch, though no direct rival regimen is disclosed in current public sources. The nearest adjacent signal is the MSK-sponsored Phase 1/2 NCT07535437, now recruiting, which pairs ivonescimab with Dato-DXd or osimertinib — read-through, not head-to-head threat ^[7]. The bigger collision sits upstream in HARMONi-3, which tests ivonescimab + chemo against pembrolizumab + chemo in 1L NSCLC; no data have read out ^[4].

What to watch — with dates

• November 14, 2026 — PDUFA goal action date. The binary event ^[1].
• An upcoming, unnamed medical conference — where the detailed June 2026 numbers (median OS, CIs, and the p-value that settles significance) are due ^[1][8].
• HARMONi-3, HARMONi-7, HARMONi-GI3 (CRC) readouts — part of the broader Phase III program, no dates disclosed ^[4].
• Registry lag — ClinicalTrials.gov NCT06396065 still shows ACTIVE_NOT_RECRUITING and has_results=false as of the March 24, 2026 snapshot; the 8-K is more current than the registry for the OS update ^[6].

The now-what

For a biotech operator with an EGFRm NSCLC program or a PD-1/VEGF bispecific in development, three options are live:

1. Bank the maturity lesson. A regional subgroup with immature follow-up can converge on the ITT estimate as events accrue — Western-heavy confirmatory designs should price that follow-up dependency into powering and timelines up front, rather than reading an early regional miss as a real effect gap.
2. Model both November worlds. An OS-significant approval sets a stat-sig OS precedent in 2L+ EGFRm that lifts your own regulatory bar; a delay or complete response letter keeps the setting open with no OS-positive incumbent. Position your program for both.
3. Gate PD-1/VEGF diligence on the conference numbers. The class thesis hinges on whether 0.76 is significant, not just directionally clean — hold any BD read-through until the median, CIs and p-value are on a slide.

Verdict: threat level moderate and rising for anyone in 2L+ EGFRm NSCLC, but unresolved. The point estimate moved the right way; the significance question — the one the FDA said it requires — is still undisclosed. Window to act: before the November 14, 2026 PDUFA and the conference disclosure that should precede it.