Witfire Elite  ·  Oncology Risk Brief  ·  FDA Complete Response

The FDA signed its rejection of ivonescimab on 13 November and put the letter online five days later. Summit’s announcement gave one reason. The letter gives four, and the first is about where the patients were enrolled.

The ivonescimab complete response letter has twelve numbered items, and the first one turns on a fraction. HARMONi, the trial behind the application, was called positive on 275 progression events. The FDA counted 257 of them in patients enrolled in China.

Summit Therapeutics announced the rejection on Sunday 15 November. Its release gave a single reason: overall survival had not reached statistical significance. That is accurate, and it is the letter’s second point. The first point, the geography, is not in the release, and neither is the remedy the FDA wrote down.

Both documents are public now. So are the trial registry and Summit’s own slides, and between them they show how a global trial ended up with 93% of its events in one country.

At a glance
4
clinical deficiencies in the FDA letter · Summit’s release names 1
257 / 275
progression-free survival (PFS) events in patients enrolled in China · FDA letter
18
PFS events outside China at the primary analysis · computed
7.9%
of patients behind five positive Phase 3 trials enrolled outside China · computed

01The ivonescimab complete response letter lists four deficiencies

The letter is dated 13 November 2026, a day before the review deadline, and its signature is time-stamped 2:08 pm. Summit told the market two days later, so 15 November is the date many readers will have seen. We follow the letter. The FDA’s public record of it was updated on 18 November, and that is how the full text came out.

One decision, two documents
Summit release, 15 Nov
Reasons given  1
Events by country  not stated
Earlier FDA meetings  not stated
Remedy  not stated
FDA letter, 13 Nov
Clinical deficiencies  4
Events in China  257 of 275
Meetings cited  Mar 2023, May 2025
Remedy  new multiregional trial

In the agency’s order, the four clinical deficiencies are these.

  1. The PFS result cannot be assumed to hold for American patients, because 257 of the 275 events came from China.
  2. The gain in median PFS was modest, the final survival analysis was not significant, and response rates were not much different.
  3. The longer follow-up Summit submitted is exploratory, and its survival figures compare regions at different cut-off dates.
  4. The PFS gain does not outweigh the added toxicity of ivonescimab on top of chemotherapy.

Items five to twelve are the standard request for a safety update. Unlike the letter that stopped Unicycive’s OLC, this one has no manufacturing content at all. That fits what Akeso’s chief executive, Michelle Xia, told investors on 15 November: the trial sites and the Guangzhou plant had passed FDA inspection.

Summit had warned about survival. Its third-quarter release in early November said the FDA had noted that “a statistically significant OS benefit is necessary to support marketing authorization” in this setting. The enrolment objection is a separate matter. The letter says the agency raised it at a Type A meeting in March 2023. It repeated the point in May 2025 at the pre-BLA meeting, the one held before filing. Summit’s 15 November release mentions neither meeting.

02257 of 275: where the events came from

93%of PFS events at the primary analysis were in patients enrolled in China (257 of 275, FDA letter)

A progression-free survival result is counted in events: a scan that shows growth, or a death. HARMONi’s primary PFS analysis had 275. Ivonescimab plus chemotherapy cut the risk of progression or death by 48%, a hazard ratio of 0.52, and lifted median PFS from 4.4 months to 6.8.

The letter attributes 257 of those events to China. The rest is subtraction.

Derivation
PFS events, primary analysis  275  (129 + 146, Summit slides; FDA letter)
Events in China  257  (FDA letter)
275 − 257  =  18 events in North America and Europe
18 / 275  =  6.5% of events
Western patients  165 / 438  =  37.7% of those randomised
37.7% of the patients supplied 6.5% of the events

Summit’s slides show why. The primary PFS analysis did not include all 438 patients. It included 345. All 273 Asian patients had been randomised by November 2022, six months before the first Western site opened, so every one of them is inside the 345. That leaves 72 places for Western patients, fewer than half of the 165 who eventually enrolled.

Asian patients in the primary PFS populationAll randomised by November 2022 (Summit slides)
273
Western patients in the primary PFS population345 − 273 (computed)
72of 165 enrolled
Asian patients with a PFS event257 / 273 (computed: FDA event count over Summit’s cohort size)
94.1%
Western patients with a PFS event18 / 72 (computed)
25.0%

By the time the trial was called positive, more than nine in ten patients in the Asian cohort had progressed or died. Three in four of the Western patients in the analysis had not yet had an event.

The letter puts China at 60% of the study population. Summit’s slides give 273 of 438, or 62.3%. Our arithmetic uses the slide counts.

Summit later ran the analysis with all 438 patients. It gave a hazard ratio of 0.57 overall, 0.55 in Asia and 0.67 in North America and Europe, where the upper end of the interval sits at 1.00. The FDA classes that follow-up as exploratory.

03Two registry records, 464 days apart

ClinicalTrials.gov shows how HARMONi was put together, because it is two records. Chinese patients went in under NCT05184712, an Akeso trial that started on 25 January 2022 and enrolled 322 people. Summit sponsors the global record, NCT06396065, which starts on 4 May 2023.

464 daysfrom the start of the China record to the start of the global one (computed from registry dates)

That global record lists 74 sites in six countries. Forty-three are in the United States, 13 in Spain, six in Canada, five each in France and Italy and two in the United Kingdom. None is in China or anywhere else in Asia, where 273 of the 438 patients were enrolled.

In regulatory language, the letter describes the same thing: a study designed for China alone and widened afterwards. Its dates, with those from the registry and Summit’s slides, give this sequence.

Jan 2022
China trial starts. Akeso’s NCT05184712 opens on 25 January.
Nov 2022
China enrolment closes. 273 of these patients become HARMONi’s Asian cohort.
Mar 2023
Type A meeting. The FDA discusses its concerns about the enrolment approach.
May 2023
Western enrolment starts. The global record opens on 4 May.
Oct 2024
Last Western patient. 165 enrolled across North America and Europe.
Apr 2025
Primary survival analysis. Hazard ratio 0.79, p = 0.057. Asian follow-up ends here.
May 2025
Pre-BLA meeting. The FDA reaffirms the enrolment concern.
14 Nov 2025
Application received. The review clock runs to 14 November 2026.
13 Nov 2026
Complete response letter signed.

Look at the order. The FDA put its concern about enrolment on record in March 2023. Western enrolment began two months later. In May 2025 the agency said it again, and the application arrived that November.

Xia’s account on the Akeso call was that the FDA had agreed to the sequential design while saying from the start that survival would have to be significant. She called it “the most disappointing moment” for her personally in Akeso’s 14 years. Her account and the letter do not contradict each other. A regulator can allow a trial to run and still say what it will not prove.

04The survival update runs on two clocks

Overall survival was HARMONi’s other primary endpoint, and it missed. At the April 2025 analysis the hazard ratio was 0.79, with a 95% interval of 0.62 to 1.01 and a p-value of 0.057. Median survival was 16.8 months with ivonescimab and 14.0 without, a gap of 2.8 months.

Summit kept following the Western patients and reported twice more. By June 2026 the hazard ratio was 0.76, and the interval, 0.61 to 0.95, no longer reached 1. Summit presented that figure in September as evidence of a consistent survival benefit.

The FDA’s third deficiency is about how that figure is assembled. Chinese sites closed during 2025, so follow-up of the Asian patients stops at April 2025. Western patients were followed to June 2026. So the updated number joins one cohort frozen in time to another that kept moving for 14 more months.

Only the top row was pre-specified, and its interval crosses 1. The two Western-only estimates, the ones closest to an American patient, cross it as well: 0.53 to 1.32 in September 2025 and 0.52 to 1.10 in June 2026.

Data gap The letter gives no cut-off dates and no count of deaths by region. The dates here come from Summit’s presentations. How many Western deaths stand behind the June 2026 estimate has not been published.

052.4 months gained, 1.85 times the serious adverse events

The fourth deficiency is a judgement, and the letter makes it in a sentence: the PFS gain does not outweigh the risk. It names serious immune-related toxicity and toxicity from VEGF inhibition, and gives no rates. The rates are on Summit’s September 2025 safety slide, with 218 patients in each arm.

Treatment-related adverse events: ivonescimab plus chemotherapy vs chemotherapy

Grade 3 or higher109 vs 92 patients
50.0%vs 42.2%
Serious61 vs 33 patients · ratio 1.85 (computed)
28.0%vs 15.1%
Grade 3 or higher, immune-related21 vs 13 patients
9.6%vs 6.0%
Grade 3 or higher, VEGF-related16 vs 7 patients
7.3%vs 3.2%
Led to discontinuation16 vs 11 patients · quoted in Summit’s release
7.3%vs 5.0%
Led to death4 vs 5 patients · quoted in Summit’s release
1.8%vs 2.3%

Summit’s release quotes the bottom two rows, where the arms look alike. Xia cited the same two when she said the FDA had recognised the safety profile. The letter rests on the rows above them. Serious treatment-related events reached 61 patients on ivonescimab and 33 on chemotherapy alone, a ratio of 1.85.

On the other side of the scale is the benefit: 2.4 more months of median PFS, and a survival difference that did not pass its test. The FDA drew a similar line between a clinical signal and proof when it turned down Replimune’s RP1.

06What the FDA asked for, and what it left open

The recommendation is short. Run a new, adequate and well-controlled multiregional trial, in line with ICH E17 and the agency’s September 2024 draft guidance on oncology programmes run across regions. Meet the FDA before it starts.

The first analyst notes were written on Sunday night from Summit’s summary. BMO Capital Markets concluded that survival matters most and PFS is “a nice to have”, BioSpace reported. The letter, published three days later, is less absolute. It asks for a statistically significant and clinically meaningful gain in “overall survival, progression-free survival or another endpoint that measures clinical benefit”, large enough to outweigh the toxicity.

So PFS is still on the list. The PFS in this application was 2.4 months wide, 93% Chinese by events, and came with more serious side effects.

There is a precedent, and it was stricter. In February 2022 the FDA’s advisers voted 14 to 1 that sintilimab, a PD-1 antibody tested only in China, needed more evidence. The letter that followed asked for a multiregional trial with overall survival as its endpoint, according to Lilly. The ivonescimab complete response letter does not narrow the endpoint that way.

Europe has just read the sintilimab data differently. On 15 November, the day Summit announced its letter, Innovent said the European Medicines Agency’s CHMP committee had recommended approval on ORIENT-11, the China-only study the FDA declined. Europe is Summit’s territory for ivonescimab. Its third-quarter release mentions no European filing.

07Five positive trials, 7.9% of the patients outside China

Summit’s chairman, Robert Duggan, answered the letter by pointing to five positive Phase 3 studies between Summit and Akeso. The registry lists 18 Phase 3 or Phase 2/3 trials of ivonescimab. Five have stopped recruiting and reported a positive result, and four of those are Akeso trials run in China.

HARMONi-A · NCT05184712Akeso · China · EGFR-mutant lung cancer after an EGFR inhibitor
322
HARMONi-2 · NCT05499390Akeso · China · vs pembrolizumab, PD-L1-positive lung cancer
398
HARMONi-6 · NCT05840016Akeso · China · vs tislelizumab plus chemotherapy, squamous lung cancer
532
Biliary tract · NCT06591520Akeso · China · vs durvalumab plus chemotherapy
682
HARMONi, Western cohort · NCT06396065Summit · North America and Europe
165
Unique patientsHARMONi’s 273 Asian patients counted once, under HARMONi-A (computed)
2,099165 outside China = 7.9%

Count each patient once and the five results rest on 2,099 people. Of those, 165 were enrolled outside China, and every one of them is in HARMONi.

HARMONi-3 is the first trial that changes the ratio. It was registered as a multiregional study in October 2023, against Keytruda plus chemotherapy in first-line lung cancer. Its record lists 260 sites in 17 countries: 76 in the United States, 46 in China, and the rest across Europe, Japan, Mexico, Turkey and Canada. China’s share of sites is 17.7% (computed). Enrolment is complete in both cohorts, 600 squamous and 1,000 non-squamous patients on the registry’s numbers.

Data gap Summit has not published how HARMONi-3’s patients split by region, and site counts are a poor proxy. HARMONi’s own record lists no Chinese site, yet 62% of its patients were enrolled in Asia. The regional split of events is the first figure to look for when the squamous data arrive.

08$2.63 billion in cash, and a partner 35.6% below its entry price

Summit can afford a second attempt. It ended September with $629.8 million in cash and investments, and AstraZeneca’s $2.0 billion equity investment closed on 5 October.

Derivation
Cash and investments, 30 Sep 2026  $629.8M  (reported)
AstraZeneca equity, closed 5 Oct 2026  $2,000.0M  (reported)
Pro forma  =  $2,629.8M  before fourth-quarter spending
Operating cash used, nine months  $423.2M  (reported)  =  $141.1M a quarter
2,629.8 / 141.1 = 18.6 quarters at this year’s rate

That run-rate flatters, because spending is climbing. Non-GAAP operating expenses were $166.9 million in the third quarter, up from $103.4 million a year earlier.

AstraZeneca’s position is the uncomfortable one. Its preferred shares convert at $18.36. Summit closed at $11.82 on Monday 16 November, down 16.4% on the day, and Fierce Pharma reported a further fall of about 10% on Tuesday.

AstraZeneca conversion price108,955 preferred shares, 1:1,000 · closed 5 October
$18.36
SMMT close, 16 NovemberDown 16.4% on the day
$11.82
Gap to AstraZeneca’s price11.82 / 18.36 − 1 (computed)
−35.6%
Mark on the $2.0bn, 42 days after closing2,000 × (1 − 11.82 / 18.36) (computed)
−$712M
Akeso (9926.HK) close, 16 NovemberIntraday low HK$56.10
HK$64.10−22.1%

Other large companies have money in the class too. Six licences and acquisitions, every one for a molecule discovered in China, carry $7.29 billion in upfront and guaranteed payments (computed). AstraZeneca’s equity brings the total to $9.29 billion. We traced the pattern behind those deals in The China Biotech Pipeline Grab.

Summit / Akeso · ivonescimabDecember 2022 · upfront
$500M
BioNTech / Biotheus · pumitamigAnnounced November 2024 · upfront
$800M
Merck / LaNova · MK-2010November 2024 · upfront
$588M
Pfizer / 3SBio · tivunatamigMay 2025 · upfront
$1,250M
BMS / BioNTech · pumitamigJune 2025 · $1.5bn upfront plus $2.0bn non-contingent through 2028
$3,500M
AbbVie / RemeGen · ABBV-1480January 2026 · upfront
$650M
Upfront and guaranteed, six dealsComputed · with AstraZeneca’s equity, $9,288M
$7,288M

The best-funded rival has not copied HARMONi’s design. BMS and BioNTech sponsor seven Phase 3 or Phase 2/3 trials of pumitamig on the registry. They plan 6,039 patients, and each lists between 161 and 366 sites. Their earliest primary completion date is January 2029.

So the letter’s first deficiency, aimed at a trial assembled in two halves, lands on Summit more than on the class. Its fourth, on toxicity, is a question every PD-1 and VEGF combination will face.

09India: twice the EGFR rate, and no site on the list

EGFR-mutant disease, the setting the FDA turned down, is a larger part of lung cancer in India than in the West. A pooled series of 3,436 Indian lung adenocarcinomas from four tertiary hospitals found EGFR mutations in 30.0% (Nakra et al., 2020, via PubMed). Tata Memorial’s 907-patient study cites 10–15% for North American and European patients (Chougule et al., 2013). On those figures the Indian rate is two to three times the Western one (computed).

Ivonescimab has no route to those patients today. India falls in Akeso’s licence territory. A drug review in the Indian Journal of Medical and Paediatric Oncology recorded no approval from the Drugs Controller General in June 2025. We found no later filing.

The short cut would have been Rule 101. In August 2024 India’s regulator named six jurisdictions whose approvals let it consider waiving a local trial. They are the United States, the United Kingdom, Japan, Australia, Canada and the European Union. The waiver is discretionary and covers five categories of drug. China is not on the list, so the approval ivonescimab has held there since May 2024 opens nothing in India. The United States is the first of the six where a decision is known, and the decision was no.

The site lists tell the same story from the other end. HARMONi names 74 sites and HARMONi-3 names 260. None is in India. If Summit runs the multiregional trial the FDA has described, it will need EGFR-mutant patients outside China. Indian hospitals see that mutation at no less than twice the Western rate. We made the wider case in India’s next CRO boom.

Data gap We have no sourced figure for how many Indian patients reach a third-generation EGFR inhibitor, which was the entry condition for HARMONi’s Western cohort. Without it, the size of the eligible Indian population is unknown.

10Witfire Risk Score: 6.0 / 10

The score covers Summit’s ivonescimab franchise in the United States after the letter.

Witfire Risk Score
6.0 / 10
Elevated
Regulatory Exposure · 30%
7.5
One indication now needs a new trial. The geography objection is specific to HARMONi. The benefit-against-toxicity test will be applied again.
Competitive Displacement · 25%
6.0
The first US label has slipped. Pumitamig’s seven global trials read out from 2029. HARMONi-3’s first cohort is due by year-end.
Capital Position · 20%
2.0
$2.63bn pro forma, 18.6 quarters at this year’s operating cash use.
Evidence Integrity · 15%
7.0
7.9% of patients behind five positive trials were enrolled outside China. The survival update spans two cut-off dates.
Execution & Credibility · 10%
8.0
Filed after two FDA meetings on enrolment. The release named one deficiency of four. No investor call by 17 November.

11What to watch

  1. By 31 December 2026. Summit expects the primary PFS analysis for HARMONi-3’s squamous cohort in the fourth quarter, with an early look at survival. Read the regional split of events before the hazard ratio. SMMT closed at $11.82 on 16 November against AstraZeneca’s $18.36, and Akeso (9926.HK) closed at HK$64.10.
  2. Summit’s investor call. The company said in early November it would hold one after the FDA acted. It had not by 17 November, Fierce Pharma reported. Citi says Summit is weighing a resubmission, and the letter asks for a new trial. Which of the two management describes will show how it has read the letter.
  3. First half of 2027. A survival interim for the squamous cohort that does not depend on PFS, and the PFS events for the non-squamous cohort.
  4. 13 November 2027. The one-year limit to resubmit or ask for an extension under 21 CFR 601.3(b), computed from the letter date. A Type A meeting request would be the first public sign of a plan.
  5. January 2029. The earliest primary completion date among the pumitamig trials: BioNTech’s first-line lung cancer protocol, 1,580 patients at 366 sites (BNTX, BMY).
Witfire Elite View

The letter is narrower than Sunday’s headlines and harder than Summit’s release. PFS remains an endpoint the FDA will accept. What it turned down was a PFS result that came 93% from China, measured 2.4 months, and carried nearly twice the rate of serious adverse events.

Summit’s deeper exposure is in the patient count. Across the five trials it calls positive, 7.9% of patients were enrolled outside China, and all of them are in HARMONi. HARMONi-3 is the first data set that can move that number, and its squamous cohort reports before the year ends.

The ivonescimab complete response letter has put a first American label out of reach for now. Summit has $2.63 billion to spend on a reply, and the regulator has now put in writing what the reply needs to show.

FAQIvonescimab and the FDA: five questions

Why did the FDA reject ivonescimab?

The ivonescimab complete response letter, dated 13 November 2026, lists four clinical deficiencies. First, 257 of the 275 progression-free survival events in the HARMONi trial came from patients enrolled in China, so the result could not be applied to US patients. Second, the PFS gain was modest and overall survival was not statistically significant (hazard ratio 0.79, p = 0.057). Third, the later follow-up analyses were exploratory. Fourth, the benefit did not outweigh the added toxicity.

What does the ivonescimab complete response letter ask Summit to do?

It recommends a new, adequate and well-controlled multiregional trial consistent with ICH E17. The trial must show a statistically significant and clinically meaningful improvement in overall survival, progression-free survival or another clinical benefit endpoint, large enough to outweigh the toxicity. Summit has one year from 13 November 2026 to resubmit or ask for an extension.

Does the FDA now require overall survival for PD-1/VEGF bispecifics?

The published letter does not say so. It names overall survival, progression-free survival or another clinical benefit endpoint as acceptable for a new trial. Summit has separately reported that the FDA said a statistically significant survival benefit was necessary in the HARMONi setting.

When is the next ivonescimab readout?

Summit expects the primary PFS analysis for the squamous cohort of HARMONi-3, with an early interim look at survival, in the fourth quarter of 2026. A separate survival interim and the non-squamous PFS analysis are planned for the first half of 2027.

Is ivonescimab approved in India?

No Indian approval has been recorded. India is in Akeso’s licence territory. India’s Rule 101 waiver list covers approvals from the US, UK, Japan, Australia, Canada and the EU. It does not include China, where ivonescimab has been approved since May 2024.

Method. Every figure here traces to a primary source. Where a news report and a filing disagree, this brief follows the filing and says so. Figures marked as computed are Witfire’s own arithmetic on reported numbers. Registry records were read on 19 November 2026.
Primary sources: FDA Complete Response Letter, BLA 761524, 13 November 2026 (openFDA record updated 18 November 2026) · Summit Therapeutics Form 8-K Exhibit 99.1, 15 November 2026 · Summit third-quarter 2026 results release · Summit HARMONi presentation, World Conference on Lung Cancer, September 2025, and updated survival release, 15 September 2026 · ClinicalTrials.gov records NCT05184712, NCT06396065, NCT05899608, NCT05499390, NCT05840016, NCT06591520 and the registered pumitamig Phase 3 trials · AstraZeneca releases, 28 September and 5 October 2026 · FDA draft guidance on oncology multiregional clinical development programmes, September 2024 · FDA Oncologic Drugs Advisory Committee minutes, 10 February 2022 · Innovent Biologics release, 15 November 2026 · CDSCO order under Rule 101, 7 August 2024 · Nakra T, et al. Curr Probl Cancer 2020;44(3):100561 · Chougule A, et al. PLoS One 2013;8(10):e76164 · Reporting by Fierce Pharma, Fierce Biotech, BioSpace, South China Morning Post, pharmaphorum, Caixin Global and Healio.

Disclosure: Editorial analysis, not investment or medical advice. The Witfire Risk Score is our own framework, weighting Regulatory Exposure (30%), Competitive Displacement (25%), Capital Position (20%), Evidence Integrity (15%) and Execution & Credibility (10%) on a 1–10 scale. Scores are judgments, not forecasts.
Witfire / Commissioned Analysis

Put this framework on your own file.

The same method used in these briefs, applied privately to a decision you are actually facing — a target, a partner, a facility, a filing.

  • CMC & manufacturing risk assessment
  • Regulatory / CRL readiness review
  • Due diligence on a target or partner
  • Investor brief or market analysis
  • Commissioned editorial
  • Something else

Replies come from the editor, usually within two working days. Nothing you send is published.

Witfire / Executive Briefing

The signal before the market catches up.

A low-noise briefing on the pharma events that change strategy, risk and value. No more than one a week.

Work email only. Unsubscribe in one click.