SMMT ivonescimab

EGFR-mutated lung cancer · Summit · BLA

Of 100 applications in this same position, about 45 get approved.

Until the FDA deadline40d

Base rate76%

Gap−31pts

PDUFA target date 2026-11-14the FDA may act earlier or later

Summary

We put approval in this review cycle at 45%, below the ~76% base rate for small/mid-cap original applications. The dominant factor is that the FDA told Summit a statistically significant overall-survival benefit is necessary in this setting, and HARMONi's prespecified OS analysis missed (HR 0.79, p=0.057). Against that: a clearly positive PFS result, a consistent OS trend with longer follow-up (HR 0.76, descriptive), significant OS in the China-only HARMONi-A study, and a PFS-based precedent in the same setting. Manufacturing in China adds secondary inspection risk. Confidence is low: this is a judgment call the FDA could make either way.

Recorded 2026-10-04 19:58 UTC · v1Proof ↓

Evidence · 6 items

Why 45%, not 76%

Indication

Ivonescimab + platinum-doublet chemotherapy, EGFR-mutated locally advanced or metastatic non-squamous NSCLC after a third-generation EGFR TKI (HARMONi)

76% of comparable applications were approved in the past: the opponent every forecast has to beat. Below are the reasons this forecast differs.

▲ supports approval, ▼ counts against it. More squares, more weight.

01▼ AgainstLarge FDA's stated approval bar was not met in the prespecified analysis

Summit's 10-K says the FDA noted that a statistically significant overall survival benefit is necessary to support marketing authorization in this setting. HARMONi's primary OS analysis missed it: HR 0.79 (95% CI 0.62–1.01), p=0.057.

Source: SEC 10-K filed 2026-02-23; 8-K 2026-07-22
02▲ SupportsMedium The totality of evidence is supportive: strong PFS, an OS trend that holds with longer follow-up, and a positive OS readout in the China-only study

PFS was statistically significant (primary endpoint). An updated June-2026 OS cut gives HR 0.76 in the ITT population and in western patients, but it is descriptive and does not restore the alpha spent. HARMONi-A (China) showed significant OS with HR 0.74. The FDA has historically discounted single-country China data.

Source: SEC 8-K 2026-07-22, 8-K 2026-09-15 (WCLC 2026), 10-K 2026-02-23
03▲ SupportsSmall The FDA accepted the BLA for filing despite the known OS miss

The BLA was submitted Q4 2025 and accepted January 2026, with no refuse-to-file. Acceptance means the file is complete, not that it is approvable.

Source: SEC 8-K 2026-09-15; OncLive / CancerNetwork BLA acceptance coverage
04▲ SupportsSmall A PFS-based precedent exists in the same post-osimertinib setting

Amivantamab + carboplatin/pemetrexed (MARIPOSA-2) was approved on PFS without mature OS. The FDA's drug-specific statement to Summit outweighs the general precedent. PD-1 agents have failed to show OS benefit in EGFRm NSCLC, which likely explains the FDA's higher bar.

Source: FDA approval coverage of MARIPOSA-2 (CancerNetwork, PMC11981408)
05▼ AgainstSmall Manufacturing in China and a pre-license inspection

Drug supply comes from partner Akeso in China. A pre-license inspection of the Chinese site is required. Summit's risk factors flag US–China trade risk and the BIOSECURE Act, signed December 2025. Other Chinese PD-1 antibodies (toripalimab, tislelizumab, penpulimab) were eventually approved; toripalimab's first cycle ended in a CRL that was partly inspection-related.

Source: SEC 10-K 2026-02-23 risk factors
06▼ AgainstSmall Standard review and no advisory committee announced about six weeks before the PDUFA date

An efficacy question this contested is often taken to ODAC. Deciding without one can cut either way; the lack of a priority review is mildly negative.

Source: No ODAC listing found in public sources as of 2026-10-04

What would change this forecast

01
The FDA schedules ODAC, or the company discloses late-cycle meeting outcomes or label negotiations
02
A statistically significant result from a prespecified updated OS analysis, or the FDA publicly accepting the updated OS data
03
A PDUFA extension (major amendment) would not change the forecast by itself, only its date
04
News of an inspection problem, or an import or BIOSECURE designation affecting Akeso

Any change is published as a new version; earlier versions stay exactly as they were.

Confidence

Low

Manufacturing and inspection risk

Medium

Biologic made by a Chinese partner, so a foreign pre-license inspection is needed, with policy risk from the BIOSECURE Act. Chinese PD-1 makers have since passed FDA inspections, so this is a secondary risk next to the efficacy question.

Record · version 1

Written on 2026-10-04, 41 days before the PDUFA target date.

Timeline

Versions

v1 2026-10-04 19:58 UTC 45% Current version JSON ↓.ots ↓ ○ Timestamp submitted, awaiting block confirmation

Fingerprint

SHA-256722fe54a3b333aeb3426a2bef4500d873aff14bcca0e190527a12b8c0e2586a2

The fingerprint of the v1 file. It is what the timestamp commits to: change one character of the file and the fingerprint no longer matches.

Verify it yourself

Download the JSON file above and the .ots file of the same name into one folder.
Compute the SHA-256 of the file. It should match the fingerprint above.
Verify with OpenTimestamps. Once the timestamp is in a Bitcoin block, ots verify reports the block time if you run a Bitcoin node; without one, drop both files on the verifier at opentimestamps.org, or run ots info to see the block height and look it up in any block explorer. If the block time is before the FDA’s decision, the file cannot have been written afterwards.

pip install opentimestamps-client
shasum -a 256 v1_20261004T195808Z.json
ots verify v1_20261004T195808Z.json.ots
ots info v1_20261004T195808Z.json.ots

Scoring rule

What is scored is the last version recorded at least 2 days before the PDUFA target date; if the FDA decides early, only versions recorded before the day of its decision count. For this forecast the cutoff is 2026-11-12. Its error (also called the Brier score) is the squared distance between the forecast and the outcome, set against the error of quoting the base rate only.