COGT bezuclastinib

GIST · Cogent · NDA

Of 100 applications in this same position, about 88 get approved.

Until the FDA deadline56d

Base rate76%

Gap+12pts

PDUFA target date 2026-11-30the FDA may act earlier or later

Summary

We put approval of bezuclastinib plus sunitinib for imatinib-pretreated GIST at 88%, above the ~76% base rate for small/mid-cap original applications. The randomized Phase 3 PEAK trial beat the active standard of care decisively (independently reviewed PFS 16.5 vs 9.2 months, HR 0.50, p<0.0001). The review has Breakthrough designation, RTOR and priority review, and the FDA said at filing that it saw no review issues and planned no advisory committee. Against that: the combination is more toxic (grade ≥3 treatment-related AEs 72% vs 52%, with more liver-enzyme rises), overall survival is immature, and this is Cogent's first commercial product, with a single-source API. Confidence is medium.

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

Evidence · 5 items

Why 88%, not 76%

Indication

Bezuclastinib in combination with sunitinib for locally advanced, unresectable or metastatic gastrointestinal stromal tumors (GIST) after prior imatinib (PEAK)

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▲ SupportsLarge A randomized Phase 3 trial against the active standard of care met its primary endpoint by a wide margin

PEAK: PFS by blinded independent central review was 16.5 vs 9.2 months, HR 0.50 (95% CI 0.39–0.65, p<0.0001), for bezuclastinib + sunitinib vs sunitinib alone in imatinib-resistant or intolerant GIST. ORR was 46% vs 26%, and the benefit held across KIT-mutation subgroups. Because the comparator is sunitinib alone, the trial directly shows what bezuclastinib adds to the combination. The FDA has previously approved GIST drugs on PFS (e.g. ripretinib).

Source: SEC 8-K 2026-08-10 (Q2 2026 results, EX-99.1); ASCO 2026 abstract 11500 (JCO 44:16_suppl); CancerNetwork PEAK coverage
02▲ SupportsMedium Every expedited-review signal is present, and the FDA flagged no review issues at filing

Breakthrough Therapy Designation (January 2026) and Real-Time Oncology Review. The NDA was submitted in March 2026 and accepted with priority review in May 2026, with an FDA-assigned PDUFA date of 2026-11-30. In the acceptance release, Cogent said the FDA had no plan to hold an advisory committee and had not identified any potential review issues. No PDUFA extension as of 2026-10-04.

Source: SEC 10-Q for Q2 2026 (filed 2026-08-10); Cogent press release 2026-05-28; Cogent press release 2026-09-15 (restates GIST PDUFA 2026-11-30)
03▼ AgainstSmall The combination is more toxic, and overall survival is still immature

Grade ≥3 treatment-related AEs: 71.6% with the combination vs 52.4% with sunitinib. Discontinuation for treatment-related AEs: 7.4% vs 3.8%. Grade ≥3 ALT/AST elevation: 10.8% vs 1.4% (all resolved; no grade 4; 1.5% discontinued). No treatment-related deaths on the combination. OS has not been reported. The FDA's 2025 draft guidance treats OS as a safety readout, so it may ask for interim OS showing no harm. This is mainly a labeling question rather than a likely cause of a CRL.

Source: Cogent ASCO 2026 press release (BioSpace, 2026-05-30); CancerNetwork; SEC 8-K 2026-08-10 ('Data for overall survival remains immature')
04▼ AgainstSmall First commercial product for a small company, made by contract manufacturers including a single-source API supplier

Cogent has never had an approved product. The tablets use a spray-dried dispersion, a moderately complex formulation. On 2026-09-01 Cogent signed a 5-year commercial supply agreement with Hovione for the dispersion and the tablets. Hovione is an established spray-drying CDMO; its published FDA inspection history (Loures, Cork, New Jersey, Macau) shows no warning letters, and recent inspections were clean. The API manufacturer is not named in filings, so its inspection status cannot be judged. The same CMC package also supports the NonAdvSM NDA (PDUFA 2026-12-30).

Source: SEC 8-K 2026-09-02 (Item 1.01, Hovione supply agreement); Cogent 10-K risk factors (sole-source API and drug product suppliers); Hovione inspections-history page
05▲ SupportsSmall The company is behaving as if it expects to launch

Cogent has hired and onboarded its full field commercial and medical team and signed the commercial supply contract. It reports pro forma cash of $865.9M, enough to fund operations into late 2028, and says pre-approval manufacturing costs will be capitalized after approval. This is consistent with management confidence, but it is not independent evidence about the FDA's decision.

Source: SEC 8-K 2026-08-10 (EX-99.1); SEC 10-Q Q2 2026

What would change this forecast

01
A PDUFA extension (major amendment) would by itself only move the date; it would matter more if it followed an FDA request for OS or safety data
02
Disclosure of interim OS data, especially any trend against the combination
03
Any FDA 483, OAI classification or import alert at Hovione or at the unnamed API site
04
A late-cycle meeting disclosure, an advisory committee being scheduled, or a problem in the parallel NonAdvSM review (shared CMC package)

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

Confidence

Medium

Manufacturing and inspection risk

Low

This is an oral small molecule, and the tablets and spray-dried intermediate are made at Hovione, an experienced CDMO with a clean published FDA inspection record. The residual risks are that this is Cogent's first commercial product and that the API comes from a single unnamed supplier whose inspection history cannot be checked. I cannot judge that site, which is the main unquantified uncertainty.

Record · version 1

Written on 2026-10-04, 57 days before the PDUFA target date.

Timeline

Versions

v1 2026-10-04 19:58 UTC 88% Current version JSON ↓.ots ↓ ○ Timestamp submitted, awaiting block confirmation

Fingerprint

SHA-256c5d29ad60f63aec91c913ce0941836cccac93eb350acfaf0c9e64f3b138ec569

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_20261004T195814Z.json
ots verify v1_20261004T195814Z.json.ots
ots info v1_20261004T195814Z.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 2026-11-28. 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.