Lumakras withdrawal seems unlikely... for now

Given panellists’ damning appraisal at yesterday’s US adcom of the handling of Lumakras’s confirmatory Codebreak-200 trial it might come as a surprise that the Amgen drug’s immediate withdrawal isn’t on the cards. Still, it’s safe to assume that Lumakras won’t be getting a full regulatory nod any time soon either.

That’s good news for Mirati, whose rival KRAS inhibitor Krazati, also backed by an accelerated approval, will thus not face an extra threat. Mirati surged 45% yesterday, though that was mostly down to Bloomberg reporting that Sanofi was considering buying the company. Either way, the adcom offers harsh advice on how a clinical trial shouldn’t be run – sobering for a company of Amgen’s scale.

While KRAS blockade was in the spotlight, any takeover interest might not be driven by Krazati but by Mirati’s PRMT5 inhibitor MRTX1719. Notably, another PRMT5 player, Tango Therapeutics, surged 14% yesterday. Coincidentally, Amgen is in the game here too: first-in-human data for its AMG 193 are to be presented at this month’s Triple Meeting.

Systemic bias

As for Lumakras, it was already obvious from adcom briefing documents that the FDA had identified deficiencies in Codebreak-200, and it was because of these that it sought advice on the reliability of its positive PFS result. Lumakras had been approved for second-line, KRAS G12C-mutated NSCLC based on the single-cohort Codebreak-100 trial, and Codebreak-200 compared it versus docetaxel.

Panellists were scathing about Codebreak-200 yesterday, outlining one obvious reason why the PFS result was unreliable, and highlighting three key pieces of evidence suggesting that the trial was systemically biased.

Chi Song, the FDA’s primary reviewer for Amgen’s filing, said 23 patients randomised to docetaxel in Codebreak-200 dropped out before getting treatment, versus just two on Lumakras, suggesting a bias against the control arm. There was also concern that investigator assessment of progression, favouring Lumakras, was at odds with blinded central review, and many docetaxel patients crossed over to Lumakras before even having their disease progression confirmed.

This appears in line with prevailing enthusiasm over the promise of KRAS blockade, and suggests that doctors were keen to get as many patients as possible onto a treatment that had shown promise. Some bias is unavoidable since Codebreak-200 was open-label (because of route of administration and toxicity differences), though independent review of imaging data to determine disease progression must be blinded.

A more fundamental point was raised about the observed median PFS benefit of five weeks; because this was shorter than the six weeks between patient scans any difference might be due to inherent measurement error. The FDA and Amgen did a sensitivity analysis hypothesising progression at any time between imaging assessments and not at the end, and this suggested that Codebreak-200’s mPFS benefit could actually have been as short as five days.

The trial was notable for not having shown an OS benefit, a fact thought to be down to 26% of docetaxel patients ultimately crossing over to Lumakras. However, the FDA reviewers said OS wasn't likely to have been affected by such crossover.

What now?

The adcom wasn’t being asked whether Codebreak-200 should be used to convert Lumakras’s accelerated approval. Rather, the voting question asked whether the trial’s PFS result could be reliably interpreted given the “multiple signals of potential bias”.

The answer was a resounding “no”, by 10 votes to two. Though the FDA isn’t bound to follow an adcom’s advice it seems logical that, if the PFS result is unreliable, Codebreak-200 can’t back Lumakras’s full approval. The question then becomes, what happens to Lumakras in the meantime?

One precedent set by an April 2021 adcom to review non-converted approvals of anti-PD-(L)1 drugs is that, as long as some other potentially confirmatory trial remains ongoing, accelerated approval won’t immediately be rescinded. For Lumakras such a study exists: Codebreak-202 tests a Lumakras chemo combo versus Keytruda plus chemo first line.

To underline the point Harpreet Singh, a director at the FDA’s division of oncology, stated: “It is not our intent to immediately withdraw a drug that has – quote unquote – failed a confirmatory trial.”

This all offers a lesson about the pitfalls Mirati must avoid when running its own confirmatory Krazati study, Krystal-12, but if the Bloomberg story has legs then the company won’t have to worry about this for much longer. 

After this story was published Bristol Myers Squibb announced a $4.8bn deal to acquire Mirati.