Roche stakes its KRAS claim

With Amgen now facing a US adcom, and Mirati undergoing more C-suite turnover, Roche has staked a claim to having a viable rival KRAS G12C inhibitor. Phase 1 data published in the NEJM yesterday suggest that activity of the Swiss firm’s divarasib is at least as good as that of Amgen’s Lumakras or Mirati’s Krazati on a cross-trial basis.

The publication is somewhat unusual in drawing a direct comparison between the three projects, and making a cautious claim about divarasib’s possible edge. For its part, however, Roche remains guarded, with a focus on divarasib’s recently initiated phase 2/3 B-Fast trial, which does not read out until next year, and which will not result in a planned regulatory filing until at least 2025.

That focus presumably arises from the fact Amgen has already generated confirmatory data for Lumakras in the Codebreak-200 trial, meaning that divarasib might have no accelerated US approval path. Hence, seeking randomised, controlled data, Roche began B-Fast in the fourth quarter of 2022. 

Clinical splash

Divarasib first made a clinical splash at last year’s World Lung congress, where its first-in human trial yielded unconfirmed and confirmed ORRs of 53% and 46% respectively among 57 NSCLC patients. Now the same trial features in NEJM, updating the results and adding a colorectal cancer cohort.

It is the colorectal data in particular where divarasib appears to stand out, with ORR of 36% or 29% (confirmed/unconfirmed) looking impressive for a KRAS-directed monotherapy. All the way back at ESMO 2021 Mirati had shown Krazati to have more promise in colorectal cancer than Lumakras, but on a cross-trial basis divarasib looks better still.

Surprisingly, Mirati has yet to take Krazati’s colorectal data before the US regulator, and both incumbents are focusing on combinations with anti-EGFR MAbs – Lumakras plus Vectibix and Krazati plus Erbitux. As these would be expected to add toxicity the possible advantage of a monotherapy is obvious.

Cross-trial comparison of single-agent KRAS G12C inhibitors

Note: ORR numbers exclude unconfirmed responses. Source: product labels, NEJM & Lancet Oncology.

The NEJM authors say up front that the clinical benefit of Lumakras and Krazati “remains limited”, and that “divarasib has been shown to be five to 20 times as potent and up to 50 times as selective in vitro” respectively. Divarasib appears to show numerically more responses in either NSCLC or colorectal cancer than the marketed drugs, they state, and the data bear this out. 

However, they do caution that cross-trial comparisons are unreliable, for instance because of differences in patients’ demographic features. Perhaps the most obvious stumbling block to a fair comparison is that divarasib responses were assessed by investigators, implying more possible bias versus the blinded independent central review used in registrational Lumakras and Krazati data.

That said, this is a strong dataset, and it can only be guessed – especially in light of doubts surrounding Amgen’s confirmatory Codebreak-200 results – whether it might lead to Roche accelerating its filing plans.