Krascendo 1: divarasib beats sotorasib/adagrasib head-to-head with PFS HR 0.58 and OS HR 0.56, challenging first-generation KRAS G12C inhibitors

Data as of 2026-10-26 (Taipei time). [Fact] statements are sourced; [Inference] statements are this article's judgment.

1. One-line conclusion

[Fact] In Krascendo 1, divarasib vs sotorasib or adagrasib gave PFS HR 0.58 (95% CI 0.44–0.76; p<0.0001) and OS HR 0.56 (0.41–0.78; p=0.0005) (Roche). [Inference] It is a rare "same target, next generation vs first generation" head-to-head trial to win OS. If divarasib is approved and becomes the standard, comparators and effect-size assumptions for later-line KRAS G12C trials will all need resetting; before approval, sponsors should run sensitivity analyses on divarasib's data (NEJM).

For sponsors: effect-size assumptions for later-line KRAS G12C trials should be stress-tested against mPFS 9.8 months and mOS 17.7 months (NEJM), even though divarasib is not yet approved.

2. What happened

[Fact] Krascendo 1 (NCT06497556) is a global, randomized, open-label, multicenter Phase 3 trial in previously treated KRAS G12C-mutated advanced NSCLC; 338 patients were randomized to divarasib once daily (n=167) or sotorasib once daily / adagrasib twice daily (n=171); the primary endpoint was BICR-assessed PFS, with OS as a key secondary endpoint; it was presented at the ESMO 2026 Presidential Symposium and simultaneously published in NEJM (Roche). ClinicalTrials.gov lists actual enrollment of 338 and primary completion on 2026-05-19 (ClinicalTrials.gov). The NEJM abstract: of 171 control patients, 156 received sotorasib and 15 adagrasib; divarasib was 400 mg once daily; median follow-up was 14.3 months (NEJM).

[Fact] mPFS 9.8 vs 5.7 months (HR 0.58, 0.44–0.76, p<0.0001), 12-month PFS 44% vs 24%; mOS 17.7 vs 12.0 months (HR 0.56, 0.41–0.78, p=0.0005); confirmed ORR 58% vs 32%, mDOR 11.9 vs 6.9 months (ORR and DOR in the response-evaluable population, n=328) (Roche). Discontinuation for TRAEs 4.8% vs 11.7%, serious TRAEs 7.8% vs 15.2%, grade 3–4 TRAEs 24.6% vs 34.5% (NEJM); time on treatment 11.3 vs 5.0 months (Roche).

For sponsors: the later-line KRAS G12C mPFS benchmark moves from about 5.7 months to 9.8 months; a new drug going head-to-head in this population should base its effect-size assumptions on divarasib.

3. Why it succeeded

[Inference] Mechanism: Roche says divarasib is more potent and more selective than sotorasib/adagrasib preclinically (Roche); greater potency may give more complete target inhibition, reflected in ORR 58% vs 32% (Roche). Better tolerability kept patients on treatment longer (11.3 vs 5.0 months, Roche), which may contribute as much to OS as potency itself.

[Inference] Biomarkers: the OS and PFS benefit was consistent regardless of PD-L1 or KEAP1 mutation status (Roche). KEAP1 co-mutation is a known group with poorer response to KRAS inhibitors, and subgroup HRs in the full NEJM paper are worth a close look; we could not find the subgroup HR figures.

[Inference] Endpoints and statistics: the trial was open-label but PFS was assessed by BICR (Roche), and OS is not subject to assessment bias, so the conclusion is credible. The control arm allowed either drug, but 156 of 171 patients (about 91%) received sotorasib and only 15 adagrasib (NEJM); the trial therefore mainly shows superiority over sotorasib, the evidence of an advantage over adagrasib is thin, and we could not find HRs for each comparator. Subsequent-therapy data were also not found.

For sponsors: when citing Krascendo 1, say "beats sotorasib", not "beats all first-generation inhibitors"; if a new drug is to be compared with adagrasib, the current evidence is not enough to skip a head-to-head trial.

4. Regulatory and commercial implications

[Fact] The FDA granted divarasib Breakthrough Therapy Designation in 2022 and Orphan Drug Designation in 2026 (Roche). [Inference] A global head-to-head trial with significant OS leaves a low bar for an FDA filing, and the later-line market for sotorasib and adagrasib would be directly displaced. For the NMPA, we found no Roche China filing plan; sites registered on ClinicalTrials.gov span 26 countries/regions, including Hong Kong and Taiwan but not mainland China (ClinicalTrials.gov API). Competitively, the bigger battle is first line: Krascendo 2 (NCT06793215) compares divarasib plus pembrolizumab with chemotherapy plus pembrolizumab (Roche).

For sponsors: until approval, sotorasib/adagrasib remain the approved comparators available in later-line KRAS G12C; if divarasib is to launch in mainland China, it may need additional Chinese data because no mainland sites are registered (ClinicalTrials.gov API).

5. Recommendations for the next trial and OncoLattice

[Inference] First, divarasib is not approved and cannot be used directly as a comparator; only if the FDA approves it and it becomes the standard will later-line KRAS G12C trials need to switch their comparator to divarasib. Until then, new trials should use approved drugs as comparators while running effect-size sensitivity analyses on divarasib's mPFS of 9.8 months (NEJM), or move to first line. Second, sponsors of pan-KRAS or KRAS(ON) inhibitors should prepare to enroll a post-divarasib population and record the type of prior KRAS inhibitor at screening. Third, OncoLattice site strategy: KRAS G12C occurs in about one in eight NSCLC patients (Roche), so screening volume is high; prioritize sites with in-house NGS and fast reporting.

For OncoLattice: if divarasib is approved, the later-line KRAS G12C population becomes "post-divarasib", and screening and enrollment design for this new population should be planned early.

6. Open questions and what to watch

Event	Timing	Where to look
Subgroup HRs (KEAP1, STK11, brain metastases), HR for each comparator, subsequent therapy	Published (2026-10-25)	NEJM full text and appendix (NEJM); we did not obtain the full text
Mature OS	Estimated study completion 2029-11-30 in the registry	NCT06497556
First-line divarasib plus pembrolizumab readout (dual primary endpoints PFS and OS, estimated 600 patients)	Estimated primary completion 2028-11-30 in the registry	Krascendo 2, NCT06793215
FDA / NMPA filings	Not announced	Roche media releases (Roche)

For sponsors: the first-line KRAS G12C competitive window lasts at least until the end of 2028 (NCT06793215); sponsors starting first-line trials before then can still use chemotherapy plus IO as the comparator, but should build in a clause to switch comparators after Krascendo 2 reads out.

The next thing to watch is Krascendo 2's first-line readout: if a chemotherapy-free combination beats chemotherapy plus IO, the first-line KRAS G12C standard will be rewritten.

Evidence limitations

The data here come from Roche's press release and the NEJM abstract (NEJM); we did not read the full NEJM paper, and the HRs are as listed in both (Roche).

Advice for readers: before citing, check subgroup and per-comparator data in the full NEJM paper, then use them to set comparator assumptions for the next trial.

Note: differences from the original relay

PFS HR 0.58 and OS HR 0.56 match the source (Roche). Additions: the control-arm split of 156 sotorasib / 15 adagrasib, grade 3–4 TRAEs of 24.6% vs 34.5% and median follow-up of 14.3 months come from the NEJM abstract (NEJM); mOS was 17.7 vs 12.0 months (the press release headline says "approximately 18 vs 12 months").