Divarasib (GDC-6036)

Divarasib (GDC-6036) is Roche/Genentech's covalent KRAS G12C inhibitor, now in three Phase 3 trials for KRAS G12C-mutated non-small cell lung cancer (NSCLC), a target population of roughly 14,000-16,000 newly diagnosed US patients per year (KRAS G12C appears in ~13% of lung adenocarcinomas). The two most consequential trials are a head-to-head superiority study against the two approved G12C drugs, sotorasib and adagrasib, in previously treated patients (NCT06497556), and a frontline combination with pembrolizumab (an anti-PD-1 antibody that blocks a brake cancer uses to evade T-cell attack) versus pembrolizumab plus chemotherapy (NCT06793215) [1][2]. Roche is the third major Western player into the KRAS G12C market and is betting that a deeper, longer Phase 1 response signal translates into a clear win over Amgen's Lumakras (sotorasib) and Bristol Myers Squibb's Krazati (adagrasib) [3]. A fourth competitor, glecirasib (Jacobio Pharma, partnered with Johnson & Johnson), is in Phase 3 primarily in China and adds APAC-side pricing pressure [9].

Divarasib is a novel small molecule with no approvals anywhere. The most advanced trial, NCT06497556, is a randomized Phase 3 comparing divarasib monotherapy against sotorasib or adagrasib in pre-treated KRAS G12C-positive metastatic NSCLC (n=338, active, not recruiting) [1]. The frontline combination trial NCT06793215 is recruiting toward 600 patients, testing divarasib plus pembrolizumab against the current standard of care (SoC - the established best-approved regimen, here pembrolizumab plus pemetrexed plus platinum chemotherapy) [2]. A Phase 2 perioperative cohort (NCT04302025) tests divarasib in resectable Stage IB-III NSCLC with major pathologic response as the readout [4]. Long-term Phase 1 follow-up was published in the Journal of Clinical Oncology in 2025 [3]. No FDA Breakthrough, Fast Track, or Orphan designations have been publicly disclosed for the lead NSCLC program. Roche has not guided to a specific filing date, but the head-to-head data is the gating event and is likely a 2026-2027 readout based on trial status.

KRAS is a protein that sits just inside the cell membrane and acts like an on/off switch for cell growth signals coming in from outside. Normally it cycles between an active state (bound to GTP, signal on) and an inactive state (bound to GDP, signal off). The G12C mutation, found in about 13% of lung adenocarcinomas, swaps a glycine for a cysteine at position 12 and biases the switch toward the on position, so cells keep getting a 'divide' signal they should not be getting [5]. Divarasib is a covalent inhibitor that latches permanently onto that cysteine, but only when KRAS is in the inactive GDP-bound state, locking the switch off. The mechanism is genetically validated and clinically validated: sotorasib (Amgen, approved 2021) and adagrasib (Bristol Myers Squibb via Mirati, approved 2022) both work by the same trick, and both produced 28-43% objective response rates in pre-treated NSCLC across Phase 1/2 and Phase 3 trials [6][8]. The open question is not whether the target works but whether divarasib's higher selectivity and slower off-rate produce meaningfully better depth and duration of response in patients.

The pivotal frontline trial is NCT06793215, an open-label randomized Phase 3 in previously untreated KRAS G12C-positive non-squamous metastatic NSCLC. Patients are randomized to divarasib plus pembrolizumab versus pembrolizumab plus pemetrexed plus carboplatin or cisplatin. Primary endpoint is progression-free survival by investigator assessment. Target enrollment is 600 patients across global sites, currently recruiting, sponsored by Hoffmann-La Roche [2]. The design choice is aggressive: putting a targeted plus IO (immuno-oncology, meaning checkpoint inhibitor) doublet up against the established IO-chemo triplet that is current SoC, with no chemo in the experimental arm, is a clean efficacy comparison but raises the bar versus a simpler IO-chemo-targeted quadruplet. The second Phase 3, NCT06497556, is a randomized superiority study against investigator's choice of sotorasib or adagrasib in pre-treated patients, n=338, PFS primary, currently active and not recruiting [1]. That is the more interpretable trial for ranking the three G12C drugs, and the smaller, faster-enrolling design suggests Roche prioritized speed to a label-differentiating comparison.

The model estimates a 33% chance this drug is eventually approved. That number starts from a historical baseline of about 48% for Phase 3 drugs in this area, then adjusts based on ten facts about the trial and sponsor. The estimate is pushed up by the trial's light or open-label blinding, more secondary endpoints than usual, and the sponsor's strong record of getting drugs approved, but pulled down by weak or limited earlier-phase results. The remaining factors fall near average for this stage and leave the final estimate close to where the base rate started.

Efficacy risk is real but bounded: the head-to-head trial NCT06497556 is a PFS superiority design against active G12C comparators, and Phase 1 numbers, while strong, came from a single-arm setting that can flatter response and duration estimates. If PFS is statistically positive but the magnitude is small, payers may not differentiate divarasib from incumbents priced at roughly $21,000/month for Lumakras and $22,000/month for Krazati (approximate US WAC, 2024-2025) [10]. Safety risk centers on on-target hepatotoxicity and gastrointestinal effects that have dogged the class, with sotorasib carrying transaminase warnings; divarasib's Phase 1 profile looked cleaner than sotorasib's, but Phase 3 patient volume can surface lower-frequency signals [3]. Combination toxicity with pembrolizumab in NCT06793215 (immune-mediated hepatitis stacked on G12C-inhibitor hepatic signal) is a specific watch item. Resistance is the long-tail clinical risk: even responders eventually escape via secondary KRAS mutations (Y96D being the canonical example), KRAS allele amplification, or bypass signaling through receptor tyrosine kinases (MET, EGFR, FGFR), and STK11/KEAP1 co-mutations predict primary resistance - no divarasib trial yet addresses post-progression strategy. Indication gap: sotorasib and adagrasib both have approved CRC labels (in combination with cetuximab/panitumumab; KRAS G12C appears in ~3-4% of colorectal cancer), but Roche has not publicly disclosed a divarasib CRC program, leaving that revenue line unaddressed. Competitively, glecirasib (Jacobio/Johnson & Johnson) showed ORR ~48% in a Phase 2b NSCLC study and is now in Phase 3 in China; its global label path is unclear but it adds APAC pricing pressure regardless [9]. The dominant commercial risk is Revolution Medicines' daraxonrasib (RMC-6236), a RAS(ON) multi-selective inhibitor that hit both PFS and OS in RASolute 302 in second-line pancreatic cancer, with reported median OS of 13.2 vs 6.7 months (HR 0.40) in a late-breaking ASCO 2026 Plenary presentation (LBA5) with simultaneous NEJM publication [11]. Daraxonrasib does not directly cannibalize G12C NSCLC volume since G12C is the dominant lung mutation, but a pan-RAS narrative could compress G12C-selective pricing power and accelerate next-gen RAS(ON) entrants including Revolution's own G12C-selective RMC-6291.

Worth watching. The signal moment is the NCT06497556 head-to-head readout: this is the first time any KRAS G12C drug will be tested against the approved competition in a randomized Phase 3, and the result decides whether divarasib leapfrogs Lumakras and Krazati or arrives third into a market that daraxonrasib's pan-RAS data has already partially reframed. The frontline combination trial NCT06793215 is the bigger commercial prize, but it reads out later and the comparator (IO-chemo) is a tougher bar. Roche carries this comfortably from a balance sheet standpoint (~$65B in 2024 group sales), so execution risk is minimal and the program will run to readout regardless of interim noise. Check back when NCT06497556 reports topline PFS, expected 2026-2027 based on current trial status, and watch for any updated divarasib + pembrolizumab combination safety data presented at AACR or ASCO 2026-2027. If the head-to-head misses, the program likely retrenches to combinations rather than a standalone label fight.