Divarasib: Head-to-Head Results in KRAS G12C Lung Cancer
In this article
On 25 October 2026, the full results of the phase 3 Krascendo 1 trial, the first head-to-head comparison of divarasib with a first-generation KRAS G12C inhibitor (sotorasib or adagrasib), were presented at the Presidential Symposium of the ESMO 2026 Congress in Madrid (LBA5) and published the same day in the New England Journal of Medicine. In 338 previously treated patients with KRAS G12C-mutated advanced non-small cell lung cancer (NSCLC), median progression-free survival was 9.8 vs 5.7 months (HR 0.58) and median overall survival at the first interim analysis was 17.7 vs 12.0 months (HR 0.56). The result looks strong, but the trial was open-label, the overall survival analysis is an interim one, and divarasib is not yet approved in any country.
Who is this article for? For patients diagnosed with KRAS G12C-mutated lung cancer and their families, for health professionals who follow lung cancer, and for anyone curious about how to read a clinical trial. This article is not medical advice; no treatment decision should be made on the basis of this text.
This news arrived in two stages. On 1-2 July 2026, Roche announced that the trial had met both its primary endpoint (PFS) and its key secondary endpoint (overall survival), but gave no numbers. The data presented today are the first time that announcement has been seen with numbers attached. In other words, this is not a new study; it is the content behind July's "positive result" sentence.
Who is behind the "new standard" and "first" statements? Strong claims circulate quickly in news like this; it is worth separating who said what.
- "The first and only global randomized phase 3 head-to-head comparison" is a statement by the trial's lead investigator, Dr. Ferdinandos Skoulidis (MD Anderson). We could not verify this claim independently.
- "A new standard of care" is also Skoulidis's comment during the presentation. It is not a regulatory decision; because the drug is not approved by any authority, there is no label or guideline recommendation either.
- "Best-in-class" appears in the headline of the manufacturer's (Roche) press release; it is company language, not an independent assessment.
Background: what did first-generation KRAS G12C inhibitors deliver?
KRAS G12C is a mutation found in roughly 14% of non-small cell lung cancers (according to Roche's announcement). The first two drugs to target this mutation directly were sotorasib and adagrasib; both were tested in previously treated patients against docetaxel (standard chemotherapy). We covered KRAS G12C and the origin story of these drugs in our KRAS G12C article (in Turkish); you can find the types of lung cancer in our article on the differences between small cell and non-small cell lung cancer (in Turkish).
| Trial | Comparison | Patients | Median PFS | Response rate | Overall survival |
|---|---|---|---|---|---|
| CodeBreaK 200 | Sotorasib vs docetaxel | 345 | 5.6 vs 4.5 months HR 0.66 (0.51–0.86) |
28.1% vs 13.2% | 10.6 vs 11.3 months; HR 1.01 (0.77–1.33): no difference |
| KRYSTAL-12 | Adagrasib vs docetaxel | 453 | 5.5 vs 3.8 months HR 0.58 (0.45–0.76) |
32% vs 9% | Could not be verified for this article |
| Krascendo 1 | Divarasib vs sotorasib or adagrasib | 338 | 9.8 vs 5.7 months HR 0.58 (0.44–0.76) |
58.4% vs 32.3% | 17.7 vs 12.0 months; HR 0.56 (0.41–0.78), first interim analysis |
A caution when reading this table: The comparator arms of the three trials differ (docetaxel in two, another KRAS inhibitor in the third), and so do the patient populations and follow-up times. You cannot conclude that "whichever HR is lower is better." All the table shows is that first-generation drugs extended progression-free survival by a few months against docetaxel and, at least in CodeBreaK 200, did not change overall survival. The fact that median PFS in Krascendo 1's control arm (5.7 months) is close to these drugs' earlier results suggests the control arm performed as expected; that is an inference, not proof.
What is divarasib?
Divarasib is a new-generation inhibitor that targets the KRAS G12C mutation and is taken by mouth once a day. Before this trial, phase 1 data had been encouraging: at the 400 mg dose the confirmed response rate was reported as 53.4% and median PFS as 13.1 months (NEJM 2023; because that was a small, single-arm study, these figures should be read as a "promise," not as proof). The US Food and Drug Administration (FDA) granted divarasib Breakthrough Therapy designation in 2022 and Orphan Drug designation for KRAS G12C-mutated NSCLC in 2026. These are not approvals; they are statuses that speed up development.
Krascendo 1 at a glance
| Trial | Krascendo 1 (NCT06497556): global, open-label, randomized phase 3 |
| Publication | Skoulidis F, et al. N Engl J Med, 25 October 2026; doi:10.1056/NEJMoa2605703. Presented the same day at ESMO 2026, LBA5 (Presidential Symposium) |
| Title | "Divarasib as Compared with Sotorasib or Adagrasib in KRAS G12C–Mutated NSCLC" |
| Sponsor | F. Hoffmann-La Roche / Genentech |
| Patients | 338 patients, 151 sites, 26 countries; 1:1 randomization |
| Arms | Divarasib 400 mg once daily (n=167) vs investigator's choice: sotorasib 960 mg once daily (n=156) or adagrasib 600 mg twice daily (n=15); control arm n=171 |
| Primary endpoint | Progression-free survival assessed by blinded independent central review (BICR) |
| Key secondary endpoints | Overall survival, confirmed response rate, duration of response, safety |
| Stratification | History of central nervous system (CNS) metastases, which drug was chosen for the control arm, KEAP1 mutation status |
| Crossover | Not permitted |
| Median follow-up | 14.3 months |
Who was enrolled?
- Locally advanced or metastatic KRAS G12C-mutated NSCLC.
- Previously received 1-3 lines of therapy, including platinum-based chemotherapy and a PD-1/PD-L1 inhibitor.
- Had not previously received a KRAS G12C inhibitor; ECOG performance status 0-1; measurable disease by RECIST 1.1.
- Patients with active or untreated brain metastases were excluded.
| Characteristic | Divarasib | Control |
|---|---|---|
| Median age | 66 | 66 |
| Current or former smoker | 98% | 96% |
| Adenocarcinoma | 95% | 96% |
| ECOG 1 | 61% | 60% |
| One prior line of therapy | 72% | 69% |
| Prior immunotherapy | 99% | 100% |
| KEAP1 mutation | 21% | 22% |
| History of CNS metastases | 29% | 30% |
| PD-L1 <1% | 34% | 41% |
| PD-L1 ≥50% | 26% | 19% |
The arms are balanced on most characteristics. The one visible difference is in PD-L1 distribution: the divarasib arm has slightly more patients with PD-L1 ≥50% (26% vs 19%) and slightly fewer with PD-L1 <1% (34% vs 41%). Whether this difference affected the result cannot be assessed, because the sources do not give PD-L1 subgroup numbers.
Results
Median times (months)
Colored bar: divarasib; gray bar: control (sotorasib or adagrasib).
For overall survival, the upper limit of the confidence interval for the divarasib arm's median was reported as "not estimable" (95% CI 16.3–NE); the figure is therefore uncertain and belongs to an interim analysis.
| Endpoint | Divarasib | Control | Comparison |
|---|---|---|---|
| Median PFS, BICR (primary) | 9.8 months (8.1–12.9) | 5.7 months (4.2–6.8) | HR 0.58 (0.44–0.76); p<0.0001 |
| 12-month PFS rate | 43.9% | 24.0% | — |
| Median PFS, investigator assessment | 11.1 months (8.8–14.1) | 5.6 months (4.4–7.0) | HR 0.57 (0.43–0.75); consistent with the primary analysis |
| Median overall survival (first interim analysis) | 17.7 months (16.3–NE) | 12.0 months (10.0–15.2) | HR 0.56 (0.41–0.78); p=0.0005 |
| 12-month overall survival rate | 67.1% | 49.2% | — |
| Confirmed response rate (328 patients evaluable for response) | 58.4% (50.4–66.1) | 32.3% (25.3–40.0) | Difference 25.8 points (15.6–36.0) |
| Complete response / partial response | 4.3% / 54.0% | 3.6% / 28.7% | — |
| Median duration of response | 11.9 months (9.7–NE) | 6.9 months (5.3–8.3) | — |
How to read this: In absolute terms, the difference in medians is +4.1 months for PFS and +5.7 months for overall survival. In relative terms, the risk of progression or loss of life is 42% lower for PFS, and the risk of loss of life is 44% lower for overall survival. When explaining this to a patient, absolute months should come before relative percentages, because "44% reduction" on its own does not say how much time was gained.
Subgroups: what was said, and what was not?
The publication summarizes the PFS and overall survival benefit as "generally consistent" in favor of divarasib regardless of CNS metastasis history, liver metastasis, KEAP1 mutation and PD-L1 status. However, we could find numerical data in the sources for only two of these groups:
| Subgroup | Divarasib | Control | HR (95% CI) |
|---|---|---|---|
| KEAP1-mutated (35 vs 37 patients) | 4.6 months | 2.8 months | 0.56 (0.33–0.92) |
| History of CNS metastases (48 vs 51 patients) | 6.8 months | 5.7 months | 0.74 (0.47–1.17): the confidence interval includes 1 |
| PD-L1, STK11, prior immunotherapy status | No numerical data found | ||
These subgroup figures need careful reading. KEAP1 mutation is associated with a worse course; in this group median PFS is short in both arms (4.6 and 2.8 months) and the numbers are small (about 35 + 37). In patients with a CNS metastasis history, the confidence interval of the 0.74 HR crosses 1, so a difference in this subgroup is not statistically certain. In addition, the trial excluded patients with active brain metastases, so these data say nothing about patients whose brain metastases are active. Because subgroups were not the trial's primary question, all of them are "hypothesis-generating."
Safety: fewer serious side effects, more gastrointestinal problems
| Adverse event | Divarasib | Control |
|---|---|---|
| Treatment-related adverse event of any grade | 92.8% | 80.1% |
| Diarrhea | 66% | 38% |
| Nausea | 53% | 20% |
| Vomiting | 39% | 8% |
| ALT elevation | 15% | 23% |
| AST elevation | 14% | 21% |
| Liver toxicity (pooled) | 26.9% | 39.2% |
| Treatment-related QT prolongation | 0 patients | 2 patients |
| Pneumonitis | 0 patients | 3 patients |
| Grade 5 (resulting in loss of life) treatment-related adverse event | None | None |
| Dose reduction / interruption rates | Not reported in the sources | |
The table points in two directions. In the divarasib arm, serious adverse events, treatment discontinuation and liver enzyme elevations were less frequent; on the other hand, diarrhea, nausea and vomiting were clearly more frequent (most of these were grade 1; in the divarasib arm grade 1 diarrhea was 35% and nausea 32%). Liver toxicity is one of sotorasib's known problems, and it is worth remembering that about 91% of the patients in the control arm received sotorasib (156/171).
A misleading point in the adverse event comparison: Median treatment duration in the divarasib arm (11.3 months) was more than twice that of the control arm (5.0 months). That the serious adverse event rate came out lower in the group exposed to the drug for longer is a positive sign; on the other hand, because the sources give no exposure-adjusted rates such as "per 1,000 patient-months," we cannot compare the adverse event burden of the two arms on exactly equal terms. In addition, patient-reported quality-of-life results (reported to be one of the trial's pre-specified key secondary endpoints) are not included in these summaries.
How robust is this result?
Six points to keep in mind while reading:
- Open-label trial. Both patient and physician know which drug is given. Because PFS was assessed by blinded independent central review (BICR), this bias is partly limited for PFS; overall survival and adverse event reporting do not have the same safeguard. Investigator-assessed PFS was also consistent with the primary analysis (11.1 vs 5.6 months).
- Overall survival comes from the first interim analysis. The number of events and the follow-up specific to overall survival are not given in the summaries we could reach. The upper limit of the confidence interval for the divarasib arm's median overall survival is also uncertain. The figure may therefore change.
- The control arm was almost entirely sotorasib. Of 171 control patients, 156 received sotorasib and only 15 received adagrasib. The result is therefore strong for "divarasib versus sotorasib" and says nothing about "divarasib versus adagrasib" (an arm of 15 patients cannot support a comparison).
- There was no crossover. Patients in the control arm could not switch to divarasib on progression. This keeps the treatment sequences of the arms clean; however, the summaries do not say which treatments patients received after progression, so we cannot tell how much of the overall survival difference comes from the first-line effect and how much from later treatments.
- A selected patient group. Patients with ECOG 0-1, without active brain metastases and without prior KRAS G12C inhibitor use. In real life, some patients do not fit these criteria.
- Sponsored trial and secondary sources. The trial was run by Roche/Genentech. We also could not access the NEJM full text; the numbers were compiled from the manufacturer's announcement and from the congress/publication summaries of OncLive and CancerNetwork, which are consistent with each other.
What happens next?
The FDA granted divarasib Breakthrough Therapy designation.
Phase 1 results were published online in NEJM (response rate 53.4% at the 400 mg dose).
Roche announced, without giving numbers, that Krascendo 1 had met its PFS and overall survival goals.
The full results were presented at ESMO 2026 (LBA5) and published in NEJM.
Roche says it will submit the data to health authorities and is pursuing the approval process. No approval date or submission date has been announced.
Other ongoing programs for divarasib:
- Krascendo 2 (NCT06793215): divarasib + pembrolizumab versus pembrolizumab + chemotherapy in previously untreated patients.
- Krascendo 170 (NCT05789082, phase 1b/2): in the chemotherapy-free combination, a confirmed response rate of 72.9% was reported in the PD-L1-positive cohort (small, single-arm data); a separate cohort is enrolling patients with untreated brain metastases.
- Krascendo 3 (NCT07541170): after surgery (adjuvant), divarasib versus pembrolizumab or nivolumab.
The situation in Turkey. Divarasib is currently not approved in any country; it therefore cannot be licensed or reimbursed in Turkey. Access to a once-daily oral drug outside a clinical trial could for now only be possible through special routes such as expanded access programs; we could not verify whether such a program exists. The licensing and reimbursement status of sotorasib and adagrasib in Turkey could not be verified in this work (the TİTCK database could not be accessed); we did not speculate. Whether Turkey is among the trial's 26 countries could not be verified either. FDA approval, TİTCK licensing and SGK reimbursement are three separate processes; we covered them in our articles on FDA-approved new cancer drugs and the Turkey access map and unlicensed cancer drugs (both in Turkish). We covered molecular testing, the first step of targeted therapy, in our article on next-generation sequencing (NGS) (in Turkish); taking part in clinical trials (in Turkish) can be one way to reach new drugs.
DROZDOGAN Academy Commentary
This result is the strongest sign so far of what second-line treatment of KRAS G12C lung cancer might look like. But a drug being approved, a drug being given to a patient, and a drug adding to a patient's life are three separate steps.
⚡ Why it matters
First-generation KRAS G12C inhibitors extended progression-free survival by a few months against docetaxel; at least in CodeBreaK 200 there was no overall survival difference. Krascendo 1 appears to be the first large randomized comparison in which a new-generation drug showed superiority over these drugs in both PFS and (at the first interim analysis) overall survival. However, the word "first" is the investigator's own and has not been independently verified.
✅ What is strong
A randomized, multicenter phase 3 with 338 patients; the primary endpoint was assessed by independent central review and came out consistent with investigator assessment. The effect sizes (PFS HR 0.58; overall survival HR 0.56) are clearly away from 1 together with their confidence intervals. Serious adverse events and discontinuation rates stayed low in the divarasib arm, even though treatment duration was twice as long.
⚠️ What is weak
An open-label design; overall survival comes from the first interim analysis and the number of events is not in the summaries; because the control arm was almost entirely sotorasib, no adagrasib comparison can be made; the results so far are compiled from the sponsor's announcement and secondary press. Gastrointestinal side effects (diarrhea 66%, nausea 53%, vomiting 39%) are marked and can affect daily life; dose reduction rates and quality-of-life results are not yet visible.
🩺 What it means in practice for patients and physicians
Nothing changes in practice today: divarasib is not approved and no treatment decision can be made on the basis of this data. But for a patient in whom a KRAS G12C mutation has been detected, the concrete meaning of this data is to talk with the oncologist about which options are available at the next line and about open clinical trials. For patients who have not been tested for the mutation, testing has to come first.
❓ Open questions
Will the difference hold as overall survival matures? Would the result change if the adagrasib arm had been large enough? How much of the difference remains once post-progression treatments are accounted for? Will results of divarasib in the first line (with pembrolizumab) and in the adjuvant setting change this picture? And the most practical question: when will approval come, and when will it reach Turkey?
Bottom line: In previously treated KRAS G12C-mutated lung cancer, divarasib showed a strong advantage over first-generation inhibitors both in delaying progression and (at the interim analysis) in survival time. This is strong but not yet complete evidence: mature overall survival, safety details and the approval decision remain to be seen. "New standard" is, for now, the investigator's interpretation, not a regulatory decision.
Frequently asked questions
- What is divarasib?
- A new-generation cancer drug that targets the KRAS G12C mutation and is taken by mouth once a day. It is currently used only in clinical trials; it is not approved in any country.
- What did the Krascendo 1 trial show?
- In previously treated, KRAS G12C-mutated advanced non-small cell lung cancer, divarasib achieved a median progression-free survival of 9.8 vs 5.7 months and, at the first interim analysis, a median overall survival of 17.7 vs 12.0 months compared with sotorasib or adagrasib. The response rate was 58% vs 32%.
- Is divarasib better than sotorasib and adagrasib?
- In this trial it gave better results mainly against sotorasib. Of the 171 patients in the control arm, 156 received sotorasib and only 15 received adagrasib, so the evidence is insufficient for a direct comparison with adagrasib. A verdict of "better" also requires mature overall survival and regulatory assessment.
- Can this drug be used in Turkey?
- No; because the drug is not approved in any country, it has no license or reimbursement. Clinical trials or special access programs can only be assessed by the oncologist who manages the treatment.
- How do I find out whether I have a KRAS G12C mutation?
- Through molecular testing on tumor tissue (sometimes on blood). Next-generation sequencing (NGS) is the most comprehensive method. Your oncologist decides which treatments and clinical trials are suitable based on the test result.
- What are the side effects of divarasib?
- In the trial the most common were diarrhea (66%), nausea (53%) and vomiting (39%); most were mild. Serious treatment-related adverse events (7.8%) and discontinuation (4.8%) were less frequent than in the control arm. Dose reduction and quality-of-life data are not yet visible.
- Is the overall survival difference certain?
- No, it comes from the first interim analysis. Although the difference was found statistically significant, event counts and follow-up time are not in the summaries; the figures may change in later analyses.
Glossary
- KRAS G12C
- A mutation in the KRAS gene in which the glycine at position 12 of the encoded protein is changed to cysteine; it leaves on a signal that keeps pushing the cell to grow.
- KRAS G12C inhibitor
- The class of drugs that directly target this mutant protein: sotorasib, adagrasib, divarasib.
- PFS (progression-free survival)
- The time from the start of treatment until the disease progresses or loss of life occurs.
- Overall survival (OS)
- The time until loss of life from any cause; the gold-standard endpoint in oncology.
- HR (hazard ratio)
- The ratio of the rate at which an event occurs in two groups. Below 1 favors the experimental arm; if the confidence interval includes 1, the difference is not statistically certain.
- BICR
- Blinded independent central review: assessment of the imaging by independent experts who do not know which drug the patient received.
- Interim analysis
- An early analysis done before the trial ends, when part of the planned number of events has occurred. The result may change in later analyses.
- Crossover
- Patients in the control arm being able to switch to the experimental drug when their disease progresses. It was not permitted in this trial.
- Open-label
- A trial design in which the patient and the physician know which drug is given.
- KEAP1
- A gene that is mutated in some lung cancers and linked to a worse course and treatment resistance.
Transparency note. Verified: The trial's NCT number, design, arms, patient numbers, primary and secondary result numbers, safety numbers and presentation/publication details (consistent between Roche's announcement and the OncLive and CancerNetwork summaries); the phase 3 results of the first-generation drugs (from the CodeBreaK 200 and KRYSTAL-12 publication abstracts); divarasib's 2022 Breakthrough and 2026 Orphan designations; that it is not approved in any country. Not verifiable, and therefore not written: the NEJM full text (not accessible; numbers are from secondary sources); overall survival event counts and OS-specific follow-up; PD-L1, STK11 and prior immunotherapy subgroup numbers; dose reduction and interruption rates; patient-reported outcomes; post-progression treatments; the overall survival result of KRYSTAL-12; the Turkish licensing and reimbursement status of sotorasib and adagrasib; whether the trial has a site in Turkey; the date of Roche's approval submission. A small discrepancy: in its table, OncLive's control-arm complete plus partial response total (33%) appears one point higher than the reported 32.3%; we used CancerNetwork's decimal values (3.6% + 28.7% = 32.3%). Source note: the statements "first and only," "new standard" and "best-in-class" belong to the investigator or the manufacturer and are attributed as such in the text. The evaluations and comments belong to this article. This article is not medical advice and no treatment decision should be made on the basis of this text.
References
- Skoulidis F, et al. Divarasib as Compared with Sotorasib or Adagrasib in KRAS G12C–Mutated NSCLC. N Engl J Med. 25 October 2026. doi:10.1056/NEJMoa2605703. ESMO 2026, LBA5.
- Roche. Roche's divarasib demonstrated best-in-class efficacy, advancing the standard of care in previously treated KRAS G12C-mutated lung cancer. 25 October 2026. roche.com
- Roche. Divarasib shows superiority in head-to-head phase III trial against approved KRAS G12C inhibitors in NSCLC (first announcement, July 2026). roche.com
- OncLive. Divarasib Improves Survival and Paves Way for New SOC in Pretreated KRAS G12C–Mutated NSCLC. 25 October 2026. onclive.com
- CancerNetwork. Divarasib Improves Survival vs Sotorasib or Adagrasib in KRAS G12C+ NSCLC. cancernetwork.com
- Targeted Oncology. Divarasib Improves Survival vs Approved KRAS G12C Inhibitors in NSCLC (July 2026; phase 1 context). targetedonc.com
- CancerNetwork. Sotorasib Improves Outcomes vs Docetaxel in KRAS G12C NSCLC (CodeBreaK 200). cancernetwork.com
- Adagrasib in KRYSTAL-12 has not broken the KRAS G12C enigma code. Lung Cancer: Targets and Therapy. dovepress.com
- ClinicalTrials.gov. NCT06497556 (Krascendo 1); NCT06793215 (Krascendo 2); NCT05789082 (Krascendo 170); NCT07541170 (Krascendo 3).
Was this article helpful?
Your feedback helps us improve our content.
