ASCO 2026: Immunotherapy Significantly Improves Survival in Advanced Endometrial Cancer
At the annual conference of the American Society of Clinical Oncology (ASCO), data from the phase 3 NRG-GY018 trial were presented. Adding pembrolizumab to chemotherapy for primary advanced or recurrent endometrial cancer led to sustained survival improvement, especially in patients with proficient mismatch repair, where median survival increased by 9.3 months.
ASCO 2026: 9.3 Months of Life — Why Keytruda's Victory Over Endometrial Cancer Was 'Pyrrhic'
[The Gist]: What's Really Happening
At ASCO 2026, Ramez N. Eskander from the University of California, San Diego, presented the long-awaited overall survival data from NRG-GY018 after 4 years of follow-up. The media will write: 'Pembrolizumab reduces the risk of death by 44% in dMMR and provides +9.3 months in pMMR.' But I'll tell you what's really happening.
The 9.3 months in the pMMR cohort is a number that masks a crisis of statistical significance. The hazard ratio was 0.86 with p=0.1072 — it did not reach the conventional significance threshold (p<0.05). Formally, if this were the primary endpoint, the trial would have failed.
But the real insight you won't see in press releases: these 9.3 months represent the difference against a backdrop of catastrophic 'contamination' of the control group. Between 81% and 93% of patients in the placebo group received immunotherapy after progression. That is, pembrolizumab didn't beat 'chemotherapy,' but 'chemotherapy with a 6-9 month delayed start of immunotherapy.' The difference was clinically meaningful but fell short of statistical significance precisely because the control group effectively received the same drug, just later.
Timeline and Context
The race in first-line endometrial cancer is a duel between Merck (Keytruda) and GSK (Jemperli). Key dates:
- March 2023 — NRG-GY018 and RUBY present first PFS data at SGO. Keytruda shows HR 0.30 in dMMR and 0.54 in pMMR, Jemperli shows 0.30 in dMMR and 0.76 in pMMR. Analysts immediately say Merck won the pMMR battle.
- June 2024 — FDA approves Keytruda + chemotherapy for all patients with advanced/recurrent endometrial cancer regardless of MMR status.
- 2025-2026 — GSK tries to catch up: their RUBY part 2 with niraparib is still ongoing, while Jemperli remains only for dMMR.
- May 2026 — ASCO 2026: 4-year OS data.
Importantly, in the dMMR cohort, the result was dramatic. 48-month survival: 78.6% vs 60.4% (HR 0.56; p=0.0124). A 44% reduction in risk of death is a blockbuster. Median OS was not reached even after 4 years. But dMMR accounts for only 25-30% of all endometrial cancer patients. The main market is pMMR, and there the statistics fell short.
Who Wins and Who Loses
Winners:
- Merck & Co — Keytruda generated $17.2 billion in 2025. Approval in first-line endometrial cancer expands an already massive market. Analysts estimate additional peak sales from this indication at $1.2-1.8 billion by 2028. Even with a non-significant p-value in pMMR, the FDA won't withdraw approval — the regulator already made its decision based on PFS and interim OS.
- Eisai — their lenvatinib (Lenvima) in combination with Keytruda is approved for pMMR patients in the second line. Lenvima sales grew 18% after KEYNOTE-775 showed 16.7% 5-year survival vs 7.3% with chemotherapy. Eisai receives royalties on every prescription.
- dMMR patients — for them, the result is revolutionary. Risk of death reduced by 44%, and that's despite 93% of the control group eventually receiving immunotherapy. 'Early introduction' provides maximum benefit — this is Eskander's key thesis.
Losers:
- GSK (Jemperli) — their drug showed in RUBY for pMMR an HR of 0.76 vs 0.54 for Keytruda. The efficacy difference is a 40% relative advantage in favor of Merck. GSK is trying to compensate with a combination with niraparib, but results are not yet available. Jemperli sales were only $7 million in 2025 — peanuts compared to Keytruda.
- Chemotherapy as monotherapy — the standard carboplatin + paclitaxel no longer exists in first-line for dMMR, and for pMMR its days are numbered. Survival 35.1 months vs 44.4 — a difference of nearly a year.
- Shareholders of companies developing new PD-1 drugs for this niche — the entry barrier has become insurmountable. Keytruda is already approved, OS data published, doctors are accustomed to the drug. A new player would need to show not just non-inferiority but superiority over Keytruda — nearly impossible with current HRs.
What the Media Isn't Saying
First and most important: The pMMR cohort in NRG-GY018 is a heterogeneous mix. The TCGA molecular classification divides endometrial cancer into 4 subtypes: POLE-mutant (hypermutant, excellent response to immunotherapy), dMMR, p53-aberrant (aggressive, 'cold'), and NSMP (no specific molecular profile). POLE and dMMR are 'hot' tumors. p53-abnormal and NSMP are 'cold.' The HR of 0.86 in pMMR means that within this group, some patients responded fantastically (those with actual POLE mutations misclassified as pMMR), while others barely responded at all. Without stratification by molecular subtype in the OS analysis, these numbers are useless for decision-making in a specific patient.
Second — a non-obvious insight: 93.2% of patients in the dMMR control group received immunotherapy after progression. Of these, 63.5% received PD-1 monotherapy, 17.6% received a combination with lenvatinib. This means the advantage of early-start pembrolizumab in dMMR persisted despite nearly everyone in the control group eventually receiving the same drug. This is strong evidence that immunotherapy should be in the first line, not delayed. But for pMMR, such confidence is lacking — 81.1% received ICI after progression, and early start gave only 9.3 months difference, which was not statistically significant.
Third: Eskander's presentation was given on May 29-30, 2026. But the OS data were frozen on April 14, 2026. The information fraction in dMMR was only 43% — meaning less than half of the patients needed for the final conclusion had died by the time of analysis. This is an early analysis, and median OS in dMMR has still not been reached. The next cut in 1-2 years may show an even greater divergence — or, conversely, a convergence of curves.
Fourth — an insight within an insight: The combination of pembrolizumab + lenvatinib in second-line pMMR gives a median OS of 17.4 months, while in first-line NRG-GY018 achieved 44.4 months. A 2.5-fold difference! But this is an incorrect comparison — different lines, different patient conditions. However, GSK is building its strategy on this: if their combination of Jemperli + niraparib in first-line yields 60+ months, they could claim their regimen is better. But data are not yet available.
Forecast: Next 30 Days and 90 Days
Next 30 days (through end of June 2026):
- Eskander's presentation will appear in the Journal of Clinical Oncology (JCO) as a full article. It will include detailed subgroup analyses by histological type — carcinosarcoma, serous, clear cell. These data will show who actually benefits from Keytruda and who does not.
- GSK will issue a press release announcing updated RUBY data with niraparib. Expect a 'spin': 'Our regimen is the only one to show benefit in carcinosarcoma.'
- Analysts at Bernstein or Leerink will revise their Keytruda sales models. Current consensus is peak sales of $22-24 billion in 2027-2028. After these data, an upward revision of 5-7% is possible.
Next 90 days (through end of August 2026):
- The New England Journal of Medicine will publish an editorial comparing NRG-GY018 and RUBY. Spoiler: the conclusion will be 'both work, but Keytruda is better in pMMR, while Jemperli is better studied in aggressive histotypes.' GSK will spend millions marketing this difference.
- The European Medicines Agency (EMA) may expand Keytruda's approval for first-line endometrial cancer based on these data (currently they have approval, but not in all countries). EMA is more conservative than FDA and may request additional analysis for pMMR.
- The first NCCN clinical practice guidelines version 3.2026 will appear, where Keytruda + carboplatin/paclitaxel becomes the preferred regimen for dMMR and recommended for pMMR (with a note 'in the absence of contraindications').
Longer-term forecast (12-18 months):
NRG-GY018 will continue follow-up. The final OS analysis, when the information fraction reaches 80-90% in dMMR (this will occur in 2-3 years), will show the true magnitude of benefit. My forecast: in dMMR, median OS in the pembrolizumab group will never be reached even after 8-10 years — patients will die from other causes. In pMMR, the difference will remain at 10-12 months but will still be statistically non-significant due to control group contamination.
What does this mean for practice? Doctors will still prescribe Keytruda to all patients with advanced endometrial cancer. Data are sufficient, side effects manageable, price (about $150,000 per course in the US) covered by insurance. They will just know now: for pMMR, those 9.3 months are real, but not guaranteed. And the main lesson of ASCO 2026: in oncology, the winner is not the one with the better p-value, but the one who first got FDA approval. Merck won this race back in 2024. ASCO 2026 simply confirmed that the victory was deserved.
— Editorial Team
