Four randomized controlled trials. Four distinct PD-1 or PD-L1 inhibitors. Hazard ratios for progression-free survival in the dMMR subgroup across the four trials cluster in a similarly dramatic range, with RUBY reporting 0.28. The data package assembled over the past three years for this patient population is, by any measure, extraordinary. The RUBY trial’s 4-year follow-up shows a 72.8% overall survival rate with dostarlimab plus carboplatin-paclitaxel versus 40.3% in the control arm, with median overall survival not yet reached in the immunotherapy group. Oncologists are using the word “cure” with increasing confidence. But every one of these trials was designed around the same foundational assumption, an assumption nobody actually tested: that carboplatin and paclitaxel still belong in the regimen at all.

That is not a small question. It is the question that determines whether tens of thousands of dMMR patients worldwide receive six cycles of myelosuppressive chemotherapy they may not need, alongside an immunotherapy that may already be doing all the work.

Four Trials, One Untested Premise

Pull up the RUBY protocol. Then NRG-GY018. Then AtTEnd. Then DUO-E4. Each one randomizes patients to chemotherapy plus immunotherapy versus chemotherapy plus placebo. Each one demonstrates, convincingly, that adding immunotherapy to the backbone produces dramatic benefit in the dMMR subgroup. What none of them does is include a third arm: immunotherapy alone, without chemotherapy. The control experiment that would actually answer whether carboplatin-paclitaxel contributes anything meaningful in this molecularly distinct population was never run.

This matters enormously for how we interpret the survival numbers. When RUBY reports that 4-year overall survival jumps 32.5 percentage points with dostarlimab added to chemotherapy, that result tells us the combination outperforms chemotherapy alone. It does not tell us how much of that benefit is immunotherapy-driven and how much, if any, is contributed by the cytotoxic backbone. The HR of 0.28 for progression-free survival is stunning. It could be entirely immunotherapy. The trial design cannot distinguish between these possibilities.

The NRG-GY018 trial, which enrolled 809 patients and presented updated overall survival data at ASCO 2026, confirms sustained benefit with pembrolizumab added to carboplatin-paclitaxel. The AtTEnd trial showed median progression-free survival was not estimable in the dMMR arm at primary analysis, a remarkable ceiling effect. The DUO-E trial, published in the Journal of Clinical Oncology, rounds out a quartet of consistent, convergent evidence. The replication is real. The gap in the evidence base is equally real.

What KEYNOTE-C93 Started to Answer

Now consider what happens when you redesign the question. The Phase 3 KEYNOTE-C93 trial (also registered as GOG-3064/ENGOT-en15) tested pembrolizumab monotherapy in dMMR advanced or recurrent endometrial cancer patients who had not previously received systemic chemotherapy for advanced disease. That trial design, pembrolizumab without any cytotoxic backbone, is the experiment the four combination trials never ran. Results showing progression-free survival benefit with monotherapy create a direct evidential problem for the assumption baked into RUBY and its peers: if pembrolizumab alone moves the needle in chemotherapy-naive dMMR patients, the burden of proof for retaining carboplatin-paclitaxel in the combination becomes substantially heavier.

The common assumption in oncology protocol design is that combining cytotoxics with checkpoint inhibitors produces additive or synergistic benefit, so removing the chemotherapy backbone risks efficacy. That assumption is reasonable in mismatch repair-proficient disease, where PD-1 inhibitors have far more modest single-agent activity. But dMMR tumors are biologically distinct. They carry high mutational burden, generate abundant neoantigens, and mount robust T-cell infiltration. The molecular rationale for why carboplatin-paclitaxel remains essential in this specific context has never been formally established. The assumption traveled from the pMMR world into the dMMR trial designs without being stress-tested.

Which raises a question that should be uncomfortable for anyone who sat on those protocol committees: were these trials designed to answer the right question for the dMMR subgroup, or were they designed to answer the fastest approvable question for the overall endometrial cancer population, with dMMR a fortunate beneficiary of the statistical harvest?

The Toxicity Debt Accumulating in the Interim

While the trial design debate remains academic, the toxicity burden on actual patients is not. The GOG0209 trial, the foundational phase III study that established carboplatin-paclitaxel as the standard backbone for advanced endometrial cancer, enrolled 1,381 women between 2003 and 2009. That trial documented the toxicity profile of carboplatin-paclitaxel comprehensively: peripheral neuropathy, myelosuppression, alopecia, fatigue, and the long-tail neurotoxicity that persists years after treatment ends. For patients who might be cured by dostarlimab or pembrolizumab alone, every cycle of carboplatin-paclitaxel delivered on top of that immunotherapy represents unnecessary harm.

The 4-year RUBY data suggest that some dMMR patients are being cured. That word carries clinical weight. A patient cured by immunotherapy who also received six cycles of carboplatin-paclitaxel experienced toxicity that served no oncologic purpose. Multiply that across the roughly 25% of endometrial cancers that are dMMR, a figure cited in the literature, and the aggregate toxicity burden of an untested assumption becomes a concrete ethical problem, not a theoretical one.

Sponsors and trial designers tend to frame the chemotherapy backbone as the conservative, safe choice. Retaining it protects against the possibility that removing it would cost efficacy. But conservative from whose perspective? For the regulatory agency trying to minimize approval risk, retaining chemotherapy is conservative. For the patient sitting across from a gynecologic oncologist who is about to prescribe a treatment that may cause permanent peripheral neuropathy on top of a curative immunotherapy, conservative means something entirely different.

The Protocol Design Failure and What Comes Next

The more instructive comparison for trial designers is what adaptive platform trial methodology could have achieved here. A seamless Phase 2/3 adaptive design with pre-specified rules for de-escalating the chemotherapy backbone in the dMMR subgroup, triggered by interim biomarker and efficacy data, would have allowed the evidence to accumulate while simultaneously testing the chemotherapy question. The technology for this kind of design exists. The RUBY, NRG-GY018, AtTEnd, and DUO-E trial teams were not operating without tools. They made a choice, and that choice prioritized regulatory certainty over scientific completeness.

The practical consequence for sponsors now designing the next generation of dMMR endometrial cancer trials is that the bar for a chemotherapy de-escalation study has actually been raised by the combination trial results. Regulators will ask: why remove a component from a regimen with 72.8% 4-year overall survival? The answer requires either a head-to-head randomized comparison of immunotherapy monotherapy versus the combination in the dMMR population, or a biomarker-stratified design that can identify within the dMMR group which patients derive zero additional benefit from the cytotoxic backbone. Neither is a trivial undertaking. Neither was designed prospectively when the data to plan them was most readily available.

Protocol teams now designing dMMR endometrial cancer studies will likely face questions from regulators about how their designs address the unresolved chemotherapy question, given that the RUBY, NRG-GY018, AtTEnd, and DUO-E datasets are part of the public record. A sponsor who arrives without a de-escalation hypothesis or a biomarker-stratification strategy will be designing around yesterday’s evidence in a space where the clinical question has already moved forward.

The 72.8% 4-year overall survival figure in RUBY’s dMMR population is one of the most striking numbers in recent gynecologic oncology. It deserves the follow-up question it was never designed to answer: how much of that number belongs to carboplatin and paclitaxel, and how much belongs to dostarlimab? Until a trial actually asks, every dMMR patient receiving combination therapy is living inside an assumption that four landmark trials collectively chose not to test.

References

  1. The Lancet, “Is chemotherapy still needed in all patients with mismatch repair-deficient advanced or recurrent endometrial cancer?”
  2. ASCO Post, “Four-Year RUBY Data Support Curative Intent with Dostarlimab in dMMR/MSI-H Endometrial Cancer”
  3. NRG Oncology, “Addition of Immunotherapy Continues to Show Survival Benefit in Prolonged Follow-Up in NRG-GY018”
  4. ASCO Post, “Recurrent or Advanced Endometrial Cancer: Atezolizumab Plus Chemotherapy (AtTEnd Trial)”
  5. Journal of Clinical Oncology via PubMed, DUO-E Trial Primary Publication
  6. Focus on Oncology, “Pembrolizumab Monotherapy Extends PFS in dMMR Endometrial Cancer: KEYNOTE-C93”
  7. PMC, GOG0209 Phase III Trial: Carboplatin-Paclitaxel Toxicity in Endometrial Cancer
  8. PMC, dMMR Prevalence in Endometrial Carcinoma: Epidemiologic Data
Website | + posts

Moe Alsumidaie, MBA, MSF, is founder and Chief Editor of Vanguard Publications, which publishes Clinical Trial Vanguard, Pharma Vanguard and BullScope, and Head of Research at CliniBiz. He has two decades in clinical trial operations and data science, with earlier roles at Genentech, Abbott Vascular and Stanford University Medical Center, and is a guest lecturer in clinical trial sciences at Rutgers University.