MMR-Proficient Endometrial Cancer: Tempted to Add a Third Drug the Trial Never Tested
Pembrolizumab's benefit in mismatch-repair-proficient endometrial cancer is real but far more modest than in mismatch-repair-deficient disease — and older data suggests bevacizumab helps too, in a population that overlaps but was never the same one. The pull to combine three active drugs runs ahead of any trial that has actually tested doing so.
Beatrice O., 68, keeps three beehives behind her garage, a hobby her late husband started and that she took over rather than let lapse, and she scheduled her hysterectomy for a week she knew would fall between honey harvests. Pathology returned Stage IIIC endometrioid endometrial carcinoma, and mismatch-repair immunohistochemistry showed intact staining across all four proteins — mismatch-repair-proficient disease, the molecular subgroup that makes up the clear majority of advanced endometrial cancer but responds to immunotherapy far less dramatically than the mismatch-repair-deficient minority.
NRG-GY018's own pMMR subgroup, added chemotherapy plus pembrolizumab against chemotherapy alone, found a real hazard ratio near 0.54 for progression or death — a genuine, clinically meaningful benefit, just a fraction of the near-0.30 effect seen in the trial's mismatch-repair-deficient patients. RUBY's own pMMR subgroup found a more modest, less statistically secure result, a divergence that matters because neither trial's pMMR subgroup was the study's primary powered comparison — both were secondary analyses, which means comparing their two effect sizes directly carries more uncertainty than comparing the trials' headline dMMR results did. Separately, and from an older and considerably weaker body of evidence, GOG-0086P tested adding bevacizumab (not immunotherapy) to chemotherapy in advanced or recurrent endometrial cancer without any biomarker selection. Its design is the part that matters here: it was a randomized phase II with no concurrent chemotherapy-alone arm, measured instead against historical controls drawn from GOG-0209, and its primary endpoint of progression-free survival was not significantly improved in any of its three experimental arms. The bevacizumab arm's longer overall survival was a secondary observation against that historical comparison, not a trial result of the kind GY018's is. No trial has combined all three drugs together in a pMMR population specifically, which leaves the actual decision in front of Beatrice's team as whether the mechanistic case for combining a validated but modest immunotherapy benefit with an older, non-biomarker-selected antiangiogenic benefit is strong enough to act on before a trial has tested the combination itself.
Beatrice's own medical history includes well-controlled hypothyroidism, itself autoimmune in origin, which the team flagged as a modest but real consideration before starting pembrolizumab — checkpoint inhibitors can unpredictably worsen an existing autoimmune condition, though data specific to that interaction remains too thin to change today's decision, only to shape how closely her thyroid function gets followed once treatment starts.
Two real benefits, no trial that combined them
I'd add bevacizumab alongside chemotherapy and pembrolizumab. GOG-0086P saw longer overall survival in its bevacizumab arm in an unselected endometrial cancer population, and checkpoint blockade and antiangiogenic therapy work through complementary mechanisms, not redundant ones. I don't think we should withhold a drug with its own positive trial simply because the specific triple combination hasn't been formally tested yet.
I'd stop at chemotherapy plus pembrolizumab, exactly as GY018 tested it in the pMMR subgroup. That hazard ratio of roughly 0.54 is real and it's specific to this biomarker group. GOG-0086P is a genuinely different, older, unselected trial, and it isn't even a clean positive — it missed its own primary endpoint, and its survival signal comes from a comparison against historical controls rather than a randomized arm. Reaching for that to justify a third drug is combining two results the way you'd combine two reagents that happen to sit near each other on a shelf, not the way an actual combination trial would tell us to.
"They work through complementary mechanisms" is a real biological argument, but it's the same argument that's justified plenty of combinations that turned out to add toxicity without adding benefit once someone actually tested them together.
There's a caution underneath this whole disagreement worth naming directly: neither RUBY's nor GY018's pMMR subgroup was the trial's primary powered comparison, and the two trials' pMMR results diverge more than their dMMR results did. That's a reason to be careful about treating GY018's 0.54 as a fixed, load-bearing number — and it's an even stronger reason not to stack an older, differently-selected trial's drug on top of it before anyone has tested the combination.
Agreed: carboplatin-paclitaxel with pembrolizumab per NRG-GY018's pMMR-subgroup regimen, without bevacizumab.
Not fully agreed: the medical oncologist's underlying instinct that complementary mechanisms justify combining two independently positive trials wasn't disproven, just judged insufficiently tested to act on today — left explicitly open as a question for a future line of therapy or a clinical trial, rather than settled as wrong.