7/ Also at #ESMO26: the KEYNOTE-B21 ctDNA analysis in dMMR (Signatera, tumor-informed). Only ~20% were ctDNA+ after surgery (and importantly BEFORE chemo not after), and in stage I/II it was 5% and 18% (n=44). That’s lower than I expected for resected high-risk disease.

@DrKatyMoore Still need no treatment arm here. Excellent response rates for recurrent maybe avoid RT altogether for 80%.

11/ Putting it together: in a GY020-type population, postop ctDNA would flag maybe 1 in 10 pts. If we want ctDNA to select who needs IO in early dMMR EC, we need more sensitive assays and testing that extends past 2-3 yrs. #gyncsm

10/ Serial draws mostly ended around wk 96, so late recurrences in ctDNA− pts had no sample positioned to catch them. dMMR tumors carry thousands of variants, and a 16-plex panel may leave sensitivity unused.

9/ The harder finding: in the placebo arm, 15 of 19 DSFS events occurred in pts who were ctDNA− at baseline (sensitivity 21%, NPV 80%). A negative postop test isn’t ready to drive de-escalation. (figure made with claude ai)

8/ Pembro + chemo cleared ctDNA: 21%→2.6% at C2D1→0% at EOT and follow-up. The placebo arm stayed 7-10% positive. ctDNA+ pts on pembro still did well (2 DSFS events in 19). ctDNA+ dMMR disease looks rescuable with IO.

6/ GY020 doesn’t show that IO does nothing in early dMMR EC, and it doesn’t show that it helps. The question is still open. The next trial needs a realistic effect size and an endpoint that doesn’t count dropout as recurrence.

5/ A caveat in the other direction: grade 3 was 46% on RT alone vs 27% on pembro, and LVSI+ was 62% vs 55%. That imbalance favors the pembro arm, so the HR may be optimistic. OS is immature (9 events).

4/ Time-to-event RFS: 88% vs 81% at 3 yrs, HR 0.63 (0.30-1.34). With 28 events, power to detect an HR of 0.63 was ~20-30%. Calling it would take ~160 events.

3/ The binary endpoint counted consent withdrawal and lost follow-up as failures. Withdrawal was 11% on pembro vs 5% on RT. Among pts with known status, success was 88% vs 80%. The composite cut the observed gap roughly in half.

2/ GY020 was powered for 80%→95% 3-yr RFS. That’s a 75% reduction in failures, roughly HR 0.23, for single-agent IO added to RT in early-stage disease where VBT already controls most disease. A big ask for 168 pts. (figure made with claude ai)