Nivolumab for mismatch-repair-deficient or hypermutated gynecologic cancers: a phase 2 trial with biomarker analyses

Endometrial and ovarian cancers are common and deadly malignancies in women. Molecular profiling has identified subtypes of endometrial cancer, including microsatellite instability (MSI) and mismatch repair deficiency (dMMR), which comprise roughly 30–35% of cases and have elevated mutation burden. PD-1 inhibitors are approved for MSI-H/dMMR tumors based on durable responses, yet robust predictors of response remain unclear. Prior studies suggest that etiology of dMMR (genetic versus epigenetic), mutations in JAK1 or B2M, and tumor microenvironment (TME) features (PD-L1 expression, CD8+ T cells) may influence outcomes, but their predictive value in gynecologic dMMR cancers is not established. This study evaluates the efficacy of nivolumab in dMMR/MSI-H advanced or recurrent endometrial or ovarian cancers and explores genomic and TME biomarkers predictive of benefit to PD-1 blockade.

Single-arm studies of pembrolizumab in MSI-H/dMMR solid tumors demonstrated durable responses (~50%), leading to tissue-agnostic FDA approval. Dostarlimab received accelerated approval for dMMR endometrial cancer post-platinum therapy. Phase 3 trials (RUBY, NRG-GY018) showed PFS benefit adding PD-1 blockade to chemotherapy in advanced/recurrent endometrial cancer, with substantial benefit in dMMR/MSI populations. Biomarker data from other cancers indicate PD-L1 expression and CD8+ TILs can be predictive, and T cell functional states (exhaustion/dysfunction markers such as PD-1 and TOX) may outperform cell counts. In endometrial cancer, prior reports linked genetic dMMR and certain mutations (e.g., JAK1, B2M) to resistance or response, but predictive biomarkers specific to dMMR gynecologic cancers remain undefined.

Design: Single-center, investigator-initiated, single-arm, phase 2 trial (MSK; NCT03241745). The trial adhered to ICH-GCP and the Declaration of Helsinki; IRB-approved; informed consent obtained. Trial accrual: first consent 09/27/2017; last consent 05/24/2021; completed 07/01/2022. Participants: Adults (female) with metastatic/recurrent uterine cancers (endometrial carcinoma, carcinosarcoma, clear cell carcinoma, leiomyosarcoma, undifferentiated sarcoma, high-grade endometrial stromal sarcoma) or endometrioid/clear cell carcinomas of ovary/fallopian tube/peritoneum not amenable to curative therapy. Required MSI-H or dMMR by IHC (loss of MLH1, MSH2, MSH6, PMS2) or MSI-H by MSK-IMPACT (MSIsensor) or hypermutated (≥20 nonsynonymous mutations on MSK-IMPACT). Additional criteria: ≥1 prior cytotoxic regimen for advanced disease; measurable disease by RECIST v1.1; ECOG 0–1; adequate organ function; no CNS metastases; key exclusions included active autoimmune disease requiring systemic therapy, prior checkpoint inhibitor therapy, significant uncontrolled comorbidity, active infections (HBV/HCV/HIV per protocol), and pregnancy. Intervention: Nivolumab 240 mg IV q2w until progression/toxicity; amended 10/30/2018 to 480 mg IV q4w per updated brochure. Radiologic assessments at baseline and every 12 weeks. Toxicities graded by CTCAE v5.0. Endpoints: Co-primary endpoints were ORR (RECIST v1.1) and PFS at 24 weeks (PFS24). Secondary endpoints: PFS, OS, duration of response (DOR), disease control rate (DCR), and safety. Exploratory endpoints: correlations between genomic/TME biomarkers and efficacy (ORR, PFS24). Sample size/statistics: Simon two-stage minimax design for both ORR (null 5%, alternative 25%) and PFS24 (null 25%, alternative 50%); type I error 0.025, type II error 0.05 (ORR) and 0.09 (PFS24). Stage I enrolled 23 eligible patients; prespecified thresholds triggered continuation. Final planned n=40; trial closed early at n=35 after meeting efficacy threshold. ORR per RECIST in evaluable patients (≥1 post-baseline assessment). PFS from start of treatment to progression/death/last follow-up; OS to death/last follow-up; DOR from first response to progression/death/last follow-up; Kaplan–Meier used for time-to-event estimates. Exploratory analyses used Mann–Whitney, Fisher’s exact tests; no multiple-testing adjustments. Correlative assays: Multiplex immunofluorescence (mpIF) on archival FFPE from 25 patients to quantify TME cell populations and spatial interactions in tumor compartment. Markers included CD8, PD-1, TOX, PD-L1, FOXP3, PAX8, DAPI. Spatial proximity analyses defined interactions within 50 µm. Whole-exome sequencing (WES) on tumors with matched germline (33/34 evaluable) for TMB, MSI by MSIsensor, mutational signatures (COSMIC v3.1), and pathogenic somatic mutations; germline status via MSK-IMPACT when consented.

• Enrollment and baseline: 35 treated patients; median age 64 (36–87); majority endometrioid endometrial cancer (83%). One patient was non-evaluable (progression and death before first assessment).
• Co-primary endpoints: Among 34 evaluable patients, ORR 58.8% (97.5% CI: 40.7–100%), including 7 complete responses and 13 partial responses. PFS24 rate 64.7% (97.5% one-sided CI: 46.5–100%), surpassing predefined thresholds.
• Secondary efficacy: DCR 73.5% (95% CI: 55.6–87.1%). At data cutoff, median follow-up 42.1 months (8.9–59.8). Median PFS 21.6 months (95% CI: 4.9–not evaluable). Median OS not reached; 1-year OS 79% (95% CI: 60.9–89.4%). Median DOR among responders not reached (median follow-up for responders 28 months).
• Safety: Any-grade TRAEs in 32/35 (91%); grade 3–4 TRAEs in 10/35 (29%); no grade 5 events. Common TRAEs (any grade): arthralgia 29%, fatigue 29%, pain 29%, pruritus 29%. Immune-mediated grade 3–4 events included myocarditis (3%), optic neuritis (3%), hemolysis (3%), and type 1 diabetes (3%); all managed per guidelines; non-endocrine grade 3/4 TRAEs led to discontinuation.
• Tumor microenvironment biomarkers (mpIF, n=25): Higher CD8+ T cell infiltration associated with clinical benefit (PFS ≥24 weeks) (P=0.026). Increased dysfunctional CD8+PD-1+ (P=0.006) and terminally dysfunctional CD8+PD-1+TOX+ T cells (P=0.001) strongly associated with benefit. PD-L1 expression alone (tumor or combined) not associated with benefit. Spatial proximity of PD-L1+ cells within 50 µm of CD8+ or CD8+PD-1+ T cells strongly associated with benefit. A multivariable model using percentage of CD8+PD-1+TOX+ (of CD8) and PD-L1+ proximity achieved AUC 0.897 (P≈0.0007) for predicting benefit.
• Genomics (WES, n=33): Frequent alterations observed in PTEN (76%), PIK3CA (48%), ARID1A (82%), JAK1 (24%), CTNNB1 (15%), Hedgehog pathway genes including MEGF8 (18%) and PTCH1 (18%). 30% had pathogenic somatic MMR gene mutations. MSI-H in 79% (26/33); 21% MSS. Dominant dMMR mutational signatures in 79% overall. No significant differences in TMB (median 18.1 vs 14.4 mut/Mb; P=0.24), MSI status (79% vs 79% MSI-H; P=1), or dominant dMMR signature (84% vs 71%; P=0.374) between benefit vs no-benefit groups. Mechanism of dMMR (genetic vs epigenetic, n=32) not associated with benefit (P=0.43). Mutations in MEGF8 (32% vs 0%; P=0.027) and SETD1B (58% vs 14%; P=0.015) were enriched in patients with PFS ≥24 weeks; SETD1B mutations associated with higher CD8+PD-1+TOX+ infiltration. No strong associations of PIK3CA, PTEN, JAK, CTNNB1 with benefit or immune recognition.

Nivolumab demonstrated robust activity in dMMR/MSI-H recurrent endometrial and ovarian cancers, meeting co-primary endpoints with high ORR and durable disease control, and an acceptable safety profile. Responses occurred irrespective of histologic subtype or dMMR mechanism. Traditional biomarkers such as PD-L1 expression and TMB did not predict benefit in this selected dMMR population. In contrast, T cell functional states—specifically the abundance of dysfunctional/terminally dysfunctional CD8+PD-1+ and CD8+PD-1+TOX+ cells—and their spatial proximity to PD-L1+ cells within the TME correlated strongly with PFS24, suggesting that markers of pre-existing antitumor T cell activity better capture immunogenicity and likelihood of response. Exploratory genomic analyses identified enrichment of MEGF8 and SETD1B mutations among patients with durable benefit, potentially linking Hedgehog signaling modulation and chromatin remodeling to heightened immune recognition. These findings refine biomarker strategies for PD-1 blockade in dMMR gynecologic cancers and support integrating functional/spatial T cell metrics into patient selection, while challenging the utility of TMB and PD-L1 alone in this context.

In this single-arm phase 2 study, nivolumab achieved a 58.8% ORR and 64.7% PFS24 in dMMR/MSI-H endometrial or ovarian cancers, with median DOR and OS not reached and a manageable safety profile. Predictive signal was not observed for PD-L1 expression or TMB; instead, dysfunctional/terminally dysfunctional CD8+ T cells and their spatial interactions with PD-L1+ cells associated with clinical benefit. Genomic alterations in MEGF8 and SETD1B were enriched among beneficiaries. These results suggest practical biomarker strategies based on limited-marker immunohistochemistry (CD8, PD-1, TOX, PD-L1) and motivate validation in larger, multi-center cohorts, alongside mechanistic studies of Hedgehog pathway and chromatin remodeling influences on antitumor immunity.

Single-center, single-arm design and small sample size limit generalizability and preclude comparative efficacy/safety conclusions. Archival tissue availability was limited, constraining correlative analyses. The cohort had a high proportion of high-grade disease, potentially not reflective of the broader dMMR endometrial cancer population. The number of patients with Lynch syndrome was small (n=5), limiting subgroup inference. Exploratory biomarker analyses were not adjusted for multiple comparisons, increasing risk of type I error.