Main

The addition of programmed cell death protein 1 (PD-1) inhibitors to chemotherapy has reshaped first-line treatment for metastatic gastroesophageal cancer, expanding therapeutic options and improving outcomes1,2. However, durable benefit remains limited, with a median overall survival (OS) of approximately 14 months3,4,5,6 and only about 17% of patients living beyond 3 years7, highlighting the need for strategies that can further extend and deepen benefit.

T cell immunoreceptor with immunoglobulin and ITIM domains (TIGIT) is an inhibitory checkpoint expressed on activated T cells and natural killer cells across multiple cancer types8. TIGIT suppresses T cell activation by outcompeting the activating receptor CD226 for the shared ligand CD155 (ref. 9). TIGIT expression correlates with PD-1, particularly in tumor-infiltrating T cells10, and the two pathways have distinct, nonredundant roles in regulating antitumor immunity11. Anti-TIGIT agents demonstrate promising antitumor activity across solid tumors when combined with anti-PD-1 therapies, with and without chemotherapy8,10,12.

Domvanalimab is an anti-TIGIT monoclonal antibody that, when combined with anti-PD-1 therapy, enhances immune cell activation compared with anti-PD-1 alone13. Domvanalimab is engineered to be Fc silent to avoid inducing antibody-dependent cellular cytotoxicity (ADCC), thereby preserving peripheral regulatory T cells (Tregs) that are critical for maintaining immune homeostasis14. This Fc-silent design may reduce autoimmune toxicities and immune-mediated side effects relative to Fc-enabled anti-TIGIT antibodies14. Zimberelimab, a fully human anti-PD-1 monoclonal antibody with high affinity binding for PD-1 (ref. 15), has demonstrated safety and efficacy across multiple tumor types13,16,17.

Chemotherapy can prime the tumor microenvironment by enhancing antigen release, increasing immune cell infiltration and reducing immunosuppressive cell populations18. In preclinical studies, chemotherapy followed by dual checkpoint blockade with domvanalimab and zimberelimab potentiated a robust and durable antitumor immune response14,19. In two phase 2 trials in patients with advanced non-small cell lung cancer, first-line treatment with domvanalimab and zimberelimab demonstrated improved outcomes versus zimberelimab13 and versus zimberelimab or chemotherapy20. Building on this rationale, we report results from arm A1 of the phase 2 EDGE-Gastric study (NCT05329766) of domvanalimab and zimberelimab in combination with oxaliplatin, leucovorin and fluorouracil (FOLFOX) as first-line treatment for patients with locally advanced unresectable or metastatic HER2-negative gastric (GC), gastroesophageal junction (GEJC) or esophageal adenocarcinoma (EAC).

Results

Study population and patient disposition

Between 30 August 2022 and 3 March 2023, 41 patients were enrolled across 20 sites (including hospitals, academic medical centers and clinical research units) in the USA, France and South Korea (Fig. 1a). No data were excluded from the arm A1 analyses. This study included both male and female human participants (n = 24 male and n = 17 female) aged 30–82 years.

Fig. 1: Trial profile and clinical response.
figure 1

a,b, Patient dispositions (a) and best percent change from baseline in sum of target lesions in patients with confirmed response (b). aOne patient did not receive leucovorin owing to institutional standard practice, but they did receive all other assigned study treatments. bOne patient is ongoing on leucovorin and fluorouracil but has discontinued domvanalimab and zimberelimab due to completion of the 2-year maximum treatment duration. The dashed reference lines indicate a 20% increase or 30% decrease from baseline in the sum of target lesions.

Full size image

Twenty-six (63%) patients had gastric adenocarcinoma, 29 (71%) patients were programmed cell death ligand 1 (PD-L1) positive (tumor area positivity (TAP) score ≥1%), 16 (39%) patients were TAP ≥5% and 11 (27%) patients were TAP <1%; there was one patient with known MSI-H status (Table 1). There was correlation between TAP score and combined positivity score (CPS) (r = 0.83, P < 0.0001). The overall percent agreement was 76% for TAP ≥5% versus CPS ≥5 and 87% for TAP≥1% versus CPS ≥1 (Extended Data Fig. 1).

Table 1 Baseline demographics and clinical characteristics
Full size table

At data cutoff (3 March 2025), median study follow-up was 26.4 months. All patients received at least one dose of each study drug except one, who did not receive leucovorin per institutional practice. Median duration of treatment exposure was 49.4 weeks (range <1–117 weeks).

Efficacy

The confirmed objective response rate (ORR) was 59% (90% confidence interval (CI) 44.5–71.6%), including 3 complete responses (7%) and 21 partial responses (51%) (Fig. 1b and Table 2). The disease control rate (DCR) was 93%. The median duration of response (DOR) was 12.4 months (90% CI 10.2–15.4 months) among the 24 patients with confirmed response.

Table 2 Tumor response
Full size table

Tumor responses were higher in TAP ≥1% and TAP ≥5% subgroups: ORR was 62% (90% CI 45.1–77.1%) and 69% (90% CI 45.2–86.8%), respectively, with DCR of 97% and 100%, respectively. Median DOR was 12.4 months (90% CI 10.2–15.9 months) and 15.4 months (90% CI 10.9–21.0 months), respectively. Tumor response was lower in patients with TAP <1%, with ORR of 46% (90% CI 20.0–72.9%).

Median progression-free survival (PFS) was 12.9 months (90% CI 9.8–14.6 months), with a 24-month PFS rate of 26% (90% CI 14.8–38.5%) (Fig. 2a). By PD-L1 status, median PFS was 13.2 months (90% CI 11.3–15.2 months) in TAP ≥1%, 14.5 months (90% CI 11.3 months to not estimable (NE)) in TAP ≥5% and 6.8 months (90% CI 3.0–13.8 months) in TAP <1% (Fig. 2b,c).

Fig. 2: Kaplan-Meier estimates of PFS.
figure 2

a–c, PFS in all treated patients (a), in patients who were TAP ≥1% and TAP <1% (b) and in patients who were TAP ≥5% and TAP <5% (c). One patient had no tissue available for central laboratory evaluation of PD-L1 expression; local laboratory results showed the patient’s tumor was PD-L1 low according to the 22C3 assay.

Full size image

Median OS was 26.7 months (90% CI 18.4 months–NE), with a 24-month OS rate of 50% (90% CI 36.3–62.6%) (Fig. 3a). Median OS was 26.7 months (90% CI 19.5 months–NE) in TAP ≥1%, not reached (90% CI 17.4 months–NE) in TAP ≥5% and 18.4 months (90% CI 12.2 months–NE) in TAP <1% (Fig. 3b,c). Corresponding 24-month OS rates were 54% (90% CI 37.3–67.7%), 56% (90% CI 33.9–73.6%) and 33% (90% CI 10.8–58.1%), respectively.

Fig. 3: Kaplan-Meier estimates of OS.
figure 3

a–c, OS in all treated patients (a), in patients who were TAP ≥1% and TAP <1% (b) and in patients who were TAP ≥5% and TAP <5% (c). One patient had no tissue available for central laboratory evaluation of PD-L1 expression; local laboratory results showed the patient’s tumor was PD-L1 low according to the 22C3 assay.

Full size image

Safety

All patients experienced at least one treatment-emergent adverse event (TEAE) (Table 3), most commonly nausea (59%) and a decrease in neutrophil count (44%) (Supplementary Table 1). Grade ≥3 TEAEs occurred in 30 (73%) patients, including neutrophil count decrease (37%), anemia (17%) and neutropenia (15%) (Supplementary Table 2). Grade ≥3 TEAEs attributed to domvanalimab and/or zimberelimab were reported in seven (17%) patients.

Table 3 Safety summary
Full size table

TEAEs leading to discontinuation of domvanalimab and/or zimberelimab were reported in four (10%) patients. The discontinuation of domvanalimab and/or zimberelimab was due to one event each of grade 2 blood alkaline phosphatase increased, grade 2 anxiety, grade 3 peripheral neuropathy and grade 3 ileus. Serious TEAEs were reported in 15 (37%) patients, none related to domvanalimab or zimberelimab. TEAEs that led to death occurred in one (2%) patient; the event term was listed as ‘death’ and was assessed as related to disease progression, not related to any study treatment.

Immune-related TEAEs were reported in 11 (27%) patients, including 9 (22%) related to domvanalimab and/or zimberelimab. The most common were hypothyroidism (12%), adrenal insufficiency (5%) and pneumonitis (5%). No grade ≥3 immune-mediated TEAEs were reported.

Infusion-related reactions were reported in 12 (29%) patients, including three (7%) related to domvanalimab and/or zimberelimab. Infusion-related reactions reported in more than one patient by preferred term were pyrexia (17%) and infusion-related reaction (7%). One grade 3 event (dyspnea) led to oxaliplatin discontinuation without further dose modifications.

Discussion

In the multicenter, international EDGE-Gastric study arm A1, we demonstrate durable disease control and long-term survival with dual TIGIT and PD-1 blockade using domvanalimab and zimberelimab plus FOLFOX as first-line therapy for advanced HER2-negative GC/GEJC/EAC. After a median study follow-up of 26.4 months, outcomes were encouraging, with an ORR of 59%, median DOR of 12.4 months, median PFS of 12.9 months and median OS of 26.7 months. Clinical activity was observed across PD-L1 subgroups. The safety profile was similar to that expected for anti-PD-1 therapy plus platinum-based chemotherapy1,3,4. These observations are being further investigated in the ongoing, randomized, phase 3 STAR-221 trial (NCT05568095), comparing domvanalimab, zimberelimab and chemotherapy with nivolumab and chemotherapy21.

Domvanalimab is an Fc-silent anti-TIGIT antibody, which may confer a differentiated safety profile compared with Fc-enabled designs. Fc-enabled anti-TIGIT antibodies can trigger ADCC, depleting activated T cells and Tregs in the tumor microenvironment and/or circulation14. While these effects may enhance immunity in some contexts, they can also reduce the pool of activated effector T cells and increase the risk of immune-mediated toxicities22,23. By contrast, the Fc-silent anti-TIGIT antibody domvanalimab preserves peripheral Tregs critical for maintaining immune homeostasis and is not associated with increased ADCC, which may mitigate autoimmune toxicities and provide durable antitumor activity when combined with PD-1 inhibition14,19,23.

Results from the EDGE-Gastric study arm A1 compare favorably with established first-line regimens of chemotherapy plus anti-PD-1 agents. Across pivotal phase 3 trials with nivolumab, pembrolizumab and tislelizumab, ORRs were 47–60%, median PFS was less than 8 months and median OS was 15 months or less1,3,4. By contrast, EDGE-Gastric arm A1 achieved a median PFS exceeding 12 months and a median OS approaching 27 months, suggesting that dual TIGIT and PD-1 blockade may extend the benefit of immunotherapy beyond current standards.

The findings are consistent with other early phase trials of anti-TIGIT, anti-PD-1 and chemotherapy in patients with advanced HER2-negative GC/GEJC, such as the phase 1b AdvanTIG-105 study and the phase 2 GEMINI-Gastric study, which reported ORRs of 53–57% overall and ~63% in patients who were PD-L1 high24,25. Together, these data suggest that patients who are PD-L1 high may derive particular benefit from TIGIT inhibition added to PD-1 blockade. These were small, early phase studies, and the combination of anti-TIGIT, anti-PD-1 and chemotherapy is being explored further in the ongoing phase 3 STAR-221 trial.

EDGE-Gastric is a proof-of-concept, open-label phase 2 study, with limitations related to sample size, study design and assay heterogeneity. FOLFOX was selected as the chemotherapy regimen in the EDGE-Gastric study arm A1 as it was widely regarded as a standard of care for patients with gastroesophageal cancer at the time of study design. It was not possible to determine the contribution of components of domvanalimab, zimberelimab and FOLFOX in arm A1, but the efficacy outcomes with the combination therapy exceed expectations from the existing data for standard of care, with a manageable toxicity profile.

In the EDGE-Gastric study, PD-L1 status was assessed by the VENTANA PD-L1 (SP263) assay with TAP score, which was recently used to support the approval of tislelizumab for the treatment of advanced PD-L1-positive GC/GEJC adenocarcinoma by the US Food and Drug Administration3,26. We observed concordance between the VENTANA SP263 assay with TAP scoring at ≥1% and ≥5% cutoffs, and the Dako PharmDx 28-8 PD-L1 assay with CPS scoring. This aligns with studies in GC that have shown high concordance in PD-L1 testing between three major PD-L1 assays (28-8, 22C3 and SP263) with TAP (≥1% and ≥5%) and CPS (≥1 and ≥5) scoring27,28,29. However, there are potential sources of variability for the concordance rate from sources other than the scoring algorithms. Namely, the EDGE-Gastric study employed multiple pathologists, and the same pathologist was not requested to perform both SP263 and 28-8 evaluations for each patient. Previous research has shown high interobserver variability of CPS scoring30. The 28-8 assay was performed retrospectively, and although serial sections adjacent to those used for the SP263 assay were taken, tissue sample heterogeneity cannot be ruled out as a potential source of variability.

Domvanalimab, zimberelimab and FOLFOX achieved durable responses and long-term survival with manageable toxicity in patients with previously untreated advanced HER2-negative gastroesophageal cancer. The Fc-silent design of domvanalimab may further optimize the therapeutic window by balancing efficacy with safety in the context of PD-1 blockade and chemotherapy. These findings provide the rationale for continued investigation of domvanalimab, zimberelimab and chemotherapy in advanced GC/GEJC/EAC in the ongoing phase 3 STAR-221 trial.

Methods

Study design

EDGE-Gastric (NCT05329766) is an ongoing, phase 2, open-label, multicenter clinical study with three cohorts, one in the first-line setting (cohort A) and two in the second-line or greater setting (cohorts B and C) (Extended Data Fig. 2). Cohort A comprises two nonrandomized (A1 and A2) and two randomized arms (A3 and A4). In arm A1, presented here, patients with no prior systemic treatment for locally advanced or metastatic GC/GEJC/EAC received intravenously administered domvanalimab 1,600 mg and zimberelimab 480 mg once every 4 weeks with FOLFOX (oxaliplatin 85 mg m−2, leucovorin 400 mg m−2, fluorouracil 400 mg m−2 on day 1 and fluorouracil 2,400 mg m−2 on days 1 and 2 (continuous 46–48-h infusion)) once every 2 weeks. The study is open label, so investigators were not blinded to the allocation of enrolled patients. EDGE-Gastric (NCT05329766) was registered on 8 April 2022 with ClinicalTrials.gov, https://clinicaltrials.gov/study/NCT05329766. Additional details are provided in the study protocol (online only).

Patients

Eligible patients were aged ≥18 years, had an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0 or 1, and had not received previous systemic treatment for locally advanced or metastatic disease. Patients with known HER2-positive status and patients with untreated, symptomatic or actively progressing central nervous system brain metastases were excluded. Complete eligibility criteria are presented in Supplementary Table 3.

Sex was recorded as a binary variable based on self-reported biological characteristics. Gender identity was not collected or analyzed in this study. No analyses stratified by sex were conducted.

Ethics approval and consent

The study was conducted in full conformance with the Declaration of Helsinki, the Council for International Organizations of Medical Sciences International Ethical Guidelines, institutional review board regulations and all other applicable local regulations. The study protocol was approved by the local ethics committee at each site (Supplementary Table 4). All patients provided written informed consent before any study procedures; patients were not compensated monetarily for participation in this trial.

Procedures

Lesions were assessed at screening and every 6 weeks thereafter through week 48 or end of treatment (whichever occurred first), then every 12 weeks thereafter until disease progression, initiation of a subsequent anticancer therapy, loss to follow-up, withdrawal of consent, study termination or death—whichever occurred first. Treatment was continued until disease progression, unacceptable toxicities, initiation of a subsequent anticancer therapy, physician or patient decision to withdraw, completion of a maximum treatment duration of 2 years or death—whichever occurred first. With the patient’s agreement, investigators could continue treatment beyond initial disease progression at their discretion based on the patient’s risk–benefit profile. Patients who experienced unacceptable toxicity, symptomatic deterioration due to disease progression or confirmed disease progression were discontinued from study treatment.

Tumor tissue was evaluated for PD-L1 expression at a central laboratory. Tumor samples were stained using the VENTANA PD-L1 (SP263) companion diagnostics assay (Roche Diagnostics). The VENTANA PD-L1 (SP263) assay is approved in the USA as a companion diagnostic to determine PD-L1 expression at a ≥1% tumor cell cutoff in patients with non-small cell lung cancer. Pathologists assessed PD-L1 expression in tumor cells and tumor-associated immune cells using the TAP score. The TAP score is defined as the total percentage of the tumor area (tumor and any desmoplastic stroma) covered by tumor cells with membranous PD-L1 staining at any intensity and tumor-associated immune cells with PD-L1 staining at any intensity, according to visual estimation. As previously published, TAP scores of ≥1% and ≥5% were used as cutoffs31. Patients with TAP score ≥1% were considered PD-L1 positive and <1% were considered PD-L1 negative. Patients with TAP score ≥5% were considered PD-L1 high and <5% were considered PD-L1 low. In the assessment of concordance between the SP263 assay with TAP score and the 28-8 assay with CPS, the 28-8 assay was performed retrospectively on serial sections to minimize effects from tissue heterogeneity.

Endpoints and assessments

The dual primary endpoints of safety and investigator-assessed ORR were evaluated in patients who enrolled and received any study treatment (treated analysis population). ORR was defined as the percentage of patients with a confirmed best overall response (BOR) of complete response or partial response, per RECIST v1.1. Patients who discontinued before completing postbaseline tumor assessments were considered nonresponders.

Secondary endpoints were OS, PFS, DCR and DOR overall and by PD-L1 expression, as well as ORR by PD-L1 expression. OS was defined as the time from first dose until death due to any cause. PFS was defined as the time from first dose until first documentation of progressive disease or death due to any cause. DCR was defined as the percentage of patients with a confirmed BOR of complete response, partial response or stable disease. DOR was defined as the time from date of initial response (complete response or partial response) until the date of first documented disease progression or death due to any cause (in confirmed responders only).

Adverse events (AEs) were assessed in the treated analysis population and coded using the Medical Dictionary for Regulatory Affairs v25.0. Severity was assessed according to the National Cancer Institute Common Terminology Criteria for Adverse Events, version 5.0. Investigators assessed whether an AE was related to study treatment. Infusion-related reactions were defined as AEs that occurred ≤1 day after the end of study drug infusion administration (within 24 h if time was available), were ≤2 days in duration and were in the custom AE preferred term search list (infusion-related reaction, pyrexia, chills, rigors, hypotension, dyspnea, wheezing, urticaria, flushing, back pain, abdominal pain, drug hypersensitivity, anaphylactic reaction, hypersensitivity, type 1 hypersensitivity, pruritus or rash). Immune-related AEs were defined as all AEs of any grade in the custom PD-1 immune-related AE search list, except for preferred terms containing “PD-1 skin toxicities”, for which only grade ≥3 AEs were included.

Statistical analysis

The EDGE-Gastric study was prespecified to include three cohorts (A, B and C) and cohort A was prespecified to include four arms (A1–A4) (Extended Data Fig. 2). All cohorts and arms have a prespecified enrollment plan and statistical design. The planned enrollment in arm A1 was approximately 40 patients, of which approximately 50% would have PD-L1 high expression. The sample size justification was based on an estimation framework, and the study was designed for descriptive statistical analysis rather than formal statistical hypothesis testing with Type I error and power considerations.

Median OS, median PFS, median DOR, PFS rate at 24 months and OS rate at 24 months were estimated using the Kaplan–Meier method. For PFS, OS and DOR, the 90% CI was based on the log–log transformation of the corresponding survival function. Exact binomial Clopper–Pearson 90% CIs were constructed for ORR. The median duration of study follow-up was calculated as the time from cycle 1, day 1, until the data cutoff date, regardless of events or dropout. Statistical analyses were performed using SAS software, version 9.4.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.