Efficacy and safety of neoadjuvant toripalimab plus chemotherapy in localized deficient mismatch repair/microsatellite instability-high gastric or esophagogastric junction adenocarcinoma (NICE): a multicentre, single-arm, exploratory phase 2 study

Summary

Background

In locally advanced gastric or gastroesophageal junction adenocarcinoma (GC/EGJC), deficient mismatch repair/microsatellite instability-high (dMMR/MSI-H) tumors exhibit high responsiveness to immunotherapy. The synergistic efficacy of neoadjuvant immunotherapy combined with chemotherapy in dMMR/MSI-H GC/EGJC remains uncertain.

Methods

The NICE trial is a multicentre, single-arm, exploratory phase 2 study conducted at six hospitals in China, evaluating the safety and efficacy of toripalimab in combination with CapeOX as perioperative therapy for locally advanced GC/EGJC across three biomarker-defined cohorts. This report presents findings from cohort C. Eligible patients were aged 18–75 years with histologically or cytologically confirmed GC/EGJC, confirmed dMMR/MSI-H status, and clinically staged as cT3–4aNxM0 or cT2N + M0 (AJCC 8th edition) based on contrast-enhanced CT or MRI, upper endoscopy, diagnostic laparoscopy, and peritoneal lavage cytology. Patients received four cycles of neoadjuvant toripalimab (240 mg IV every 3 weeks) plus CapeOX (capecitabine 1000 mg/m² orally twice daily on Days 1–14 and oxaliplatin 130 mg/m² IV on Day 1), followed by curative-intent surgery and up to four cycles of the same regimen as adjuvant therapy. The primary endpoint was the major pathological response (MPR) rate, defined as ≤10% residual viable tumor cells in the tumor specimen resected after neoadjuvant therapy. All patients who received at least one dose of treatment were included in the efficacy and safety analyses. The trial is registered with ClinicalTrials.gov, NCT04744649.

Findings

Between March 12, 2021 and June 1, 2024, twenty-two patients were screened, with sixteen meeting the inclusion criteria and undergoing treatment. Tumor stages were cT2N1 (n = 1), cT3N0-3 (n = 3), and cT4aN1-3 (n = 12). Fifteen patients completed four cycles of therapy preoperatively, while one patient completed two cycles due to adverse events. None of patients experienced disease progression. One patient achieved a complete clinical response as indicated by radiology and endoscopy and consequently refused surgery, while the remaining fifteen patients underwent resection. The R0 resection rate was 100% (15/15). The MPR rate was 93.3% (14/15), and the pathological complete response (pCR) rate was 80% (12/15). Six patients (37.5%, 6/16) experienced grade 3/4 treatment-related adverse events. One patient died of COVID-19 287 days post-surgery without relapse. No disease relapse was observed in any patient.

Interpretation

Given the small sample size and limited population diversity, these findings should be interpreted with caution. Nonetheless, neoadjuvant toripalimab combined with the CapeOX regimen is feasible for localized advanced dMMR/MSI-H GC/EGJC, demonstrating high MPR and pCR rates without unexpected adverse events.

Funding

Noncommunicable Chronic Diseases-National Science and Technology Major Project, Beijing Hospitals Authority Clinical Medicine Development, Beijing Natural Science Foundation, Key Clinical Technique of Guangzhou, Key Areas Research and Development Programs of Guangdong Province, National Natural Science Foundation of China, Natural Science Foundation of Guangdong Province, Clinical Research Program of Nanfang Hospital.

Research in context.

Evidence before this study

We conducted a PubMed search using the terms (“gastric cancer” or “gastro-esophageal junction cancer”) and (“deficient mismatch repair” or “microsatellite instability-high” or “dMMR” or “MSI-H”) and (“neoadjuvant therapy” or “perioperative treatment”) and (“immunotherapy” or “immune checkpoint inhibitors” or “PD-1/PD-L1 blockades” or “anti-PD-1/PD-L1” or “CTLA-4”). The search was limited to clinical trials, with no language restrictions from database inception to June 1, 2024. Two pivotal phase II trials, NEONIPIGA and INFINITY, investigated neoadjuvant immunotherapy in resectable dMMR/MSI-H gastric or gastroesophageal junction adenocarcinoma (GC/EGJC). NEONIPIGA, a French single-arm study, reported a 58.6% pathological complete response (pCR) rate with nivolumab plus low-dose ipilimumab and no unexpected immune-related adverse events. INFINITY, a multicohort trial, evaluated tremelimumab plus durvalumab, showing a 60% pCR and 80% major-complete pathological response (<10% viable cells) in Cohort 1. However, the synergistic efficacy of neoadjuvant immunotherapy combined with chemotherapy in dMMR/MSI-H GC/EGJC remains uncertain.

Added value of this study

To our knowledge, our study is the first phase II trial reporting the efficacy and safety of adding an anti-PD-1 monoclonal antibody to perioperative doublet chemotherapy for dMMR/MSI-H GC/EGJC. Our results demonstrated that combining toripalimab with neoadjuvant chemotherapy provided encouraging antitumor activity with a pCR rate of 80% and an acceptable safety profile in dMMR/MSI-H GC/EGJC.

Implications of all the available evidence

Our findings, together with prior evidence, suggest that patients with localized advanced dMMR/MSI-H GC/EGJC may benefit from neoadjuvant immunotherapy-based approaches, with encouraging pCR rates and manageable safety profiles. However, the absence of a control arm, small sample size, lack of long-term survival data, and the China-only multicentre setting limit the generalizability and interpretation of efficacy. These results provide a rationale for further validation in larger, randomized trials across diverse populations.

Introduction

Gastric or gastroesophageal junction adenocarcinoma (GC/EGJC) ranks as the fifth most common malignancy and the fourth leading cause of cancer-related mortality.¹ Patients diagnosed with GC/EGJC often present at advanced stages, typically resulting in a poor prognosis. Recently, immune checkpoint inhibitors (ICIs) targeting programmed cell death protein 1 (PD-1) or programmed death ligand 1 (PD-L1) in combination with chemotherapy achieved superior outcomes for advanced GC in phase III randomized controlled trials.^2,3 ICIs combined with chemotherapy have now become the standard treatment for advanced GC/EGJC.

For patients with locally advanced GC/EGJC, the advent of immunotherapy represents a new frontier in treatment options, with drugs targeting PD-1/PD-L1 expected to significantly enhance prognosis. In KEYNOTE-585 study, pathological complete response (pCR) rate in the pembrolizumab combined with cisplatin-based chemotherapy group was significantly higher than in the placebo combined with cisplatin-based chemotherapy group (12.9% vs. 2%, P < 0.0001).⁴ Preliminary results from the MATTERHORN study reported that the combination of the PD-L1 inhibitor durvalumab with the FLOT regimen significantly increased the pCR rate in resectable GC/EGJC compared to the FLOT regimen alone (19% vs. 7%, P < 0.0001).⁵ Current ongoing studies indicate that perioperative immunotherapy combined with chemoradiation and/or targeted therapy significantly improves tumor shrinkage and pCR rates in locally advanced GC/EGJC predominantly characterized by microsatellite stability (MSS), with pCR rates ranging approximately from 10% to 36.4%.6, 7, 8

The synergistic efficacy of immunotherapy combined with chemotherapy in the treatment of deficient mismatch repair/microsatellite instability-high (dMMR/MSI-H) GC/EGJC remains uncertain and necessitates further elucidation. The dMMR/MSI-H phenotype has been widely recognized as a distinct molecular subclass of tumors, characterized by unique biological features and specific clinical manifestations. These tumors typically exhibit a high mutation burden, express cancer-specific neoantigens, and are responsive to PD-1/PD-L1 blockade.^9,10 The dMMR/MSI-H phenotype occurs in approximately 5%–20% of patients with GC/EGJC and has become a significant predictive biomarker for assessing the efficacy of ICIs.^11,12 In patients with locally resectable dMMR/MSI-H GC/EGJC, neoadjuvant dual immunotherapy has significantly increased the pCR rate, as demonstrated in the phase II trials NEONIPIGA (58.6%) and INFINITY (60%).^13,14 Existing studies have demonstrated that chemotherapy may synergize with ICIs by modulating tumor cell immunogenicity, promoting antigen presentation, and ameliorating the immunosuppressive tumor microenvironment.¹⁵ Efficacy of immunotherapy with chemotherapy as standard first-line treatment against advanced gastric cancer has been firmly established, primarily based on robust evidence from landmark clinical trials including CheckMate-649, RATIONALE-305, and ORIENT-16.^2,3,16, 17, 18 However, the efficacy of the combined regimen as neoadjuvant modality remains to be fully elucidated for locally advanced dMMR/MSI-H GC. To address this unmet clinical need, we conducted the NICE study, a prospective phase II clinical trial evaluating the clinical value of perioperative toripalimab (a PD-1 inhibitor) combined with chemotherapy in resectable dMMR/MSI-H GC/EGJC.

Methods

Study design and patients

This phase 2, open-label, single-arm, multicenter study was conducted across China (NCT04744649), with the aim of investigating the safety and efficacy of toripalimab combined with the CapeOX regimen as perioperative treatment for localized advanced GC/EGJC in three cohorts: PD-L1 CPS ≥5 (cohort A), EBV-positive (cohort B), and dMMR/MSI-H (cohort C). In present study, we report the findings of Cohort C. Eligible patients were aged 18–75 years with histologically confirmed, resectable GC/EGJC (cT3/4aNxM0 or T2N + M0, according to AJCC 8th edition, determined by CT/MRI, diagnostic laparoscopy and peritoneal lavage cytology before treatments). dMMR/MSI-H status was confirmed by immunohistochemistry or genetic/transcriptional profiling. Additional inclusion criteria were: Eastern Cooperative Oncology Group (ECOG) performance status 0–1, expected survival ≥12 weeks, and adequate hematologic, hepatic, and thyroid function. Patients were required to test positive for at least one biomarker (PD-L1 CPS ≥5, EBV, dMMR, or MSI-H). Key exclusion criteria included unresectable or metastatic disease, prior systemic therapy for gastric cancer, HER2-positive status, recent use of immunosuppressants or investigational drugs, autoimmune or immunodeficiency disorders, and active infections or other uncontrolled comorbidities. The complete inclusion and exclusion criteria can be found in the protocol (Supplementary Data S1). The enrollment period was from March 12, 2021 to July 21, 2023, with the final follow-up date being June 1, 2024.

The study protocol was approved by the Ethics Committee of Nanfang Hospital (NFEC-2021-016), Sun Yat-sen University Cancer Center (SL-B2022-375-01), Guangdong Provincial Hospital of Chinese Medicine (BF2021-069-01), Peking University Shenzhen Hospital, (P2021-048), Zhongshan City People's Hospital, (K2021-052), The Sixth Affiliated Hospital, Sun Yat-sen University (2021ZSLYEC-368). Written informed consent was obtained from all participants prior to study enrollment. This trial was conducted in compliance with the Declaration of Helsinki and Good Clinical Practice guidelines.¹⁹

Procedures

Participants received four cycles of neoadjuvant therapy consisting of toripalimab plus CapeOX. The regimen included toripalimab (240 mg, intravenously, Day 1 of each 21-day cycle); capecitabine (1000 mg/m², orally, twice daily on Days 1–14); and oxaliplatin (130 mg/m², intravenously, Day 1). Surgery was performed four weeks after the completion of neoadjuvant therapy. The type of surgery depended on the location and extent of the primary tumor. Postoperative adjuvant therapy was initiated within 6 weeks after surgery, provided the patient had adequate recovery and tolerance, followed by standard chemotherapy cycles. Baseline and preoperative tumor staging for patients was determined by contrast-enhanced abdominal CT, diagnostic laparoscopy, and peritoneal lavage cytology. Treatment-related adverse events as assessed by National Cancer Institute-Common Terminology Criteria for Adverse Events (NCI-CTCAE) v4.03 throughout the study. Surgical complications were classified according to the Clavien-Dindo grading system.²⁰ The pathological assessment of primary tumor and lymphatic tissues following neoadjuvant therapy were graded using the Tumor Response Grading (TRG) system outlined in the NCCN guidelines for gastric cancer and the Becker TRG grading.^21,22 Major pathological response (MPR) was defined as ≤10% residual viable tumor cells on microscopic examination of resected tumors after neoadjuvant therapy. And all pathological assessments were independently reviewed by two blinded gastrointestinal pathologists at central laboratory (Nanfang Hospital), with discrepant cases resolved by consensus discussion.

Outcomes

The primary endpoint was the MPR rate (proportion of patients achieving MPR in the resected primary tumor). Secondary endpoints included pCR rate, R0 resection rate, event-free survival (EFS), overall survival (OS), treatment-related adverse events, and exploratory tumor microenvironment correlates. Gut microbiota analysis was planned as an exploratory endpoint; however, due to quality control failures in some patient samples, this analysis was not performed.

Statistical methods

This exploratory, single-arm study enrolled patients with MSI-H/dMMR GC/EGJC (cohort C), with a planned sample size of 15 and without formal statistical power calculations. The study was non-randomized and open-label. Biomarker analyses and hematoxylin and eosin (H&E) staining were performed on each independent sample. Continuous and categorical variables were summarized using the median (range) and frequency (percentage), respectively. EFS and OS estimates were obtained using the Kaplan–Meier method. The median follow-up time with a 95% confidence interval was calculated using the reverse Kaplan–Meier method. Statistical analyses were conducted using R software (version 4.0.3; R Foundation for Statistical Computing, Vienna, Austria).

Role of the funding source

The study funders and Shanghai Junshi Biosciences (provider of toripalimab for the study) had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; nor the decision to submit the manuscript for publication.

Results

Patients characteristics and preoperative treatment

Between March 12, 2021 and June 1, 2024, 22 patients with localized advanced dMMR/MSI-H GC/EGJC were screened at six participating centers (Fig. 1). Of these, six patients failed to meet the inclusion criteria and were consequently excluded, leaving a total of 16 eligible patients enrolled who were all included in the per-protocol analysis. Detailed baseline characteristics of the enrolled patients are presented in Table 1. In this group, the median age was 68.5 years (range 51–73), 11 patients (68.8%, 11/16) were male and 5 patients were female (31.2%, 5/16). Fifteen patients (93.8%, 15/16) were diagnosed with GC, and one patient (6.3%, 1/16) was diagnosed with GEJC. Tumor stages included cT2N1 (n = 1), cT3N0–3 (n = 3), and cT4aN1–3 (n = 12). Thirteen patients had dMMR involving hMLH1 and/or hPMS2, and two patients had dMMR involving hMSH2 and/or hMSH6. In one patient, IHC testing did not reveal dMMR, but next-generation sequencing identified MSI-H. Regarding the preoperative treatment cycles, 15 patients completed all four cycles. One patient underwent only two preoperative cycles due to renal impairment and subsequently recovered following dose reduction and discontinuation of oxaliplatin.

Fig. 1.

Flow chart. Abbreviations: dMMR/MSI-H, deficient mismatch repair/microsatellite instability-high; CapeOX, capecitabine and oxaliplatin.

Table 1.

Baseline patient characteristics.

Characteristic	Patients (N = 16)
Age, years, median (range)	68.5 (51–73)
Sex, n (%)
Male	11 (68.8)
Female	5 (31.2)
ECOG PS score of 0, n (%)	15 (93.8)
Primary tumor site, n (%)
Gastric	15 (93.8)
Gastroesophageal junction	1 (6.3)
Clinical T category, n (%)
T2	1 (6.3)
T3	3 (18.8)
T4a	12 (75.0)
Clinical N category, n (%)
N0	1 (6.3)
N1	3 (18.8)
N2	7 (43.8)
N3	5 (31.3)
Histologic grade, n (%)
G2	4 (25.0)
G3	11 (68.8)
Unknown	1 (6.3)
MMR status (IHC), n (%)
Loss of hMLH1 and/or hPMS2	13 (81.3)
Loss of hMSH2 and/or hMSH6	2 (12.5)
pMMR	1 (6.3)
MSS status (NGS), n (%)
MSI-H	8 (50.0)
MSS	2 (12.5)
Unknown	6 (37.5)
HER2 status, n (%)
0	10 (62.5)
1+	4 (25.0)
2+ (FISH−)	2 (12.5)
EBV status, n (%)
Negative	13 (81.3)
Unknown	3 (18.8)
PD-L1 CPS, n (%)
CPS < 5	7 (43.8)
CPS ≥ 5	8 (50.0)
Unknown	1 (6.3)

Abbreviations: ECOG PS, eastern cooperative oncology group performance status; T, tumor; N, node; IHC, immunohistochemistry; MMR, mismatch repair; pMMR, proficient mismatch repair; MSI-H, microsatellite instability-high; MSS, microsatellite stable; NGS, tumor tissue-based next generation sequencing; HER2, human epidermal growth factor receptor 2; EBV, epstein barr virus; PD-L1, programmed cell death-ligand 1; CPS combined positive score.

Surgical and pathological outcomes

Following neoadjuvant therapy, all patients were deemed eligible for surgical intervention. One patient, whose tumor response was classified as complete response (CR), declined surgical treatment. A total of 15 patients underwent surgical treatment, of which 4 patients (26.7%, 4/15) underwent total gastrectomy with D2 lymphadenectomy and 11 patients (73.3%, 11/15) underwent distal gastrectomy with D2 lymphadenectomy. The median interval between the date of medication initiation and the date of surgery was 3.6 months (range: 2.6–4.2 months). Details regarding the types of surgical procedures, digestive tract reconstruction, number of lymph nodes, and intraoperative bleeding are presented in Table 2. Fifteen surgical patients underwent laparoscopic procedures, achieving an R0 resection rate of 100%.

Table 2.

Characteristics of surgery and pathological results.

Characteristic	Patients (N = 15)
Type of surgery, n (%)
Total gastrectomy	4 (26.7)
Subtotal gastrectomy	11 (73.3)
D2 lymphadenectomy, n (%)	15 (100.0)
Digestive tract reconstruction, n (%)
Billroth II	1 (6.7)
Roux-en-Y	12 (80.0)
Overlap	2 (13.3)
No. of lymph nodes examined, median (rang)	49 (23–72)
R0 resection, n (%)	15 (100.0)
Intraoperative bleeding, ml, median, (range)	80 (50–500)
ypT stage, n (%)
ypT0	12 (80.0)
ypT2	2 (13.3)
ypT3	1 (6.7)
ypN stage, n (%)
ypN0	15 (100.0)
TRG NCCN, n (%)
TRG 0 (complete response): no viable cancer cells	12 (80.0)
TRG 1 (near-complete response): single cells or rare small groups of cancer cells	2 (13.3)
TRG 2 (partial response): residual cancer cells with evident tumor regression, but more than single cells or rare small groups of cancer cells	1 (6.7)
TRG Becker, n (%)
TRG 1a: complete tumor regression without residual tumor	12 (80.0)
TRG 1b: 10% residual tumor per tumor bed	2 (13.3)
TGR 2: 10%–50% residual tumor	1 (6.7)
MPR, n (%)	14 (93.3)

Abbreviations: yp, pathological; T, tumor; N, node; yp, pathological; TRG, tumor regression grade; MPR, major pathological response.

Fig. 2A presents the pathological regression of all surgical patients. The pCR in the primary tumor and lymph nodes was observed in 12 patients (80%, 12/15). The percentage of residual tumor in the 3 non-pCR patients was 2.9%, 7.5%, and 33.8%, respectively. MPR was attained in 14 patients (93.3%, 14/15). Postoperative pathological T staging was categorized as follows: ypT0 in 12 patients (80%, 12/15), ypT2 in 2 patients (13.3%, 2/15), and ypT3 in 1 patient (6.7%, 1/15). All 15 resected patients had postoperative pathological N stage ypN0. Additionally, in some lymph nodes from each patient, evidence of post-treatment tumor cell necrosis and fibrosis was observed. Notably, the patient who was dMMR at baseline and identified as MSI-H by NGS also achieved a pCR on final pathology.

Fig. 2.

Tumor regression and survival. (A) Waterfall plot of tumor regression by pathology (n = 15). (B) Swimming plot showing events during treatment and follow-up (n = 16). Abbreviations: MMR, mismatch repair; IHC, immunohistochemistry; NGS, tumor tissue-based next generation sequencing; MSI-H, microsatellite instability-high; MSS, microsatellite stable; CPS, combined positive score; T, tumor; MPR, major pathological response; TRG, tumor regression grade; NA, unknown.

TRG was quantified using the tumor response scoring system of the NCCN Gastric Cancer guidelines, with the following results: grade 0 in 12 cases (80%, 12/15), grade 1 in 2 cases (13.3%, 2/15), and grade 2 in 1 case (6.7%, 1/15). TRG quantified by the Becker criteria were: grade 1a in 12 cases (80%, 12/15), grade 1b in 2 cases (13.3%, 2/15), and grade 2 in 1 case (6.7%, 1/15). The changes in T-stage and N-stage at baseline, preoperative, and postoperative pathology of all patients are shown in Supplementary Fig. S1. The concordance between T- and N-stage at preoperative and surgical pathology was only 6.7% (1/15).

Postoperative treatment

In the cohort of 15 patients who underwent surgery, 13 patients (86.7%, 13/15) received adjuvant therapy combining toripalimab with the CapeOX regimen. The median interval from the date of surgery to the initiation of postoperative treatment was 1.3 months (range: 0.8–2.3 months). Two patients with pCR refused postoperative adjuvant therapy. As of June 1, 2024, all patients had completed postoperative adjuvant therapy and were under regular observation and follow-up. Specifically, 7 patients underwent four cycles of postoperative adjuvant therapy, 2 patients completed three cycles, and 4 patients completed two cycles. Complete information for all patients is recorded in Supplementary Material S1.

Adverse events

Table 3 presents the drug-related adverse events and their classifications observed in participants throughout the neoadjuvant therapy and subsequent postoperative adjuvant therapy stages. Overall, the treatment was well tolerated, and the study met the primary endpoints of safety and feasibility, with no deaths caused by adverse events during the treatment. Among the 16 patients, 12 (75%, 12/16) experienced treatment-related events of any grade. Six patients (37.5%, 6/16) experienced grade 3–4 adverse events. One patient (P3) with renal impairment showed recovery after dose reduction and discontinuation of oxaliplatin while maintaining capecitabine therapy. Renal function normalized subsequently, and the patient achieved pCR following two preoperative treatment cycles. In addition, a total of four patients (25%, 4/16) developed immune-related adverse events (irAEs), including hypothyroidism (3 patients, 18.8%, 3/16) and rash (2 patients, 12.5%, 2/16). All irAEs were mild to moderate (Grade 1–2), except for one case of renal impairment (6.3%, 1/16) potentially associated with immune-related mechanisms, which resolved after dose adjustment. No grade 5 adverse events or treatment-related deaths were observed.

Table 3.

Treatment-related adverse events.

Adverse events	All patients (N = 16)
	Any grade, n (%)	Grade 3–4, n (%)
Treatment-related adverse events	12 (75.0)	6 (37.5)
ALT/AST increased	9 (56.3)	2 (12.5)
Neutropenia	7 (43.8)	1 (6.3)
Leukopenia	6 (37.5)	0 (0.0)
Thrombocytopenia	6 (37.5)	1 (6.3)
Decreased appetite	3 (18.8)	0 (0.0)
Vomiting	2 (12.5)	0 (0.0)
Hypokalemia	1 (6.3)	1 (6.3)
Hyperglycemia	1 (6.3)	0 (0.0)
Constipation	1 (6.3)	0 (0.0)
Diarrhea	1 (6.3)	0 (0.0)
Dizziness	1 (6.3)	0 (0.0)
Osteoporosis	1 (6.3)	0 (0.0)
Immune-related adverse events	4 (25.0)	1 (6.3)
Hypothyroidism	3 (18.8)	0 (0.0)
Rash	2 (12.5)	0 (0.0)
Renal impairment	1 (6.3)	1 (6.3)
Bilirubin increased	1 (6.3)	0 (0.0)

Abbreviations: ALT, Alanine aminotransferase; AST, Aspartate aminotransferase.

Using the Clavien-Dindo grading system to evaluate perioperative treatment-related adverse events, Supplementary Table S1 presents the incidence and grade, and no adverse events leading to death were identified. Thirteen patients (53.3%, 8/15) experienced adverse events of any severity. Two patients (13.3%, 2/15) experienced grade 3 adverse events.

Survival

By the deadline of June 1, 2024, the median duration of follow-up (time from the first dose to the cutoff date) was 22.5 months (range: 10.5–39 months). Among the 16 enrolled patients, no cases of tumor recurrence or progression were observed. One patient who achieved clinical complete response (cCR) (endoscopic biopsy without tumor cells) after four cycles of neoadjuvant therapy and subsequently declined surgery did not undergo any further treatment. During the follow-up period, no tumor recurrence was observed in this patient. One patient died 287 days post-surgery due to COVID-19 (a non-tumor related cause), resulting in a one-year overall survival rate of 93.8% (15/16). Kaplan–Meier curves for EFS and OS are shown in Supplementary Fig. S2A and B, respectively. Fig. 2B summarizes the survival of patients with different treatment responses.

Biomarker analysis

The relationship between PD-L1 CPS (<5 vs. ≥5) and TRG is summarized in Supplementary Table S2. However, no significant correlation was observed between PD-L1 CPS (<5 or ≥5) and TRG (1a or 1b/2) (P = 0.57). To further investigate potential biomarkers associated with the efficacy of neoadjuvant immunochemotherapy, we utilized mIF technique to examine the tumor immune microenvironment in biopsy specimens obtained from four patients prior to neoadjuvant treatment (pCR: P1, P5, P6; non-pCR: P10). Typical mIF images show the infiltration of immune cells in the tumor and stroma in patients (P1, P5, P6) who achieved pCR (Fig. 3A and Supplementary Fig. S3) and P10 who did not (Fig. 3B). The biopsy specimens from the three pCR patients were enriched with a substantial number of CD3+ T cells, CD8+ T cells, and PD-1+CD8+ T cells in both the tumor and stroma regions (Supplementary Fig. S4A–C). The infiltration levels of CD3+CD4+ T cells and CD3+CD4+FoxP3+ T cells were similar in P5, P6, and P10 (Supplementary Fig. S4D and E). Regarding tumor-associated macrophages, P1, P5, and P6 showed an enrichment of PD-L1+CD68+ macrophages and CD68+CD163− macrophages (M1 macrophages) in the tumor (Supplementary Fig. S4F and G). Additionally, P1 and P5 had a lower abundance of CD68+CD163+ macrophages (M2 macrophages) in both the tumor and stromal regions (Supplementary Fig. S4H). Regarding NK cells, P1, P5, and P6 exhibited a significant enrichment of CD56^dim NK cells in both tumor and stromal regions, whereas P10 had a markedly lower presence of CD56^bright NK cells, CD56^dim NK cells, and CD56^brightCD56^dim NK cells in the tumor parenchyma (Supplementary Fig. S4I–K). In the tumor of P10, CD20+ B cells were exceedingly scarce, with a cell number of only 6.73 cells/mm² (Supplementary Fig. S4L). All source data are available in Supplementary Material S2.

Fig. 3.

Representative multiplex immunofluorescence (mIF) of tumor immune microenvironment. (A) The expression levels of all markers in tissue slides of P6 achieving pCR. (B) The expression levels of all markers in tissue slides of P10 achieving non-pCR.

Discussion

The role of immunotherapy combined with chemotherapy in the neoadjuvant treatment of dMMR/MSI-H GC/EGJC is not yet well-established. In this study, we evaluated the efficacy and safety of neoadjuvant toripalimab combined with the CapeOX regimen followed by surgery in patients with localized advanced dMMR/MSI-H GC/EGJC, which resulted in a 100% R0 resection rate, an 80% pCR rate, and a 93.3% MPR rate in 15 patients with resectable cT3/4aNxM0 or T2N + M0 GC/EGJC. Over a median follow-up period of 22.5 months, there were no observed cases of tumor recurrence or progression.

Across various tumor types, irrespective of disease stage, the MSI-H phenotype is generally considered a prognostic factor. Interestingly, reported results suggest that in dMMR/MSI-H GC/EGJC, the MSI-H phenotype tends to be an adverse prognostic factor for neoadjuvant chemotherapy.^23,24 The pCR rate for patients with dMMR/MSI-H GC/EGJC on neoadjuvant chemotherapy ranged from 3% to 38%.25, 26, 27 The dMMR/MSI-H phenotype is a crucial predictive factor for a favorable response to immunotherapy, as confirmed in subgroup analyses of several large prospective clinical studies.^2,12,28 Several recently published studies have confirmed the high efficacy and safety of immunotherapy in patients with dMMR/MSI-H GC. Shannon et al. reported neoadjuvant immunotherapy in dMMR gastric cancer: 5 patients (50%, 5/10) underwent surgery, 3 patients (60%, 3/5) achieved pCR, but only 2 had exclusive immunotherapy, both with pCR.²⁹ A multicenter, single-arm, phase I study in Japan evaluated the efficacy of neoadjuvant treatment with nivolumab monotherapy in locally advanced GC; 4 of 7 MSI-H patients had tumor regression to MPR after treatment (MPR: 57.1%, 4/7).³⁰ In the phase II NEONIPIGA trial, nivolumab plus ipilimumab (dual PD-1/CTLA-4 inhibition) for neoadjuvant treatment of MSI-H/dMMR GC/EGJC achieved a pCR of 58.6% (17/29).¹³ Similarly, in the phase II INFINITY study, the tremelimumab plus durvalumab (dual PD-L1/CTLA-4 inhibition) regimen had similar efficacy in neoadjuvant therapy, with a pCR rate of 60% (9/15). Two patients died after surgery for other reasons than disease or adverse events.¹⁴ The efficacy of immunotherapy plus chemotherapy in dMMR/MSI-H GC/EGJC remains unclear. Based on the significant efficacy of FLOT in the neoadjuvant treatment of GC/EGJC confirmed in previous studies, a randomized, open-label phase II trial (DANTE) explored the efficacy of neoadjuvant treatment with perioperative atezolizumab in combination with three-drug FLOT chemotherapy regimen. Interim analyses revealed significant tumor regression in patients with resectable GC/EGJC, with a pCR rate of 50% in the MSI-H subgroup.³¹ In Asia, dual-agent chemotherapy (e.g., platinum plus fluoropyrimidine) is the standard neoadjuvant treatment regimen for locally advanced GC.¹⁷ However, the precise efficacy of CapeOX combined with anti-PD-1 monoclonal antibody as a neoadjuvant treatment for dMMR/MSI-H GC remains unknown. In the present study, our findings suggest that the neoadjuvant regimen of CapeOX plus toripalimab is associated with improved tumor remission in patients with dMMR/MSI-H GC/EGJC. This regimen demonstrated a high pCR of 80% (12/15) and a MPR of 93.3% (14/15), indicating its potential benefits for these patients. The notably high pCR rate may reflect chemo-immunologic synergy: oxaliplatin/fluoropyrimidines elicit immunogenic cell death and type-I-IFN signaling, transiently up-regulating tumor-cell PD-L1 and expanding the pool of PD-1-targetable clones,³² 5-fluorouracil–induced DNA damage activates cancer-cell-intrinsic cGAS–STING, boosting CXCL9/10 production and attracting cytotoxic CD8⁺ T cells and oxaliplatin concurrently depletes myeloid-derived suppressor cells, relieving immunosuppression and converting the tumor micro-environment from “cold” to “hot”.^33,34

Laparoscopic exploration and peritoneal lavage cytology are reliable methods for diagnosing occult metastasis in GC. According to literatures, the sensitivity of peritoneal cytology for detecting occult peritoneal metastasis ranges from 26% to 70.8%.^35,36 Additionally, the American Joint Committee on Cancer (AJCC) classifies peritoneal lavage cytology positive without peritoneal metastasis as distant metastasis (M1).³⁷ In this study, all patients underwent contrast-enhanced chest and abdominal CT to determine tumor staging (cT2N1 (n = 1), cT3N0-3 (n = 3), and cT4aN1-3 (n = 12)), and laparoscopy combined with peritoneal lavage cytology confirmed the absence of peritoneal metastasis and negative peritoneal cytology. Notably, 10 patients (83.3%, 10/12) with T4 stage tumors achieved pCR, which contrasts sharply with previous studies on neoadjuvant immunotherapy for dMMR/MSI-H GC/EGJC. In the INFINITY study, the pCR rate for T4 tumors was 16.7% (1/6), while the pCR rate for T2-3 tumors was 88.9% (8/9) (P = 0.011).¹⁴ The NEONIPIGA study included only T2-3 patients, with a ypT0 rate of 65.5% (19/29).¹³ This discrepancy may reflect both the enhanced staging rigor in our study—combining CT with laparoscopy and cytology to exclude occult peritoneal spread—and the use of combination therapy. Unlike INFINITY and NEONIPIGA, which used dual immune checkpoint blockade alone, our study employed toripalimab in combination with chemotherapy. Chemotherapy may potentiate the immunogenicity of tumors by promoting immunogenic cell death, enhancing antigen presentation, and reducing immunosuppressive elements in the tumor microenvironment, such as myeloid-derived suppressor cells and regulatory T cells.^38,39 For locally advanced dMMR/MSI-H GC/EGJC with deeper tumor infiltration, toripalimab combined with chemotherapy may enhance tumor regression. Future clinical trials are warranted to validate this approach by directly comparing chemoimmunotherapy with dual immunotherapy in GC patients with more advanced disease stages.

Notably, we found significant discrepancies between post-treatment radiological staging and laparoscopic findings versus pathological staging, with only 6.7% (1/15) concordance. This may reflect the unique biology of dMMR tumors. After immunotherapy, tumors often show extensive CD8⁺ T-cell infiltration replacing tumor cells, forming inflammatory pseudotumors. CT may show stable disease or minimal shrinkage, sometimes mimicking progression, while pathology reveals lymphocyte infiltration and necrosis.⁴⁰ In the NICHE-2 trial, 67.6% (75/111) of patients achieved pCR, while only 2.7% (2/75) attained radiological response, highlighting that imaging assessment may underestimate the true efficacy of neoadjuvant immunotherapy in dMMR tumors.⁴¹ Moreover, preoperative CT demonstrates limited accuracy in predicting pT and pN staging for localized MSI-H/dMMR colorectal cancer.⁴² Whether this similarly applies to dMMR/MSI-H GC remains unclear, necessitating a reassessment of the benefit-risk profile of neoadjuvant treatment strategies in this population. Additionally, all surgically resected patients in this study achieved ypN0 status, with post-treatment tumor cell necrosis/fibrosis observed in partial lymph nodes—predominantly within the stromal regions. Similarly, prior studies reported that in MSI-H colorectal cancer, immune therapy induces nodular aggregates of lymphocytes in the stromal areas of metastatic lymph nodes, which correlate with pathological complete response.⁴³ This suggests that preoperative N staging may not reliably predict pCR in dMMR gastric cancer. In summary, future approaches should incorporate functional imaging and liquid biopsies to improve the accuracy of post-neoadjuvant immunotherapy staging in dMMR/MSI-H GC.

Regarding the distinct subtype of gastric cancer characterized by dMMR/MSI-H in GC/EGJC, the NEONIPIGA trial documented a one-year DFS rate of 94% in patients treated with a neoadjuvant dual PD-1/CTLA-4 inhibition regimen, with an impressive pCR rate of 58.6%.¹³ Similarly, our study demonstrated a remarkable pCR rate of 80% and a one-year disease-free recurrence rate of 100%, with one patient dying from COVID-19. These exceptional outcomes, combined with the high pCR rates observed following immunotherapy and chemotherapy, strongly suggest that organ preservation strategies may be feasible for select patients. The potential for non-surgical management is further supported by Cercek et al.'s findings in dMMR rectal cancer, where all 12 patients treated with dostarlimab achieved cCR and avoided surgery during follow-up.⁴³ Similarly, one patient in our study who achieved cCR declined surgery and remained recurrence-free for over one year, with no evidence of tumor activity on imaging or endoscopy. However, critical questions remain before organ preservation can be widely adopted. The correlation between cCR and pCR requires validation, particularly given gastric cancer's spatial heterogeneity. While ctDNA has shown promise in predicting pCR (as seen in the INFINITY trial, where all pCR patients had negative preoperative ctDNA), its role in gastric cancer remains uncertain. Future studies should prospectively evaluate non-surgical approaches, incorporating rigorous multimodal assessments (imaging, endoscopy, ctDNA, and targeted biopsies) to refine patient selection and surveillance protocols.

The immune microenvironment of pre-treatment specimens may be associated with treatment efficacy. CD8⁺ T cells are central to tumor cell elimination and have been associated with immunotherapy responses across various tumor types.44, 45, 46, 47 In our study, pre-treatment tissues from patients who achieved pCR showed apparent enrichment of CD8⁺ T cells. Similarly, higher infiltration of M1 macrophages—typically linked with anti-tumor activity⁴⁸—was observed in pCR cases. NK cells can directly kill tumor cells or mediate tumor cell lysis through the antibody-dependent cellular cytotoxicity effect.⁴⁹ CD56^dim NK cells represent the final maturation stage of NK cells and possess higher cytotoxic capabilities compared to CD56^bright NK cells and high levels of NK cell infiltration are associated with better tumor prognosis.^50,51 The pCR-patients exhibited particularly rich infiltration of both CD56^dim and CD56^bright NK cells in this study. While these findings support the established correlation between immune cell infiltration and treatment response, we must acknowledge that the small number of non-pCR cases in our study limits the statistical power of microenvironment analyses. Larger prospective cohorts are required to validate these preliminary observations and to integrate such immune-microenvironmental features—together with emerging dynamic biomarkers such as early ctDNA clearance—into algorithms for identifying patients most likely to benefit from chemo-immunotherapy.

Immunotherapy in clinical practice has increased the risk of immune-related adverse events, potentially due to the disruption of immune checkpoint pathways and modulation of the immune microenvironment.^52,53 For advanced dMMR/MSI-H GC/EGJC, the NO LIMIT study reported the favorable safety profile of the nivolumab plus low-dose ipilimumab regimen, with the incidence of any-grade TRAEs being 93.1%, grade 3 TRAEs being 37.9%, and grade 4 TRAEs being 3.4%, with no treatment-related deaths.⁵⁴ In the NEONIPIGA study, the nivolumab plus low-dose ipilimumab regimen had an incidence of any-grade TRAEs of 75%, with grade 3–4 TRAEs occurring at a rate of 19%.¹³ The INFINITY study reported an incidence of grade 3 or higher immune-related adverse events of 16.7%.¹⁴ In this study, the incidence of any-grade TRAEs was 75% (12/16). Six patients (37.5%, 6/16) experienced grade 3–4 TRAEs. Of these six patients, five recovered after conservative symptomatic treatment, and one patient with renal impairment recovered following dose reduction and discontinuation of oxaliplatin. Overall, our study confirmed that toripalimab and CapeOX-based neoadjuvant therapy is feasible and associated with no un-expected toxicity for patients with dMMR/MSI-H GC/EGJC.

As an exploratory study, several important limitations should be noted. First, no formal sample size calculation was performed, as the primary goal was to obtain preliminary evidence of efficacy and safety for this novel combination in dMMR/MSI-H GC. Given the rarity of this molecular subtype, a conventionally powered sample size was not feasible at this stage. Nevertheless, the sample of 16 patients falls within the typical range for exploratory phase II oncology studies and represents one of the largest reported cohorts for this molecular subtype to date. Additionally, the exclusively Chinese study population may limit generalizability to other ethnic groups, particularly given the known differences in gastric cancer characteristics between Asian and Western populations. While the preliminary efficacy data are promising, the lack of long-term survival outcomes represents another limitation that will be addressed in our ongoing follow-up research. In summary, these initial findings require validation through larger, statistically powered clinical trials.

In conclusion, in this multicenter, single-arm trial, toripalimab combined with chemotherapy showed good efficacy and safety in the perioperative treatment of locally advanced dMMR/MSI-H GC/EGJC. Further large-scale randomized clinical trials are needed to confirm the survival benefit.

Contributors

LZ, YH, HL, SY, WW, JJ, and GL contributed to the conception, design, and planning of the study. LZ, YH, HL, JY, SY, WW, JJ, ZC, XC, FL, TL, MZ, TC, and HC contributed to the acquisition of data. LZ, HL, LY, HYL, LL, XZ, and YZ contributed to the analysis of data. LZ, HL, LY, HYL, LL, XZ, YZ, and GL contributed to the interpretation of results. The data reported in the manuscript were accessed and verified by LZ, HYL, and HL. LZ, HL, and HYL drafted the paper. All the authors have read and approved the final version of the manuscript. LZ and GL had final responsibility for the decision to submit for publication.

Data sharing statement

All requests for data will be reviewed by the leading clinical site, Nanfang Hospital, Southern Medical University, Guangzhou, Guangdong, China, to verify whether the request is subject to any intellectual property or confidentiality obligations. Requests for access to the patient-level data from this study can be submitted via email to gzliguoxin@163.com with detailed proposals for approval and will be responded to in two weeks. A signed data access agreement with the sponsor is required before accessing shared data. Correspondence and requests for materials should be addressed to Guoxin Li.

Declaration of interests

The authors declare no competing interests.