Neoadjuvant PD-(L)1 blockade with or without chemotherapy versus chemotherapy alone in mismatch repair-deficient, potentially resectable stage III-IV gastric cancer patients: a single-center retrospective study.

Introduction
Gastric cancer is a common malignancy worldwide with a dismal prognosis [1]. Microsatellite instability-high (MSI-H) or mismatch-repair deficiency (dMMR) gastric cancer has been identified as a distinct subtype according to The Cancer Genome Atlas [2]. The incidence of MSI-H/dMMR status in gastric cancer patients varies from 2.6 to 35.3% worldwide [3]. Gastric cancer patients with MSI-H/dMMR status are characterized by a high tumor mutational burden, high neoantigen load, and dense cytotoxic T-cell immune infiltration [4], and immunotherapy has been proven to be more beneficial in patients with MSI-H/dMMR gastric cancer owing to its distinct pathological characteristics and biological behavioral patterns [5]. Due to high heterogeneity, treatment strategies and tumor responses exhibit wide discrepancies among gastric cancer patients, highlighting the importance of personalized treatment.
Locally advanced gastric cancer (LAGC) was defined as clinical tumor stage T3-T4 and/or regional lymph node-positive disease. LAGC patients may benefit from neoadjuvant chemotherapy through the elimination of residual tumor tissue [6] and micrometastasis [7]; the phase III MAGIC trial [8] laid the foundation for the use of neoadjuvant chemotherapy as a standardized preoperative treatment strategy in patients with LAGC. Disappointingly, LAGC patients with MSI-H/dMMR status cannot benefit substantially from perioperative chemotherapy [9, 10], which prompted oncologists to explore optimal treatment options for these patients.
The excellent breakthrough in the application of immunotherapy in various solid tumors has also shed light on the neoadjuvant setting in LAGC patients. Several recent prospective studies have concentrated on the neoadjuvant setting of PD-1/PD-L1 blockade in patients with LAGC, and some of them have shown breath-taking results [11, 12]. The interim analysis results of the phase III KEYNOTE-585 trial [13] and MATTERHORN trial [14] both demonstrated a significantly improved pCR rates after neoadjuvant treatment with PD-(L)1 blockade plus chemotherapy in contrast to neoadjuvant chemotherapy alone, with a mild and tolerable safety profile.
For gastric cancer patients with MSI-H/dMMR status, the phase III KEYNOTE-062 trial demonstrated that in first-line treatment of unresectable gastric cancer patients, patients in the MSI-H cohort obtained more satisfactory benefits from immunotherapy than did their microsatellite stable (MSS) counterparts, with prolonged overall survival (OS) and progression-free survival (PFS) [15]; the same trend was observed in the phase III CheckMate-649 study [16]. In the neoadjuvant setting, neoadjuvant nivolumab PD-1 blockade plus ipilimumab CTLA-4 blockade induced an encouraging pCR rate of 58.6% in patients with MSI-H/dMMR gastric or esophagogastric junction adenocarcinoma in the phase II GERCOR NEONIPIGA trial [17]. However, the clinical efficacy and safety of neoadjuvant PD-(L)1 blockade immunotherapy plus chemotherapy in locally advanced gastric cancer patients with MSI-H/dMMR status have not been fully elucidated.
Our current study was designed to retrospectively evaluate the clinical efficacy including tumor response and EFS, and safety profile of neoadjuvant PD-(L)1 blockade with or without chemotherapy compared with neoadjuvant chemotherapy alone in patients with locally advanced MSI-H/dMMR gastric cancer.

Methods

Study design
This retrospective study received ethical approval from the Institutional Review Board of the Affiliated Hospital of Qingdao University (Approval No. QYFYWZLL27969, Qingdao, China) and was performed at our center. We retrospectively collected clinical data from treatment-naïve, stage T3 − 4aN+M0 (stage III), or potential resectable T4bNanyM0 (stage IVA) disease after multidisciplinary discussions, and MSI-H/dMMR gastric cancer patients, who were treated at our center from January 2019 to June 2023.
The major inclusion criteria for patients were as follows: (a) diagnosed with clinical stage T3 − 4bN+M0 (stage III-IVA) histologically proven gastric cancer with MSI-H/dMMR status; (b) not receiving previous antitumor treatment; (c) an Eastern Cooperative Oncology Group performance status (ECOG PS) score of 0 to 1; and (d) sufficient vital organ function. The exclusion criteria for patients were as follows: (a) distant metastatic lesions; (b) had received previous antitumor therapy; (c) had inadequate vital organ functions or autoimmune systematic diseases; and (d) had malignancies other than gastric cancer.
Based on the different treatment regimens used, the included patients (treated with or without PD-1/PD-L1 blockade) were categorized into two groups: the neoadjuvant PD-(L)1 blockade-based group and the neoadjuvant chemotherapy group. Figure 1 shows the flow chart of this study.

Treatment
Patients in the PD-(L)1 blockade-based treatment group received PD-(L)1 blockade monotherapy, or PD-(L)1 blockade in combination with standardized chemotherapy, and patients in the chemotherapy group received neoadjuvant chemotherapy alone. After the last dose of neoadjuvant treatment, the tumor responses were evaluated, and multidisciplinary discussions were held to evaluate the feasibility of surgical resection. The patients in whom R0 resection was considered feasible underwent standardized D2 gastrectomy, and the radiographical CR patients who were considered to be spared from surgery were continuing to be followed up. Subsequent adjuvant treatments were administered to the patients according to the outcomes of surgery.
The PD-(L)1 blockade used included one of the following regimens: nivolumab (3 mg/kg iv drip, every two weeks), pembrolizumab (200 mg iv drip, every three weeks), sintilimab (200 mg iv drip, every three weeks), camrelizumab (200 mg iv drip, every three weeks), tislelizumab (200 mg iv drip, every three weeks), serplulimab (3 mg/kg iv drip, every two weeks), and envafolimab (200 mg ih, every week). The chemotherapy regimens used in the two groups included standardized FLOT, SOX, mFLOT, and mDOS regimens.

Assessments
Contrast-enhanced computed tomography (CECT) and endoscopic ultrasound (EUS) were used to assess the primary tumor at baseline and response to neoadjuvant treatment according to the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 [18]. Tumor tissue biopsies were collected both at baseline and during surgery. Surgical samples from primary tumors and lymph nodes were staged based on the eighth edition of the Gastric Cancer Staging system of the American Joint Committee on Cancer [19].
To grade the pathological response after surgery, the tumor regression grade (TRG) was scored [20]. TRG0, TRG1, TRG2, and TRG3 were defined as no residual viable tumor cells (equal to pCR), no more than 2%, more than 2% but no more than 50%, and more than 50% residual viable tumor cells, respectively. Two board-certified pathologists majoring in tumor pathology participated in the pathological evaluations.
The mismatch repair (MMR) status was evaluated using immunohistochemical staining. Primary antibodies against MSH2, MSH6, MLH1, and PMS2 were used for immunohistochemical staining of mismatch repair-related proteins. Generally, the absence of one or more of these proteins was considered the dMMR status. The status of MSI was evaluated using real-time quantitative polymerase chain reaction by detecting the amplification of five MSI loci: BAT25, BAT26, D5S346, D17S250, and D2S123. If two or more unstable markers were observed at these five loci, MSI-H status was defined; if one unstable marker was observed, MSI-low (MSI-L) status was defined; if no unstable markers were observed, the microsatellite status was defined as MSS [2].
The expression of programmed death ligand-1 (PD-L1) was detected using immunohistochemical staining with the anti-PD-L1 antibody 22C3 (Dako, Glostrup, Denmark). The combined positive score (CPS) was calculated to evaluate the number of PD-L1-positive cells, including tumor cells, macrophages and lymphocytes. Briefly, a CPS < 1 was considered to indicate PD-L1 negativity, and a CPS ≥ 1 indicated PD-L1 positivity.
Adverse events (AEs) during neoadjuvant treatment period were evaluated based on the National Cancer Institute Common Terminology Criteria for Adverse Events version 5.0.

Outcome evaluation
The outcomes of interest were pathological tumor response and radiographic tumor response (complete response, CR; partial response, PR; stable disease, SD; or progressive disease, PD). Other outcomes included EFS (the time from diagnosis to any one of the following three events: unable to proceed surgery due to disease progression, local or distant relapsed disease, or deaths caused by any reason), and the monitored TRAEs.

Statistical analysis
Medians (interquartile ranges) or means (standard deviations) were calculated to describe continuous variables, and percentages and frequencies were used to describe categorical variables. Student’s t test was applied to evaluate the significance of differences between two groups for continuous variables, and χ2 or Fisher’s exact test was applied for categorical variables. A two-tailed P value < 0.05 was considered indicative of statistical significance. All the statistical analyses were performed with SPSS 25.0 (IBM Corp., Armonk, NY, USA) and GraphPad Prism 9.5 (GraphPad Software, Inc., San Diego, CA, USA).

Results

Enrolled patients and treatment regimens
A total of 30 patients were enrolled after strict screening, among whom 7 patients were with MSI-H status, 17 patients were with dMMR status, and 6 patients were identified with both MSI-H and dMMR status. 23 patients received neoadjuvant PD-(L)1 blockade plus chemotherapy or PD-(L)1 blockade monotherapy and were classified in the neoadjuvant PD-(L)1 blockade-based treatment group, among whom 21 patients received neoadjuvant PD-(L)1 blockade plus chemotherapy, and the other two patients were treated with PD-(L)1 blockade monotherapy. Seven patients received standardized neoadjuvant chemotherapy and were classified in the neoadjuvant chemotherapy group (Fig. 1). In the PD-(L)1 blockade-based treatment group, the PD-(L)1 blockade agents used were as follows: sintilimab in eight patients, camrelizumab in eight patients, serplulimab in two patients, envafolimab in two patients, nivolumab in one patient, tislelizumab in one patient, and pembrolizumab in one patient. The chemotherapy regimens used in the two groups included standardized FLOT, SOX, mFLOT, and mDOS regimens, as shown in Table 1 and Table S1.

Most of the patients in both groups had stage III disease, with incidences of 82.6% (19 of 23) in the PD-(L)1 blockade-based group and 85.7% (6 of 7) in the chemotherapy group. Four (17.4%) patients in the PD-(L)1 blockade-based treatment group and one (14.3%) patient in the chemotherapy group had clinical T4bNanyM0 (stage IVA) disease, and all of them were considered potentially resectable after multidisciplinary discussions. Table 1 and Supplementary Table 1 report the detailed information of the individual patients involved in this study.
A total of 87.0% (20 of 23) of patients in the PD-(L)1 blockade-based treatment group and all the patients in the chemotherapy group underwent standardized D2 radical gastrectomy after the last dose of neoadjuvant treatment. Three patients (P5, P6 and P23) in the PD-(L)1 blockade-based treatment group did not undergo radical surgery because they achieved radiographical CR as evaluated by EUS, and they were considered to be spared from surgery after multidisciplinary discussions.

Clinical activity and tumor response
After the last dose of neoadjuvant treatment, four (17.4%) of the 23 patients achieved radiographical CR confirmed by EUS, 17 (73.9%) patients reached PR, and two (8.7%) patient maintained SD in the PD-(L)1 blockade-based treatment group, with an objective response rate (ORR) of 91.3% (21 of 23, 95% CI 0.732–0.976). In the chemotherapy group, five (71.4%) of the seven patients achieved a PR, and two (28.6%) had SD, with an ORR of 71.4% (5 of 7, 95% CI 0.359–0.918). There were not any patients who had PD after neoadjuvant treatment. (Fig. 2).

Among the four patients with radiographical CR in the PD-(L)1 blockade-based treatment group, one (patient 11) still underwent surgery, while the pathological response was TRG2 (ypT2N0). In total, 87.0% (20 of 23) of patients in the PD-(L)1 blockade-based treatment group and all the patients in the chemotherapy group underwent surgery. R0 resections were achieved in all the patients who underwent surgery in both groups. Seven patients (30.4%, 95% CI 0.141–0.530) in the PD-(L)1 blockade-based treatment group and one patient (14.3%, 95% CI 0.026–0.513) in the chemotherapy group achieved pCR (P = 0.638). TRG 0/1 was achieved in 47.8% (11 of 23) of patients in the PD-(L)1 blockade-based group and 28.6% (2 of 7) patients in the chemotherapy group. The overall CR rate was 43.5% (10 of 23, 95% CI 0.239–0.651) in the PD-(L)1 blockade-based treatment group and 14.3% (1 of 7, 95% CI 0.026–0.513) in the chemotherapy group (P = 0.215). Although the differences in pCR and overall CR between the two groups were not statistically significant, there were potential trends toward an improved pathological response in the PD-(L)1 blockade-based treatment group (Table 2).
Notably, among the patients who had undergone surgical operation, pathological lymph node downstaging to ypN0 was observed in 85.0% (17 of 20) of patients in the PD-(L)1 blockade-based treatment group. In the PD-(L)1 blockade-based treatment group, lymph node downstaging to N0 of was observed in 87.0% (20 of 23) of patients, including three EUS-confirmed CR patients who had been spared from surgery; there were 42.9% (3 of 7) of patients in the chemotherapy group who had lymph node downstaging to N0 (P = 0.033*). Tables 2 and 3 show the details of the tumor response and surgical outcomes.
Next, we analyzed the correlations between tumor response and PD-L1 expression (according to three thresholds, CPS < 1, ≥ 1, and ≥ 5) in the PD-(L)1 blockade-based treatment group. In total, there were three patients in the CPS < 1 subgroup, one with pCR (P16), one with TRG1 (P17) and one with TRG2 (P10). Among the 16 patients with PD-L1 positivity (CPS ≥ 1), the pCR rate was 37.5% (6 of 16), and the overall CR rate was 50.0% (8 of 16). Among the 10 patients with a CPS ≥ 5, 33.3% (3 of 10) achieved pCR, and an overall CR rate of 50.0% (5 of 10) was achieved.

Survival outcomes
By the last follow-up, the median follow-up time was 28.0 months (range, 9.3 months to 53.1 months) in the PD-(L)1 blockade-based treatment group and 32.5 months (range, 4.8 months to 58.3 months) in the chemotherapy group. The median EFS time in the two groups was not reached.
In the PD-(L)1 blockade-based treatment group, two patients (P3 and P12) experienced disease relapse, with EFSs of 4.5 months and 25.5 months, respectively. Notably, 5 patients had been followed up for more than three years, and they were still alive without recurrence except P3. In addition, none of the three patients who had radiological CR and spared from gastrectomy (P5, P6 and P23) had developed disease relapse. In the chemotherapy group, two patients (P24 and P29) suffered from disease relapse, with EFSs of 21.8 months and 4.8 months, respectively (Fig. 3A). No statistical significance of probabilities of EFS was observed between the two groups (P = 0.188) (Fig. 3B).

Safety and feasibility
Overall, 82.6% (19 of 23, 95% CI 0.629–0.930) of patients in the PD-(L)1 blockade-based treatment group and 85.7% (6 of 7, 95% CI 0.487–0.974) of patients in the chemotherapy group experienced neoadjuvant TRAEs, with a decreased neutrophil count, increased aspartate aminotransferase and most frequent anemia. Most TRAEs were Grade 1 to 2 and could be managed by symptomatic treatment, while TRAEs of Grade 3 or worse still occurred in 26.1% (6 of 23) and 42.9% (3 of 7) of the patients in the two groups, respectively. There was no delay in surgery or death caused by TRAEs. Table 4 describes the details of the TRAEs.

Discussion
Regarding neoadjuvant therapy in patients with LAGC, although several recent studies have demonstrated the promising efficacy of the combination of immunotherapy plus chemotherapy [11–13, 21], the efficacy of this combination in patients in the MSI-H/dMMR subgroup is unclear; whether MSI-H/dMMR can be a reliable predictor of tumor response to immunotherapy has not been determined. To date, only a few case series studies have reported the efficacy and safety of PD-(L)1 blockade-based neoadjuvant treatment in patients with MSI-H/dMMR gastric cancer [22, 23]. In our previous study observing the efficacy and safety of neoadjuvant PD-1 blockade plus chemotherapy versus chemotherapy alone in LAGC patients, we observed a significantly improved pCR rate in the PD-1 blockade plus chemotherapy group (23.5% vs. 4.7%, P = 0.019); in addition, among the patients who had received neoadjuvant PD-1 blockade plus chemotherapy, patients with MSI-H/dMMR status had a greater pCR rate than did their MSS counterparts (4 of 5, 80.0% vs. 4 of 30, 13.3%, P = 0.006) [24]. According to the results of the current study, a promising tumor response to PD-(L)1 blockade-based neoadjuvant treatment was observed in LAGC patients with MSI-H/dMMR status, with a pCR rate of 30.4% (7 of 23) and an overall CR rate of 43.5% (10 of 23).
Notably, in the PD-(L)1 blockade-based treatment group, nodal downstaging from preoperative cN+ to postoperative ypN0 was observed in 85.0% (17 of 20) of patients who had undergone surgery. As previously reported, nodal downstaging to ypN0 could indicate the promising efficacy of neoadjuvant treatment and satisfactory survival in patients with LAGC [25]. In LAGC patients with MSI-H/dMMR status, nodal downstaging was particularly significant, possibly owing to the stronger activity of cytotoxic T cells activated by high neoantigen loads and TMB [4, 5]. Moreover, in the PD-(L)1 blockade-based treatment group, there were three patients with available TMB data (P6, 51.06 muts/Mb; P8, 60.99 muts/Mb; P9, 136.71 muts/Mb), and all of them achieved CR (P6 with cCR and P8 and P9 with pCR) regardless of PD-L1 CPS. To date, the relationships between the PD-L1 expression profile and the patient prognosis, MSI status, TMB and efficacy of neoadjuvant immunotherapy have been controversial. Explorations of the underlying mechanisms and novel hallmarks of LAGC patients with MSI-H/dMMR status are needed to guide clinical practice.
HER2-positive gastric cancer is a distinct subtype of cancer that is invasive and highly metastatic [26]. In the neoadjuvant setting, treatment strategies for LAGC patients with concurrent HER2 positivity and MSI-H/dMMR status are debatable. In our current study, one patient (P4) had HER2 amplification, and PD-1 blockade monotherapy was used as a neoadjuvant treatment. The surgical outcome was TRG2, which was not satisfactory. A recent study reported that the combined neoadjuvant treatment regimen of PD-1 blockade tislelizumab, trastuzumab, and chemotherapy had promising outcomes in LAGC patients with HER2 positivity, with a pCR rate of 58.3% and a major pathological response rate of 66.7% [27], indicating the satisfactory efficacy of anti-HER2 treatment in HER2-positive LAGC patients. However, the optimal neoadjuvant treatment options for LAGC patients with concurrent HER2 positivity and MSI-H/dMMR status need to be further explored.
Reportedly, MSI/MMR heterogeneity can be observed in approximately 8.9% of MSI-H/dMMR GC patients, and its occurrence is correlated with an unsatisfactory tumor response to immunotherapy [28]. Moreover, the conversion from MSI-H/dMMR to heterogeneous MSS or mismatch-repair proficiency (pMMR) status could cause treatment resistance during immunotherapy [22]. In our study, we observed two patients (P4 and P12) with dMMR status at baseline who had postoperative pMMR status, and neither of them achieved a TRG0/1, with pathological tumor responses of TRG2 and TRG3, respectively. Even though these two patients had not experienced relapse before the last follow-up, longer follow-up is needed to draw robust conclusions. Additionally, it should be noted that in the PD-(L)1 blockade-based group, one patient (P19) experienced pathological transformation after neoadjuvant treatment with PD-1 blockade plus chemotherapy, with adenocarcinoma at baseline and postoperative large-cell neuroendocrine carcinoma. Neither radiographical nor pathological downstaging was observed, indicating poor tumor response to PD-1 blockade plus chemotherapy in this patient. In summary, the factors causing unsatisfactory tumor responses to neoadjuvant PD-(L)1 blockade-based immunotherapy in MSI-H/dMMR patients might be but not limited to MSI/MMR heterogeneity or pathological type transformation, and an in-depth understanding is essential for more accurate treatment.
RECIST v1.1 has been widely applied in the evaluation of neoadjuvant treatment efficacy in several solid tumors, but it seems inefficient in treating LAGC. In our study, we also observed inconsistent radiographical and pathological responses in several patients. On the one hand, the pathological tumor response of one radiographical CR patient (P11) was evaluated as TRG2 (ypT2N0); on the other hand, none of the patients with pCR had radiological tumor regression rates greater than 90%. The dissimilar assessments of radiological and pathological evaluations might lead to discrepancies, especially in the neoadjuvant treatment of LAGC patients. Recently, positron emission tomography and computed tomography (PET/CT)-based response criteria in solid tumors (PERCIST) have been applied in several studies [11, 29] and have shown promising prospects in detecting tiny lesions and quantifying metabolism. Admittedly, there are no generally acknowledged perfect criteria till now, and the surrogate measurement still warrants further validation.
Several limitations in this study should be noted. Because of the nonrandomized nature of the study, potential inclusion biases were inevitable; and the relatively limited sample size and insufficient follow-up time might introduce lack of power. In addition, both the agents used for PD-(L)1 blockade and the regimens used for chemotherapy varied. High-quality prospective clinical trials with larger sample sizes and longer follow-up times are urgently needed to further validate the efficacy and safety of neoadjuvant PD-1/PD-L1 blockade plus chemotherapy in MSI-H/dMMR LAGC patients.

Conclusion
In conclusion, neoadjuvant PD-(L)1 blockade-based treatment regimens presented promising results in potentially resectable gastric cancer patients with MSI-H/dMMR status, with favorable overall CR rates and mild safety profiles.

Electronic supplementary material
Below is the link to the electronic supplementary material.