Adjuvant Chemotherapy for Gastric Cancer Patients with Mismatch Repair Deficiency or Microsatellite Instability: Systematic Review and Meta-Analysis

Mismatch repair-deficient (dMMR) or microsatellite instability-high (MSI-H) status serves as a predictor of a poor response to adjuvant chemotherapy among stage 2 colon cancer patients. This study aimed to investigate the efficacy of adjuvant chemotherapy in dMMR/MSI-H gastric cancer (GC). Clinical studies comparing adjuvant chemotherapy and surgery alone in dMMR/MSI-H GCs through June 2021 were retrieved to assess the survival of patients managed with both treatments. Two approaches were used to pool the hazard ratio (HR) of survival: (1) if Kaplan-Meier curves and number of patients at risk were provided, individual patient data were extracted. Cox models were used to calculate the HR with its 95% confidence interval (CI); (2) for study-level data, pooled HR was estimated using fixed/random-effects models. Seven clinical studies were assessed. For dMMR/MSI-H versus mismatch repair-proficient (pMMR)/microsatellite stable (MSS)/microsatellite instability-low (MSI-L) status, the estimated 5-year disease-free survival (DFS) rate was 74.2% versus 51.5% (HR, 0.44; 95% CI, 0.32–0.62; P < 0.001) and the estimated 5-year OS rate was 60.5% versus 49.1% (HR, 0.71; 95% CI, 0.60–0.85; P < 0.001). The study-level data showed pooled HRs of 0.42 for DFS (95% CI, 0.31–0.57; P < 0.001) and 0.65 for OS (95% CI, 0.38–1.11; P = 0.114). For adjuvant chemotherapy versus observation of dMMR/MSI-H, the estimated 5-year DFS rate was 76.1% versus 73.3% (HR, 0.72; 95% CI, 0.45–1.15; P = 0.171) and the estimated 5-year OS rate was 73.5% versus 59.7% (HR, 0.62; 95% CI, 0.46–0.83; P = 0.001). Significant survival differences also were observed at study level. The study findings confirm the benefit of adjuvant chemotherapy for dMMR/MSI-H GC patients.

CI, 0.46-0.83; P = 0.001). Significant survival differences also were observed at study level. Conclusions. The study findings confirm the benefit of adjuvant chemotherapy for dMMR/MSI-H GC patients.
More than 1 million new gastric cancer (GC) cases and 783,000 deaths were estimated to occur globally in 2018, ranking GC as the fifth most frequently diagnosed cancer and the third leading cancer-related death worldwide. 1,2 In Asia, adjuvant chemotherapy with S1 3,4 or capecitabine plus oxaliplatin 5,6 is the current standard of care to prevent recurrence and improve survival after curative D2 gastrectomy for stages 2 and 3 GC patients. 7 However, approximately 40% of patients with primary resected GC receiving adjuvant chemotherapy still experience relapse, metastases, or both. 4,6,8 Thus, predictive tools able to identify patients who would possibly benefit from adjuvant chemotherapy are urgently needed.
Cheong et al. 9 reported that four classifier genes (GZMB, WARS, SFRP4, and CDX1) could serve as a prognostic and predictive tool to predict survival and adjuvant chemotherapy response after surgery for patients with stages 2 and 3 GC. Recently, Wang et al. 10 established a novel score system based on six immune checkpoints to assist adjuvant chemotherapy selection among GC patients.
Tumors deficient in DNA mismatch repair (dMMR) lose the ability of cells to recognize and repair spontaneous DNA mutations. 11 The distinct characteristics of dMMR tumors include high tumor mutation burden and hypermutator phenotype, termed microsatellite instability-high (MSI-H). 12,13 Previous studies have indicated that dMMR/ MSI-H correlates with improved survival in resectable colon cancer and serves as a predictor of poor response to adjuvant chemotherapy for stage 2 colon cancer patients. 14,15 Findings have shown the frequency of dMMR/MSI-H to be relatively high in GC (about 9-22% of all GC cases), 12,16 raising the debatable issue whether dMMR/MSI-H could serve as a predictor to identify patients who would respond to adjuvant chemotherapy.
Both the post hoc of the CLASSIC study 17 and individual patient data (IPD) from four randomized trials (MAGIC, CLASSIC, ARTIST and ITACA-S) 18 showed that patients with MSI-H did not benefit from adjuvant chemotherapy. However, the sample of patients (n = 61) with MSI-H was small in the MAGIC and CLASSIC trials, which could have largely reduced statistical power. In addition, resected rather than biopsy specimens were examined in the chemotherapy arm of MAGIC study. These considerations indicate that the pathologic major or complete responders were not included in these trials, suggesting a certain level of bias. Subsequently, a recent large cohort study comprising 162 MSI-H patients with GC showed that adjuvant chemotherapy was associated with longer disease-free survival (DFS) and overall survival (OS) than surgery alone. 19 Thus, the significance of adjuvant chemotherapy in dMMR/MSI-H GC is yet to be fully elucidated.
In this study, we performed a systematic review and meta-analysis at the individual and study levels to investigate whether resectable GC patients with dMMR/MSI-H could benefit from adjuvant chemotherapy. The prognostic value of dMMR/MSI-H versus mismatch repair-proficient (pMMR), microsatellite-stable (MSS), and microsatellite instability-low (MSI-L) conditions in gastric cancer also was explored.

Search Strategy and Selection Criteria
Three investigators (R.C.N., G.M.C., and S.Q.Y.) searched the PubMed, Web of Science, and Embase databases until 13 June 2021 for clinical studies that reported on the survival outcomes of adjuvant chemotherapy versus surgery alone for dMMR or MSI-H GC patients after radical resection. The key words used were ''gastric cancer,'' ''adjuvant chemotherapy,'' and ''dMMR'' or ''MSI-H.'' The detailed search terms are displayed in online supplemental Box 1. This study was conducted in compliance with the PRISMA statement guidelines for systematic reviews. 20 We excluded studies that were reviews, conference abstracts, case reports, or abstracts. If duplicated studies were reported, we included only their most recent reports that had complete data. Disagreements were resolved by consulting a senior investigator (Y.F.L.).

Data Extraction and Quality Assessment
The following information was extracted (by R.C.N. and G.M.C) from the included studies if available: study type, study period, number of dMMR/MSI-H gastric cancer cases, and number of adjuvant chemotherapies versus surgeries alone for dMMR/MSI-H patients. Subsequently, the study also extracted survival rates and hazard ratios (HRs) with 95 % confidence intervals (CIs) of adjuvant chemotherapy versus surgery alone for the dMMR/MSI-H patients and/or dMMR/MSI-H patients versus the pMMR/ MSS/MSI-L patients. The detailed information of MSI status, chemotherapy, and survival in the study of Wang et al. 10 is shown in the Supplementary Data.
The qualities of randomized controlled trials (RCTs) and retrospective studies were assessed using the Cochrane risk of bias tool 21 and the Newcastle-Ottawa scale, 22 respectively. A score of 0 to 9 was allocated to each retrospective study. All RCTs and retrospective studies with a score higher than 6 points were considered as high quality.

Statistical Analysis
The primary end points of this study were DFS and OS differences between adjuvant chemotherapy and surgery alone after surgery for dMMR/MSI-H patients. We also assessed the survival differences between the dMMR/MSI-H and pMMR/MSS/MSI-L patients. This study included relapse-free survival in the definition of DFS.
Two different approaches were used to pool the HR of survival. (1) If the Kaplan-Meier curves and number at risk information were provided, IPDs were extracted using graph digitizer software (Engauge Digitizer 12.1) and the ''reconstructKM'' package in R (https://github.com/ryanrs un/reconstructKM). Then, the survival curves were estimated using the Kaplan-Meier method, and Cox proportional hazard regression models were used to estimate the HR. (2) For study-level data, the HR was estimated by pooling the HRs of each eligible study. The odds ratio (OR) was used to explore the associations between dMMR/MSI-H and clinicopathologic characteristics. The Cochrane Q test and the I 2 statistic were used to assess the level of heterogeneity among studies, with a P value lower than 0.10 and an I 2 value greater than 50% considered to indicate significant heterogeneity. The random-effects models were chosen if heterogeneity was observed. Otherwise, fixed-effects models were used.
A two-sided P value lower than 0.05 was considered statistically significant. All analyses were performed using statistical software R version 3.6.1 (R Foundation for Statistical Computing; http://www.r-project.org).

Study Characteristics
A literature search of databases resulted in the retrieval of 354 records. Subsequently, the study excluded 133 records because of duplication, 205 records after screening titles and abstracts, and 9 studies after testing the full-text articles. Finally, seven studies 10,17,19,[23][24][25][26] were identified for quantitative analysis (Fig. 1).
Overall, the meta-analysis comprised six retrospective studies and one RCT. The percentage of dMMR/MSI-H ranged from 6.8 % to 26.3 %. Kim et al. 19 reported two cohorts, one of which compared survival between MSI-H and MSS/MSI-L. The other cohort was included to analyze the efficacy of adjuvant chemotherapy for MSI-H patients. The chemotherapy regimens of the seven studies in the current series were different, mainly including platinum/ fluorouracil-based regimens. Three of these seven studies were identified as high quality (online Table S1).

Survival of the dMMR/MSI-H Patients
The survival curves for dMMR/MSI-H and pMMR/ MSS/MSI-L are shown in Fig. 2
Finally, the efficacy of adjuvant chemotherapy for the dMMR/MSI-H patients with study-level data was explored (Fig. 5). The pooled results were assessed by the fixedeffects model because the Cochrane Q test and the I 2 statistic showed homogeneity. The forest plot showed that the pooled HRs for adjuvant chemotherapy after surgery versus surgery alone indicated that dMMR/MSI-H patients could benefit from adjuvant chemotherapy, with estimated HRs of 0.56 (95% CI, 0.36-0.87; P = 0.010; Fig. 5a) for DFS and 0.62 (95% CI, 0.45-0.83; P = 0.002; Fig. 5b) for OS.

DISCUSSION
To the best of our knowledge, this study is the first systematic review and meta-analysis to investigate the efficacy of adjuvant chemotherapy in GC with dMMR/ MSI-H. Using the published data from seven studies, our pooled analysis suggested that patients with dMMR/MSI-H had superior DFS and OS compared with those with pMMR/MSS/MSI-L. Notably, DFS and OS were longer for the dMMR/MSI-H patients treated with adjuvant chemotherapy than for those treated with surgery alone. These findings indicated that GC patients with dMMR/ MSI-H still could benefit from adjuvant chemotherapy after radical resection.
One of the four major subtypes in gastric cancer is MMR deficiency or MSI-H. 27 Tumors with dMMR/MSI-H have distinct clinical and pathologic features. A previous meta-analysis 28 showed that dMMR/MSI-H occurred more frequently in tumors of female and elderly patients, tumors were located in the lower one third of the stomach with intestinal Lauren classification, and tumors at an early TNM stage. Our study indicated the coincidental results. Individual-and study-level data confirmed the positive prognostic effect of dMMR/MSI-H status with respect to DFS and OS, consistent with previous studies. 18,28,29 A higher percentage of Lauren intestinal subtype and early stage in dMMR/MSI-H could indicate that dMMR/MSI-H may serve as a confounder rather than prognostic factor. However, IPD showed that the pooled HRs of dMMR/MSI-H for DFS (0.42) and OS (0.65) were quite low. Kim et al. 19 also reported that MSI-H status was a prognostic  18 reported the IPD results from four randomized trials. Their results showed that MSI-H GCs could not benefit from adjuvant chemotherapy. Based on these results, a 2020 clinical guideline of the Chinese Society of Clinical Oncology (CSCO) for the diagnosis and treatment of gastric cancer recommends that clinicians not encourage MSI-H patients to receive adjuvant chemotherapy after radical surgery, but to observe or involve immunotherapy in clinical trials. 30 However, the small sample size in the study of Pietrantonio et al. 18 significantly reduced its statistical power, limiting the guidance of adjuvant chemotherapy based on the status of MMR or MSI. Nevertheless, even in colorectal cancer, Cohen et al. 31 reported that adding oxaliplatin to fluoropyrimidine significantly improved the OS of MSI-H stage 3 patients, emphasizing the importance of platinumbased adjuvant chemotherapy.
Our study had the largest cohort of individual-and study-level data of dMMR/MSI-H patients. Our findings showed that adjuvant chemotherapy could improve the DFS and OS of dMMR/MSI-H patients. Therefore, we should not justify a change in the clinical practice at this time. Based on our findings, we suggest that adjuvant chemotherapy rather than observation should be recommended to dMMR/MSI-H patients after radical surgery.
One new dilemma that could be raised is whether adjuvant chemotherapy is the optimal therapeutic strategy for GC patients with dMMR/MSI-H. It is recognized that tumors with dMMR/MSI-H often are associated with many somatic mutations that encode ''non-self'' immunogenic neoantigens. These tumors are accompanied by infiltrative immune cells and a high tumor mutation burden, and thus are susceptible to immunotherapy. Programmed death 1 (PD-1) blockade has demonstrated a clinical benefit for patients with dMMR/MSI-H regardless of tumor types. [32][33][34] Based on the results of KEYNOTE-059 that showed an effective objective response rate for GC patients with dMMR/MSI-H (OR: 57%, 4 of 7 patients), 35  To date, three prospective clinical trials (NCT03257163, NCT04556253, and NCT04744649) have been performed to explore the therapeutic efficacy of neoadjuvant anti-PD-1 treatment for GC patients with dMMR/MSI-H (online Table S3). \ST3[ Therefore, prospective clinical trials to investigate the efficacy of adjuvant PD-1 blockade versus chemotherapy for dMMR/MSI-H GCs are urgently needed.
A major limitation of this study was that the eligible studies were retrospective studies or post hoc analyses of clinical trials. The heterogeneity among the studies could have affected our results to a certain extent. Thus, randomeffects models were performed when heterogeneity was observed, and the large sample size used improved the statistical power and reliability of our analysis. Despite this, a head-to-head prospective randomized control trial still is needed to clarify the efficacy of adjuvant chemotherapy for patients with dMMR/MSI-H. In this era of immunotherapy, we believe that prospective trials should compare the efficacy between adjuvant PD-1 blockade and chemotherapy for patients with dMMR/MSI-H.
In conclusion, our meta-analysis suggests that GC patients with dMMR/MSI-H had better survival than those with pMMR/MSS/MSI-L. Adjuvant chemotherapy could prolong the survival of dMMR/MSI-H patients compared with observation alone. Thus, chemotherapy should not be omitted for dMMR/MSI-H patients after surgery. In this era of immunotherapy, adjuvant PD-1 blockade for dMMR/MSI-H patients should be investigated prospectively, but chemotherapy still should be set as the control arm of the trial.