Efficacy of PD-1 Inhibitors in First-Line Treatment for Advanced Gastroesophageal Junction and Gastric Cancer by Subgroups: A Systematic Review and Meta-Analysis

Gastric cancer is one of the malignant tumors with the highest morbidity and mortality worldwide [1]. Most patients with gastric cancer are in the advanced stage of the tumor when they are first diagnosed [2]. With the continuous exploration of molecular genetics by researchers, the roles of various molecular targets, such as VEGF and HER2, in the progression of gastric cancer have gradually emerged, and the treatment of gastric cancer has continued to improve [3‒5]. The treatment strategy for advanced gastric cancer has evolved from chemotherapy monotherapy, such as cisplatin and anthracycline, to a 5-FU-based regimen and platinum-containing regimen and then developed from a combination regimen containing various new chemotherapy drugs to a combination of chemotherapy and targeted therapy [6‒8]. Over the past 40 years, the median overall survival (OS) of advanced gastric cancer has effectively improved from 4 months to 13.8 months under the continuous development of chemotherapy regimens. Unfortunately, the development of new targeted drugs frequently fails, and the median OS stagnates at the 1-year mark. The innovation of gastric cancer treatment is stuck in a bottleneck, and a breakthrough is urgently needed.

The emergence of immunotherapy brings new hope for the treatment of advanced gastric cancer. Due to the lack of antigen expression in tumor cells and the existence of an immunosuppressive environment, malignant tumors often exhibit immune evasion. Although gastrointestinal tumors are not as highly immunogenic as melanoma and nonsmall cell lung cancer, the infiltration of lymphocytes around tumors is closely related to the efficacy of tumor immunotherapy and long-term prognosis [9, 10]. In recent years, immune checkpoint inhibitors targeting CTLA-4, PD-1/PD-L1, and other immune checkpoints have achieved promising results in various advanced malignant tumors, including melanoma and nonsmall cell lung cancer [11‒14].

Thus, immunotherapy has been initially explored in gastric cancer, from the back line to the front line and from monotherapy to the combination strategy. A number of clinical trials of advanced gastric and gastroesophageal junction adenocarcinoma (G/GEJAC) suggest that ICI therapy can improve the efficacy with controllable toxicity and side effects [15‒18]. The KEYNOTE-059 trial confirmed that the use of pembrolizumab in recurrently advanced gastric cancer resulted in a higher response rate in patients with a positive combined positive score (CPS) of PD-L1 [19, 20]. Based on this result, pembrolizumab was approved by the US Food and Drug Administration (FDA) for advanced G/GEJAC patients with a CPS ≥1. Due to the affected functions, such as eating and digestion, patients with advanced gastric cancer are often too weak to suffer drug stacking. In addition to being unable to tolerate multiple rounds of treatment, their immune systems are also more damaged. Therefore, first-line therapy may be a better time to treat. After moving the number of treatment lines forward, the phase III randomized first-line study CHECKMATE-649 found that ICI-combined chemotherapy can significantly improve the survival benefit for HER2-negative advanced G/GEJAC patients compared with chemotherapy alone [21]. On April 16, 2021, the FDA first approved nivolumab combined with chemotherapy for the first-line treatment of all advanced G/GEJAC based on the results of CHECKMATE-649.

However, not all first-line clinical trials have yielded completely consistent results. The ATTRACTION-4 results suggested that the combined strategy of ICI did not significantly prolong OS compared with chemotherapy alone [22, 23]. Not all gastric cancer patients have a clear response to ICIs, and the screening of ICI-predominant populations is still difficult. In gastric cancer tissues, the expression level of PD-L1 is distinct and associated with many clinicopathological features with poor prognosis. KEYNOTE-062, which only enrolled patients with a CPS ≥1, confirmed that pembrolizumab-combined chemotherapy improved objective response rate (ORR) [24]. However, OS (CPS ≥1 or CPS ≥10) and progression-free survival (PFS) (CPS ≥1) in the pembrolizumab-combined group were not superior to those in the chemotherapy group. The distinctive role of a CPS ≥1 in patients undergoing ICI-combined chemotherapy for G/GEJAC remains unclear. Although microsatellite instability (MSI) has a good predictive effect on the efficacy of ICIs in many solid tumors, it has not been clearly indicated in G/GEJAC. Therefore, it is very important to explore the efficacy of ICI-combined chemotherapy in different subgroups and to identify biomarkers for the population benefiting from ICI therapy. This article aims to evaluate the efficacy of anti-PD-1/PD-L1 therapy in the whole population and subgroups of patients through a meta-analysis of clinical trials for first-line treatment in advanced G/GEJAC.

Clinical trials and randomized controlled trials published in PubMed, Embase, Cochrane database, and Clinical Trials (www.ClinicalTrials.gov) before January 2023 were retrieved for primary screening. The search strategy was utilized as follows: (((advanced gastric cancer [MeSH Terms]) OR gastroesophageal cancer [MeSH Terms]) OR esophageal cancer [MeSH Terms]) AND ((immunotherapy OR PD-1 inhibitor OR PD-L1 inhibitor OR anti-PD-1 OR anti-PD-L1 OR pembrolizumab OR nivolumab OR opdivo OR keytruda OR atezolizumab OR tecentriq OR toripalimab OR JS001 OR sintilimab OR tislelizumab OR immune checkpoint)). To avoid missing the latest research progress in some clinical trials, we also searched conference proceedings in the American Society of Clinical Oncology (ASCO), European Society of Medical Oncology (ESMO), ASCO Gastrointestinal, and ESMO Gastrointestinal during the past 5 years (from 2018 to 2022). Duplicate articles were eliminated from our initial screening by reading titles and abstracts. For the rest of the literature, we intensively read the full text to evaluate the content of the articles and determine whether they met the inclusion and exclusion criteria. In addition, we also searched relevant references manually to find and supplement other potentially relevant reports.

The included literature met the following criteria: (1) randomized clinical trials; (2) metastatic or advanced adenocarcinoma of the gastro-esophagus; (3) comparison of the therapeutic effect and long-term outcome of anti-PD-1/PD-L1 immune checkpoint therapy with standard therapy; (4) first-line therapy; and (5) specific data on OS, PFS, and ORR, including hazard ratios (HRs), survival curves, and the number of clinical remission and the total number of patients in each group. The exclusion criteria were as follows: (1) retrospective phase I study or single-arm phase II study; (2) early-stage gastroesophageal cancer; (3) first-line maintenance therapy, second-line and third-line therapy; (4) lack of specific data on survival and clinical efficacy.

All data used in this study were obtained from publications, relevant online supplementary materials, and presentations or abstracts of conference proceedings. Data were independently reviewed and extracted by two reviewers (Jun Wang and Jing Chen). If there was a disagreement, a consensus solution was obtained through consultation with the third reviewer Wenxuan Wu. When multiple papers or literature referred to the same data, the latest data corresponding to the desired study endpoint were included in the review, and these data were cross-checked with information mentioned in other literature.

The comparative data of long-term survival and short-term clinical efficacy between standard treatment and ICI-combined treatment were collected, including HRs and 95% confidence intervals (CIs) of median OS and median PFS, ORR, and number of patients in clinical remission. In addition, OS, PFS, and ORR data for populations with MSI or various PD-L1 expression levels were collected. We also extracted the NCT number, ICI type and agents, number of patients included in each group, treatment regimen given to the experimental and control groups, endpoint of the study, cutoff time and median follow-up time.

Meta-analyses were performed using Review Manager 5.4.1 software (the Cochrane Collaboration, Copenhagen, Denmark). Figures including forest plots and funnel plots were automatically drawn by Review Manager 5.4.1 software. All reported p values are two sided, with p values <0.05 defined as statistically significant. We used I² to assess heterogeneity across studies. Statistically significant heterogeneity was defined when I² was greater than 50%, while the statistical analysis was considered to have no significant heterogeneity when I² was less than 50%. When the heterogeneity was significant, the random-effects model was used for analysis; when there was acceptable heterogeneity, the fixed-effects model was used for analysis.

For long-term outcomes, HRs from Cox proportional-hazards models were performed in clinical trials to compare survival differences between standard chemotherapy and ICI-combined treatment. We extracted HRs and 95% CIs for OS and PFS assessments in the overall population of the included studies. Meta-analyses were performed to obtain pooled HRs, 95% CIs, and p values via the generic inverse variance method, comparing the effects of different treatments on OS and PFS. For short-term clinical efficacy, clinical trials usually perform imaging assessments of tumor regression in patients according to RECIST criteria. We extracted ORR for the overall population in the included studies, as well as the total number and the number of tumor regressions in each arm. Odds ratios (ORs), 95% CIs, and p values were meta-analyzed to assess whether there were significant differences in clinical efficacy among different arms. For further subgroup analysis, we additionally collected the above OS, PFS, and ORR information in the MSI and PD-L1 subgroups.

The process of literature selection is shown in the flowchart (Fig. 1). We retrieved a total of 197 results from online databases, while manual searches identified 35 studies from other sources. After a rough screening to eliminate duplicates, 127 studies were left for further retrieval. By completely reading the titles and abstracts of the above studies, 92 records were excluded due to inconsistencies in research purposes, cancer types, treatment regimens, or lack of comparable data. Then, full-text articles of the remaining 35 studies were thoroughly assessed. Consequently, 30 studies were excluded due to noncompliant therapeutic lines, phases of trials, treatment regimens, or unreported efficacy, and outcome data. Ultimately, 5 randomized clinical trials were included in this study.

The basic characteristics of the included clinical trials are shown in Table 1. The five clinical trials finally included are all for first-line treatment of advanced G/GEJAC. Three arms were set in the KEYNOTE-062 trial, including a chemotherapy group, pembrolizumab monotherapy, and chemotherapy combined with pembrolizumab. To compare the difference in efficacy between ICI-combined and single chemotherapy, this study only included data from the control arm of chemotherapy and the experimental arm of pembrolizumab-combined therapy. The KEYNOTE-062 researchers published a paper in 2020 releasing ORR and partial survival data [25] and updated the final analysis results at the ASCO meeting in 2022 [26]. Therefore, ORR data for this study were derived from papers published in 2020, while survival data were extracted from results presented at the 2022 ASCO meeting.

CHECKMATE-649 lays the foundation for nivolumab combined with chemotherapy as a first-line treatment for advanced HER-2-negative gastric cancer. It also has three treatment arms, including chemotherapy, nivolumab combined with chemotherapy, and nivolumab combined with ipilimumab [24]. This study only extracted the latest efficacy and survival data of the chemotherapy arm and nivolumab-combined chemotherapy arm. The phase III portion of the ATTACTION-4 trial compared the efficacy and safety of nivolumab plus chemotherapy or placebo plus chemotherapy as first-line therapy in Asian patients with HER2-negative advanced G/GEJAC. Efficacy and survival data for treatment groups in ATTRACTION-4 were extracted for this meta-analysis. The KEYNOTE-811 global cohort compared the efficacy and safety of patients receiving pembrolizumab or placebo in combination with trastuzumab and chemotherapy as standard first-line therapy for HER2-positive advanced G/GEJAC. The latest published results of interim analysis for KEYNOTE-811 provided ORR-related data for this study. The ORIENT-16 trial is the first phase III study to report the efficacy and safety of sintilimab-combined chemotherapy versus placebo-combined chemotherapy in the first-line treatment of advanced G/GEJAC. This meta-analysis included the latest ORR and survival results of ORIENT-16 presented in an oral presentation at the ESMO Congress 2021. All trials included were randomized, prospective, and multicenter clinical trials. We assessed the risk of bias and found that all included studies were of acceptable quality (online suppl. Fig. 1, 2; for all online suppl. material, see https://doi.org/10.1159/000531457).

In the intention-to-treat cohort, as Figure 2a shows, treatment with ICIs prolonged the median OS overall compared with chemotherapy alone (HR = 0.82, 95% CI: 0.76–0.88, p < 0.00001). Integrated analysis of 4 trials (ATTRACTION-4, CHECKMATE-649, KEYNOTE-062, and ORIENT-16) supported that median PFS acquired the same benefit as median OS (Fig. 2b). Compared to the chemotherapy group, the median PFS was significantly longer in the ICI-based group (HR = 0.75, 95% CI: 0.69–0.82, p < 0.00001). Collecting the ORR data, we evaluated the short-term clinical responses of all included trials (n = 5), as shown in Figure 2c. Meta-analysis showed that when combined with ICIs, the objective response significantly improved compared to accepting chemotherapy alone (OR = 0.63, 95% CI: 0.55–0.72, p < 0.00001).

OS was first assessed to evaluate different MSIs (Fig. 3a). As the forest plot shows, the median OS was prolonged in the ICI-combined group compared with the control group in both MSI-H patients (HR = 0.38, 95% CI: 0.20–0.70, p = 0.002) and MSS patients (HR = 0.78, 95% CI: 0.70–0.87, p < 0.00001). Although subgroup differences met statistical significance (p = 0.02), high heterogeneity existed in the subgroups (I² = 80.7%). Regarding the ORR as the endpoint of short-term response, we assessed the cumulative HRs for the ORR of the MSI-H and MSS subgroups (Fig. 3b). Similarly, ICI-combined treatment did not improve ORR significantly when compared with chemotherapy alone in MSI-H patients (OR = 0.44, 95% CI: 0.18–1.08, p = 0.07). However, in MSS patients, ICI-combined treatment was related to a higher ORR than in the control group (OR = 0.60, 95% CI: 0.48–0.74, p < 0.00001).

CPS was frequently discussed as the most promising biomarker for ICI-based treatment. Thus, OS, PFS, and ORR data were sorted for the subgroup analysis. To accurately analyze the subpopulation who acquire benefits most from ICI-based strategies, 1, 5, and 10 were most commonly set as the cutoff value of CPS. As the main endpoint of survival, the median OS data of various subgroups were adequate (Fig. 4). Analysis of the survival data in the CPS ≥1 subgroup suggested that ICI-combined treatment prolonged the median OS compared with chemotherapy (HR = 0.76, 95% CI: 0.70–0.84, p < 0.00001), while the median OS did not significantly differ between the ICI-combined arm and chemotherapy arm in the CPS <1 subgroup (HR = 0.95, 95% CI: 0.73–1.24, p = 0.70). Similarly, the HR for the median OS benefit with the ICI-combined arm compared to the chemotherapy arm was 0.69 (95% CI: 0.62–0.78, p < 0.00001) in the CPS ≥5 subgroup, while the HR in the CPS <5 subgroup was 0.94 (95% CI: 0.79–1.12, p = 0.49). The subgroup difference (p = 0.002) between CPS ≥10 (HR = 0.66, 95% CI: 0.58–0.75, p < 0.00001) and CPS <10 (HR = 0.91, 95% CI: 0.78–1.06, p = 0.23) patients was similar to that of subgroups when the CPS cutoff was 5 (p = 0.004).

As another survival endpoint, evaluation of PFS demonstrated similar results as OS (Fig. 5). In the CPS ≥1 subgroup, patients who received ICI-combined therapy had longer PFS than those who received chemotherapy alone (HR = 0.75, 95% CI: 0.68–0.84, p < 0.00001). In the CPS <1 subgroup, patients who received chemotherapy or ICI-combined therapy had indistinctive PFS (HR = 0.93, 95% CI: 0.69–1.25, p = 0.63). However, the subgroup difference did not approach statistical significance (p = 0.19). When the cutoff was set as 5, the magnitude of benefit for ICI-combined therapy was higher compared to chemotherapy alone in the CPS ≥5 subgroup (HR = 0.68, 95% CI: 0.61–0.77, p < 0.00001) versus in the CPS <5 subgroup (HR = 0.93, 95% CI: 0.76–1.14, p = 0.48) and met statistical significance with p = 0.01 for subgroup difference.

For short-term response, we calculated the ORs for ORR differences in patients under ICI-combined therapy and chemotherapy in the CPS ≥1 subgroup (OR = 0.55, 95% CI: 0.45–0.66, p < 0.00001) and CPS <1 (OR = 0.82, 95% CI: 0.21–3.22, p = 0.78) (Fig. 6a, subgroup differences p = 0.56, I² = 0%). When the subgroups were divided by a cutoff of 5, ICI-combined treatment benefitted more than chemotherapy alone in the CPS-high subgroup (OR = 0.56, 95% CI: 0.45–0.69, p < 0.00001), while it benefited the same in the CPS-low subgroup (OR = 0.82, 95% CI: 0.21–3.22, p = 0.78), with no significance between subgroups (Fig. 6b, p = 0.57, I² = 0%).

In the subgroup analysis by PD-L1 expression in tumor cells, only median OS data were fully reported in the CHECKMATE-649 and ATTRACTION-4 trials and sequentially collected for our meta-analysis (Fig. 7a). The median OS was not distinct between the ICI-combined group and the control group in patients with ≥1 PD-L1-positive tumor cell (HR = 0.74, 95% CI: 0.38–1.42, p = 0.36). For patients with PD-L1-positive tumor cells <1, the HR for the median OS benefit in the ICI-combined group compared to the chemotherapy group was 0.85 (95% CI: 0.77–0.93, p = 0.0009). In total, the difference between subgroups was not statistically significant (p = 0.68), even with low heterogeneity (I² = 0%).

Since the ICI agents used above were both PD-1 inhibitors, we assessed the clinical response of different agents (Fig. 7b). Subgroup analysis suggested that the combination of pembrolizumab (OR = 0.50, 95% CI: 0.30–0.83, p = 0.007) and nivolumab (OR = 0.64, 95% CI: 0.53–0.76, p < 0.00001) could both improve the ORR compared to the chemotherapy group. However, the effect of sintilimab on ORR was rare, which might be due to the limited number of trials, and only ORIENT-16 was included for assessment (OR = 0.75, 95% CI: 0.55–1.04, p = 0.08). Although the heterogeneity of this subgroup assessment was as low as I² = 0%, the difference between subgroups of various ICI agents did not approach statistical significance (p = 0.38).

PD-1 is a negative regulatory receptor mainly expressed on the surface of activated T cells. It can inhibit T cells by binding to the ligand PD-L1 expressed on the surface of tumor cells, making it unable to attack tumor cells and escape immune surveillance. At present, a variety of drugs targeting the PD-1/PD-L pathway have been successfully validated in trials and have been approved for clinical applications in a variety of solid tumors [13, 25, 28, 29, 30]. In the field of advanced G/GEJAC, many large multicenter clinical trials of various therapeutic phases are underway, and some have revealed exciting results [21, 23, 24, 26, 27, 31]. CHECKMAKE-649 and ORIENT-16 consolidate the position of ICI-combined chemotherapy in the first-line treatment of advanced G/GEJAC. However, the population that is most sensitive to ICI treatment and acquires benefit is still unclear. As KEYNOTE-811 announced part of the efficacy evaluation data, we are the first to retrospectively analyze five phase III clinical trials of advanced G/GEJAC for first-line treatment. This study conducted a meta-analysis of five randomized controlled trials to clarify the prognostic value of ICI-combined chemotherapy in advanced G/GEJAC for first-line treatment and tried to explore the population who could benefit most from ICI-combined chemotherapy through subgroup analysis.

In this meta-analysis, by analyzing short-term efficacy and long-term survival data in the overall population, we found that the combination of ICI and chemotherapy was effective in improving ORR, OS, and PFS compared with chemotherapy alone. Although ATTRACTION-4 and KEYNOTE-062 did not show superior results in long-term survival data, the results of the meta-analysis affirmed the effect of ICI-combined chemotherapy on optimizing OS (HR = 0.77, p < 0.00001) and PFS (HR = 0.75, p < 0.00001). Among the different agents, pembrolizumab (HR = 0.50, p = 0.007) and nivolumab (HR = 0.50, p < 0.00001) both effectively improved the ORR, while sintilimab nonsignificantly improved the ORR (HR = 0.75, p = 0.08). This may be because ORIENT-16 was the only trial of the sintilimab-combined study that had published results with a limited sample size. This result still needs to be supplemented by more clinical trials with larger sample sizes to provide new agents for ICI combination therapy.

To determine the population that could benefit most from ICI combination therapy, we performed subgroup analysis of common biomarkers. It is well known that a high tumor mutational burden is associated with better ICI efficacy in various solid tumors. Tumors with high microsatellite instability (MSI-H) or mismatch repair deficiency showed higher mutation rates and methylation, leading to enhanced expression of neoantigens. Thus, MSI-H tumors are often accompanied by high infiltration of CD8⁺ T cells and are more sensitive to ICIs. Through sequencing analysis of primary tumors in G/GEJAC, previous researchers also found that MSI-H tumors were resistant to chemotherapy and were more likely to have durable responses to immunotherapy [44]. This meta-analysis suggested that ICI-combined chemotherapy could prolong OS in both MSI-H (HR = 0.38, p = 0.002) and MSS (HR = 0.78, p < 0.00001) populations, and there was a significant difference between the groups (χ² = 5.17, p = 0.02). In terms of improving the ORR, ICI-combined chemotherapy also had a benefit in the MSS group (HR = 0.60, p < 0.00001) but not in the MSI-H group (HR = 0.44, p = 0.07). However, there was no significant difference between groups (χ² = 0.41, p = 0.52). This result suggests that MSI has an inaccurate discriminative ability in distinguishing the efficacy of ICIs in patients with G/GEJAC. In the future, combined evaluation of MSI-H and other clinical or pathological features may be required to improve the sensitivity and accuracy of identifying potentially beneficial populations.

In addition to MSI, the expression level of PD-L1 is also a common marker of immunotherapy. In vivo, tumor cells, immune cells such as lymphocytes, macrophages, and interstitial cells also express PD-L1. The common evaluation criteria for PD-L1 expression level include CPS score, TPS score, IPS score, and so on. Different clinical trials have adopted various evaluation criteria for PD-L1 levels. In addition to evaluating PD-L1 expression only on tumor cells, more studies currently use CPS to comprehensively evaluate PD-L1 expression on tumor cells, immune cells, and interstitial cells in tumor lesions. At present, in different cancer types, the cutoff value of the CPS score is different. In the included trials of this study, CPS and TPS were mainly used for the evaluation. EMA and NCCN guidelines have recommended ICI-combined chemotherapy for advanced G/GEJAC with a CPS ≥5. However, the population that the FDA approved is the whole population regardless of CPS. Therefore, the discriminative role of the PD-L1 expression level in G/GEJAC remains unclear. Our subgroup analysis suggested that combination therapy with ICIs was more effective than chemotherapy alone in prolonging OS in the subgroup with a high CPS, regardless of the CPS cutoff for PD-L1. However, when the cutoff of the CPS was 1, the difference between subgroups did not reach statistical significance (χ² = 2.36, p = 0.12), while the benefit ratio of the MSI-H group when the cutoff was 10 (χ² = 8.16, p = 0.004) was higher than when the cutoff value was 5 (χ² = 9.85, p = 0.002).

This study also has some limitations. Data on additional biomarkers, such as TMB and TCR diversity, could not be extracted from the included literature due to the paucity of published clinical trial data. It is hoped that a more in-depth stratified analysis can be carried out in the future to determine the groups that could benefit from immunotherapy. Second, heterogeneity existed in the included studies. Thus, this article uses more random-effects models for analysis rather than fixed-effects models.

Overall, this meta-analysis suggests that for the first-line treatment of advanced G/GEJAC, ICI-combined chemotherapy is superior to chemotherapy with better short-term efficacy and higher long-term survival. In addition, the ability to distinguish the best benefit population of MSI is poorer than that of the CPS, and the subgroup of patients with a CPS ≥10 may have more significant benefits. A CPS ≥10 has the potential to become an accurate marker of the dominant population of immuno-combined therapy.

We greatly thank the Department of Gastroenterology Surgery, the Second Affiliated Hospital, Zhejiang University School of Medicine, for technical advice.

An ethics statement is not applicable because this study is based exclusively on published literature.

All authors state that there is no conflict of interest.

This work was supported by Zhejiang Provincial Key Project of Research and Development (2019C03043), Health Science and Technology Plan of Zhejiang Province (2022RC165), and Clinical Research Fund of Zhejiang Medical Association (2021ZYC-A68).

Conceptualization: Shengqi Fei, Jun Wang, and Yu Lu; data review and extraction: Jun Wang, Jian Chen, and Wenxuan Wu; software: Jing Chen and Jia Qi; original draft preparation: Shengqi Fei and Beidi Wang; review and editing: Yaxuan Han, Kefan Wang, and Jian Chen; and supervision: Xiaying Han, Haiyan Zhou, and Jian Chen. All authors have read and agreed to the published version of the manuscript.

Shengqi Fei and Yu Lu contributed equally to this work.

All data generated or analyzed during this study are included in this article and its online supplementary material files. Further inquiries can be directed to the corresponding author.