In 2022, gastric cancer (GC) was the 5th leading cause of cancer death in the Russian Federation, with 38% of cases presenting with metastatic disease at the time of the diagnosis [1]. The median overall survival (mOS) of previously untreated patients with metastatic gastric cancer (mGC), who received platinum agents with fluoropyrimidines, is 8.8–10.7 months [2]. Recent advances in addition of targeted therapy to standard doublet chemotherapy resulted in increasing of long-term outcomes of patients with mGC, whose tumors expressed human epidermal growth factor receptor 2 (HER2), programmed death-ligand 1 (PD-L1), and claudin-18 isoform 2 (CLDN18.2). In the phase III ToGA trial, mOS of patients with HER2-positive mGC who received trastuzumab with chemotherapy was 13.8 months compared to 11.1 months in those assigned to chemotherapy alone [3]. Novel monoclonal antibody zolbentuximab prolongs both progression-free survival (PFS) and OS in patients with CLDN18.2-positive GC up to 8.2 and 14.4 months, respectively [4]. The addition of anti-PD-L1 antibodies to standard oxaliplatin-containing chemotherapy in patients with PD-L1 combined positive score (CPS) ≥ 5 and CPS ≥ 10 showed similar results [5, 6].

Unfortunately, the efficacy of the second line in patients with mGC is lower. A combination of ramucirumab and paclitaxel demonstrated an objective response rate (ORR) of 28% and median PFS (mPFS) of 4.4 months in the RAINBOW trial [7]. In the RAMIRIS trial, the mPFS of patients who received ramucirumab with FOLFIRI was 3.9 months [8]. Approximately 38% of patients who failed second-line therapy, received subsequent treatment for mGC [9]. Regorafenib might be a treatment choice in such patients with mOS 5.8 months, however, 67% of grade 3–5 adverse events (AEs) have been observed [10]. Similarly, mOS in heavily pretreated patients, who received trifluridine/tipiracil for mGC, was 5–6 months with higher toxicity grade—up to 84% grade 3–5 AEs [11].

About half of the patients received combined regimens in the second-line treatment for mGC [12], from which nearly 30% had Eastern Cooperative Oncology Group (ECOG) 2 performance status at the beginning of chemotherapy [12]. Immune checkpoint inhibitors (ICIs) seem to preserve the patient’s quality of life and improve long-term outcomes in certain subsets [13–15].

PD-L1 expression and high microsatellite instability (MSI-H) are well-known predictive biomarkers for ICIs in patients with mGC [16, 17]. Nivolumab and pembrolizumab in combination with oxaliplatin-containing chemotherapy are standard first-line therapy of advanced GC in patients with PD-L1 CPS ≥ 5 and CPS ≥ 10, respectively [6, 14]. However, optimal PD-L1 CPS cut-off for GC immunotherapy in patients who failed the first- and second lines of therapy remains challenging. A slight increase in survival rates was observed in patients who received nivolumab for mGC, regardless of PD-L1 expression: mOS was 5.3 months [18]. Similar results were obtained from the KeyNote-059 study with ORR of 15.5% and mOS of 5.6 months in patients with PD-L1 positive mGC [19].

Another predictive biomarker for ICIs is MSI-H: combined analyses of three trials revealed durable responses to anti-PD-L1 antibodies and high survival outcomes in patients with mGC [16]. The prevalence of MSI-H GC is 22%, according to The Cancer Genome Atlas (TCGA) Group [20], however, real-world data suggests the prevalence of MSI-H GC is up to 3–10% [21]. Molecular analyses of gastric adenocarcinoma conducted by TCGA showed a high number of immune signaling pathways not only in MSI-H tumors but in adenocarcinomas with Epstein-Barr virus (EBV), which predicts high sensitivity to ICIs [22, 23]. In Chinese retrospective analyses of patients with mGC who received ICIs, the mPFS of patients with EBV+/MSS was 8.2 months versus 2.0 months in patients with EBV−/MSS (p < 0.001), and mOS in the first group was not reached (p = 0.002) [24].

Tumor mutational burden (TMB) also predicts response to anti-PD-L1 antibodies [25]. In the Korean study, the mOS of patients who received immunotherapy for mGC with high TMB (TMB cut-off was 14.3 Mut/Mb) was 22.4 months compared to 3.6 months in those with low TMB (p < 0.033) [26]. Currently, TMB and EBV determination in tumors is not routinely performed.

Сlinical factors associated with response to ICIs are actively studied in addition to tumor biomarkers [27, 28]. Inflammatory signaling pathways resulting from genomic instability are one of the “Hallmarks of cancer”, and its tumor-induced inflammatory response has been confirmed as an effective prognostic biomarker in many cancers [29]. Neutrophil-lymphocyte ratio (NLR), platelet-lymphocyte ratio (PLR), and lymphocyte-monocyte ratio (LMR) are the most well-studied systemic inflammatory markers with both prognostic and predictive role [30]. Most studies show the prognostic significance of NLR and PLR in patients who received immunotherapy for non-small-cell lung cancer and melanoma treatment [31–34]. In the recent two years, meta-analyses showed an association between low NLR and PLR levels and favorable outcomes with durable responses to ICIs in patients with mGC [35]. Some prognostic scales for patients with mGC have been investigated and validated, however, none include immunotherapy [36, 37].

In real-world data, the mPFS and mOS of patients, who received ICIs for mGC, were 2 and 6 months, respectively, with no statistically significant difference in OS according to microsatellite or PD-L1 status [38, 39]. Given the mentioned above survival rates and lack of prognostic factors, the administration of immunotherapy in pretreated patients with mGC seems challenging. The aim of our study is to assess the feasibility of ICIs in patients with advanced GC, especially in heavily-pretreated populations, and evaluate the prognostic and predictive role of clinical characteristics and tumor biomarkers to identify patients with the most significant benefit from receiving ICIs for the treatment of mGC.

We have analyzed the outcomes of 122 patients who received ICIs for mGC in Russia. Research has shown higher efficacy of anti-PD1 inhibitors in Asian patients compared to European ones [18]. However, real-world data suggested comparable survival outcomes in the Asian population: mOS and mPFS were 5.8 months (95% CI 5.29–7.00) and 1.8 months (95% CI 1.71–1.97), respectively [39]. DCR was 39.4%. Similarly, mOS and mPFS of patients who received nivolumab in the European study were 6.3 months (95% CI 3.3–9.3) and 2.1 months (95% CI 1.4–2.8), respectively [38]. In our study, most patients had PD-L1 positive mGC, so we can not support the benefit of anti-PD-L antibodies in patients with positive PD-L1 CPS. A trend toward better OS was observed in patients with PD-L1 CPS ≥ 5: mOS was 15 months compared to 6 months in patients with PD-L1 CPS СPS < 5. Our study is probably underpowered to detect statistically significant differences in survival rates according to PD-L1 CPS expression. The results of phase III randomized controlled trials strongly suggest the benefit of combining ICIs with standard first-line chemotherapy in patients with PD-L1 CPS ≥ 5–10 mGC [6, 14].

Pembrolizumab is currently the optimal treatment for mGC with PD-L1 CPS ≥ 10: the mOS of patients who received pembrolizumab monotherapy was seven months higher (17.4 months) compared to those who received chemotherapy [13], while the mOS of patients, who received a combination of ramucirumab with chemotherapy for second-line mGC, usually does not exceed 8–9 months [7, 8]. Concerning patients who failed the second line of treatment, in the ATTRACTION-2 trial, no differences in survival outcomes were observed according to PD-L1 TPS [18], due to lower predictive significance for ICIs compared to PD-L1 CPS [17, 40]. As is demonstrated in published data, we revealed statistically significant differences in OS in patients according to MSI status, and a trend towards better PFS in patients with MSI-H compared to MSS was observed: 10 months versus three months, respectively. Almost half of patients with MSI-H were still alive after two years of treatment: the survival curve of PFS reaches a plateau at around 40%, as in several studies of anti-PD-L1 inhibitors [16]. The administration of ICIs in patients with MSI-H alone, with chemotherapy, or via dual blockade in first-line treatment of mGC is still controversial.

In 2021, subgroup analyses of CheckMate-649 revealed that mOS in patients with MSI-H who received nivolumab and ipilimumab was not reached [5]. In updated results of the three-year follow-up of patients with MSI-H, mOS in patients who received nivolumab and oxaliplatin-containing chemotherapy were 38.7 months [14]. Currently, there are no randomized trials assessing the efficacy of anti-PD-L1 inhibitors in combination with chemotherapy compared to chemotherapy alone. However, the mPFS of patients with MSI-H in subgroup analyses of the Keynote-062 study was not reached compared to 11.2 months in patients who received paclitaxel alone [13]. 1-year and 2-year OS were also higher in the pembrolizumab group. In our study, five patients with MSI-H who received ICIs were treatment-naïve, and one of them was alive at the data cut-off. The mean number of immunotherapy cycles was 31. In our study, 18.2% of patients who failed immunotherapy received at least one line of treatment. That is slightly higher than in real-world data [38]. Obtained results indicate acceptable tolerability of ICIs despite issues with data collection.

Our analyses of 71 patients revealed ascites, NLR ≥ 1.8, and administration of anti-PD-L1 inhibitors in III and subsequent lines were prognostic factors for poor OS in patients with MSS mGC.

Peritoneal carcinomatosis is identified in 15%–25% of mGC cases, of which half of patients presented with ascites [41]. Ascitic tumor cells have an extremely immunosuppressive microenvironment [42, 43], resulting in lower survival rates in such patients [43]. Supporting data was obtained in analyses of 59 patients with MSI-H GC: mOS in patients with ascites was six months compared to 29 months in patients with peritoneal carcinomatosis only (p < 0.006) [43]. We did not find any differences in OS in patients with MSI-H according to the presence of ascites (p = 0.592), probably due to a small number of patients (n = 27). The prognostic role of NLR in patients with mGC was first demonstrated in 2019: in this study mOS of patients with low NLR (< 2.83) was 17.1 months compared to 9 months in patients with high NLR (p < 0.001) [44]. Currently, there is no standardized cut-off value of NLR: in existing literature, NLR cut-off levels ranged from 2.5 to 4 [45–47]. Our NLR cut-off (1.8) was detected using ROC analysis for predicting death with the largest square under the ROC-curve. mOS and mPFS in patients with low NLR were not reached compared to four months and two months in patients with high NLR (p < 0.001). However, patients with low NLR have ECOG 0–1 in 75% of cases, and 72% have 1–2 metastatic sites. Those factors were associated with favorable OS in univariate analyses of patients with MSS, which resulted in high outcomes in patients with low NLR. Similar to available data, ORR and DCR in such patients were significantly higher (p < 0.002 and p < 0.027, respectively) compared to those with high NLR [37, 48]. In one study, NLR was not correlated with survival, however, ORR was numerically higher in patients with low NLR [47]. In patients with mGC who received immunotherapy, low TLR is also associated with better survival rates and depths of response [35, 37].

Appropriate selection of pretreated patients with mGC who will benefit from ICIs is needed, considering the increasing data on clinical and biological predictive and prognostic factors. The novel prognostic index was developed in Russian patients with mGC with three prognostic groups based on three risk factors: ECOG 0–1/≥ 2, Hgb level < 10/≥ 10, and progression-free interval of < 5/≥ 5 months after the completion of the first-line chemotherapy [12]. mOS were 13.5 months, 6 months, and 2.9 months according to prognostic groups (p < 0.001). In the modified JCOG prognostic index, which included 608 Asian patients with mGC, risk factors for poor OS were ECOG ≥ 1, presence of primary tumor, high serum alkaline phosphatase level, diffuse Lauren type, and NLR ≥ 4 [36]. However, this study included patients who have failed only first-line treatment of mGC, with low tumor volume (0 to 1 metastatic site in 72% of patients), and with good performance status (ECOG 0–1 in 97% of patients) [36]. The mOS of patients in favorable, intermediate, and unfavorable risk groups were 20.5 months, 13.5 months, and 10.2 months, respectively (p < 0.001) [36].

Limitations of our study are the heterogeneous sample and its small size, inappropriate response assessment, insufficient data on lab tests, and retrospective trial design, resulting in difficulties in distinguishing prognostic and predictive factors. However, we determined an unfavorable prognostic group with mPFS of 1 month, which might help in patient selection for the treatment of mGC with ICIs.

The results of our retrospective real-world analyses of ICIs for mGC are similar to existing data. The low efficacy of immunotherapy was observed in heavily pretreated patients with ascites and NLR ≥ 1.8. More patients need to be included in further analyses with those who did not receive ICIs in the latter lines to determine the prognostic significance of the developed prognostic index.