1 Introduction

The addition of immunotherapy in the treatment of advanced gastric cancer can prolong the overall survival (OS) compared to chemotherapy [1], in which the survival time is poor with a median OS of less than one year [2]. According to checkmate 649, the combination of nivolumab with chemotherapy can have beneficial outcomes with PFS and OS regardless of Programmed cell death ligand 1 (PD-L1) expression [3]. However, not all patients can benefit from immunotherapy. Hence, the selection of biomarkers for predicting the efficacy of treatment or survival prognosis is of great interest. Well-recognized biomarkers include the expression of PD-L1 [4, 5], the tumor mutation burden (TMB) [6, 7], and microsatellite instability (MSI) [8]. However, these methods still have limitations.

Furthermore, system inflammation indexes (SII) such as the neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR) in peripheral blood samples had been explored as biomarkers in various cancers treated with immunotherapy, including melanoma, esophagus cancer, non-small cell lung cancer (NSCLC), and renal cell cancer. Additionally, NLR changes were also found to be prognostic factors in retrospective studies of NSCLC [9] and renal cell carcinoma patients treated with ICIs. Previously, in patients with metastatic gastric cancer (MGC) in the third line of nivolumab monotherapy, NLR changes could be used to predict outcomes. Meanwhile, there are no reports exploring the prognosis value of dynamic changes on the NLR before and after Programmed cell death-1(PD-1) inhibitors were given in combination or as a monotherapy for patients with advanced or recurrent GC. Therefore, in the present study, we explored the prognostic value of NLR changes in MGC patients treated with PD-1 inhibitors.

2 Methods

2.1 Patients

We performed a retrospective analysis of patients diagnosed with metastatic gastric cancer (MGC) under anti-PD-1/PD-L1-based treatment regimens from October 1st, 2015 to December 31st, 2020 at the Chinese PLA General Hospital. Medical and pathological records were obtained from electric medical records. This study followed the Declaration of Helsinki and was approved by the Ethics Committee of the Chinese PLA General Hospital (No. S2019-200–01). The selection criteria included: confirmed pathologic diagnosis as adenocarcinoma gastric cancer; had at least one measurable lesion; received at least two cycles of anti-PD-1 inhibitor drugs besides anti-PD-L1 antibodies; had at least one imaging response assessment through the treatment period.

2.2 Data collection

The data collected included age, gender, PD-L1 expression status, Eastern Cooperative Oncology Group Performance Status (ECOG PS), tumor location, differentiation, metastasis organs, therapy line, and treatment regime. The complete blood count, including leukocytes, neutrophils, and lymphocytes, was evaluated before PD-1 inhibitor administration and after two cycles of PD-1-based treatment. The baseline NLR values were defined as the absolute neutrophil count divided by the absolute lymphocyte count at the beginning of the PD-1 treatment. The ΔNLR was defined as the difference between the NLR after two cycles of treatment and the baseline NLR. The anti-cancer response was also derived from medical data. Based on the Response Evaluation Criteria in Solid Tumor version 1.1, response states were classified into CR (complete response), PR (partial response), SD (stable disease), and PD (progression disease). The survival data, including the OS calculated from the first PD-1 inhibitor treatment until death or the last follow-up (censored), and the PFS derived from the first PD-1 inhibitor treatment until the first documented disease progression (PD), death, or the last follow-up (censored) was also retrieved.

2.3 Treatment regimen

Baseline characteristics of treatment regimens were divided into two parts: monotherapy or combination with chemotherapy. The type of PD-1 inhibitor was decided according to the physician’s choice and patient's willingness, which contain Pembrolizumab (200 mg), and nivolumab at a dose of 3 mg/kg, 200 mg sintilimab, or 240 mg toripalimab intravenously once every 3 weeks. The combination chemotherapy included oxaliplatin-based regimen or docetaxel-based with cisplatin or fluorouracil and other combinations. For treatment, all patients signed informed consent forms.

2.4 Patients groups

Considering the dynamic NLR changes, we classified patients into ΔNLR ≥ 0 or ΔNLR < 0 groups. To further explore the prognostic value of the combination of baseline NLR and ΔNLR, patients were also stratified into three groups regarding the baseline NLR cut-off (optimal cutoff of baseline NLR were 3.23 according to our previous study [10]) and ΔNLR ≥ 0; Group 1: NLR < 3.23 and ΔNLR < 0; Group 2: NLR ≥ 3.23 and ΔNLR < 0 or NLR < 3.23 and ΔNLR ≥ 0; Group 3: NLR ≥ 3.23 and ΔNLR ≥ 0.

2.5 Statistical analyses

Statistical analyses were performed using SPSS for Windows v 20.0 (SPSS, Inc., Chicago, IL). Differences in category variables among groups were compared by χ2 tests. The objective response rate (ORR) was defined as the number of CR plus PR/all populations, and the disease control rate (DCR) was defined as the number of CR + PR + SD/populations. Survival curves of PFS and OS were estimated by the Kaplan–Meier method. The log-rank test was used to compare OS and PFS between groups when the patients were stratified into different ΔNLR groups or different combination groups (baseline NLR and ΔNLR groups). Prognostic factors for OS and PFS were evaluated by multivariate analysis. The level of significance was set to p < 0.05.

3 Results

3.1 Patient characteristics

A total of 137 patients were enrolled in this study. The baseline NLR group (cutoff of 3.23) is presented in Table 1. Then, we categorized patients into decreased (ΔNLR < 0) and increased (ΔNLR ≥ 0) post-treatment NLR groups. Sixty-seven (48.9%) and 70 patients (50.1%) presented in t ΔNLR < 0 and ΔNLR ≥ 0, respectively. No significant differences were detected in baseline characteristics between baseline NLR or NLR change groups. The median OS and PFS of the whole population were 10.70 (95% CI: 8.67–12.72 months) and 5.20 months (95% CI: 3.75–6.64 months), respectively.

Table 1 Patient demographic and clinical characteristics

3.2 ΔNLR and survival

The median OS and PFS were 12.3 (95% CI: 10.7– 14.4 months) and 7.8 months (95% CI: 5.4–11.3 months) for patients with ΔNLR < 0, respectively, and 7.5 (95% CI: 4.6–8.7 months) and 4.3 months (95% CI: 3.0–4.7 months) for patients with ΔNLR ≥ 0, respectively (p = 0.012 for OS and p = 0.038 for PFS) (Figs. 1 and 2). The multivariate analysis showed that patients with ΔNLR ≥ 0 and receiving anti-PD-1 inhibitor in the second or further lines were significantly associated with worse PFS and OS (Table 2). Notably, increased NLR levels after treatment were independently associated with increased risk of death (HR = 1.615, 95% CI: 1.103–2.365, p = 0.014) and disease progression (HR = 1.466, 95% CI: 1.016–2.116, p = 0.041) (Table 3). Meanwhile, the ORR (41.8% vs. 24.3%) was higher in the decreased post-treatment NLR (ΔNLR < 0) group compared with the increased post-treatment NLR (ΔNLR ≥ 0) group (p = 0.045). On the other hand, there were no statistically difference regarding the DCR between the two groups (89.6% vs. 77.1%, p = 0.068) (Table 4).

Fig. 1
figure 1

Overall survival (OS) of patients stratifed by change in NLR

Fig. 2
figure 2

Progression free survival (PFS) of patients stratifed by change in NLR

Table 2 Multivariate analysis of the associations of NLR change and survival
Table 3 Multivariate analysis of the associations of combined NLR and survival
Table 4 Tumor response between different NLR groups

3.3 Baseline NLR, ΔNLR, and survival

To further evaluate the prognostic value of the combination of baseline NLR and ΔNLR, patients were stratified based on baseline NLR and ΔNLR into three groups: 34 patients with NLR < 3.23 and ΔNLR < 0 (Group 1); 80 patients with either NLR ≥ 3.23 and ΔNLR < 0, or NLR < 3.23 and ΔNLR ≥ 0 (group 2); 23 patients with NLR ≥ 3.23 and ΔNLR ≥ 0 (group 3).

The median OS was 14.4 (95% CI: 13.6–21.9) versus 9.2 (95% CI: 7.8–13.7) versus 4.6 (95% CI: 4.3–6.3) months for patients in groups 1, 2, and 3, respectively (Fig. 3). Moreover, the median PFS was 10.4 (95% CI: 10.1–11.6) versus 5.1 (95% CI: 3.1–7.3) versus 2.2 (95% CI: 1.7–3.2) months (both p < 0.001), respectively (Fig. 4). The multivariate analysis showed that patients with elevated NLR and ΔNLR (group 3) had a significantly increased risk of death (HR: 2.349, 95% CI: 1.701 – 3.243, p < 0.001) and disease progression (HR: 2.297, 95% CI: 1.666 – 3.167, p < 0.001) compared to patients with NLR < 3.23 and ΔNLR < 0 (group 1) (Table 5). Patients with either NLR ≥ 3.23 and ΔNLR < 0 or NLR < 3.23 and ΔNLR ≥ 0 (group 2) showed an intermediate risk for both death (HR: 2.626, 95% CI: 1.660 – 4.155, p < 0.001) and disease progression (HR: 2.554, 95% CI: 1.590 – 4.101, p < 0.001) (Table 5). Detailed information of the multivariate analysis is provided in Table 3. The ORR for groups 1, 2, and 3 was 50.0, 28.8, and 21.7%, respectively (p = 0.040) and the DCR was 91.2, 85.0, and 65.2%, respectively (p = 0.029) (Table 4).

Fig. 3
figure 3

Overall survival (OS)of patients stratifed by both NLR and ΔNLR

Fig. 4
figure 4

Progression free survival (PFS) of patients stratifed by both NLR and ΔNLR

Table 5 Multivariate analysis of the associations of combined NLR with survival

4 Discussion

Previous studies have demonstrated that changes in the NLR during nivolumab monotherapy were associated with GC survival, as well as combined with PD-1 inhibitor [11]. Here, we reported the prognostic value of baseline NLR and early NLR variations and its utility in MGC patients receiving anti-PD1 with different treatment combinations.

The relationship between high baseline NLR values and poor survival outcomes has been recognized for some tumors treated with checkmate inhibitors [12,13,14,15,16]. As reported in our center, the baseline NLR is associated with the survival of MGC patients using the anti-PD-1 drugs. The optimal cut-off of 3.23 was derived from our previous study [10]. Based on that cut-off value, we not only explored the NLR changes (ΔNLR) after two weeks of treatment but also the combined value of baseline NLR and ΔNLR.

We showed that the NLR increased after two cycles of anti-PD-1 treatment and was associated with worse outcomes. The NLR variations could be used to predict the OS and the PFS. High baseline NLR and increased post-treatment NLR were also independent factors of worsened survival outcomes. Furthermore, the combination of low baseline NLR and decreased post-treatment NLR was associated with higher ORR and DCR and especially favorable prognoses. Compared to current prognostic factors, such as the expression of PD-L1 [17] and TMB [18], routine lab tests are convenient and easy to monitor. Meanwhile, although PD-L1 or TMB are optimal candidate indicators as demonstrated by different clinical studies, they also presented controversial outcomes in some types of cancer [3, 5, 19, 20]. The inexpensive and readily available nature of the NLR allows it to be conveniently monitored over time. Additionally, in the present study, combining baseline NLR and NLR variants was better to stratify patients who might benefit from immunotherapy, consistent with the results reported for other types of cancer [11, 21,22,23]. Since inflammations are associated with tumor growth and progression, many inflammatory system markers are used as surrogates of systemic inflammatory response and as prognostic indicators [24,25,26,27]. However, the mechanisms by which NLR relates to ICI activity and OS remain unknown. An earlier report suggested associations between the NLR and the levels of pro-inflammatory cytokines and interleukin (IL)-6 that induce cancer progression, and GC patients with high levels of these cytokines exhibited poor prognoses [28,29,30].

Our current study also has some limitations. First, this was a retrospective single-center work and the sample size was small. Hence, external validation in a prospective study is required. Second, although we applied the cut-off value confirmed by our previous study [10], the optimal cut-off values for NLR remain unknown. Third, neither the expression of PD-L1 nor tumor mutation burden was analyzed in our clinical routine. So it is hard to provide these information of enrolled patients in the manuscript and compare your findings (ΔNLR) to these biomarkers in terms of sensitivity, specificity and outcome predictive capacity. Fourth, the mechanisms of action are unknown, and both neutrophil and lymphocyte number are sensitive to systemic nutrition, immune and infection status, further studies evaluating the relationship between NLR and tumor micro-environment, nutrition status, line of therapy, and immune-related adverse events might help delineate the mechanisms between baseline NLR and NLR changes and OS.

5 Conclusion

In summary, we showed that the changes in NLR in the early stages of different immunotherapies for MGC patients can be used to predict outcomes, especially combined with the baseline NLR. Finally, our current results might help clinicians decide whether to continue immunotherapy in MGC patients.