Introduction

Advanced cancer poses a significant global health threat, with its incidence on rise1. Immunotherapy has revolutionized the landscape of cancer treatment2,3, offering unprecedented survival advantages4. Despite its efficacy, immunotherapy is effective only for select patients, with many individuals failing to respond5. Tumor mutational burden, programmed death ligand-1, tumor-infiltrating lymphocytes, and microsatellite instability are gaining importance in the selection of treatment options6. However, these biomarkers have some limitations, such as being unsuitable for dynamic monitoring, high cost, and inconsistent viability7,8. Thus, the urgent need arises for identifying biomarkers that are readily available at a low cost for routine clinical use without specialized genomic technologies. The neutrophil-to-lymphocyte ratio (NLR), defined as the absolute counts of neutrophils and lymphocytes, has been reported to be a predictive biomarker of mortality in all conditions, including cancers9,10,11,12. NLR has been an emerging marker of the association between immune system and diseases13. Considering the interplay among systemic inflammation, immune system, and immunotherapy, NLR may be an attractive biomarker for predicting the efficacy of immunotherapy in cancer patients. Despite some studies14,15,16,17 and meta-analyses18,19,20 have examined the clinical application of NLR for immunotherapy efficacy in patients with advanced cancer, the results are inconsistent. With the development of immunotherapy, and updated clinical data, the performance of the NLR value in prognosticating immunotherapy efficacy deserves further exploration. To obtain objective and comprehensive results, we performed meta-analysis to evaluate the prognostic significance of NLR in patients with advanced cancer receiving immunotherapy. These results may be helpful in guiding the option of immunotherapy in advanced cancer patients.

Materials and methods

Search strategy

Based on the defined search strategy, appropriate literatures were searched from Embase, PubMed, Web of Science, and Cochrane Library databases with the keywords of “neutrophil-lymphocyte ratio,” “NLR,” “immunotherapy,” “immunological therapy,” “immune checkpoint inhibitors,” “neoplasm,” “advanced cancer,” and “tumor.” Keywords were combined with “OR” in the same category and with “AND” in different categories. Both subject headings and free-text terms were used. The retrieval strategy was adjusted based on the database characteristics. Supplementary Table 1 presents the retrieval steps for PubMed. Articles published until November 29, 2024 and limited to the English language were included. Additionally, eligible studies were identified by screening the references of both reviews and included articles.

Literature screening

The inclusion criteria were as follows: (1) Patients with histologically or pathologically confirmed or medical records recorded advanced cancer and received immunotherapy (immune checkpoint inhibitors, ICIs); (2) The study investigated the association of pretreatment NLR levels with overall survival (OS), progression-free survival (PFS), objective response rate (ORR), and disease control rate (DCR) and efficacy evaluation was on the basis of the Response Evaluation Criteria in Solid Tumors criteria; (3) The research type was retrospective or prospective cohort study; (4) The hazard ratio (HR) or odds ratio (OR) (95% confidence interval (CI)) of univariate or multivariable adjustment was provided or could be converted according to the frequency and sample size. The exclusion criteria were as follows: (1) non-literary research, such as meeting abstracts, reviews, and comments; (2) studies in which patients had received immunotherapy prior to the study; (3) NLR during and after treatment; and (4) for multiple studies with the same data or repeated publications, the study with the most comprehensive data.

Data extraction and quality evaluation

Two investigators (Jialin Su and Yuning Li) independently screened the relevant literature. Data extraction was conducted independently after confirming the inclusion criteria. Information included publication year, first author, region, study type, basic features (sample size, sex, and age), ICI regimens, follow-up time, NLR threshold, and clinical outcomes. Following data extraction, discrepancies were resolved through discussion between the two investigators until a consensus was reached. The quality of the included studies was assessed using the Newcastle-Ottawa Scale (NOS), which evaluated selection, comparability, and exposure with eight scoring items and a maximum score of 921. A score of 7–9 was considered high quality, 4–6 as medium quality, and < 4 low quality.

Evidence quality evaluation

GRADE (Grading of Recommendations, Assessment, Development and Evaluation) approach is widely used for rating the quality of evidence in meta-analysis and review. In this study, the evidence quality of included studies was assessed based on GRADE system22 by using GRADE Profiler (GRADEpro) Guideline Development Tool (GDT) online tool.

Statistical analyses

To analyze the association between NLR and immunotherapy efficacy in patients with cancer before treatment, patients were divided into high and low NLR group according to the NLR cutoff value of 4 according to the previous description23,24. In addition, our data showed that the heterogeneity was minimal, when NLR was 4. OR and 95% CI served as effect size indices to evaluate whether there were statistically significant differences in ORR and DCR between high NLR vs. low NLR group. HR and 95% CI were used to analyze the association between NLR and survival risk.

Due to the methodological heterogeneity of these studies, a random-effects model was applied to pool the effect values of the meta-analysis. Heterogeneity was tested using the Cochran’s Q and I2 test25. P < 0.05 or I2 > 50% indicated significant heterogeneity among studies. If P > 0.05 and I2 ≤ 50%, no significant heterogeneity was considered. Subgroup analysis was performed according to variables, such as research region, cancer type, confounder correction, and NLR cutoff. A one-by-one exclusion test was used to assess whether a single study significantly affected the results26. Egger test27 and Begg’s test were used to test publication bias. If significant publication bias was present, the stability of the pooled results was analyzed using the trim-and-fill method28. The above statistical analyses were completed using Stata12.0 software.

Results

Literature search

Figure 1 illustrates the literature search process and the PRISMA checklist is listed in Supplementary File 1. A total of 5711 articles were searched (1397 from PubMed, 2497 from Embase, 154 from The Cochrane Library, and 1663 from the Web of Science). After eliminating 2337 duplicate articles, 3374 remained. After browsing the titles and abstracts, 3165 studies that did not meet the inclusion criteria were excluded. Finally, 80 of the 209 studies were eliminated after full-text reading (Supplementary File 2), and finally 129 studies14,24,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154,155,156,157,158 were included in this analysis.

Fig. 1
figure 1

Selection process for literature included in the meta-analyses.

Full size image

Research characteristics and quality evaluation

A total of 129 studies published between 2016 and 2024, with 18,780 cases, were included (Table 1). Most participants were patients with advanced cancers, including renal cell carcinoma, non-small cell lung cancer, urothelial carcinoma, head and neck cancer, and liver cancer. The median age of the participants ranged 50–75 years and the median follow-up time ranged 2-48.6 months. The ICI types and grouping thresholds of the NLR are shown in Table 1. The results of the quality evaluation are presented in Supplementary Table 2. The NOS scores were 5–8. Methodological quality was moderate. The main types of biases in the included studies were recall and confounding factors.

Table 1 Characteristics of the included studies.
Full size table

Meta-analysis

The forest plot of OS risk in high NLR vs. low NLR before treatment is shown in Fig. 2A. A total of 107 articles were included, and the heterogeneity test results showed I2 = 87.5% (P < 0.001), indicating significant statistical heterogeneity. The pooled result was HR (95% CI) = 2.26(2.02, 2.53) (P < 0.001), indicating that a high level of pretreatment NLR was related with significantly reduced OS. The forest plot of risk of PFS between high NLR and low NLR groups is shown in Fig. 2B, and the heterogeneity test result of the included studies was I2 = 57.8% (P < 0.001). The pooled result was HR (95% CI) = 1.83 (1.69, 1.98), (P < 0.001), indicating that a high pretreatment NLR level was related to a significantly reduced PFS. Thus, an increased pretreatment NLR level might be served as a biomarker for poor prognosis in patients with advanced cancer receiving immunotherapy.

Fig. 2
figure 2

The forest plot of risk of (A) OS and (B) PFS in high NLR vs. low NLR before immunotherapy.

Full size image

A comparison of the ORR between high and low NLR groups before treatment is shown in Fig. 3A. A total of 38 articles were included, and the heterogeneity test results showed I2 = 0% (P = 0.657), indicating no significant heterogeneity among the studies. Pooled results were OR (95%CI) = 0.54 (0.47, 0.62) (P < 0.001), indicating an obvious correlation between the NLR and ORR. Patients with a high NLR before immunotherapy showed a significantly decreased ORR. A forest plot of the DCR is presented in Fig. 3B, and the heterogeneity test result of the included studies was I2 = 45.2% (P = 0.001). The pooled result was OR (95% CI) = 0.36 (0.29, 0.43) (P < 0.001), indicating that a high pretreatment NLR was related to a significantly decreased DCR. Thus, an increased pretreatment NLR may predict decreased immunotherapy efficacy.

Fig. 3
figure 3

The forest plot of (A) ORR and (B) DCR in high NLR vs. low NLR before immunotherapy.

Full size image

Subgroup analysis

The subgroup analysis results for OS, PFS, ORR, and DCR are shown in Table 2. For OS, the differences between subgroups were not statistically significant after grouping by region, study type, cancer type, multivariate adjustment and NLR cutoff (two studies40,141 did not report the cutoff and were not included in the subgroup analysis) (P > 0.05). Except for metastatic MCC (Merkel cell carcinoma), the pooled results of the other subgroups were in accordance with the original pooled results (P < 0.05). However, none of the above grouping factors were the sources of significant heterogeneity (Figure S1A-1E). For PFS, no significant differences were observed among the subgroups after grouping by region, study type, and multivariate adjustment (P > 0.05); however, subgroup analysis of the cancer type and NLR threshold showed statistically significant differences (P < 0.05). Pooled results for PCS (prospective cohort study), aCRC (advanced colorectal cancer), and metastatic MCC were not significant (P > 0.05), whereas the pooled results for the other subgroups were significant (P < 0.05). Similarly, no significant source of heterogeneity was identified in the subgroup analysis (Figure S2A-2E).

Table 2 Results of subgroup analyses.
Full size table

For ORR, although significant results were found in both Asian and non-Asian subgroups (P < 0.001), significant difference was found in the pooled results with regard to region (P = 0.005). There was no significant difference between subgroups based on the study type, cancer type, or NLR cutoff (P > 0.05). In addition, the pooled results of the PCS, aMelanoma (advanced melanoma), aGC (advanced gastrointestinal/gastric Cancer), aUC (advanced urothelial cancer) and UGI (upper gastrointestinal cancer) subgroups were not significant (P > 0.05), and the pooled result of the other subgroups were significant (P < 0.05) (Figure S3A-3D). For DCR, there were no significant differences among the subgroups by region, study type, cancer type, and NLR cutoff (P > 0.05). The pooled results of the PCS, UGI cancer, and aCRC subgroups were not significant (P > 0.05), whereas the pooled result of other subgroups were significant (P < 0.05). None of the above grouping factors was a significant source of heterogeneity (Figure S4A-4D). Thus, an NLR cutoff value of 4 was related to PFS and OS.

Sensitivity analysis and publication bias test

As shown in Table 3, the pooled results for the four indicators showed good stability. Excluding any one study, the pooled results of the other studies were significant (P < 0.05). The Egger results are listed in Table 3 and indicate a significant publication bias in the included studies for OS, PFS, ORR, and DCR. The trim-and-fill method revealed that after including 18 virtual studies on OS (Figure S5A), the pooled result was HR (95% CI) = 2.01 (1.81, 2.24) (P < 0.001), indicating that the original pooled result was stable. For PFS, after including 11 virtual studies (Figure S5B), the pooled results were changed to HR (95% CI) = 1.71 (1.57, 1.86) (P < 0.001), indicating the stability of the original pooled result. For ORR and DCR, no virtual negative results were used to enhance the symmetry of the funnel plot, and the results did not change, suggesting that publication bias in ORR and DCR may be caused by small sample bias. Begg’s test achieved the consistent results (Table 3).

Table 3 Outcomes of the sensitivity analysis and test of publication bias.
Full size table

Certainty of evidence

The included studies were observational studies with a low or moderate risk of bias. At the same time, the degree of heterogeneity of studies was high. The GRADE of the evidence was categorized as low quality. The certainty of evidence was low for ORR and was very low for OS, PFS, and DCR (Supplementary Table 3).

Discussion

Our study collected available evidence from 120 articles with 17,969 cases of advanced cancer. The results revealed that the pretreatment NLR was significantly related to OS, PFS, ORR, and DCR. Subgroup analyses by research region, cancer type, study type, and confounder correction remained unchanged. To the best of our knowledge, this is the latest, most comprehensive, and largest meta-analysis on the relationship between the NLR and the efficacy and prognosis of immunotherapy.

Hematological parameters are the most common and easily available for routine clinic monitoring159. The NLR reflects host inflammatory processes, and the relationship between inflammation and human cancer has been extensively explored160. It was gradually found that inflammation is implicated in cancer initiation, invasion, and metastasis161. Inflammation affects host immune response and plays an important role in immunotherapy162. The function of neutrophils in tumor microenvironment is controversial because they contribute to tumor growth, angiogenesis, immune tolerance, and metastasis163. Tumor-associated neutrophils can decrease CD4+/CD8 + T cell and suppress the generation of IFN-γ and TNF-α, leading to an immunosuppressive environment164. Lymphocytes are involved in the antitumor immune response; therefore, elevated lymphocyte infiltration is related to a good prognosis in patients with cancer165. Neutrophils reflect a response to systemic inflammation, whereas decreased lymphocytes represent cell-mediated immune impairments166. The present meta-analysis confirmed that a high NLR before immunotherapy was significantly associated with decreased clinical efficacy and poor prognosis. The cutoff value of the NLR has been reported to be between 2 and 5167. However, the optimal cutoff value remains undetermined. In our study, a cutoff value of 4 was related to prognosis and might serve as a prognostic biomarker. Further large-scale studies are required to validate the clinical value.

Several published meta-analyses have analyzed the prognostic value of pretreatment NLR in patients with advanced cancer168,169,170,171. Templeton18 included 100 articles with 40,559 patients for a meta-analysis, and the pooled results showed that a high NLR was related to poor OS in many solid cancers. Mei et al.20 analyzed 66 studies involving 24,536 patients. The pooled results revealed that an elevated pretreatment NLR correlated with poor clinical outcomes in advanced cancers. However, only these two studies analyzed the prognostic role of the pretreatment NLR in advanced cancers, and it remains unclear whether the NLR has a value in prognosticating immunotherapy efficacy in these patients. A recent study by Jiang19 analyzed the predictive and prognostic value of pretreatment NLR for immunotherapy in patients with advanced cancers. The pooled results revealed that an elevated NLR was significantly associated with poor OS and PFS. However, only 11 studies were included in their meta-analysis. In our study, we analyzed the correlation of pretreatment NLR with not only survival risk but also the efficacy of immunotherapy in patients with advanced cancer. The pooled results indicated that a high NLR before immunotherapy was significantly correlated with poor prognosis and decreased clinical efficacy. Importantly, our study collected available evidence from 120 articles with 17,969 patients, making it the largest and most comprehensive meta-analysis in this field. Overall, combined with previous meta-analyses, more clinical studies are necessary to validate the critical role of the pretreatment NLR in patients with advanced cancer receiving immunotherapy. It is important to determine prognosis and predict biomarkers for immunotherapy efficacy. An elevated pretreatment NLR may be related to clinical outcomes. As a result, it can provide a basis for clinicians to make reasonable medical decisions. This meta-analysis proposes NLR as a new prognostic indicator for patients receiving immunotherapy. The strengths of this meta-analysis include the following: (1) Despite significant heterogeneity in the included studies, the strength of the pooled effect size was large, and NLR was significantly associated with ORR, DCR, OS, and PFS. (2) The included studies were of medium or high quality. Methodological quality evaluation revealed that although the included studies had a certain degree of recall bias and confounding bias, the control for measurement, selection, and withdrawal biases was reasonable. (3) The results of statistical analyses, such as the publication bias test, one-by-one exclusion method, and trim-and-fill method, suggest that the pooled results were reliable and stable.

This study had some limitations. First, the heterogeneity was statistically significant, and no significant source of heterogeneity was identified through subgroup analysis. Heterogeneity may result from clinical and methodological heterogeneity among the studies. Unfortunately, the descriptions of this information in the included studies were not comprehensive or uniform, and their impact on heterogeneity could not be explored quantitatively. Second, the results of the meta-analysis were mostly based on univariate analyses. Although some studies reported no significant differences in age, sex, and earlier treatment regimens, the possibility of incorrect estimation, which exaggerated the association among NLR, ORR, and DCR, could not be ruled out. Additionally, for certain cancer types, there were fewer included studies, and more studies with larger sample sizes are necessary to verify the results. The grouping thresholds of the NLR were inconsistent; therefore, it is recommended to explore and formulate a unified grouping threshold in future studies to facilitate better extrapolation of the research results. Finally, we did not register the protocol for this study. The protocol register may minimize the reporting bias and reduce unplanned duplication172. Thus, we will pay attention to the protocol register in the following review and meta-analysis.

In summary, this meta-analysis indicated that an elevated NLR before immunotherapy was significantly associated with the prognosis of patients with advanced cancer. Pretreatment NLR may be served as a promising marker for immunotherapy efficacy.