Prognostic value of neutrophil-lymphocyte ratio in gastroenteropancreatic neuroendocrine neoplasm: a systematic review and meta-analysis

Literature search

We adopted the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) statements (Moher et al., 2009), registered at the International Prospective Register of Systematic Reviews (registration number CRD42020187679). This registration record has undergone automated eligibility checks and is published exactly as submitted. PubMed, Embase, and Web of Science electronic databases were searched for publications until July 2024. Combined text and MeSH terms were used for searching and the detailed search strategies are described below:

Embase: (‘neuroendocrine tumor’/exp OR ‘neuroendocrine tumors’/exp OR ‘tumor, neuroendocrine’ OR ‘tumors, neuroendocrine’) AND (‘neutrophil to lymphocyte ratio’/exp OR ‘neutrophil-to-lymphocyte ratio’/exp OR ‘neutrophil/lymphocyte ratio’/exp OR ‘NLR’). PubMed: ((Neuroendocrine Tumors) OR (Neuroendocrine Tumor) OR (Tumor, Neuroendocrine) OR (Tumors, Neuroendocrine)) AND (“neutrophil lymphocyte ratio” or “neutrophil to lymphocyte ratio” or “neutrophil-to-lymphocyte ratio” or “neutrophil/lymphocyte ratio” or “NLR”). Web of Science: ((AB= (Neuroendocrine Tumors) OR AB= (Neuroendocrine Tumor) OR AB= (Tumor, Neuroendocrine) OR AB= (Tumors, Neuroendocrine)) AND ((AB= (neutrophil-lymphocyte ratio)) OR AB=(NLR)) OR AB= (neutrophil-to-lymphocyte ratio).

We include all possible eligible studies and consider them for review before screening, regardless of their primary outcome. We collected articles that were only written in English. Ethical approval was unnecessary because the meta-analysis is based on published clinical trial data.

Inclusion and exclusion criteria

Two investigators (YW and BW) independently assessed all the potentially qualified articles at different times. If there are different views on inclusion or exclusion, we resolve them through discussion or consultation with other authors (YZ and KD). Titles and abstracts were reviewed first to determine whether studies were related to the theme. Then, full articles were judged according to the inclusion and exclusion criteria. If studies satisfied the inclusion criteria, they were used for detailed analysis and data extraction.

The inclusion criteria for qualified studies were as follows: (1) Patients included in the study were pathologically confirmed GEP-NEN in any stages and classification; (2) Assessment of OS, RFS, and PFS as outcomes; (3) The text reported the data of adjusted hazard ratio (HR) with 95% confidence intervals (CI), or any other type of survival information that can be calculated into HR with 95% CI; (4) The text reported an exact cut-off value of the NLR; (5) Articles published in English.

The exclusion principles were as shown below: (1) Lack of distinction of GEP-NENs from other NENs; (2) Duplicated studies or overlapping cohorts from the same centers; (3) Conference abstracts, reviews, case reports, mete-analysis, letters, animal studies, or laboratory studies; (4) Lack of necessary data.

Data extraction and quality assessment

We extracted the following related information from each included study: name of the first author, published year of the article, study area, maximum tumor size, the period of the research, gender and sex ratio, age, study design type, intervention methods, the cut-off value of the NLR, HR and 95% CI of overall survival (OS), recurrence-free survival (RFS), and progression-free survival (PFS). In the meta-analysis, we preferred the adjusted HR values to maintain data consistency.

The OS time was figured out from the date of treatment initiation to the date of death. If the patient were still alive at the last follow-up, the endpoint is the date of the last follow-up. The RFS time was calculated as the number of months from the date of treatment to the date of confirmation of disease recurrence or the date the endpoint was realized. The PFS is the time from treatment to the first occurrence of disease progression or death from any cause. All the time is measured in a month for the unit. The NLR was calculated based on pretreatment laboratory data using the white blood cell differential counts by dividing the neutrophil count by the lymphocyte count. OS, RFS, and PFS outcomes were expressed as hazard ratio (HR) (and 95% CI) for patients with high NLR versus patients with low NLR.

We performed the quality assessment for the included studies according to the Newcastle–Ottawa scale (NOS). The full score is nine points, and studies with over five points were regarded as high-quality studies.

Statistical analysis

We used the software of Review Manager (RevMan 5.3) for all statistical analyses. The pooled HR with its 95% CI was utilized to quantitatively assess the prognostic function of the NLR for GEP-NEN patients with the method described by Parmar (Parmar, Torri & Stewart, 1998). Cochrane Q and I² tests were used to evaluate the heterogeneity among the included studies. P-value < 0.10 for the Q test or I² > 50% indicates apparent heterogeneity, and we will choose the random-effects model. Otherwise, if P-value ≥ 0.10 for the Q test or I²  ≤  50%, the fixed-effects model will be taken. We performed a subgroup analysis based on the geographical location of the studies, and tumor site if necessary. The stability of the results was confirmed by sensitivity analysis. p < 0.01 was considered statistically significant.

Characteristics of the included studies

The initial search contained 530 studies. After the removal of duplicates, 136 studies were excluded. Of the remaining 394 studies, 288 were eliminated by reading results (titles and abstracts) for apparent irrelevance. 46 full-text articles were downloaded to assess their eligibility, of which 33 were excluded because they were not English language (n = 6), ineffective NLR or OS data collected (n = 17), without adjusted HR data (n = 5), review (n = 4), and not GEP-NEN (n = 1). Ultimately, 13 studies (Abdelmalak et al., 2021; Arima et al., 2017; Cao et al., 2017; Gaitanidis et al., 2018; Harimoto et al., 2019; Luo et al., 2017; Miura et al., 2021; Panni et al., 2019; Pozza et al., 2019; Yang et al., 2024; Yucel et al., 2013; Zhang et al., 2019; Zhou et al., 2017) published between 2013 and 2024 were included in this meta-analysis. The sample sizes ranged from 48 to 620. The total sample size of our meta-analysis was 2,040 (Fig. 1).

The characteristics of the included studies are summarized in Table 1; Among them, 12 studies (Abdelmalak et al., 2021; Arima et al., 2017; Cao et al., 2017; Harimoto et al., 2019; Luo et al., 2017; Miura et al., 2021; Panni et al., 2019; Pozza et al., 2019; Yang et al., 2024; Yucel et al., 2013; Zhang et al., 2019; Zhou et al., 2017; Arima et al., 2017; Harimoto et al., 2019; Zhang et al., 2019) are retrospective cohort studies, one (Gaitanidis et al., 2018) is a prospective study. In terms of the research area, five studies (Luo et al., 2017; Yang et al., 2024; Zhou et al., 2017; Zhang et al., 2019; Cao et al., 2017) were conducted in China, two (Gaitanidis et al., 2018; Panni et al., 2019) in the USA, three (Arima et al., 2017; Harimoto et al., 2019; Miura et al., 2021) in Japan, one (Abdelmalak et al., 2021) in the UK, one (Pozza et al., 2019) in Italy, and one (Yucel et al., 2013) in Turkey. Regarding tumor site, eight studies concentrated on pancreatic neuroendocrine neoplasm (P-NEN), one study enrolled patients with gastric neuroendocrine neoplasm (G-NEN), three studies one study enrolled patients with gastrointestinal and pancreatic neuroendocrine neoplasms (GEP–NEN), one study only included enteric neuroendocrine neoplasms (E-NEN). A total of 13 studies selected 1.79, 1.9, 2.2, 2.28, 2.3, 2.4, 2.4, 2.4, 2.62, 2.63, 3.41, 5 and 5 as the cut-off value of the NLR, respectively.

Relationship between NLR with OS, RFS, and PFS

Nine studies embraced HR and 95% CI for OS. As shown in Fig. 2, a higher NLR was significantly associated with worse OS in the overall population with a pooled HR of 2.75 (95% CI [2.16–3.85], p < 0.00001). The heterogeneity analysis among the studies showed an I² value of 0% (P = 0.58), which indicated a low heterogeneity. Eight studies included risk ratio (HR) and 95% CI for RFS. Overall, higher NLR results in poorer RFS regardless of the effect model. In the random effects model, the pooled HR was 3.70 (95% CI [2.13–6.43], p < 0.00001) with high heterogeneity (I² value of 67%, P = 0.006). Only one study reported HR and 95% CI data for PFS.

Subgroup analysis of the relationship between NLR and OS, RFS and PFS based on race

Since this meta-analysis was mainly based on Asian studies, we conducted OS subgroup analysis between Asians and Caucasians (Fig. 3). The subgroup analyses showed that high NLR was associated with a poor OS for patients in Asians (HR = 2.82, 95% CI [1.92–4.13], p < 0.00001) with a low heterogeneity (I² value of 0%, P = 0.72) and Caucasians (HR = 2.54, 95% CI [1.25–5.16], p = 0.01) with a low heterogeneity (I² value of 40%, P = 0.19).

The subgroup analysis demonstrated that an elevated NLR indicated a poor RFS in Asians (HR = 3.13, 95% CI [1.45–6.78], p = 0.004) with heterogeneity (I² value of 59%, P = 0.06) and Caucasians (HR=4.47, 95% CI [2.24–8.90], p < 0.00001) with heterogeneity (I² value of 55%, P = 0.11).

Subgroup analysis of the relationship between NLR and OS, RFS and PFS in p-NEN and g-NEN

Since gastric neuroendocrine tumors (g-NEN) and pancreatic neuroendocrine tumors (p-NEN) accounted for the majority of the included literature, we conducted a subgroup analysis specifically for the prognostic analysis of NLR with g-NEN and p-NEN. The subgroup analysis pooled results showed that higher NLR was associated with worse OS (HR = 3.98, 95% CI [1.92–8.25], p = 0.0002), and RFS (HR = 2.27, 95% CI [1.63–3.16], p < 0.00001) of p-NEN (Figs. 4A–4B). The same results were found for NLR’s prediction of g-NEN’s OS (HR = 2.63, 95% CI [1.72–4.01], p < 0.00001) and RFS (HR=4.81, 95% CI [1.83–12.68], p = 0.001) (Figs. 4D–4E).

Subgroup analysis of the relationship between NLR and OS, RFS and PFS in the surgery group and non-surgery group

We also examine the effect of NLR in surgical and non-surgical procedures. The subgroup analysis pooled results showed that higher NLR was associated with worse OS (HR = 2.93, 95% CI [2.01–4.28], p < 0.00001), RFS (HR = 4.29, 95% CI [2.97–6.18], p < 0.00001) in patients under surgery procedure (Figs. 5A–5B).

Sensitivity analysis

We conducted a sensitivity analysis for the primary outcome of OS by individually excluding each study to assess the impact of each study on the overall results. We found that the removal of any single study did not significantly affect the final meta-analysis results or the heterogeneity results (see Fig. 6). Similarly, we removed each study individually to conduct a sensitivity analysis to assess each study’s effect on the overall primary result for RFS. We found an apparent descending heterogeneity (I² value of 39% with P = 0.15) when we removed Panni’s study, which might be the source of high heterogeneity (Fig. 7).

Study quality

According to the NOS score, all the retrospective studies were high quality, ranging from 6 to 8 (Table 2).

Publication bias

The shape of the funnel plots (Figs. 8A–8B) showed asymmetry and indicated significant publication bias in OS, RFS. We found that the funnel plot for publication bias in the OS outcome was symmetrical, indicating that there was essentially no publication bias. However, in the funnel plot for the RFS outcome, the study by Harimoto deviated from the symmetrical funnel, suggesting that this study had a significant publication bias.