Prognostic impact of neutrophil-to-lymphocyte ratio in gliomas: a systematic review and meta-analysis

The electronic databases, PubMed, Cochrane Library, and Embase, were searched from the time of their conception until May 2019. The databases were searched using the following words: (‘glioblastoma’ OR ‘glioma’) AND (‘neutrophil to lymphocyte ratio’ OR ‘neutrophil lymphocyte ratio’ OR ‘neutrophil-to-lymphocyte ratio’ OR ‘NLR’) AND (‘survival’ OR ‘mortality’ OR ‘outcome’ OR ‘prognostic’ OR ‘prognosis’). We manually screened the references of the related articles to expand the search range.

The inclusion criteria in our meta-analysis were as follows: (1) patients pathologically confirmed with gliomas, (2) the prognostic significance of peripheral blood NLR for gliomas was assessed, (3) cutoff of NLR was provided, and (4) hazard ratios (HRs) and 95% confidence intervals (CIs) for NLR on overall survival (OS) were available. The following studies were excluded: (1) case reports, letters, conference abstracts, non-clinical studies, and reviews without available data; (2) studies with insufficient information to evaluate HRs and 95% CIs; and (3) duplicated publications.

Two investigators independently selected the studies that fulfilled our inclusion criteria and extracted the relevant information. The related information was extracted as follows: first author’s surname, country, sample size, age of the study population, publication year, histology, duration, treatment, cutoff value of NLR, sampling time, and HR and 95% CI for OS. Any disagreement was resolved through discussion.

Two reviewers used Newcastle–Ottawa quality assessment scale (NOS) [15] to evaluate the quality of studies. Using the NOS, the studies are evaluated on three ways, namely comparability, selection, and outcome confirmation. Each parameter also has subitems. The maximum score is nine stars, and NOS scores ≥ 5 is considered of high quality [15].

The collected data from the included studies were combined using Review Manager 5.3 (The Cochrane Collaboration, Copenhagen, Denmark). Forest plots were constructed to assess the predictive role of NLR in gliomas. HRs and 95% CIs for OS were synthesized with a random effect model. A random effect model or fixed effect model was employed depending on the heterogeneity of the studies [16]. The heterogeneity was evaluated with the I² statistic. The data were synthesized using a fixed effect model with I² < 25%. In case of I² > 25%, a random effect model was used for data synthesis. The sources of heterogeneity were evaluated by subgroup analysis. Sensitivity analysis was used to appraise the stability of the outcome. Funnel plots were constructed to evaluate publication bias. Statistical difference was defined as P value < .05.

A flow diagram based on the PRISMA statement (Additional file 1) summarizing the process of study retrieval is illustrated in Fig. 1. A total of 16 articles published between 2013 and 2019 were incorporated in our study [17‐32]. The data of 2275 patients in whom the prognostic significance of NLR was assessed were included. The demographic data of the patients in the included trials is shown in Table 1. The NOS scoring details are presented in Additional file 2. There were 2 studies from USA, 8 from China, 1 from Canada, 1 from Russia, 2 from Turkey, 1 from Singapore, and 1 from Portugal. All trials were retrospective ones. The cutoff values ranged from 2.5 to 7.5 in the included trials, with an average value of 4.03. Eleven studies used NLR from the preoperative blood sample, whereas 2 used NLR from the postoperative blood sample. Thirteen of the 15 trials applied multivariate analysis. The NOS scores ranged from 5 to 7. The average number of NOS scores was 5.375.

A high preoperative NLR was connected with unfavorable OS (HR: 1.43, 95% CI: 1.27–1.62, P < 0.00001; Fig. 2) in patients with gliomas. The heterogeneity analysis among the studies showed an I² value of 83% (P < 0.00001), which indicated obvious heterogeneity. A subgroup analysis was conducted on the basis of the latent confounding factors, such as histology, cutoff value of NLR, analysis method, ethnicity, NOS score, and sampling time. On stratification by ethnicity in the subgroup analysis, a low NLR predicted a positive prognosis in the Asian (HR: 1.64, 95% CI: 1.28–2.10), but not in the Caucasian (HR: 1.26, 95% CI: 0.92–1.72). Stratification by histology revealed that a low NLR predicted longer OS in trials with patients with gliomas of various grades (HR: 1.68, 95% CI: 1.40–2.01) and in those patients with GBM (HR: 1.29, 95% CI: 1.13–1.47). Furthermore, the subgroup analysis according to the cutoff value of NLR indicated that a high NLR was connected with negative OS in patients with gliomas in trials with cutoff value of NLR = 4 (HR: 1.55, 95% CI: 1.22–1.97) and in those in trials with cutoff values of NLR ≠ 4 (HR: 1.52, 95% CI: 1.06–2.19). Results of the subgroup analysis on the basis of the NOS score suggested that high NLR was connected with poor OS when the NOS score was ≤ 5 (HR: 1.46, 95% CI: 1.17–1.83) and NOS score was > 5 (HR: 1.72, 95% CI: 1.09–2.72). Results of the subgroup analysis on the basis of the analysis method showed that a low NLR represented good prognostic significance in both the univariate analysis (HR: 1.69, 95% CI: 1.06–2.67) and multivariate analysis (HR: 1.31, 95% CI: 1.16–1.48). Finally, analysis on the subgroup of sampling time indicated that an elevated NLR was connected with negative OS in gliomas with preoperative blood sampling (HR: 1.29, 95% CI: 1.14–1.45), but not in those with postoperative blood sampling (HR: 1.36, 95% CI: 0.69–2.70) (Table 2).

To appraise the impact of each research on the overall outcome (HR) of OS, a sensitivity analysis was conducted. As for the HR on overall survival, we removed each study individually and the HR value or degree of significance did not substantially change.

The shape of the funnel plots showed asymmetry and indicated significant publication bias in OS (Fig. 3).