Prognostic significance of neutrophil-to-lymphocyte ratio in cervical cancer: A systematic review and meta-analysis of observational studies

Background and aims The prognostic role of neutrophil-to-lymphocyte ratio (NLR) in cervical cancer are controversial. We conducted this meta-analysis to obtain a more accurate assessment of prognostic significance of NLR in cervical cancer. Results A total of 9 studies, consisting of 2,804 patients, were selected in this meta-analysis. Our pooled results showed that high pre-treatment NLR level was significantly associated with poorer overall survival (HR: 1.88, 95% CI 1.30–2.73) and shorter progression free survival (HR 1.65, 95% CI 1.18–2.29). Additionally, increased NLR was also significantly correlated with tumor size (OR 2.05, 95% CI 1.14–3.65), advanced FIGO stage (OR 2.12, 95% CI1.28–3.49) and lymph node involvement (OR 2.24, 95% CI 1.65–3.04). Materials and Methods We conducted a systematic literature search using the electronic databases PubMed, Web of Science, and Embase up to May 2016.Statistical analysis was performed using Stata 10.0. Conclusions Elevated pretreatment NLR could serve as a predicative factor of poor prognosis for cervical cancer patients.


INTRODUCTION
Cervical cancer is the second most common female malignancy, accounting for 500,000 new cases and over 200,000 deaths worldwide annually [1,2]. Although tumor size, lymph node status, International Federation of Gynecology and Obstetrics (FIGO) staging criteria as well as pretreatment hemoglobin level were reported to be independent prognostic factors for cervical cancer [3][4][5][6], these factors are not routinely used in clinical settings due to insufficient specificity and sensitivity among patients. To further increase predictive accuracy, more prognostic parameters are still warranted.
Neutrophil to lymphocyte ratio (NLR), is a marker for evaluating the systemic potential balance between neutrophil-dependent pro-tumor inflammation and lymphocyte-associated anti-tumor immune response [7][8]. A higher level of NLR could represent a trend towards increased pro-tumor inflammation and decreased antitumor immune capacity.

Selection and characteristics of included eligible studies
A flowchart for the selection of eligible studies is demonstrated in Figure 1. A total of 233 studies was retrieved and screened by title and abstract. 176 studies were excluded after the initial assessment of title and abstract.

Research Paper
Among the remaining 57 articles, 36 were further excluded because they were letters, comments, editorials, or reviews. The full texts of the remaining 21 articles were evaluated. A total of 12 full-text articles were excluded, including 7 without available data and 5 without NLR category.
The basic information of the selected studies is summarized in Table 1.

Association of pre-treatment NLR with overall survival
The association between NLR and OS was assessed in 7 studies consisting of 2,804 patients. The pooled estimate indicated a significantly shorter OS in cervical cancer patients with high NLR compared to those with low NLR (HR: 1.88, 95% CI 1.30-2.73) ( Figure 2). Significant heterogeneity was observed among these studies (I 2 = 72.8%, P < 0.001).

Association of pre-treatment NLR with progression free survival
The association between NLR and PFS was evaluated in 5 studies including 2,607 patients. Those with high pre-treatment NLR had a significantly poorer PFS than those with low NLR (HR 1.65, 95% CI 1.18-2.29) ( Figure 3). Significant heterogeneity was observed among these studies (I 2 = 85.4%, P < 0.001)

Subgroup analysis
Results of subgroup meta-analyses are summarized in Table 2.

Publication bias analysis
The funnel plots showed a low probability of publication bias (Supplementary Figure 1 and 2). Consistently, the Egger's and Begger's regression tests

Sensitivity analysis
We performed a sensitivity analysis for every analysis by sequential omission of the individual study. The pooled HRs for OS and PFS were not significantly changed, which suggested the robustness of the results ( Supplementary Figures 3 and 4).

DISCUSSION
Accumulating evidence demonstrated that inflammation exerted an essential role in cancer formation, development, and progression [26][27][28][29][30]. Neutrophils have been considered to be the primary source of circulating VEGF, which play a critical role in tumor-associated angiogenesis through producing many inflammatory cytokines such as tumor necrosis factor, interleukin 1, and providing a favorable micro-environment for tumor [28]. Conversely, lymphocytes exert a critical role in cancer- 111 1999-2010 20.6 (0.6-40.6) 2.5 IB2-IIB pretreatment retrospective 8 OS: overall survival; PFS: progression-free survival; DFS: disease-free survival;"-": not reported; Nonsurgery was defined as chemotherapy, radiotherapy or chemoradiotherapy. specific immune response [29]. It has been shown that an increased infiltration of lymphocytes into tumor tissue is associated with good prognosis [30]. In the present study, we found that an elevated NLR was associated with poorer OS and shorter PFS in cervical cancer patients, which was in accordance with the results from studies with several other cancer types.
Previous studies reported that higher pretreatment NLR had a stronger predictive effect in cancer at a more advanced stage [31,32]. Yodying H et al. [31] evaluated the prognostic role of NLR in esophageal cancer and indicated that NLR was associated with tumor invasion and lymph node metastasis. Meanwhile, Xue TC et al. [32] found that NLR was associated with vascular invasion in hepatic carcinoma. In the present study, our results showed that NLR was positively related to tumor size and significantly correlated with lymph node involvement as well as advanced tumor stage (FIGO). These observations indicated that inflammation severity might significantly affect intrinsic tumor characteristics in patients with cervical cancer.
Heterogeneity was observed in this meta-analysis. This heterogeneity may be partially caused by geographic area, statistical methods, sample size, NLR cut-off value, and the follow-up duration. In order to explore the source of heterogeneity in this meta-analysis, we performed a subgroup analysis. The subgroup analysis results demonstrated that the prognostic value of NLR was unaffected by the confounders mentioned above in the analysis. Moreover, the sensitivity analysis indicated that our results were relatively conclusive. There are several limitations that need to be addressed in this study. First, most of the studies selected in this meta-analysis were retrospective, observational studies, and no prospective cohort study was identified. Moreover, most studies only contained information of NLR predictive value in the same stages or in the same therapy. Therefore, they may be more susceptible to bias in data analysis and the predictive value of NLR in the same stages and the same therapy need to be explored in the future. Second, a previous systemic review demonstrated that increased NLR predicted poor PFS with prostate cancer only in Asians, but not in Caucasians, which could be attributed to the ethnicity heterogeneity [33]. In this meta-analysis, majority of included articles came from Asian countries. Therefore, our current conclusions may not be suitably applied to other populations. Third, only 9 studies were included and the cut-off value for defining high NLR in each individual study was different, ranging   1.9 to 5, which could hinder the application of this ratio in the clinical setting. In the future, original research is required to determine the most accurate cut-off value of NLR in cervical cancer patients.
Despite these limitations, our meta-analysis also has some strengths. To our knowledge, this meta-analysis is the first to evaluate the prognostic role of a pretreatment peripheral blood NLR in cervical cancer. Moreover, our results showed a significantly positive correlation between NLR and the clinical features of cervical cancer, such as tumor size, FIGO stage, and lymph node involvement. Thus, NLR could have a wider clinical application regarding the prognostic assessment of cervical cancer, may be useful in stratifying patients, and in determining individual treatment plans in the future.
In conclusion, our meta-analysis of currently available clinical evidence demonstrated that NLR could serve as a promising prognostic marker of cervical cancer,  because it is available from routine blood tests in daily clinical practice, which are convenient, low cost, and reproducible. In addition, considering that NLR level was associated with prognosis of cervical cancer, it would be interesting to explore whether decreasing the inflammatory conditions, such as lowing NLR level could serve as an adjuvant therapy and therefore prolong the survival of cervical cancer patients in the near future.

MATERIALS AND METHODS
This systematic review and meta-analysis was performed following the guidance provided in the Cochrane Handbook and was reported according to the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines [34].

Search strategy
We conducted a systematic literature search using the electronic databases PubMed, Web of Science, and Embase up to May 2016. Search terms included "neutrophil to lymphocyte ratio", "NLR", "cervix", "cervical" and "tumor, cancer, neoplasm, carcinoma or malignancy". The titles and abstracts of potential references were scanned carefully to exclude irrelevant articles. The remaining studies were assessed to identify the topic of interest, and full texts were then reviewed comprehensively.

Selection criteria
A study was included if it met the following criteria: (1) included patients with cervical cancer diagnosed histopathologically; (2) provided pre-treatment and/or post-treatment NLR and cut-off values, (3) evaluated the associations between pre-treatment and/or post-treatment NLR and survival outcomes. Exclusion criteria were (1) review articles, editorial comments, letters, expert opinion, conference abstracts, or case reports; (2) insufficient data for estimating hazard ratios (HRs) and 95% confidence intervals (CIs); or (3) full text unavailable and non-English article.
All assessments were conducted independently by two reviewers to assure accuracy of inclusive studies. Multivariate data were preferred to univariate data if both were provided. However, univariate data were acceptable if no multivariate results were presented.

Data extraction
Two investigators independently gathered information from each eligible study. Data was extracted as follows: surname of first author, study country, year of publication, sample size, cancer stage, treatment method, cut-off value defining elevated NLR and HRs with 95% CI for overall survival (OS) and progress-free survival (PFS)/recurrence-free survival (RFS). Disagreements in data extraction were resolved through discussion.

Assessment of methodological quality
Two independent investigators assessed the quality of each included study using the Newcastle-Ottawa Quality Assessment Scale (NOS) [35]. On a score scale from 0 to 9, a study with 7 or more stars was considered as high-quality.

Statistical analysis
The HR and corresponding 95% CI were used to evaluate the prognostic efficiency of NLR on cervical cancer. In addition, the relationship between NLR and clinical-pathological features were reported as ORs and 95% CIs. Cochran's Q test and Higgins I-squared statistic were adopted to test the heterogeneity of pooled data. I 2 < 50% and p > 0.1 indicated no significant heterogeneity and fixed-effects model was applied to combine the effective value [36]. Otherwise, a random-effects model was adopted. All statistical analyses were performed using STATA 12.0 software (StataCorp LP, TX, USA).
Subgroup analyses were performed to investigate the associations of NLR with clinical features in relation to geographic area, statistical methods, sample size, cancer stage, lymph node involvement, NLR cut-off value, and follow-up duration. Moreover, a sensitivity analysis was performed to examine the robustness of the pooled results.