Prognostic value of pretreatment neutrophil-to-lymphocyte ratio in renal cell carcinoma: a systematic review and meta-analysis

Background Numerous studies show that the pretreatment neutrophil-to-lymphocyte ratio (NLR) is associated with the prognosis of patients with RCC. However, their findings are inconsistent, urging us to explore the prognostic value of NLR in RCC patients. Methods This study was pre-registered in PROSPERO (CRD42020167131). Two reviewers independently performed a systematical search of PubMed, Web of Science, EMBASE, and Cochrane Library databases for prospective or retrospective cohort studies investigating the prognostic value of pretreatment NLR. Hazard ratios with 95% confidence intervals for overall survival (OS), disease-free survival (DFS), progression-free survival (PFS), cancer-specific survival (CSS), and other useful clinicopathological features were extracted and analyzed with fixed or random-effect models by using Review Manager 5.3 and Stata 12.0 software. Heterogeneity was estimated on the basis of Cochran’s Q test and I2 value. Sensitivity analyses and subgroup analyses were also performed to explore the potential sources of heterogeneity. Publication bias was assessed with funnel plots and precisely assessed by Egger’s tests. The quality of the evidence was evaluated in accordance with the Grading of Recommendations Assessment, Development, and Evaluation (GRADE). Results Overall, 6461 RCC patients from 24 retrospective studies and 1 prospective study were included. In overall population, elevated pretreatment NLR was associated with poorer OS (pooled HR = 1.90, 95% CI = 1.56–2.30, p < 0.001; I2 = 87%), DFS/PFS (pooled HR = 2.09, 95% CI: 1.49–2.94, p < 0.001; I2 = 99%), and CSS (pooled HR = 2.31, 95% CI: 1.61–3.33, p < 0.001; I2 = 14%). Furthermore, this negative association was further confirmed in patients with nonmetastatic and metastatic RCC patients, respectively. We also investigated the predictive role of NLR in metastatic RCC patients treated with immune checkpoint inhibitors (ICIs). The results indicated that the level of NLR was significantly associated with OS (pooled HR = 3.92, 95% CI: 2.00–7.69, p < 0.001; I2 = 0%) and PFS (pooled HR = 2.20, 95% CI: 95% CI: 1.61–3.01, p < 0.001; I2 = 20%). Conclusions This study demonstrated that elevated pretreatment NLR was significantly associated with poor prognosis of RCC patients. NLR could be helpful as a potential prognostic biomarker to guide clinical decision-making and select individualized treatment strategies for RCC patients.


Background
Renal cell carcinoma (RCC) is a common malignant cancer of the urinary system; its morbidity and mortality have been increasing in recent years [1,2]. When RCC is identified early by imaging conducted for other reasons serendipitously, long-term survival is generally excellent. When RCC is detected with symptoms, the prognosis is poor [3]. However, reliable biomarkers suitable for clinical application remain undiscovered worldwide, and presently except for imaging examination, no method is effective for the early diagnosis and prognosis of RCC. Therefore, in order to improve the prognosis of RCC patients and guide clinical decisionmaking, it is necessary to identify reliable pretreatment biomarker to diagnose, monitor, and manage this disease.
There is growing evidence indicating that immune response and systemic inflammation is the crucial component of human cancer development and progression [4][5][6]. Several studies have indicated that pretreatment neutrophil-to-lymphocyte ratio (NLR), as a systemic inflammatory biomarker, was associated with the prognosis of patients with malignancies; thus providing a new perspective for predicting the prognosis of cancer [7][8][9]. More recently, various studies evaluated the prognostic value of NLR in RCC patients, whereas their conclusions are controversial [10,11]. In 2019, a meta-analysis reported that NLR is a predictor associated with prognosis in RCC patients. However, this study did not perform sensitivity analysis and subgroup analysis to explore the potential sources of heterogeneity and assess the publication bias [12]. Therefore, the aim of this systematic review and meta-analysis was to provide a systematical and comprehensive perspective clarifying the prognostic value of pretreatment NLR for both non-metastatic and metastatic RCC patients.

Search strategy
This study was pre-registered in PROSPERO (CRD42020167131) and conducted in accordance with the guidelines of the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA). A comprehensive online literature search was performed to select the potential studies on PubMed, Web of Science, EMBASE and Cochrane Library databases from inception to December 2019. The main terms used in our search strategy included the following: ("renal" or "kidney") and ("carcinoma" or "neoplasms" or "cancer" or "tumor") and ("NLR" or "neutrophil-lymphocyte ratio" or "neutrophil-to-lymphocyte ratio").

Inclusion and exclusion criteria
The enrolled studies should meet the following inclusion criteria: (1) prospective or retrospective cohort studies evaluating the association between the pretreatment NLR and overall survival (OS), disease-free survival (DFS), progression-free survival (PFS), cancer-specific survival (CSS) of the patients with RCC; (2) patients in these studies did not receive any treatment before obtaining samples; (3) NLR was collected within 30 days before treatment; and (4) the study directly provided hazard ratios (HRs) with 95% confidence intervals (CIs) or had sufficient data to calculate these statistics. If the data in the studies were duplicated, only the data from the most recent study were used. The studies were excluded based on the following exclusion criteria: (1) studies without sufficient survival data for further analysis, (2) duplicated studies or publications, and (3) expert opinions, meeting abstracts, editorials, case reports, letters, reviews or meta-analysis.

Date extraction
For each eligible study, two separate authors independently extracted the following items: study characteristics (first author's name, recruitment region, year of publication, type of study, and sample size); patient information (gender, age, and race), pathological characteristics (TNM stage and histology subtype), disease setting (localized or metastatic), NLR cut-off values (the number and/or percentage of patients with high NLR versus those with low NLR), clinical features (treatment strategy, patient's survival outcome, and follow-up duration), OS, DFS, PFS, and CSS outcomes expressed as HRs (and 95% CI) for RCC patients with high pretreatment NLR versus patients with low pretreatment NLR. In the case of disagreements between individual judgments, the consensus was achieved by discussion with the third investigator.

Quality assessment
The quality of each enrolled study was assessed using the Newcastle-Ottawa Scale (NOS), which consists of three factors: selection, comparability and exposure [13].
The highest score is 9 points and studies with scores 7 or more, 4-6, and lower than 4 were respectively considered to have a low, moderate, and high risk of bias. Any disagreement was resolved by discussion with the third reviewer.

Statistical analysis
Coprimary end-points of the present meta-analysis were OS, DFS, PFS, and CSS in all patients and in patients with nonmetastatic or metastatic RCC. When the included studies directly reported the survival analysis, the HRs and 95% CIs were extracted and used to calculate pooled HRs; otherwise, the Engauge Digitizer software (version 4.1) was used to compute and estimate these data from the Kaplan-Meier survival curves [14,15].
Cochran's Q test and I 2 statistic were used to assess the heterogeneity among the included studies. If significant heterogeneity existed (I 2 > 50% and/or P < 0.10), the pooled HRs and 95% CIs were calculated by a randomeffect model; otherwise, the fixed-effect model was performed (I 2 < 50% and/or P > 0.10) [16]. Sensitivity analyses were conducted to assess the stability of the results by sequentially omitting a single study at a time. Subgroup analyses were conducted to explore the potential sources of heterogeneity. In addition, funnel plots and Egger's tests were used to assess the risk of publication bias. Egger's test and the trim and fill method were performed with Stata 12.0 software (STATA Corporation, College Station, TX, USA). Other statistical analyses were conducted using Review Manager 5.3 software (Cochrane Collaboration, Copenhagen, Denmark). All p-values were two-sided, and a statistically significant difference was defined as p < 0.05.

Quality of evidence
The quality of the evidence of the predictive value of pretreatment NLR for the prognosis in RCC patients was assessed according to the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) [17].

Included literature
Based on our search strategies, we identified 1039 potentially relevant studies. After removing duplicates, we viewed the titles and abstracts of the remaining 942 records. Subsequently, we assessed the full text for 76 articles. Finally, 25 studies were enrolled in the present meta-analysis [10,11,. The study selection process is presented as a flowchart in Fig. 1.
To explore whether a single study affected heterogeneity and conclusions, we performed a sensitivity analysis by sequentially excluding each single study in turn. After removing Viers' study, heterogeneity among studies was reduced (I 2 = 48%, p = 0.07) in nonmetastatic RCC. Similarly, after removing Silagy's study, heterogeneity was reduced (I 2 = 46%, p = 0.05) in metastatic RCC. Overall, the results of sensitivity analyses did not affect the conclusions described above and validated the robustness of our findings.

NLR and DFS, PFS in RCC
When we investigated the association between NLR and DFS/PFS, 18 studies with 2735 patients were selected.
Consequently, sensitivity analyses were also performed in nonmetastatic RCC patients. The results showed that the omission of any one study, except Viers' study, did not significantly affect heterogeneity. However, heterogeneity was still relatively significant after the removal of Viers' study. Since different study features were involved, we further performed subgroup analyses to explore the source of heterogeneity (Table 2). In the subgroup analysis based on sample sizes, elevated pretreatment NLR was significantly associated with poorer DFS in both sample sizes ≥200 (pooled HR = 1.57, 95% CI: 1.09-2.26, p = 0.02; I 2 = 69%) and sample sizes < 200 (pooled HR = 1.85, 95% CI: 1.17-2.92, p = 0.008; I 2 = 60%). Moreover, histology type, mean age, cut-off value of NLR, treatment strategy, the mean follow-up months, and other study features did not affect the relationship between NLR and DFS in nonmetastatic RCC patients. Interestingly, when stratified by race of patients, heterogeneity was significantly reduced and results demonstrated that elevated NLR was significantly associated with pooer DFS in both Asian and Caucasian RCC patients (pooled HR = 2.31; 95% CI: 1.70-3.14, p < 0.001; I 2 = 0%; pooled HR = 1.21, 95% CI: 1.08-1.36, p = 0.001; I 2 = 0%, respectively).

NLR and CSS in RCC
Four studies, comprising 2314 patients, provided data on the association of NLR with CSS. The forest plot indicated that higher pretreatment NLR was significantly associated with worse CSS (pooled HR = 2.31, 95% CI: 1.61-3.33, p < 0.001, I 2 = 14%; Fig. 2). Considering that the patients enrolled in these four studies were nonmetastatic RCC, we did not further investigated the association between NLR and CSS in metastatic RCC patients. Furthermore, a sensitivity analysis was performed to explore whether a single study affected heterogeneity and conclusions. After removing Zheng's study, heterogeneity among studies was markedly changed (I 2 = 0%, p = 0.47) in nonmetastatic RCC. However, the pooled HR recalculated did not affect the conclusion described above, which validated the strength of our results.

Subgroup analyses based on histology type
Considering that histology type may be the source of heterogeneity, we especially performed subgroup analyses based on histology types of RCC patients. The results revealed that histology types could change heterogeneity significantly. As shown in Table 3

Publication bias
Publication bias was assessed, respectively for OS, DFS/PFS, and CSS without considering the stage of RCC patients. For both OS and DFS/PFS, the funnel plots were asymmetric (Fig. S1). Disappointingly, the Egger's test also showed some degree of publication bias (both p < 0.001). Therefore, the trim and fill method was carried out to examine the funnel plot's asymmetry by using hypothetically non-published studies. The recalculated results indicated that elevated NLR was significantly associated with OS (HR: 1.51, 95%CI: 1.28-1.79, p < 0.001) and DFS/PFS (HR: 1.96, 95%CI: 1.42-2.72, p < 0.001), indicating the stability of the results (Fig. S2). For CSS, the funnel plot was relatively symmetric (Fig. S1). Moreover, the Egger's test showed that there was no obvious publication bias (p = 0.285).

Quality of evidence
The assessment of the quality of evidence was performed for OS, DFS/PFS, and CSS which were critical in showed that the quality of evidence of OS and DFS/PFS was both "very low" because of observational studies, some degree of publication bias, and significant heterogeneity. However, the quality of evidence of CSS was "low" due to observational studies (Table S2).

Discussion
Prognostic markers have multiple applications in the diagnosis, treatment, and prediction of clinical outcome and can contribute to choosing the best treatment strategy. Pretreatment NLR, as a prognostic predictor for RCC, has been evaluated by numerous studies, but they It is largely recognized that systemic inflammatory response and tumor microenvironment are essential in the development and progression of cancer [5,41]. Some researchers have shown cancer-related inflammatory response consists of cytokines, chemokines, transcription factors, and inflammatory cells, which play decisive roles at different stages of tumor development including initiation, promotion, malignant conversion, invasion, and metastasis [42,43]. Hence, several inflammatory biomarkers, such as platelet-to-lymphocyte ratio (PLR) [44], lymphocyte-to-monocyte ratio (LMR) [45], C-reavtive protein to albumin ratio (CAR) [46], C-reactive protein (CRP) [47] and modified Glasgow Prognostic Score (mGPS) [48] are the potential prognostic biomarkers in RCC patients.
Recently, various studies investigated the prognostic value of NLR in RCC patients and the function of neutrophils and lymphocytes may be responsible for the underlying mechanisms. Neutrophils are usually regarded as an important part in the acute phase of inflammation and confer resistance against microbes. Some studies showed that neutrophils were involved in cancer development. Neutrophils could directly affect tumor cells to promote cancer progression. Moreover, neutrophils may indirectly change the tumor microenvironment to promote cancer metastasis [49]. In addition, neutrophils could release tumor growth promoters or immunoregulatory mediators, including vascular endothelial growth factor (VEGF) to affect cancer progression, which is generally regarded as an important part in tumor angiogenesis and has been indicated by the inhibitory effects of anti-VEGF antibodies on tumor growth in vivo [50][51][52]. By contrast, lymphocytes reflect on cell-mediated immunity and are essential in anti-tumor immune responses. Increased infiltration of lymphocytes in the tumor region has been associated with better responsiveness to therapy and better prognosis in patients with solid tumors [53]. Moreover, lymphopenia, with a decrease in CD4 + T-cells, which are often observed in cancer patients, compromises the anti-tumor response mediated by lymphocytes [54]. In other words, NLR not only reflects the inflammatory response in patients, but it also represents the decline of anti-tumor immunity, thus bringing a new perspective in determining the outcome of RCC patients.
Several meta-analyses have discussed the relationship between NLR and prognosis in RCC patients [55,56]. Their results were similar to our results, but these metaanalyses did not perform sensitivity analysis and subgroup analysis to evaluate the source of heterogeneity and assess the publication bias. Furthermore, the advent of ICIs has changed the management of metastatic RCC. In particular, we investigated OS and PFS in metastatic RCC patients treated with ICIs, and the negative association was also confirmed in these RCC patients. Additionally, in this analysis, we found that heterogeneity was significantly reduced by dividing the study population according to the race of patients, namely, Caucasian and Asian RCC patients. Besides, we also noted that heterogeneity was significantly reduced by limiting the histology type to clear cell RCC. Therefore, when we explore the clinical role of NLR, it would be better to understand the histology type of patients. Moreover, there were several different defined methods about NLR in our studies. Some studies regarded NLR as a continuous variable, whereas some studies divided it into two groups based on ROC curve, median value, Xtile, and other methods. These different defined methods could potentially lead to heterogeneity. Of note, our study mainly evaluated the prognostic value of the pretreatment NLR. Several studies have shown that the post-treatment NLR was also an effective prognostic biomarker in RCC patients [44,57,58]. Therefore, the dynamic detection of the peripheral blood NLR level during treatment could play a more important prognostic role for RCC patients.
This study has several limitations that should be acknowledged. First, most of the included studies were retrospective and some of these studies had small sample sizes. Second, heterogeneity in some subgroup analyses was moderate or high. The possibility of selection biases or other unidentified confounders could not be completely avoided. Third, a certain degree of publication bias in our study may weaken the quality of evidence. Thus, the trim and fill method was performed and the adjusted results validated the stability of our results. Finally, there was no established cut-off value of NLR. Most scholars selected a cut-off value based on the highest sensitivity and specificity or used predefined cut-off values from other studies. Therefore, more largescale prospective studies are needed to establish the standard cut-off value of NLR and provide more evidence.