Pretreatment Neutrophil to Lymphocyte Ratio as a Prognostic Predictor of Urologic Tumors: A Systematic Review and Meta-Analysis.

The relationship between inflammation and tumor development and progression has been recognized in recent decades. NLR is an easily reproducible and widely used inflammatory response marker. The prognostic value of NLR for urologic tumors has been reported in succession. Here, we perform a systematic review and meta-analysis to summarize the association between the NLR and prognosis of urologic tumors. We conducted a computerized search of PubMed, Embase, and ISI Web of Knowledge to identify clinical studies that had evaluated the association between the pretreatment NLR and prognosis in urologic tumors. Prognostic outcomes included overall survival (OS), cancer-specific survival (CSS), recurrence-free survival (RFS), progression-free survival (PFS), and metastasis-free survival (MFS). We extracted and synthesized corresponding hazard ratios (HRs) and confidence intervals (CIs) using Review Manager 5.3 and STATA 13. We identified 34 retrospective cohort studies and conducted the meta-analysis. The results showed that all OS, CSS, RFS, PFS, and MFS risks were significantly different between patients with an elevated NLR and those with a low NLR in various urologic tumors. A high NLR portended poor prognosis. However, no significance was observed for CSS in patients with renal cell carcinoma (HR = 1.38, 95% CI: 0.96-1.99). Our meta-analysis suggests that NLR could be a prognostic predictor for urologic tumors. Patients with a high NLR were deemed to have a poor prognosis.

INSTRUCTION

The relation between inflammation and tumor development and progression has been recognized in recent decades.[1,2] As a typical representative of inflammatory reactions, C-reactive protein (CRP) has been reported to be significantly associated with the prognosis of several cancers.[3-7] Other systematic inflammation markers have been validated as predictive in various types of cancer.[8-10] The neutrophil to lymphocyte ratio (NLR) is also a widely used inflammatory marker that is defined as the absolute neutrophil count divided by the absolute lymphocyte count and can be easily acquired from complete blood cell parameters.[11] It is a cheap and easily acquired marker compared with other inflammatory markers, such as CRP. Prognostic factors are essential for the stratification of cancer risk, medical treatment, and clinical research. Hence, we aimed to conduct a systematic review and meta-analysis to reveal the predictive effect of NLR on urologic tumor prognosis. Adding NLR to the inflammation-based prognostic score model may lead to improved patient management. This study is complied with Meta-analysis of Observational Studies in Epidemiology (MOOSE).[12]

MATERIALS AND METHODS

Search and Filtration Strategy

A systematic literature search of PubMed, Embase, and ISI Web of Knowledge (Web of Science + BIOSIS Previews + MEDLINE + SciELO Citation Index + KCI-Korean Journal Database) was conducted to retrieve clinical studies up to January 2015. We used Mesh terms and text words of neutrophil, lymphocyte, renal cancer, upper tract urothelial carcinoma, bladder cancer, prostate cancer, and urinary cancer to search for related articles (http://links.lww.com/MD/A446). Citations in the retrieved articles were also searched for any relevant studies. The initial selection was performed to eliminate obviously irrelevant articles and retain potentially relevant articles about NLR or urologic tumor prognostic risk factors by an analysis of the title and abstract by 2 independent investigators (YL and D-LS). Thereafter, the full text was reviewed according to eligibility criteria. For inclusion in this analysis, studies should contain an evaluation of the relationship between pretreatment NLR and urologic cancer prognosis. Studies with the following criteria were excluded: duplicated literature, overlapping patients, or duplicated data presented in conferences; no available data; and abstract only. Meeting abstracts were not included based on their lack of sufficient detailed information to assess the methodological bias or quality before the quantitative meta-analysis. All data and analyses were based on the previous published studies; thus, ethical approval and patient consent are waived.

Quality Assessment and Data Extraction

There are no standard quality assessment tools for prognostic studies in systematic reviews. The quality assessment of included studies was independently applied using the “Newcastle-Ottawa Scale (NOS)” for cohort studies,[13] which includes 3 domains with 8 items. Each item was awarded 1 to a maximum of 2 stars, and the total possible score was 9 stars. Studies with > 5 stars were deemed as being of good quality. Data extraction and cross-checking were also performed by 2 independent investigators (YL and D-LS). Additionally, any disagreement or uncertainty was brought to a group discussion where a consensus was reached. Data extracted from these citations included the name of the first author, year of publication, tumor category, cutoff value of NLR, prognostic outcomes, sample size, region, statistic model, follow-up time, and NOS score. The data were extracted from the original articles. Situations lacking exact data were resolved in a number of ways: multivariate outcomes were preferred to univariate outcomes when both were provided, but if no multivariate results were presented, univariate outcomes were used instead; and given survival or mortality curves were used to calculate the estimated HR and 95% CI provided by Tierney et al[14] or the corresponding author was contacted to obtain the original data or results. Finally, before the meta-analysis, we rechecked the data and potential studies for overlapping patients to avoid an over-analysis.

Statistical Analysis

Review Manager 5.3 (The Cochrane Collaboration, Copenhagen) was used to carry out the synthesis analysis. Quantitative small-study effect detection with Egger regression intercept test was merely performed on outcomes with >10 studies by using STATA 13 (Stata Corp LP, College Station, TX).[15,16] Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated as the time-to-event effect estimate for the survival analysis. First, Cochran's Q test and Higgins I statistic were performed for heterogeneity.[15]P ≥ 0.1 and I ≤ 50% meant no significant heterogeneity, and thus, a fixed-effects model was used. Otherwise, a random-effects model was used to calculate pooled HRs. A sensitivity analysis was performed using the method of leave-one-out to leave-all-out of univariate or estimated outcomes to test the feasibility of the pooled results. If a small-study effect existed, a trim and fill method was also performed. A 2-tailed P < 0.05 was considered statistically significant. All the results are presented in the forest plots.

RESULTS

Eligible Studies and Quality Assessment

A total of 1017 articles were retrieved by the initial search strategy in PubMed, Embase, and ISI Web of Knowledge. Overall, 747 of 1017 articles were retained after removing duplicates. After reading the title and abstract, 550 articles were excluded for unrelated information, and the remaining 197 were screened by skimming the full-text. Then, 49 potential studies were screened and validated by reading the full-text. Finally, 34 studies[17-50] were included according to the inclusion and exclusion criteria. Two studies were excluded for lacking dichotomous NLR variables,[51,52] 5 studies had no available data,[53-57] and 7 studies[58-64] included overlapping patients with 5 other studies.[19,21,29,38,42] The screening diagram is shown in Figure 1. Table 1 tabulates the characteristics and quality assessment of the included studies. The majority of the included studies were adjusted for potential confounders using the COX proportion hazard model, but the adjusted factors did not conform to each study. Univariate and estimated outcomes were acquired from the article when no multivariate outcomes were reported. The urological tumors were defined as renal cell carcinoma, upper tract urothelial carcinoma, bladder cancer, and prostate cancer.

FIGURE 1

Literature screening flowchart.

TABLE 1

Characteristics and Quality Assessment Results of the Included Studies

Survival Outcome

Prognostic outcomes, including overall survival (OS), cancer-specific survival (CSS), recurrence-free survival (RFS), progression-free survival (PFS), and metastasis-free survival (MFS), were quantitatively synthesized. The meta-analysis results are displayed in Figures 2–6. Heterogeneity is illustrated in each forest plot. In Figure 2, OS outcomes were available from 23 studies on renal cell carcinoma, upper tract urothelial carcinoma, bladder cancer, and prostate cancer. The synthesized hazard risk for each type of cancer consistently favored the low NLR patients (pooled HR: 1.79, 95% CI: 1.61–2.00 for renal cell carcinoma; pooled HR: 2.48, 95% CI: 1.31–4.70 for upper tract urothelial carcinoma; pooled HR: 1.68, 95% CI: 1.45–1.94 for bladder cancer; and pooled HR: 1.44, 95% CI: 1.28–1.62 for prostate cancer), which meant that patients with a higher NLR had a higher all-cause mortality risk than those with a low NLR.

FIGURE 2

Overall survival based on the dichotomous NLR.

FIGURE 6

Metastasis-free survival based on the dichotomous NLR.

Figure 3 shows that 13 studies provided sufficient data on CSS outcome. Pooled results showed significant superiority of a low NLR in upper tract urothelial carcinoma (pooled HR: 2.52, 95% CI: 1.41–4.52) and bladder cancer (pooled HR: 1.70, 95% CI: 1.45–1.99) but not in renal cancer (pooled HR: 1.38, 95% CI: 0.96–1.99). No significant difference was observed in renal cell carcinoma for CSS. No study reported CSS in prostate cancer.

FIGURE 3

Cancer-specific survival based on the dichotomous NLR.

According to Figure 4, 7 studies reported RFS in renal cancer, upper tract urothelial carcinoma, and bladder cancer. Each pooled result showed a significantly higher risk of tumor recurrence in patients with a high NLR (pooled HR for renal cell carcinoma: 1.97, 95% CI: 1.37–2.84; pooled HR for upper tract urothelial carcinoma: 1.47, 95% CI: 1.11–1.95; pooled HR for bladder cancer: 1.55, 95% CI: 1.21–2.00). No study reported RFS for prostate cancer.

FIGURE 4

Recurrence-free survival based on the dichotomous NLR.

Similarly, in Figure 5, 7 studies provided PFS information in renal cancer: 1 study each in upper tract urothelial carcinoma and bladder cancer and 2 studies in prostate cancer. All the pooled HRs favored patients with a low NLR (pooled HR: 1.85, 95% CI: 1.24–2.77 in renal cell carcinoma; pooled HR: 1.70, 95% CI: 1.14–2.56 in upper tract urothelial carcinoma; pooled HR: 3.52, 95% CI: 1.33–9.33 in bladder cancer; and pooled HR: 1.29, 95% CI: 1.04–1.59 in prostate cancer).

FIGURE 5

Progression-free survival based on the dichotomous NLR.

In Figure 6, only 3 studies showed MFS in renal cancer and upper tract urothelial carcinoma, and the pooled results also favored patients with a low NLR (pooled HR: 1.60, 95% CI: 1.29–1.98 in renal cell carcinoma; pooled HR: 2.47, 95% CI: 1.16–5.29 in upper tract urothelial carcinoma). In summary, the majority of the synthesized results showed a significantly higher risk for patients with a high NLR in terms of prognostic outcome.

Small-Study Effect and Sensitivity Analysis

Quantitative small-study effect detection was conducted by Egger asymmetric test only for OS of renal cell carcinoma. The P value of the linear regression was 0.015. A significant small-study effect was observed, and the funnel plot was omitted, which potentially contributed to selective outcome reporting or publication bias. Thus, we conducted the trim and fill method to test the stability of the pooled outcome. The HR was HR = 1.63 (1.48–1.80) for OS of renal cell carcinoma (Figure 7). The significance of the results was not altered. A sensitivity analysis of the univariate and estimated outcomes was also performed manually as described, and no pooled outcomes were altered except MFS in renal cell carcinoma (results were omitted).

FIGURE 7

Trim and fill method for overall survival of renal cell carcinoma.

DISCUSSION

Many studies have revealed the correlation between cancer and inflammation.[65-67] A hypothesis was proposed that chronic inflammation promotes tumor development and suppresses immune activity.[68-70] Hanahan et al[71] summarized an important hallmark of cancer is that cancer cells evade immunological attack from lymphocytes, macrophages, and natural killer cells, etc. High NLR represents systemic and local inflammation that provides a favorable microenvironment for tumor invasion and metastasis[72] and suppresses the host immune surveillance.[73] It is also associated with high infiltration of tumor-associated macrophages (TAMs)[74] that contribute to tumor growth, invasion, and evasion.[72,75] During inflammatory procedure, neutrophilia as an important component of inflammatory response inhibits the immune system by suppressing the cytolytic activity of immune cells such as lymphocytes, activated T cells, and natural killer cells.[76,77] Additionally, neutrophils and macrophages produce tumor growth factors including epidermal growth factor (EGF), vascular endothelial growth factor (VEGF), interleukin (IL)-6, and IL-8, which promote stimulating tumor microenvironment.[71,72] In addition, these cells may produce proangiogenic and proinvasive matrix-degrading enzymes, including matrix metalloproteinase,[78] elastases,[79] cysteine cathepsin proteases, and heparanase[80,81] that promote tumor metastasis.

Nonsteroidal anti-inflammatory drug (NSAID) consumption was confirmed to reduce the risk of colorectal cancer.[82] As a systematic inflammatory marker, NLR has been recognized to be associated with solid tumor prognosis.[83] Additionally, other inflammatory markers have been reported to be significantly associated with tumor development and prognosis. For example, CRP is correlated with urologic cancer prognosis,[3,84] and ESR and PLR are associated with renal cell carcinoma prognosis.[85] The Glasgow prognostic score is an inflammation-based prognostic score that can predict the prognosis of several types of cancer.[33,86,87]

NLR is an easily reproducible and widely available marker obtained from peripheral complete blood cell counts because cell separation has been widely applied. However, the feasibility of the NLR was seldom researched in the included studies,[48] and the volatility of the pretreatment NLR is not clear. Additionally, NLR changes followed by tumor changes during anticancer management are also essential to understand for its application as an indicator of treatment efficacy.

The majority of the included studies used the dichotomous NLR to determine the prognostic value. Several studies used both continuous and dichotomous NLR variables to determine the direct prognostic value. The continuous NLR also significantly portended the prognosis of urologic tumors in various studies.[21,35,37,43,48,49] The use of a continuous variable reflected a small intrinsic effect. However, the cutoff value seemed to adjust itself in various studies. The NLR threshold was calculated in each study to acquire the most significant effect, and the final significance of the outcomes seemed to be created, not intrinsic. Additionally, it is difficult to interpret and compare different studies when different cutoff values were used. Thus, we recommend using a continuous NLR variable rather than a categorical variable in future studies.

Our study has several limitations. First, all the included studies in our meta-analysis were retrospective. In observational studies, selection bias is impossible to avoid, although a multivariate analysis can control the confounding factors to a certain extent. Second, the NLR could be affected by different conditions, especially undetected diseases such as chronic infection, chronic disease, and autoimmune disorders, such as rheumatic disease. Third, we noted that the majority of the included studies did not report cancer-specific survival, which is an essential outcome for cancer survival analysis. Fourth, the reciprocal correlation between the NLR and other systemic inflammatory response markers should be noted, which probably result in high co-linearity in a multivariate analysis and affect the parameter estimation of the Cox model. For instance, we may realize that CRP or the platelet lymphocyte ratio was correlated with NLR. Report bias was also observed while reading the full texts. Several studies did not present the univariate or multivariate HR for lack of significance; additionally, stepwise regression that would eliminate the nonsignificant factors was used, and only the significant factors were included.

CONCLUSIONS

Our meta-analysis summarized the published literature of the prognostic value of the NLR, and the pooled results suggested that the NLR was an effective prognostic predictor. Patients with a high NLR were deemed to have a poor prognosis.