Prognostic value of platelet to lymphocyte ratio in non-small cell lung cancer: a systematic review and meta-analysis

The prognostic value of the platelet-to-lymphocyte ratio (PLR) in non-small cell lung cancer (NSCLC) remains controversial. We therefore conducted a meta-analysis of published studies to determine the prognostic value of PLR in NSCLC. A systematic search was performed in PubMed, Web of Science and Embase for relevant studies. The data and characteristics of each study were extracted, and the hazard ratio (HR) at a 95% confidence interval (CI) was calculated to estimate the effect. We also performed subgroup and meta-regression analyses. A total of 2,889 patients in 12 studies were enrolled in this meta-analysis, and the pooled HR of 1.492 (95% CI: 1.231–1.807, P < 0.001) indicated that patients with an elevated PLR are expected to have a shorter overall survival (OS) after treatment. This meta-analysis indicates that a high PLR might be a predictive factor of poor prognosis in NSCLC. Further large-cohort studies are needed to confirm these findings.

Scientific RepoRts | 6:22618 | DOI: 10.1038/srep22618 and 95% CI, we were able to calculate these statistics from data provided in the article. The NOS scores of the studies ranged from 5 to 8, with a mean value of 7.

Meta-analysis.
For the overall population, the pooled HR used to evaluate PLR on OS was 1.492 (95% CI: 1.231-1.807, P heterogeneity = 0.005; Fig. 2), indicating that a high PLR was predictive of a poor prognosis in NSCLC patients. The main results of this meta-analysis are listed in Table 2. Subgroup analyses by ethnicity revealed that PLR was a negative predictor of OS for Asian (HR = 1.384, 95% CI: 1.067-1.795, P heterogeneity = 0.003) and Caucasian populations (HR = 1.682, 95% CI: 1.348-2.099, P heterogeneity = 0.432). Because the PLR cut-off values were different among the enrolled studies, we performed subgroup analyses based on different cut-off values and found that a high PLR was a negative predictor of OS both for a cut-off value of 150 or less (HR = 1.302, 95% CI: 1.028-1.648, P heterogeneity = 0.029) and a PLR greater than 150 (HR = 1.831, 95% CI: 1.502-2.233, P heterogeneity = 0.758). When considering differences in sample size, a high PLR was a poor prognostic marker for the sample sizes > 200 (HR = 1.445, 95% CI: 1.074-1.945, P heterogeneity = 0.003) and sample sizes ≤ 200 (HR = 1.546, 95% CI: 1.203-1.988, P heterogeneity = 0.152). We also analyzed the significance of a high PLR with respect to OS for patients who received different treatments. Among five surgical and seven non-surgical studies, an elevated PLR was a uniformly poor predictor of OS (surgery: HR = 1.347, 95% CI: 1.012-1.793, P heterogeneity = 0.022; non-surgery: HR = 1.624, 95% CI: 1.278-2.063, P heterogeneity = 0.104). In addition, subgroup  analyses showed that an elevated PLR predicted the prognosis for NSCLC, regardless of the HR estimation (reported vs. estimated) and NOS score (≥ 7 vs. < 7, Table 2).

Sensitivity analyses.
We conducted meta-regression analysis to investigate the potential source of heterogeneity among studies of OS. The results demonstrated that ethnicity (P = 0.710), PLR cut-off value (P = 0.399), sample size (P = 0.983), treatment (P = 0.865), HR estimation (P = 0.660) and NOS score (P = 0.627) did not contribute to the source of heterogeneity for OS. Sensitivity analyses were used to examine whether individual  studies influenced the results. When individual studies were removed one at a time from the above analyses, the corresponding pooled HRs were not markedly altered by any single study, indicating the robustness of the presented results.
Publication bias. Begg's funnel plot and Egger's test were performed to evaluate the publication bias of the included studies. The funnel plot was symmetric for OS (P = 0.631, Fig. 3). Additionally, no indication of publication bias in terms of OS was found using Egger's test (P = 0.089).

Discussion
Inflammation is a hallmark of cancer 19 . Accumulated evidence shows that the systemic inflammatory response (SIR) is related to different stages of cancer progression, including its initiation, promotion, invasion and metastasis. A component of SIR, thrombocytosis is common in patients with solid tumors 20,21 . Through their secretion of VEGF, platelet-derived growth factor (PDGF), TGF-β and FGF, platelets can trigger angiogenesis, cell proliferation and migration [22][23][24] . Platelets not only augment the growth of primary tumors but also support cancer cells in their evasion of the immune system and extravasation to second sites. Nieswandt et al. have reported that platelets might protect cancer cells from natural killer-mediated lysis to facilitate metastasis 25 . Furthermore, Labelle et al.
found that platelet-derived TGF-β promotes metastasis by activating the Smad and NF-kB pathways 26 . Based on these results, the prognostic value of thrombocytosis in various cancer patients was investigated 27 . Lymphocytes play an important role in the cell-mediated anti-tumor immune response. In colorectal and ovarian cancers, patients with a higher number of tumor-infiltrating lymphocytes (TILs) exhibit improved survival 28,29 . Kawai et al. reported that tumor-infiltrating CD8 + T cells in cancer nests have a positive effect on prognosis in cases of NSCLC 30 . Hence, platelets and lymphocytes might play a critical role in tumor progression. Because it is based on platelet and lymphocyte counts, the PLR could aid in predicting the clinical outcomes of cancer patients. Many studies have investigated the prognostic value of the PLR for various forms of cancer. However, the prognostic value of the PLR remains controversial for many ailments, including NSCLC. The aim of the current study was therefore to evaluate the prognostic value of the PLR for cases of NSCLC. To our knowledge, this is the first meta-analysis to investigate the association between the PLR and the prognosis of NSCLC patients. By integrating the results of independent studies, meta-analysis provides a useful tool for the detection of effects that may be missed by individual studies. In this meta-analysis, we included 12 eligible studies of a total of 2,889 patients to clarify the prognostic value of PLR in NSCLC. The combined results demonstrated that a high PLR was associated with a poor OS (HR: 1.492, 95% CI: 1.231-1.807, P < 0.001) in NSCLC. In addition, the significance of this association was not changed in a sensitivity analysis that removed single studies. Subgroup analyses showed that the trend of a worse OS with a higher PLR was present in both Asian and Caucasian patients. We performed subgroup analyses based on PLR cut-off values and found that patients with a low PLR had a better OS compared to those with an elevated PLR, regardless of the PLR cut-off values. When analyzed by sample size, similarly significant results were found for both large and small studies. Subgroup analysis stratified by treatment (i.e., surgical or non-surgical), HR estimation (i.e., reported or estimated) and NOS (i.e., NOS ≥ 7 or < 7) also revealed that the PLR had a negative effect on OS. Our data help to clarify the results of individual studies and to identify NSCLC patients at high risk for whom adjuvant therapy might be necessary.
There is significant heterogeneity among the studies included in this meta-analysis. To explore the possible causes of heterogeneity, we performed subgroup analyses according to ethnicity, cut-off value, sample size, treatment, HR estimation and NOS. The results revealed that the subgroups representing Caucasian ethnicity, a cut-off value of PLR > 150, a sample size ≤ 200, non-surgical treatment, estimated HR and NOS score < 7 had decreased heterogeneity. However, we could not rule out the possibility that heterogeneity was caused by these factors. For example, for PLR cut-off value, there was no universal threshold that defined a higher PLR in cancer patients. Different methods were used to calculate the PLR cut-off value in this meta-analysis, in which six studies set the cut-off value using a receiver operating curve (ROC), three by taking the median PLR, and three by using values reported in previously published literature. In addition, many factors could explain the heterogeneity observed among these studies, such as the clinical features of the enrolled patients, the method used to determine the laboratory index (such as platelet and lymphocyte counts), the pathological stage difference, or histology type.
Several limitations of this study must be carefully considered. First, significant heterogeneity was demonstrated in our results due to confounding factors such as the clinical features of the patients, ethnicity, sample size, treatment, HR estimation and PLR cut-off value. However, subgroup, meta-regression and sensitivity analyses revealed that none of the aforementioned confounders completely explained the observed heterogeneity. Thus, we propose that the genotypic diversity of NSCLC and the combined effect of the aforementioned confounders contributed to heterogeneity. Second, the HR of some studies was not provided, and we had to calculate the value from the data. There might be minor differences between the exact HR and the data extrapolated according to Tierney's method. Third, although we did not impose a language limitation, only studies in English and Chinese were enrolled in the meta-analysis. Finally, studies lacking sufficient data were also excluded from the meta-analysis. Furthermore, due to negative or null results, many studies could not be published; these studies were also omitted from the present analysis. All of these factors might contribute to the heterogeneity observed in this study.
In conclusion, a meta-analysis of published studies revealed that PLR is an unfavorable predictor of prognosis in patients with NSCLC, which could be useful for stratifying NSCLC patients and in determining individual treatment plans. In addition, as a routine test, the PLR was easy to obtain and, importantly, does not incur additional costs. Thus, PLR is a promising biomarker for use in clinical management. However, due to the heterogeneity and limitations uncovered in the present study, the results of our meta-analysis may be estimations. Future larger-scale prospective and standard investigations should be conducted to confirm our results.

Materials and Methods
Search strategy. We conducted a comprehensive of the PubMed, Web of Science, and Embase databases to identify studies for inclusion in the present meta-analysis (last search updated to December 1, 2015). The following terms and combinations were used to identify studies: "lung cancer", "lung carcinoma", "lung neoplasm", "lung malignancy", "platelet", "lymphocyte" and "ratio". To ensure that no studies were overlooked, the reference lists of relevant articles and potential related articles were manually searched to identify additional studies.
Date extraction and quality assessment. Two independent investigators (H.Z. and L.W.G.) read the titles and abstracts of all potential articles. Articles that could not be categorized based on title and abstract alone were retrieved for full-text review. The articles were read independently and checked for the inclusion criteria used in this study. For each study, the following data were collected: first author's name, publication date, country of origin, ethnicity, number of patients, stage, treatment, cut-off value, follow-up, and hazard ratio (HR) of PLR for OS. Disagreement in the data extraction phase was settled together by the two reviewers. The Newcastle-Ottawa Quality Assessment Scale (NOS) was used to assess the quality of each study by two independent investigators (H.Z. and L.W.G.). The NOS consists of three parts: selection (four points), comparability (two points), and outcome assessment (three points). NOS scores ≥ 7 are considered to indicate high-quality studies.
Inclusion and exclusion criteria. The inclusion criteria for this study were as follows: (1) patients with NSCLC in the studies were histo-pathologically confirmed; (2) the association of PLR with OS was investigated; and (3) the study reported sufficient data to estimate the HR at a 95% confidence interval (CI). The exclusion criteria were as follows: (1) reviews, letters, case reports, abstracts, conferences, or expert opinions; (2) studies in which necessary data were not provided; and (3) non-human research. We avoided the duplication of data by examining the names of all authors and medical centers involved for each article. When multiple publications using the same study population were identified or when the study populations overlapped, only the most complete and/or recently published was included.
Statistical Analysis. The impact of PLR on OS was measured by the combined HRs and their 95% CIs extracted from each eligible study. The HR and its 95% CI in each eligible study were directly extracted from the report or were indirectly estimated by methods described by Tierney 31 . The Cochrane Q test and I 2 statistic were performed to assess the heterogeneity of the pooled results. If P < 0.1 and/or I 2 > 50%, indicating the presence of heterogeneity, the random-effects model (i.e., the DerSimonian-Laird method) was used to calculate the pooled HRs 32 . In other cases, the fixed-effects model (i.e., the Mantel-Haenszel method) was used 33 . The robustness of the pooled results was confirmed by a sensitivity analysis in which the data of an individual study was removed each time. Begg's funnel plot and the Egger's linear regression tests were conducted to identify the possibility of publication bias; P < 0.05 was considered significant. All statistical analyses were performed with Stata software version 11.0 (Stata Corporation, College Station, Texas, USA), and all P values were two-sided.