Prognostic value of platelet count in lung cancer: a systematic review and meta-analysis

Background : The prognostic value of pretreatment elevated platelet count remains controversial in lung cancer patients. We performed the present meta-analysis to determine the precise role of it in these patients. Methods: We performed a multiple search strategy in PubMed database, EMBASE and Cochrane Library to identify eligible studies. Disease-free survival (DFS) /Progress-free survival (PFS)/Time to progress(TTP) and Overall survival (OS) were used as outcomes with hazard ratio (HR) and its 95% confidence intervals (CIs). Heterogeneity among studies and publication bias were also evaluated. Results : A total of 39 studies including 16696 lung cancer patients were eligible in the analysis. Overall, the pooled analysis showed that pretreatment elevated platelet count was associated with poorer OS (HR= 1.47, 95%CI: 1.31-1.66, P<0.001) and poorer DFS/PFS/TTP ((HR=1.63, 95%CI: 1.28-2.09, P<0.001) in patients with lung cancer compared with normal platelet count. In subgroup analyses, pretreatment elevated platelet count was also associated with poorer OS and DFS/PFS/TTP in most subgroups. There was no evidence of publication bias. Conclusions : This meta-analysis revealed that pretreatment elevated platelet count was an independent predictor of OS and DFS/PFS/TTP in lung cancer patients. Large scale prospective studies and a validation study are warranted.

In 1960s, RICHARD B.et al have suggested that the platelet was correlated with cancers 5 . Tumor cells could secret platelet agonists to induce platelet aggregation and result in thrombocytosis by producing thrombopoietic cytokine like interleukin-1 (IL-1), IL-3, IL-11, and particularly tumorderived IL-6 6-9 . Many studies have shown thrombocytosis played a role in cancer genesis and development 10,11 . Increasing evidence has indicated that platelet count correlates with prognosis in various malignancies such as lung, renal, gastric, colorectal and hepatocellular cancer and is considered as a hallmark of cancer [12][13][14][15][16] . Additionally, the platelet count is convenient to perform, less expensive, and easily available. However, the current opinions about the correlation between platelet count and lung cancer prognosis are controversy. Some studies identify that platelet count is a poor prognosis in NSCLC, while some suggest that platelet count has no association with lung cancer 11,[17][18][19] . Therefore, we conduct this meta-analysis to further investigate the prognostic value of platelet count on the survival in lung cancer patients.

Selection criteria
The including criteria were as follows: (1) the diagnosis of lung cancer was pathologically confirmed; (2) platelet count was measured preoperation; (3) hazard ratios (HRs) and their 95% confidence intervals (CIs) for platelet count can be obtained; (4) the cut-off value of thrombocytosis was reported;(5) the relationship between OS or DFS/PFS/TTP was evaluated.

Exclusion criteria
Articles were excluded if they meet the following criteria: (1) studies were reviews, case reports, letters, editorials, or conference abstracts; (2) articles not written in English; (3) studies with duplicate data; (4) missing key information for evaluating the HR and its 95%CI; (5) studies based on cancer cells and animal models and irrelevant studies. The candidate articles were assessed by two reviewers independently. Any disputes were resolved through their discussion.
Data extraction and quality assessment Two reviewers independently extract data and complete quality assessment from the selected literature. We extracted data including: the first author name, year of publication, sample size, the gender of patients, histological type, lymph nodes metastasis, recurrence, the cut-off value of thrombocytosis, venous thromboembolism, follow-up time, and primary outcome. The Newcastle-Ottawa Scale (NOS) scoring system was used to assess the quality of selected articles 20 . Two reviewers independently evaluated the quality of each included study. The judges include three aspects of evaluation: selection, comparability, and outcome between the case group and control group. Studies with NOS scores ≥ 6 were considered as high-quality studies.

Statistical analysis
The meta-analysis was conducted by STATA 12.0 software (Stata Corp., College Station, TX, USA). HRs and corresponding 95% CI were used to analyse the association between platelet count and lung cancer. The Cochrane's Q test and I2 statistic were used to evaluate the heterogeneity among the included studies 21 . I2 >50% or the P-value<0.05 defined heterogeneity in the studies 22,23 , and a random-effect model was adopted. Otherwise, a fixed-effect model was used. Moreover, subgroup analysis was conducted to detect the potential source of heterogeneity. A P value less than 0.05 indicated statistically significance. Publication bias was evaluated by Begg's test and Egger's regression test 24 . Additionally, a sensitive analysis was performed to check the stability of the results 25 .

Study characteristics
A flow diagram demonstrating the search procedure was illustrated in Figure 1. After the original search, 2395 records were retrieved from electronic databases. Firstly, we removed the duplications, and 1178 were left. Among them, another 1112 items were also excluded by titles and abstracts examination. Next, the remaining 66 full texts were left for eligibility. Of these, 27 studies were excluded on account of duplicate date and incomplete data. Ultimately, 39 studies including a total of 16696 participants met the criteria and were enrolled in this meta-analysis 17-19,26-61 . Simultaneously in data extraction, if the study provides both univariate analysis and multivariate analysis results, we take the results of multivariate analysis, because multivariate analysis excluded the correlated confounding factors, which are more accurate.
The characteristics of patients included are presented in Table 1. All included studies were published between 1976 and 2017. As in Table 1, 39 studies were included in the meta-analysis, 29 studies with NSCLC, 2 studies with SCLC, and 8 studies include all types. Twenty-two studies were performed in Asians and 16 in Caucasians, while one study did not report the race. In terms of the cut-off value of platelet count, 7 studies conducted a cut-off value of <300, while 18 studies took 300-400 as cut-off point, and the remaining 13 studies considered ≥400 as the cut-off point. There were 39 studies evaluating the association between OS and platelet count, while 8 studies evaluated the DFS/PFS/TTP outcome. All of the 39 studies reported the HR and 95% CI directly. Additionally, the quality of the studies was assessed by NOS as shown in Table 2.

OS
All of 39 studies including 16696 patients provide data on the prognostic role of the platelet count on OS in lung cancer. The results indicate that elevated platelet count was associated with poorer OS in lung cancer patients (HR = 1.47,95%CI: 1.31-1.66, P<0.001, Fig. 2A). Then we conducted the subgroup analysis to further investigate and the results were summarized in Table3. In the subgroup stratified by ethnicity, we observed that elevated platelet count predicted poor OS in Asian populations (HR = 1.54,95%CI: 1.32-1.8, P<0.001) while that in non-Asians had no significance (P = 0.063). Based on the clinic stage, an significant association between elevated platelet count and OS was found not only in stage I-III (HR = 1.52, 95%CI:1.22-1.89, P<0.001), but also in stage >III (HR = 1.7,95%CI: 1.26-2.29, P<0.001). The obvious association between elevated platelet count and OS was observed when integrating the data from 28 studies which OS was evaluated with multivariate analysis (HR = 1.47,95%CI: 1.31-1.66, P<0.001). In terms of the cut-off value, the subgroup analysis confirmed that increased platelet count was a negative predictor in patients with cut-off values<300 (HR = 1.64,95%CI: 1.25-2.15, P<0.1), and with cut-off values>400 (HR = 1.73,95%CI: 1.35-1.61, P<0.001) Additionally, high platelet count still predict worse OS in patients with lung cancer, regardless of subtype of lung cancer (SCLC or NSCLC).
The subgroup analysis was performed and the results were shown in Table 3. The results suggested that in the subgroup analysis such as Asian populations (P<0.001), multivariate analysis subgroup (P<0.001), stage III-IV disease subgroup (P<0.001), 300≤cut-off value< 400 subgroup (P<0.001), cutoff value>400 subgroup (P = 0.383), elevated platelet count was a negative predictor.

Publication bias and sensitivity analysis
As shown in Figure 3, the funnel plot was symmetrical. Based on the Begg's test (P = 0.866) and Egger's regression test (P = 0.376), no significant publication bias was found.
Furthermore, we performed a sensitivity analysis to evaluate the reliability of our results. The corresponding pooled HR values was not significantly impacted, indicating the robustness of our conclusions (Fig. 4).

Discussion
Cancer is undoubtedly one of the most serious public health problems. In the past few years, neuronspecific enolase, carcino-embryonic antigen, squamous cell carcinoma associated antigen and gastrin-releasing peptide precursor fragments have played an important role in the clinical diagnosis of lung cancer. However, their specificity and sensitivity in diagnosis are still not satisfactory.
Therefore, the exploration of new lung cancer markers is of great significance for clinicians to realize early detection and early treatment.
In recent years, it has been observed that some systemic inflammation indicators like neutrophil-tolymphocyte ratio (NLR) 62 , platelet-to-lymphocyte ratio (PLR) 63 , Glasgow prognostic score (GPS) 64 , PI (Prognostic Index) and PNI (Prognostic Nutritional Index) 65 played an important role in tumor genesis and development, which can be considered as predictors of prognosis. Since the 1960s, RICHARD B had observed that platelet count elevated in patients with cancers compared to those with nonmalignant diseases 5 . Accumulating evidence suggesting that elevated platelet is associated with various cancers such as colorectal cancer, lung cancer, endometrial carcinoma and so on 12,66,67 .
Platelet sustains proliferative signaling, resists cell death, and induces tumor angiogenesis 68 . Also, platelet activates TGF-β/Smad and NF-κB pathways, further promoting tumor migration and invasion 69 . Moreover, as the immune cell 70 , platelet releases TGF-β, reducing the expression of NKG2D, weakening the role of NK cells 71 . Platelet could be the prognosis predictor used in clinic.
Recently, several studies confirming the prognostic value of platelet count in lung cancer have been carried out, however the results were inconsistent. Therefore, we conduct the meta-analysis to determine the precise role.
We combine the outcomes of 39 studies with 16696 patients, suggesting that elevated platelet count is a poor predictor of OS and DFS/PFS/TTP in lung cancer patients. In our subgroup analysis, elevated platelet count is significantly associated with poor OS and DFS/PFS/TTP in diverse subgroups like in Asian populations, in stage I-III, and stage III-IV, in lung cancer patients whose score>6. However, the result is not significant in lung cancer patients score<6. The cut-off value of platelet count is disunity.
We found the elevated platelet count was significantly associated with poor OS and DFS/PFS/TTP when cut-off value of PLT was between 300 and 400, while the cut-off value of PLT >400 does not have relationship with poor DFS/PFS/TTP. Taken together, the cut-off value between 300 and 400 can separate patients well by OS and DFS/PFS/TTP, and should be used as a prognostic biomarker in clinic use., which was more precise than the findings of the previous meta-analysis 72 .Compared to the previous meta-analysis 72 ,our results is more comprehensive and accurate. On the one hand, we take 39 articles into the meta-analysis, which include more.new and important studies, increasing analytical capability of the analysis. On the other hand, more detailed subgroup analysis was performed. Except for the race and cut-off value, we also investigate the prognostic role of platelet count in different tumor stage, histology and quality score. Additionally, we discuss the association between PLT and OS and DFS/PFS/TTP, while the previous meta-analysis only studied the significance in OS.
However, there are some limitations of this study deserving mentioning. Firstly, the studies included in our meta-analysis are retrospective studies, more likely to have selection bias. Secondly, although publication bias and sensitivity analysis have confirmed the credibility, heterogeneity still existed in this meta-analysis due to several factors such as patients' characteristics, sample size, adjuvant therapy, which did not include in our analysis. Moreover, the cut-off value for definition of the elevated platelet count differed among the studies. Most of the studies used 300-400 as cut-off value, while several others use <300 or >400 as cut-off value of platelet count to assess the prognosis, which might lead to between-study heterogeneity. Last but not least, platelet count could be affected by several factors such as thrombosis, hypertension, splenic diseases, blood coagulation disorders, myeloproliferative disease, infection and drugs. Therefore, platelet count cannot play the role of prognosis if patients have these diseases above.

Conclusions
In conclusion, our meta-analysis reveals that pretreatment elevated platelet count is related to poor OS and DFS/PFS/TTP in lung cancer patients, and is an independent prognostic predictor of lung