Increased platelet-lymphocyte ratio closely relates to inferior clinical features and worse long-term survival in both resected and metastatic colorectal cancer: an updated systematic review and meta-analysis of 24 studies

Colorectal cancer (CRC) is one of the most common cancers worldwide. However, the prognostic and clinical value of platelet-lymphocyte ratio (PLR) in colorectal cancer was still unclear, which attracted more and more researchers considerable attention. We performed a systematic review and meta-analysis to investigate the relationship between PLR and survival as well as clinical features of CRC update to September 2016. The hazard ratio (HR) or odds ratio (OR) with 95% confidence interval (CI) were calculated to access the association. We included 24 eligible studies with a total of 13719 patients. Elevated PLR predicted shorter overall survival (OS) (HR=1.47; 95%CI, 1.28-1.68; p<0.001), poorer disease-free survival (DFS) (HR=1.51; 95% CI, 1.2-1.91; p=0.001), and worse recurrence-free survival (RFS) (HR=1.39; 95% CI, 1.03-1.86; p=0.03), but had nothing to do with Cancer-specific survival (CSS) (HR=1.14; 95% CI, 0.92-1.42; p=0.223). After trim and fill method, the connection between PLR and DFS disappeared (HR=1.143; 95%CI, 0.903-1.447; p=0.267). By subgroup analyze, we found that increased PLR predicated a worse OS and DFS in patients who underwent surgery, and this prognostic role also shown both in metastatic and nonmetastatic patients. In addition, elevated PLR was associated with poorly differentiated tumor (OR=1.51; 95% CI, 1.26-1.81; p<0.001), higher tumor stage (OR=1.25; 95% CI, 1.05-1.49; p=0.012), lymphovascular invasion (LVI) (OR=1.25; 95% CI, 1.09-1.43; p=0.001), and the recurrence of CRC (OR=2.78; 95% CI, 1.36-5.68; p=0.005). We indicated that pretreatment PLR was a good prognostic marker for CRC patients. High PLR was related to worse OS, RFS and poor clinical characteristics.


INTRODUCTION
Colorectal cancer (CRC) caused almost 700,000 deaths worldwidely every year, making it the world's fourth most deadly cancer [1]. It was the third most commonly diagnosed cancer in males and the second in females, with an estimated 1.4 million cases and 693,900 deaths occurring in 2012 [2]. The lifestyle changes in past low-CRC-risk countries resulted in rapid growth of colorectal cancer and the 5-year survival rate was still poor despite the progress of the treatment [3][4][5][6]. Pretreatment predicting indexes are in dire need to forecast potential of the tumor recurrence and prognosis for that the majority of the available prognostic markers are assessed postoperatively. The clinical and pathological Review www.impactjournals.com/oncotarget TNM stages, the number of resected lymph nodes (nLNs), carcino-embryonic antigen (CEA), the lymphovascular invasion (LVI), the perineural invasion, in addition to some molecular markers (eg. PinX1, RAS, BRAF, MMR and so on) have all been identified as prognostic factors [7][8][9][10][11], however several weaknesses limited their application in routine clinical practice, such as high costs, lack of standardization, low consistency, and poor reproducibility [7,12,13,14]. Therefore, finding a proper prognostic factors to assist coloractal cancer patients in guiding appropriate treatment to improve the therapeutic effectiveness is extremely urgent.
Prior studies showed that systemic inflammatory response (SIR) status played a vital role in tumor progression and therapeutic response [15][16][17][18][19]. The levels of platelets and lymphocytes represents the systemic inflammatory response (SIR), and they are easily obtained by widely used peripheral blood test. Platelet-lymphocyte ratio (PLR) as a combination of these two factors has been reported to be associated with poor prognosis in different tumor types, including CRC [20][21][22][23][24][25][26][27][28][29][30][31][32], but some other studies drew a different conclusion [33][34][35][36][37][38][39][40][41][42][43]. In brief, independent research results of the relationship between PLR and its impact on survival and clinical features were still inconsistent, partly due to limited published studies previously, various confounding factors and less detailed analysis. Recently, a lot of new studies on this issue were published continuously. Thus the aim of our study was to perform a systemic review and meta-analysis with all eligible current evidence to clarify this relationship and to

Literature search and study selection
A total of 194 relevant publications were initially retrieved. Of these, 29 duplicates were removed, 134 publications were excluded because the studies were animal experiment, literature reviews, comments, letters, or unrelated studies based on the titles and abstracts screening. After reading the full text, 7 publications were excluded due to irrelevant publications, studies with overlapping case series or lack of sufficient data for analysis, Therefore, a total of 24 publications with 13719 patients were included . All of these studies contained the required information and evaluated the correlation between PLR and the prognosis of CRC. Figure 1 presents a summary of the study selection process.

Study characteristics
The main features of the 24 selected studies are shown in Table 1. From the 24 studies, fourteen publications were originated from the Asian (eight from China, three from Korea, three from Japan), nine were performed in Caucasian population (four from the UK, one from Hungary, one from Canada, one from Australia, one from Austria, one from Italy), and one from USA with mixed races. The OS was investigated in 22 studies, the DFS rate was analyzed in 12 studies, the CSS rate was evaluated in 4 studies, and the RFS rate was reported in 2 studies. These eligible studies were published from 2012 to 2016. Thirteen of these directly provided HR in multivariate analysis, and survival data of nine studies were extracted from univariate analysis while survival data of two studies were extracted from survival curves. The extracted data in detail were presented in Supplementary  Table S1 and S2, while detailed NOS scores of each included study were presented in Supplementary Table S4.

PLR and clinical characteristics of CRC patients
In addition, we examined the association between PLR and the clinical parameters of colorectal cancer (  (Figure 4).

Sensitivity analysis and publication bias
To identify the source of heterogeneity across selected studies, sensitivity analysis was conducted by removing each study in turn from the analysis. The pooled  (Figure 5b).

DISCUSSION
This systematic review and meta-analysis, including 24 individual studies of 13719 patients, found that increased PLR was strongly associated with poor overall survival and recurrence-free survival in patient with colorectal cancer. However, PLR was unrelated to cancerspecific survival and disease-free survival after trim and fill method. The stratified analyses showed that elevated PLR was associated with poor outcome in both Caucasian and Asian population, univariate and multivariate analysis, metastatic and nonmetastatic CRC, and resected patients. However, we did not observe the significant association in nonsurgery subgroup for that the number of the included original studies in this subgroup is limited. For the negative study result subgroup, the HR of OS was 1.29 (95% CI, 1.05-1.59; p = 0.016), which meaned that after enlarging the sample size by meta-analysis, we overthrow the old conclusion that PLR had nothing to do with OS. Our finding confirmed the hypothesis that PLR was an appropriate prognostic factor for CRC patient survival.
Cancer progression and prognosis was determined not only by tumor characteristics but also by the host inflammatory response [44,45]. Using clinical, inflammatory, and molecular biomarkers as CRC prognostic factors are increasingly interesting, but there remained a lack of reliable, reproducible, and low-cost markers that can be readily incorporated into routine practice to optimally predict prognosis and guide treatment [31]. Some combinations of the inflammatory response parameters (eg. lymphocytes, neutrophils, platelets and acute-phase proteins, which are simple and easy to measure using standardized and widely used assays) including platelet-to-lymphocyte ratio (PLR), neutrophil-to-lymphocyte ratio (NLR), lymphocyteto-monocyte ratio (LMR) and albumin/globulin ratio (AGR), have been performed to evaluate the prognosis in various cancers, including CRC, and so on [27,33,[46][47][48]. There were many reasons for PLR's ideal prognostic role in CRC patient. Firstly, platelets secrete several tumor growth and angiogenic factors, which might influence tumor progression [49]. Secondly, while in antitumor reaction of the immune system, the CD8+ and CD4+ T-lymphocyte interaction among each other can induce tumor cell apoptosis, which can improve the survival of CRC patients for the chemotherapy efficacy [50]. These supported our finding that the PLR was a promising prognostic factors for the survival of CRC patients, which was consistent with previous meta-analysis [51][52][53][54], however, our study was to some extent superior to the previous studies because of much more included studied and patients, more detailed analyses and less limitation. We included all current eligible relative studies by systemic review and meta-analysis. We did subgroup analyses to explore the heterogeneity sources, besides we explored the relationship between PLR and the inferior clinical features. By analyzing clinical factors, we found the relationship between the increased PLR and the clinical characteristics of CRC patients. PLR tended to be higher in colon cancer than rectum which need further explanation. Poorly differentiated cancer always accompanied with elevated PLR, for that poorly differentiated tumour cells growing faster with angiogenic and tumour growth factors secreted by platelet cell, such as platelet factor 4 (PF4), thrombospondin, vascular endothelial growth factor (VEGF), transforming growth factor beta (TGF-β) and platelet-derived growth factor (PDGF) [36]. Moreover, platelets reflected the invasive potential of CRC and was closely associated with lymphovascular invasion (LVI) [55]. PLR was a good prognostic marker in mCRC patients, because several studies have shown that platelets induce circulating tumor cell epithelialmesenchymal transition and promote extravasation to metastatic sites [43,56,57]. Lymphocytes were involved in cancer immune surveillance which influenced the tumor recurrence to some extent [58]. Our study results indicated that the relationship between PLR and some clinical factors presented a new researching direction for future research. Moreover, the easily got PLR can be used to reflect some clinical characteristics which were difficult to obtain like tumour differentiation, lymphovascular invasion (LVI), recurrence and so on. Pretreatment blood test for PLR played a vital role in assessment of cancer characteristics and patients prognosis.
There were limitations in our systematic review. First, the included studies were almost retrospective studies and more studies with prospective design were warranted in future. Second, eleven enrolled studies applied univariate analysis only (without providing multivariate analysis data), while subgroup analysis showed the prognostic values of PLR existed in these studies. Moreover, there are significant heterogeneity existing in OS and DFS analysis. Therefore additional large cohorts of prospective studies are needed to correct for heterogeneity.
In conclusion, peripheral blood PLR was an effective prognostic marker for CRC patients. Elevated PLR was related to worse overall survival and recurrencefree survival, but not for disease-free survival and cancerspecific survival. The prognostic utility of PLR might help to guide use of individual therapies and patient counselling in future.

Search strategy
PubMed, Web of Science and Embase were searched from inception to September 2016. The search strategy used the keywords as follows: "PLR" or "platelet lymphocyte ratio" or "platelet to lymphocyte ratio" or "platelet-lymphocyte ratio" or "platelet lymphocyte" and "CRC" or "colon neoplasm" or "rectal neoplasm" or "colorectal neoplasm" or "colorectal tumor" or "colorectal cancer" or "colorectal carcinoma". There was no language restriction in our study. References of relevant studies and review articles were searched for potential eligible studies.

Inclusion and exclusion criteria
The inclusion criteria in this meta-analysis study were as follows: (1) studies investigated the relationship between PLR and colorectal cancer prognosis or clinical characteristics; (2) the PLR was obtained from a preoperative peripheral blood test; (3) adequate data were provided to measure odds ratio (OR) and hazard ratios (HRs) with 95% confidence intervals (CIs). The exclusion criteria were as follows: (1) unrelated studies, animal experiment, cell experiment, literature reviews, comments, letters, meta analysis, or case reports; (2) studies without sufficient data for analysis; and (3) duplicated publications. When studies with overlapping cases were met, the study with the larger number of patients was included. www.impactjournals.com/oncotarget

Data collection and quality assessment
Relevant datas were professionally extracted by two authors independently, and disagreements were resolved through discussion with a third author. Data collected from each study included first author, publication year, country and ethnicity of the study participants, number of patients, tumor characteristics (stage, location, size, differentiation, lymphovascular invasion, treatment, recurrence), cut off value for high or low PLR, and survival data (OS/DFS/ CSS/RFS). If some publications provided survival data by Kaplan-Meier curves indirectly, Engauge Digitizer version 4.1 was applied to extract the data. The quality of included articles were assessed using the Newcastle-Ottawa Scale (NOS) by two authors independently (Supplementary Table S2 showed the Newcastle-Ottawa quality assessment scale). The total scores of NOS ranged from 0 to 9, with higher scores indicating better quality. A high-quality study was defined as the study with ≥6 points on NOS.

Statistical analysis
According to the cut-off values, PLR was devided into high or low level groups in each study, and the hazard ratio with the 95% confidence interval (high vs low level of PLR) were used to evaluate the relationship between PLR and long-term prognosis (OS/DFS/CSS/ RFS). Odds ratio and 95%CI were pooled to access the role of PLR on clinical features of colorectal cancer. Statistical heterogeneity was evaluated by Q and I 2 tests, and if the p-value < 0.1 or I 2 > 50%, which suggested the existence of substantial heterogeneity, thus we used a random-effect model to calculate the pooled estimate. Otherwise, the fixed-effect model would be applied instead. The subgroup analyses were applied to explore the heterogeneity sources. Publication bias was evaluated using the Egger's weighted linear regression and Begg's regression method. A trim and fill method was used when significant publication bias existed. All statistical tests were two-sided, and a p value < 0.05 was considered to be statistically significant. All analyses were conducted by Stata 14.0 (STATA Corporation, College Station, TX, USA).