Neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, and their dynamic changes during chemotherapy is useful to predict a more accurate prognosis of advanced biliary tract cancer

Background and Purpose Systemic inflammation is known to promote carcinogenesis in biliary tract cancer (BTC). Neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR) are indicative of systemic inflammation. We evaluated the clinical significance of systemic inflammation measured by NLR and PLR in patients with advanced BTC. Additionally, we also co-analyzed the dynamics of NLR and PLR during chemotherapy. Methods We reviewed 450 patients with unresectable BTC receiving palliative chemotherapy. NLR and PLR were obtained before initiation of palliative chemotherapy. Changes in NLR, PLR were obtained by subtracting the initial value from the value obtained after progression of chemotherapy. Results Higher systemic inflammation status also had relation with a primary tumor site (p = 0.003) and higher levels of CEA (p = 0.038). The ROC cut-off values of NLR and PLR for predicting overall survival (OS) were 3.8 and 121, respectively. Patients with a high NLR or PLR had worse OS independently in multivariate analysis (6.90 vs. 9.80 months, p =0.002; 7.83 vs. 9.90 months, p =0.041, respectively). High NLR with increased NLR after chemotherapy is associated with worse OS and progression-free survival (PFS) (p < 0.001, p = 0.013 respectively). Results are similar for PLR. Conclusion Systemic inflammation represented by NLR and PLR, predicts the OS of patients with advanced BTC who are receiving palliative chemotherapy. In addition, considering NLR/PLR with a dynamic change of NLR/PLR during chemotherapy might help to predict a more accurate prognosis.


INTRODUCTION
Biliary tract cancers (BTCs) are tumors that are rare and have a poor prognosis. In Western countries, BTCs comprise less than 2% of all malignancies [1], but in Korea, they are more common, accounting for approximately 4% of malignancies [2]. Most patients who undergo surgical resection for BTC, generally experience a recurrence after a short disease-free period [3]. Overall survival (OS) in patients with advanced or metastatic BTC is less than 1 year [4]. The known prognostic factors include albumin, Eastern Cooperative Oncology Group (ECOG) Performance Status (PS), and serum carcinoembryonic antigen (CEA) level [5]; however, there is a lack of prognostic factors that can accurately predict the prognosis of BTC.
Chronic inflammation is known to promote carcinogenesis [6], which is triggered by cytokine or
Although many biomarkers of systemic inflammation have recently been studied, excessive costs and technique factors prevent their clinical use. The most commonly investigated indicators of inflammation are those that are clinically most cost-effective and available, such as C-reactive protein (CRP) levels, platelet counts, and blood neutrophil counts [11]. Poor prognostic factors in various cancers include high neutrophil-to-lymphocyte ratio (NLR), high platelet-to-lymphocyte ratio (PLR), increased levels of neutrophils and platelets, and decreased levels of lymphocytes [12]. However, in BTC the role of NLR and PLR has  [13].
In the case of advanced BTC, little is known about the significance of NLR and PLR.
In this study, we investigated the prognostic role of NLR, PLR and CRP in patients with advanced BTC who received palliative chemotherapy in terms of response to chemotherapy, progression-free survival (PFS), and overall survival (OS). Additionally, we also co-analyzed the dynamics of NLR and PLR during chemotherapy.
The area under the receiver operating characteristic (ROC) curve for NLR was 0.628 (95% CI: 0.573-0.683), and an NLR value of 3.8 was identified as the cut-off value for predicting OS, with a sensitivity of 52.3% and a specificity of 68.4%. The area under the ROC curve for PLR was 0.592 (95% CI: 0.536-0.648), and a PLR value of 121 was identified as the cut-off value for predicting OS, with a sensitivity of 78.4% and a specificity of 33.2%. Additionally, the area under the ROC curve for CRP was 0.552 (95% CI: 0.477-0.628), and an CRP value of 3.3 was identified as the cut-off value for predicting OS, with Abbreviations: ICC, intrahepatic cholangiocarcinoma; BTC, biliary tract cancer; GB Ca, gallbladder cancer;AoV Ca, ampulla of Vater cancer; ECOG PS, Eastern Cooperative Oncology Group performance status;CEA, carcinoembryonic antigen; CA 19-9, cancer antigen 19-9; NLR, neutrophil-lymphocyte ratio; PLR, platelet-lymphocyte ratio; CRP, C-reactive protein; ORR , objective response rate; PR, partial response; CR, complete response; SD, stable disease; PD, progression disease; OS, overall survival; PFS, progression-free survival; 95% CI, 95% confidence interval www.impactjournals.com/oncotarget 49.1% sensitivity and 50.9% specificity. The patients were separated into 2 groups according to the cut-off values for NLR and PLR (low: <3.8 or high: ≥3.8; low: <121 or high ≥121, respectively). Our study showed a strong correlation for NLR and PLR (r=0.516, p < 0.001) ( Figure  2).However, there was no correlation between NLR and CRP (p = 0.711) (Supplementary Figure S1A). Also, there was no correlation between PLR and CRP (p = 0.452) (Supplementary Figure S1B).

Comparisons between high and low NLR groups
The patients with elevated CEA, low albumin and histologic poorly differentiation were more frequently found in the high NLR group (p = 0.038, p < 0.001, p = 0.029, respectively) ( Table 2). In contrast, primary tumor site, the response rate, CRP were not different between high and low NLR groups (p =0.162, p =0.214, p = 0.683). Patients with high NLR showed worse PFS and OS compared with patients with low NLR (3.23 months vs 4.57 months, p = 0.028; 6.90 months vs 9.80 months, p < 0.001).

Comparisons between high and low PLR groups
Unlike NLR, there were differences between low PLR group and high PLR group according to primary tumors (p = 0.003) ( Table 3). Also, our study showed CEA, albumin, histologic differentiation had no differences between high and low PLR groups (p =0.075, p = 0.052, p =0.748, respectively). Patients with high PLR showed worse PFS and OS compared with patients with low PLR (4.10 months vs 4.73 months, p =0.091; 7.83 months vs 9.90 months, p = 0.012).

NLR, PLR, and other prognostic factors for PFS
In multivariate analysis, a high NLR together with an increased PLR after chemotherapy, CEA and CA 19-9 levels, and response rate were significant prognostic factors for PFS (Table 5, Figure 4).

Analysis of NLR and PLR with co-analysis of dynamics of NLR and PLR during chemotherapy
Analysis of NLR in patients in whom NLR changes had occurred (low NLR with maintained NLR, low NLR with increased NLR, high NLR with maintained NLR, high NLR with increased NLR) revealed that patients with high NLR with increased NLR had the shortest survival (p <0.001; Table 6). Compared with a median OS of 11.10 months for patients with low NLR with maintained NLR, patients with high NLR with increased NLR had the worst OS (median OS 4.73 months; HR 3.113; p < 0.001; Table  6, Figure 5). In addition, compared with a median PFS of 4.57 months for patients with low NLR with maintained NLR, patients with high NLR with increased NLR had the      Figure 6). Analyses according to PLR showed similar results.

DISCUSSION
In this study, we found that systemic inflammation represented by NLR and PLR predicts the overall survival of patients with unresectable BTC who are receiving palliative chemotherapy.
Tumor metastasis depends on interactions between tumor cells and the host microenvironment. This involves blood cells, components of the coagulation system, stromal cells, and the extracellular matrix. Cells within the bloodstream include endothelial cells, platelets, lymphocytes, macrophages, mast cells, and fibroblasts [14]. These cells produce cytokines, which are cytotoxic mediators like reactive oxygen species (ROS), tumor necrosis factor-alpha (TNF-α), and interleukins, leading to tumor progression [15]. High neutrophil counts have long been reported to negatively affect the prognosis of various other cancers [11]. The neutrophils release ROS, which causes point mutations and DNA damage [16]. Also, neutrophils release neutrophil elastase, which is related to tumor cell proliferation, VEGF-related angiogenesis [17]. Through these mechanisms, neutrophils promote proliferation, invasion, and angiogenesis in cancer. Furthermore, T cells in tumors induce an immune response to the lesion, which includes a high number of cytotoxic T lymphocytes and increased neoplastic cell apoptosis [18]. Lymphocytopenia suppresses the immune response due to a marked decrease in T4 helper lymphocytes and an increase in T8 suppressor lymphocytes [19].
Platelet count is an additional index of systemic inflammation caused by the tumor. Platelet adhesion and aggregation leads to the formation and release of platelet granules that contain active proteases, growth factors, matrix proteins, and chemokines that enhance tumor progression [20]. Also, platelets release many factors, such as angiopoietin 1, epidermal growth factor, basic fibroblast growth factor, and interleukin-1β and IL-8 cytokines,  which regulate the angiogenic process [21].
Therefore, NLR and PLR are considered to be the balance between inflammatory status and antitumor status. Patients with elevated NLR and PLR have relative lymphocytopenia, neutrophil leukocytosis, and thrombocytosis, which cause pro-tumor inflammation. Our study showed that NLR is associated with the serum albumin level. Scheede-Bergdahl et al. have reported an association between cancer cachexia and interleukin (IL)-1b levels [22], and Tisdale et al. have also reported that cachexia is associated with cytokines, including tumor necrosis factor-a (TNF-a), IL-1, IL-2, and IL-6, and interferon-r (IFN-r). In light of these findings, NLR could be used as a predictive marker of cancer cachexia. NLR is also significantly associated with CEA, which is an indicator of tumor burden [23]. This suggests that the greater the tumor burden, the greater the NLR, and this is consistent with reports that there is a significant association between higher NLR, larger tumor size, and the extent of micro-vascular invasion in cases of resectable BTC [24]. In addition, NLR is significantly associated with tumor cell differentiation.
In contrast with NLR, PLR varies in BTC according to tumor origin (p = 0.003). Compared with other primary tumors, ampulla of Vater cancer (AoV Ca) has a greater distribution of lower PLR values. Thus, differences in overall survival (OS) according to the primary tumor in BTC might be associated to some extent with PLR. PLR also had a slight association with CEA, CA 19-9, and albumin levels and with tumor cell differentiation.  Considering the above, when we look at both NLR and PLR, we can more accurately predict the prognosis of BTC.
In a previous report about NLR as a prognostic factor in advanced BTC [25], an association was reported between NLR and OS, but no association was found between NLR and PFS. In our study, an association was found between NLR OS and PFS. Interestingly, in our study, no relationship existed between changes in NLR or PLR. However, when we analyzed subgroups according to NLR and dynamic change in NRL during chemotherapy, our study reveals that high NLR with increased NLR after chemotherapy was associcated with a worse survival outcome (Table 6). There was a 6.4 month OS difference between high NLR with increased NLR group and low NLR with maintained NLR group. The results for PLR are similar. Furthermore, our study reveals that high NLR with increased NLR after chemotherapy is associcated with a worse PFS. So, in light of the above information, considering NLR and PLR along with change might help to more accurately predict heterogeneous prognoses in BTC.
Currently, no consensus exists regarding NLR and PLR cut-off values, which vary across tumor types [12,26]. In the present study, NLR and PLR cut-off values of 3.8 and 121, respectively, were selected using ROC curves, and the same procedure was used as reported in other studies [26]. McNamara et al. demonstrated a median NLR cut-off value of 3 as a prognostic marker in advanced BTC [25]. Dumitrascu et al. demonstrated a median NLR cut-off value of 3.3 as a prognostic marker in resectable BTC [13].
Recently, there have been several reports showing that circulating tumor cells are associated with prognosis [23,27], and in our previous published reports, the soluble form of programmed death-ligand 1 (PD-L1) in peripheral blood was associated with prognosis in advanced BTC [28]. Further studies to find the association of NLR with circulating stem cells might be worthy.
A limitation of the present study is its retrospective design. Another caveat is that NLR is known to be a nonspecific marker of inflammation. Therefore, the presence of another systemic disease could have influenced the NLR value.
In conclusion, systemic inflammation represented by NLR and PLR predicts the OS of patients with advanced BTC who are receiving palliative chemotherapy. In addition, considering NLR and PLR with dynamic changes of NLR and PLR during chemotherapy might help to predict a more accurate prognosis in advanced BTC.

Study population and data collection
We consecutively enrolled patients with advanced BTC who were diagnosed at the Seoul National University Hospital (Seoul, South Korea) between 2003 and 2013. All patients fulfilled the following criteria: (1) Histologically confirmed biliary tract cancer, (2) Unresectable, metastatic biliary tract cancer, and (3) Had received at least one cycle of palliative chemotherapy, (4) The recurrent biliary tract cancer patients were excluded.
The medical records of each patient were reviewed with respect to laboratory complete blood count (CBC) at the time of diagnosis, sex, age, tumor origin, localization, tumor marker, PFS, and OS.
The NLR was defined as the absolute neutrophil count in peripheral blood divided by the absolute lymphocyte count. The PLR was defined as the absolute platelet count in peripheral blood divided by the absolute lymphocyte count. Changes in NLR, PLR were obtained by subtracting the initial value from the value obtained after progression of chemotherapy.

Statistics
A receiver operator characteristic (ROC) curve was used to assess the best discriminative cut-off values of NLR, PLR, and CRP while showing an association between NLR, PLR, and OS. The chi-square test was used to compare baseline characteristics of patients between the NLR and PLR groups. Overall survival (OS) was calculated from the first day of diagnosis of advanced BTC to the date of death, and progressionfree survival (PFS) was calculated from the first day of palliative chemotherapy to the date of progression or any cause of death. We analyzed the OS of all patients in this study. The OS and PFS were assessed using Kaplan-Meier estimates. The log-rank test was used to assess the equality of the survivor function across groups. The Cox proportional hazards model with 95% confidence interval was used for the multivariate analysis to assess the effect of patient characteristics and other significant prognostic factors such as NLR and PLR. A p value of 0.05 or less was considered statistically significant. All analyses were performed using SPSS software for Windows (version 21; IBM SPSS, Somers, NY).

Ethics
The study protocol was reviewed and approved by the Institutional Review Board of Seoul National University Hospital (IRB No.H-1306-109-500). All studies were conducted according to guidelines (Declaration of Helsinki) for biomedical research.