Neutrophil-to-lymphocyte Ratio and Platelet-to-lymphocyte Ratio in the Prediction of Pancreatic Cancer Patients’ prognosis: a retrospective study CURRENT

Although neutrophil-to-lymphocyte ratio (NLR) and a high platelet-to-lymphocyte ratio (PLR) have been reported to be an inverse prognostic predictor of survival in patients with pancreatic cancer (PC), the comparison of their prognostic roles in patients with PC undergoing gemcitabine-based chemotherapy and 5-fluorouracil (5-FU) remains unclear. Methods This study was designed and performed to determine the predictive role of NLR and PLR in patients diagnosed with PC who underwent one of these two regimens. We retrospectively enrolled 95 patients diagnosed with PC undergoing supportive care, gemcitabine-based chemotherapy or 5-FU therapy from January 2015 to October 2018. Univariate and multivariate Cox regression analyses were done to identify clinicopathological predictors of time to treatment failure (TTF) and overall survival (OS), including pretreatment NLR and PLR.


Introduction
Pancreatic cancer (PC) is one of the most lethal cancers that remains a challenging medical problem for many years. The most common form is pancreatic ductal adenocarcinoma (PDA), the tenth most common solid cancer and the fourth leading cause of death from cancer in the United States.

]
Considering the poor prognostic outcome of patients with PC, efforts have been made to explore the predictive factors of this malignancy. It has been recognized that inflammatory responses play decisive roles at different stages of tumor development [ 4 ], and several inflammatory biomarkers have been proposed for the evaluation of cancer patients. Recently, pretreatment neutrophil-to-lymphocyte ratio (NLR) has been revealed to be a promising prognostic predictor for pancreatic cancer by several studies, where low NLR stands for better survival in patients with pancreatic cancer. [ 5 , 6 ] NLR is calculated by neutrophil count divided by lymphocyte count, which can be easily obtained by routine blood tests. Previous studies indicated that a high level of NLR might be significantly associated with poorer prognosis of several tumors, including colorectal cancer, lung cancer, gastric cancer, esophageal cancer and so on [7][8][9]. There are also evidences suggesting that platelet-to-lymphocyte (PLR), as another easily accessible biomarker that is calculated by platelet count divided by lymphocyte count, may also be a prognostic predictor of a variety of malignancies such as ovarian cancer [ 10 ], breast cancer [ 11 ] and non-small-cell lung cancer [ 12 ], where low PLR suggests a better outcome.
This study was to determine the association between both pretreatment NLR and PLR and survival of patients with PC treated with supportive care, gemcitabine-based chemotherapy or 5-FU therapy.

Methods
Patients and treatment Searching the database system of electronic medical charts, we collected clinical data of patients with the diagnosis of pancreatic cancer referred to Renji Hospital affiliated to Shanghai Jiao Tong University School of Medicine between January 2015 and October 2018. Patients whose diagnosis was PC were included in the study. Patients without available data or those with infection, regimens with steroids or aspirin, autoimmune disease or other conditions that might possibly confounded neutrophil count, lymphocyte count or platelet were excluded. In this study, patients underwent supportive care, gemcitabine-based therapy or 5-FU therapy. Gemcitabine-based therapy included gemcitabine monotherapy and gemcitabine combination therapy. Treatment was not terminated until tumor recurrence, tumor progression, treatment toxicity or patients' requirement for withdrawal. The doses and regimens were adjusted by corresponding physicians in accordance with adverse events and general conditions of the individual patient.
Clinical and laboratory data collection Patients included were either chemotherapy-naïve or chemotherapy-free for at least 1 month before they were referred to Renji Hospital affiliated to Shanghai Jiao Tong University. Baseline data was collected before treatment start. The history of patients was taken on the first days of hospitalization. Karnofsky performance status (KPS) was evaluated by treating physicians before the commencement of chemotherapy. Tumor location and TNM (tumor, lymph node and metastasis) staging were judged from the result of computed tomography (CT) or positron emission tomography (PET) in accordance with the 7 th edition of the AJCC Cancer Staging Manual. Biological markers, including carbohydrate antigen 19-9 (CA 19-9), carcinoembryonic antigen (CEA), carbohydrate antigen-125 (CA-125), neutrophil count, lymphocyte count and platelet count were collected within 3 weeks before the first cycle of chemotherapy as the pretreatment data.
NLR was defined by the absolute neutrophil count divided by the absolute lymphocyte count. Likewise, PLR was defined by the absolute platelet count divided by the absolute lymphocyte count. All consecutive parameters were categorized for the further analysis as follows: age (≤65 or >65 years), body mass index (BMI) (≤18.5, 18.5-24.0 or >24.0), CA 19-9 (≤1000 or >1000 U/ml), CEA (≤5 or >5 ng/ml), CA-125 (≤38 or >38 U/ml). Cut-off values were set on the basis of previous studies.[ 13 , 14 ] Statistical analysis The χ 2 test or Fisher's exact test were used to compare baseline patient characteristics that were categorized variables. The association between pretreatment NLR and the time to treatment failure (TTF) as well as overall survival (OS) was evaluated, so was pretreatment PLR. TTF was defined as the time from the date of chemotherapy initiation to the date of termination due to various reasons, including tumor recurrence, tumor progression, treatment toxicity or patients' requirement for withdrawal. If a patient had not reached the endpoint caused by any of these reasons, TTF was censored at the time of the last follow-up. Recurrence and progression were determined using CT or PET. OS was calculated from the date of chemotherapy initiation to the date of death for any reason, or censored at the date of the last follow-up if the endpoint event was not observed.
Univariate and multivariate Cox regression analyses were done to identify clinicopathological predictors of TTF and OS, including age, gender, BMI, KPS, personal history, diabetes at diagnosis, tumor location, TNM stage, CA 19-9, CEA, CA-125 and pretreatment NLR and PLR. The differences of TTF and OS were compared utilizing the Kaplan-Meier method with log-rank tests for survival plot depiction and Cox-regression analysis for the evaluation of hazard ratio (HR) and its 95% confidence interval (95% CI).
In order to set the cut-off points of both NLR and PLR, Receiver Operating Curves (ROC) were depicted. The classification variable was long vs short-term survival (>6 vs ≤6 months). For one thing, the choice of the 6 months as the division point was due to the convenience of the study, since a relevant proportion of patients (47.4%) were classified as short-term survivors and only 4 patients had a follow-up less than 6 months. For another, the majority of patients included in this study was in stage IV, whose median survival was only 4-6 months [ 15 ], so the choice of the 6 months was feasible. The effects of potential prognostic predictors were tested by Cox regression. Only variables with a statistical significance in univariate analysis were investigated in multivariate analysis.
The statistical significance of all tests was two-sided, p<0.05. Analyses and calculation were done by IBM SPSS Statistics 24.0.

Discussion
NLR has been confirmed to be a prognostic predictor in various cancers that significantly correlates with response rates, therapeutic effects and survival rates.[ 7-9 ] For patients with pancreatic cancer, high pretreatment NLR is found to be an unfavorable predictor of OS and TTF, and some studies pointed out that patients with high pretreatment NLR would tend to have a longer OS or PFS if their NLR was lowered after the initiation of chemotherapy. [ 16 , 17 ] In the present study, we demonstrated that pretreatment NLR independently predicted the prognosis of patients with advanced or localized PC, even after adjusting for potential confounding factors. But PLR failed to independently predict the prognostic outcome in term of both TTF and OS.
A number of studies have suggested the inverse association between both NLR and PLR and prognosis of patients with PC. [ 6 , 18 ] Although it has long been established that NLR and PLR both play a role in predicting the prognosis of various malignancies, the question which one is more predictive and suggestive is still debatable. The comparison between NLR and PLR was done to analyze which factor does better in predicting the prognosis of patients with PC.
Inflammation is a hallmark of various cancers.

] The association between thrombocytosis and PC is still not well understood.
Platelets can secrete tumor growth factors, such as VEGF, TGF-β, platelet-derived growth factor, and insulin-like growth factor-1, which play a critical role in cancer angiogenesis and metastasis. [ 24 ] In this study, NLR did provide independent prognostic information of survival in term of both TTF and OS, whereas PLR was not significantly associated with TTF or OS. Several studies have compared the predictive value of NLR and PLR in term of survival, response rates to treatment and recurrence after resection. NLR was recognized as a promising prognostic predictor whereas the association of PLR with patient outcome seems to be controversial. We recognize several limitations in our study. First, the research was retrospective with a relatively small sample size. Besides, the analysis of NLR and PLR for predictive values of different chemotherapeutic modalities was achieved by dividing further the population into smaller chemotherapy subgroups. Moreover, the major drawback is that treatment assignment was not randomized. Finally, blood samples were not derived during follow-up, making it infeasible to analyze whether further change of NLR and PLR would predict prognosis more accurately.

Conclusions
In conclusion, pretreatment NLR is a promising independent outcome predictor for patients with pancreatic cancer. The predictive value of PLR might not be as good as NLR.   Figure 1 Receiver Operating Curve (ROC) analysis constructed to find the best cut-off point of neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR).