Significance of baseline and change in neutrophil-to-lymphocyte ratio in predicting prognosis: a retrospective analysis in advanced pancreatic ductal adenocarcinoma

The neutrophil-lymphocyte ratio (NLR) has been reported to be associated with prognosis in several cancers. The objective of our study was to evaluate the prognostic role of baseline NLR and change in NLR (ΔNLR) in advanced pancreatic cancer underwent chemotherapy. Between January 2010 and June 2015, 132 patients underwent chemotherapy were eligible for assessment. Based on our patients’ data, the cut-off value of NLR was 2.78 according to receiver operating characteristic curve. We observed that a high level of baseline NLR (NLR > 2.78) was a poor prognostic factor for overall survival (multivariable hazard ratio [HR] = 2.648, P < 0.001). Increased NLR (ΔNLR > 0) after 2 cycles of chemotherapy was associated with higher risk compared to ΔNLR ≤ 0 (multivariable HR = 1.894, P = 0.007). Combining both NLR and ΔNLR factors, multivariate analysis showed a significant higher risk (HR = 5.817, P < 0.001) for patients with high baseline NLR and increased NLR after 2 cycles of chemotherapy compared to patients with low baseline NLR and ΔNLR ≤ 0. In conclusion, both baseline NLR and ΔNLR are independent prognostic predictors for patients with advanced pancreatic cancer underwent chemotherapy.


Results
Total 132 patients (85 male and 47 female) with histologically confirmed advanced pancreatic cancer who received chemotherapy were eligible for assessment (Fig. 1). The median age at diagnosis was 57 (95% CI: 41.6-69.7). By July 30, 2016, 116 (87.9%) patients passed away, and the median survival was 7.66 months (95% CI: 6.13-9.23). Of the 31 (23.5%) patients were locally advanced and 101 (76.5%) patients had distant metastasis. We used 1-year survival as the time point to generate the receiver operating characteristic (ROC) curve and determined the optimal cut-off value of 2.78 for baseline NLR with the area under the curve (AUC) of 0.634. We used baseline NLR and NLR after 2 cycles to determine ΔNLR and categorized patients into ΔNLR ≤ 0 group and ΔNLR > 0 group (Table 1).
Univariable and multivariable analysis was performed to investigate the prognostic role of baseline NLR and ΔNLR in pancreatic cancer patients underwent chemotherapy. Univariate analysis indicated male (P = 0.017), a high white blood cell (WBC) count after 2 cycles (P < 0.001), a high baseline NLR (P < 0.001) and ΔNLR > 0 (P < 0.001) were poor prognostic factors for overall survival (OS) in this study cohort (Figs 2,3). Age, location, stage at diagnosis, albumin, KPS, and WBC, platelets and hemoglobin at baseline, platelets and hemoglobin after 2 cycles were not significantly associated with prognosis ( Table 2).
Multivariate analysis indicated that a high level of baseline NLR was a solid poor factor for OS (hazard ratio [HR] = 2.648, 95% CI: 1.631-4.300, P < 0.001) adjusted for gender, KPS, stage at diagnosis, and WBC, platelets, and hemoglobin at baseline and after 2 cycles. Whereas HR of death was 1.894 (1.160-3.091) for patients with ΔNLR > 0 compared to ΔNLR ≤ 0. Based on our analysis, baseline NLR and ΔNLR were independent prognostic factors for pancreatic cancer patients (Table 3) (Fig. 4).

Discussion
On the basis of our data, we found the independent prognostic value of baseline NLR and ΔNLR in advanced pancreatic cancer. Several studies indicated the poor survival was associated with high baseline NLR of the cancers in lung 9, 10 , colon 11 , gastric 7 , breast 12 , ovarian 13 , prostate 14 and pancreas [15][16][17] . Our findings of baseline NLR was in consistent with previous studies. In addition, we found that HR of death were significantly higher in patients with ΔNLR > 0 compared to ΔNLR ≤ 0 (HR = 1.807, 95% CI: 1.202-2.714, P = 0.007). Baseline NLR and ΔNLR in patients underwent chemotherapy appear to be of significant clinical value. Combining both baseline NLR NLR and other similar indicators, such as C-reactive protein, albumin, the Glasgow prognostic score, thrombocytosis and PLR may reflect systemic inflammation [18][19][20] . Until now, several theories regarding these facts have been made. A high NLR goes along with either a larger neutrophil amount or a relatively lower number of lymphocytes.
First, literatures showed neutrophils can irritate angiogenesis and suppress the immune system for anti-tumour activities to induce tumour growth 2,21 . Several studies indicated that neutrophils were involved in the release of various cytokines and chemokines including vascular endothelial growth factor (VEGF) and matrix metalloproteinase (MMP) 22 Gemcitabine and S1/capecitabine 15 (11.4) Gemcitabine and nab-Paclitaxel 13 (9.8) Gemcitabine and cisplatin/oxaliplatin 6 (4.5) nab-Paclitaxel and S1 56 (42.5) al. 26 found that TGF-β within the tumour microenvironment induced a population of tumour-associated neutrophils (TAN) with a protumour phenotype. Lei Gong et al. 27 confirmed that neutrophils played a very important role in the promotion of lung cancer which was strongly mediated through the IL-8/CXCR2 pathway and release of neutrophil elastase and development of a type 2 protumour microenvironment. KRAS mutation was the most common genetic alterations in pancreatic cancer 28 . Ji et al. 29 indicated that activation of KRAS/RAF/MEK pathways could stimulate the accumulation of neutrophils, followed by intensive inflammatory response. Based on the above literatures, we think neutrophils could induce tumour growth and metastasis, and the increased number of tumour-associated neutrophils has been linked to poorer outcome in cancer patients. Similar findings in the literatures indicated that elevated neutrophils were significantly associated with larger tumour size and worse survival in patients with localized renal cell carcinoma 30 and nasopharyngeal cancer 31 .
Second, in vitro studies showed that the cytolytic activity of lymphocytes and natural killer cells was suppressed when co-cultured with neutrophils, and the extent of suppression was proportionally enhanced to the addition of neutrophils 32 . Pillay et al. 33 identified that neutrophil subpopulations could suppress T-cell proliferation by integrin Mac-1 and hydrogen peroxide. Lymphocytes were known to have a crucial role in tumour defense. Lymphocytes could induce cytotoxic cell death and inhibit tumour cell proliferation and migration 21,34 . Therefore, a reduced lymphocyte count indicated a weaker immune reaction against tumour cells. Fogar et al. 35 confirmed that lower total lymphocyte count in blood for patients with pancreatic cancer was associated with poor outcome. Therefore, the levels of baseline NLR and ΔNLR could reflect the inflammatory response and  immune status of the patients undergoing chemotherapy, and theoretically both factors could be predictors for patients' prognosis.
The limitations of this study included: (a) this was a single center retrospective study; (b) there were different chemotherapy regimens; (c) we had included limited sample size; and (d) the study was conducted only in the Chinese population.
In conclusion, we found both baseline NLR and ΔNLR were independent prognostic factors for patients with advanced pancreatic cancer underwent chemotherapy. Our results suggest that evaluation of baseline NLR and ΔNLR is helpful in prognosis prediction, drug dose adjustment and inflammatory status assessment. Further validation in a prospective study is warranted.

Methods
Patients. This was a retrospective study approved by the ethics committee of Chinese People's Liberation Army (PLA) General Hospital. From January 1, 2010 to June 1, 2015, patients with advanced pancreatic cancer admitted for chemotherapy were included for analysis. Prior to chemotherapy initiation, written informed consent was reviewed and signed by the patients or their legal guardian. All relevant blood tests and treatments were performed based on institutional guidelines and regulations. Clinical data were retrieved and collected for retrospective analysis from the medical records of PLA General Hospital database electronically.  The inclusion criteria were: (1) patients were cytological or histologically confirmed pancreatic cancer and not eligible for operation; (2) patients received at least 2 cycles of chemotherapy (Please see Supplementary Table S1 for detailed information about chemotherapy regimens); (3) sufficient bone marrow function; (4) normal hepatic and renal function; (5) without targeted therapy or other biologics; (6) patients with a KPS score of 70 or more (on a scale from 0 to 100, with higher scores indicating better performance status); (7) no history of previous chemotherapy for advanced disease or adjuvant therapy within one year; (8) no radiotherapy. Exclusion criteria: (1) incomplete data of toxicities; (2) lost follow-up. Total 132 patients were eligible for analysis. Follow-up evaluations were performed every 3 months. Dates of death were obtained from the China disease prevention and control information system or telephone calls follow-up. Medical records were reviewed, and the cause of death was decided by investigator. Lost follow-up refers to the patient who was out of contact. We followed up until July 30, 2016 to obtain clinical and outcome information.
Data collection. All relevant clinic-pathological data were retrieved from patient medical records.
Laboratory data, including neutrophil and lymphocyte, WBC, platelets, hemoglobin were obtained within 1 week before chemotherapy and after 2 cycles of chemotherapy. The absolute neutrophil count was calculated by the percentage of segmented neutrophils out of the WBC count. The NLR was determined by the absolute neutrophil count divided by the absolute lymphocyte count. ΔNLR was calculated by subtracting the baseline NLR from the NLR after 2 cycles of chemotherapy (cycle 2-cycle 0). Overall survival was defined as the time from date of treatment to death. The primary study endpoint was OS. Censoring occurred if patients were still alive at last follow-up.
Statistical analysis. Data were presented as median (interquartile) for continuous variables, and as frequency or percentage for categorical variables. Survival curves were estimated using the Kaplan-Meier method and compared by the log-rank test. Multivariate survival analyses were performed using Cox proportional hazards regression models. We used the ROC curve to determine the best cut-off values for OS with baseline NLR. All of the analyses were performed with the statistical software packages R (http://www.R-project.org, The R Foundation). Statistical significance was defined as a two-sided P < 0.05.