Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal malignancies worldwide (1). Most PDACs are unresectable at the time of diagnosis and are treated with palliative chemotherapy. Gemcitabine monotherapy was the standard treatment for PDAC until 2011; however, combination chemotherapy regimens such as FOLFIRINOX and gemcitabine plus nab-paclitaxel therapy have become standard treatments in recent years and have improved the prognosis of PDAC patients. However, it is sometimes difficult to determine the most appropriate regimen for each PDAC patient. The prediction of patient prognosis is useful for providing proper treatment.

In advanced cancers, inflammation is associated with tumor proliferative activity. Nutritional state is also important, because patients with advanced cancer tend to become malnourished due to cachexia. Malnutrition leads to poor prognosis and decreased quality of life. Some inflammatory and nutritional scores, such as neutrophil-to-lymphocyte ratio (NLR) (2–4), platelet-to-lymphocyte ratio (PLR) (5), prognostic nutritional index (PNI) (6–8), modified Glasgow prognostic score (mGPS) (9,10), and controlling nutritional status (CONUT) score (11,12), have been reported to be useful in estimating the prognosis of patients with advanced cancers. The definition of each score is shown in Fig. 1. We previously reported that NLR is useful for predicting the outcome of gemcitabine therapy in unresectable PDAC (13). However, which marker is the most reliable for predicting the prognosis of PDAC patients treated with chemotherapy, including several combination regimens, remains unknown.

The present study aimed to detect the most useful marker for predicting the prognosis of unresectable PDAC patients treated with chemotherapy.

This retrospective observational study reviewed the data of unresectable PDAC patients, including those with locally advanced and metastatic disease, who underwent palliative chemotherapy between September 2006 and August 2016 at Fukushima Medical University Hospital (Fukushima, Japan). All patients were pathologically diagnosed with PDAC, and patients with rare primary pancreatic neoplasms, including acinar cell carcinoma or neuroendocrine carcinoma, were excluded. The chemotherapy regimen (FOLFIRINOX, gemcitabine plus nab-paclitaxel therapy: GnP, gemcitabine, or S-1) was decided by each patient's physician. All clinico-pathological data were measured prior to treatment.

The progression-free survival (PFS) and overall survival (OS) times were calculated from the date of histological diagnosis to the date of disease progression and mortality by any cause, respectively. We applied receiver operating characteristic (ROC) curve analysis to determine ideal cut-off values to predict poor prognosis for the following continuous variables: age, tumor diameter, carcinoembryonic antigen (CEA), cancer antigen 19-9 (CA19-9), NLR, PLR, PNI and CONUT score. The association of each clinico-pathological parameter (age, gender, tumor diameter, disease stage (UICC7th), first-line chemotherapy, NLR, PLR, PNI, mGPS and CONUT score) with PFS and OS time was investigated. Survival analysis was performed using the Kaplan-Meier method with log-rank tests in univariate analysis. Multivariate Cox regression analysis was performed to determine the effect of clinico-pathological variables on survival time. The data were analyzed using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), a graphical user interface for R (The R Foundation for Statistical Computing, version 3.2.1) that was a modified version of the R commander (version 2.1–7). Differences with a P-value <0.05 were considered statistically significant (14). This study was approved by the Institutional Review Board of Fukushima Medical University.

We included 72 PDAC patients who met the inclusion criteria described above in the analysis. The characteristics and laboratory data of the 72 patients are shown in Table I. The median PFS and OS was 117 days (range, 10–781 days) and 244 days (range, 43–781 days), respectively.

We calculated the cut-off value of continuous variables that predicted the median OS (244 days) using ROC analysis. The cut-off value of laboratory data and inflammatory markers are shown in Table II.

Univariate analysis of PFS showed that disease stage (locally advanced vs. metastatic, P<0.001), first-line chemotherapy regimen (combination therapy vs. monotherapy, P<0.001), CEA (<5.0 vs. ≥5.0 ng/ml, P=0.043), NLR (<4.0 vs. ≥4.0, P<0.001), PLR (182.0< vs. ≥182.0, P=0.044), mGPS (0, 1 vs. 2, P<0.001) and CONUT score (3< vs. ≥3, P=0.034) were related to PFS. Among them, stage (P=0.036), first-line chemotherapy regimen (P<0.001), CEA (P=0.024), NLR (P<0.001) and mGPS (P=0.005) were independent prognostic factors for PFS in multivariate analysis (Table III).

Univariate analysis of OS showed that the stage (P<0.001), first-line chemotherapy regimen (P<0.001), CA19-9 (P=0.040), NLR (P<0.001), PLR (P=0.004), PNI (<45.2 vs. ≥45.2, P=0.013), mGPS (P<0.001) and CONUT score (P=0.016) were related to OS. Among them, first-line chemotherapy (P=0.007) and mGPS (P<0.001) were independent prognostic factors for OS according to multivariate analysis (Table IV).

The Kaplan-Meier analysis of mGPS revealed that the survival time of patients with mGPS 2 was significantly poorer than that of patients with mGPS 0 and 1 (Figs. 2 and 3).

We aimed to clarify the best inflammation-based marker to predict the prognosis of unresectable PDAC patients treated with chemotherapy. The results showed that mGPS was the most reliable inflammation-related marker to predict PFS and OS. To the best of our knowledge, this study is the first to clarify the most useful inflammatory marker to predict prognosis in cases of unresectable PDAC treated with palliative chemotherapy.

Several studies have assessed the association between inflammation-based markers and the prognosis of resectable and unresectable PDAC patients. In patients with resectable PDAC, preoperative NLR (15), PLR (16,17), GPS (18) and PNI 45 (7) were reported to be factors in early recurrence and poor prognosis after surgical resection. In unresectable PDAC, previous reports showed that patients with a high NLR had a shorter OS time (13,19). In terms of GPS, a retrospective study intended for PDAC patients treated with gemcitabine showed that GPS could predict poor prognosis (20). PNI has been shown to be related to prognosis and a systemic inflammatory response (i.e., NLR, PLR and serum TNF-α level) in PDAC (8).

There are many reports concerning the relation between each inflammation-based marker and the prognosis of PDAC, but there are few reports comparing the utility of several inflammation-based markers. Yamada et al (21) conducted a retrospective study to clarify which score could best reflect survival in resected pancreatic cancer patients and found that GPS was superior to the other markers. Additionally, a systematic review summarized past reports about PDAC and inflammatory markers and concluded that GPS and NLR were useful in predicting prognosis (22); however, unlike our study, this study included patients with various stages and treatments (e.g., surgical resection and chemotherapy).

mGPS, which most accurately reflected the prognosis of unresectable PDAC patients in our study, is based on C-reactive protein (CRP) and serum albumin levels. CRP elevation indicates systemic inflammation. Because the cytokines produced by tumors increase the inflammatory response, CRP is considered to reflect growth activity in tumors. Furthermore, the serum albumin level indicates patient nourishment state and is related to performance status. Unresectable PDAC patients often fall into an undernourished state because of a decrease in pancreatic exocrine function and cancer cachexia. This undernourished state worsens the prognosis by reducing the patient's quality of life and making the continuation of chemotherapy more difficult. In the original GPS, hypoalbuminemia (Alb <3.5 g/dl) or elevated C-reactive protein (CRP >10 mg/l) is defined as a score of 1 (23). The modified GPS, which defined hypoalbuminemia without elevated CRP as a score of 0, was proposed because the GPS score of 1 was most commonly due to elevated CRP (24). In this study, mGPS 2 patients had a significantly poorer prognosis than those with scores of 0 or 1 because of the high-grade malignancy and cachexic state of the tumors. Patients with a score of 1 tended to have a poorer prognosis than patients with a score of 0, but the difference was not significant. This result indicates that nourishment state plays a very important role in determining the prognosis of PDAC. Naturally, serum CRP levels can increase as a result of an infection rather than tumor activity. Therefore, we used blood data collected in the afebrile state just before the initiation of chemotherapy, but some cases might be complicated by occult cholangitis.

A limitation of this study was that it was a single-center study with a limited number of patients. The results should be validated in a large population across multiple clinical sites. In addition, the chemotherapy regimen was determined at the physician's discretion, which may have resulted in bias. Studies focused only on patients treated with the single regimen or on those treated with multidrug therapy are expected.

In conclusion, mGPS appears to be an independent prognostic factor for predicting the PFS and OS of unresectable PDAC patients. Predicting prognosis before the start of therapy may be helpful in choosing the optimal treatment.