Imaging Finding Analysis and Prognosis Comparison of Pancreatic Head Cancer Invading the Peripheral Nerve Plexus Among Different Groups

Hongkun Ping Department of Radiology, Shanghai Jiao Tong University A liated Sixth People’s Hospital Nianhui Yu Department of Radiology, Shanghai Jiao Tong University A liated Sixth People’s Hospital Guang Tan Department of Radiology, Shanghai Jiao Tong University A liated Sixth People’s Hospital Lipeng Yang Department of Pathology, Shanghai Jiao Tong University A liated Sixth People’s Hospital Jiaqi Yu Department of Radiology, Shanghai Jiao Tong University A liated Sixth People’s Hospital Hui Li (  drlihui@sjtu.edu.cn ) Department of Radiology, Shanghai Jiao Tong University A liated Sixth People’s Hospital


Background
Patients with pancreatic cancer have an extremely poor prognosis, with the 5-year survival rate in these patients being approximately 6% [1]. Currently, surgery is the only way to improve the overall survival of these patients. However, even for patients who underwent surgery, the median survival is only 17 months [2]. Farrow et al. and Mitsunaga et al. reported that the most important factor resulting in poor postoperative prognosis was extrapancreatic perineural invasion (EPNI) caused by the characteristics of neurotrophic growth of pancreatic cancer, leading to postoperative recurrence [3,4]. Chang et al. reported that the prognosis of patients with pancreatic head cancer and EPNI/duodenal invasion was signi cantly poorer [5].
Pathologically, the peripheral nerve plexus around the head of the pancreas is divided into four pathways ( Fig. 1a). Pancreatic plexus 1 (PLX1) arises from the right celiac ganglion (CG) and passes behind the portal vein to the uncinate process of the pancreas (a typical neuroinvasive pathway for pancreatic head cancer). Pancreatic plexus 2 (PLX2) arises from the mesenteric ganglion (MG) near the superior mesenteric artery and extends along the posterior inferior pancreaticoduodenal artery (PIPDA)/jejunal trunk (JT) to the hook process. This is the most common invasive route for pancreatic uncinate process cancer. The root of the mesenteric pathway (ROM) involves PIPDA as in the case of PLX2. However, the tumor does not extend toward the MG but grows along the tail of the nerve bers toward the small intestinal mesentery. This is the second most common invasion route for pancreatic uncinate process cancer. The anterior neural pathway (AN) extends along the gastroduodenal artery (GDA), common hepatic artery (CHA), and hepatoduodenal ligament to the right CG. The PLX1, PLX2, and ROM pathways extend along the dorsal side of the pancreas and have an intersection point, the uncinate process of the pancreas. The AN is in the ventral side of the pancreas and does not pass through the uncinate process of the pancreas.
There is a consensus that the prognosis of pancreatic head cancer with EPNI is signi cantly poor [6-8].
However, we previously found that the imaging ndings and prognosis of all patients with pancreatic head cancer and EPNI differ slightly. Therefore, in this study, we aimed to elucidate the imaging features and prognosis of patients with pancreatic head cancer involving the peripheral nerve plexus by dividing

Methods
The preoperative MSCT images, surgical records, and pathological ndings of pancreatic cancer patients who underwent Whipple surgery in our institution from December 2014 to December 2019 were reviewed.
The study was approved by the institution's ethical review committee. The inclusion criteria were as follows: (1) Patients for whom enhanced MSCT scans were performed preoperatively in our institution and in whom the tumor was diagnosed to be a resectable pancreatic head tumor (Pancreatic adenocarcinoma, version 2.2014: featured updates to the NCCN guidelines). According to our previous research [9], the CT diagnostic criteria for arterial and venous (superior mesenteric artery/vein, portal vein, celiac artery, and common hepatic artery) invasion should be dealt with slightly differently.
(2) Whipple surgery in our institution and tumor was pathologically con rmed as pancreatic cancer with EPNI ( Fig. 1b).
On the other hand, the following patients were excluded: (1) Patient with other major diseases that may affect the prognosis, such as other cancers and kidney failure; (2) patients who died of other sudden factors such as a cerebral infarction or a car accident; (3) patients (n = 6) whose MSCT scans showed simultaneous invasion noted in groups A and B owing to the large size or the special growth site of the mass. Accordingly, 93 patients were screened, of whom 43 were included in group A and 50 in group B. The mean age of the patients in groups A and B was 67.3 years (range 36-84 years) and 65.5 years (range 38-81 years), respectively. MSCT Before CT, the gastrointestinal tract was lled with 500 ml of an aqueous solution. Images were obtained using a 256-row CT system (Revolution CT, GE Healthcare). A contrast agent was injected intravenously at a rate of 4 ml/s, and the total amount injected was 100 ml. Enhanced biphasic scanning consisting of breath-holding, arterial (35 s after injection) and intravenous (60 s after injection) phase scans with thicknesses of 0.63 mm was performed.
MSCT image interpretation, surgical review, and pathological outcome assessment

Statistical analysis
Univariate log-rank test was used for statistical evaluation. The imaging ndings of pancreatic head cancer were evaluated using the chi-square test and independent sample T test, and the prognosis of pancreatic head cancer in the two groups was evaluated by performing the Kaplan-Meier survival analysis. A P-value of <.05 was considered signi cant. All statistical analyses were performed using SPSS 25.0 software.

Results
In our study, surgical exploration and postoperative pathological assessments con rmed that all cases were resectable without obvious invasion of the main peripancreatic vessels. The accuracy, sensitivity and speci city of MSCT for diagnosing EPNI were 92%, 92.5% and 85.7%, respectively.
The radiographic manifestations of pancreatic head cancer involving the peripheral nerve usually show a mass-and-strand pattern (a mass with diameter greater than 2-mm diameter mass or strand-shaped soft tissue density connecting to the pancreatic head cancer) or a coarse reticular pattern (composed of thick reticular lines with less intermingled fat density) [10]. There was no evidence of any difference in the patterns of invasion of the peripheral nerve between groups A and B (Figure 2). In group B, the tumors invading the plexus often developed along more than one pathway, as shown in Figure 2b, whereby the tumors invaded both PLX2 and ROM as the nerve plexus was crisscrossing and complex. Furthermore, the pancreatic uncinate process cancer is more prone to multi-pathway invasion (PLX1, PLX2, and ROM) due to the close anatomical location, which was the important reason for their integration into group B.

Tumor size
The two groups showed no signi cant difference in the size of pancreatic head carcinoma (P = 0.882). The mean size of pancreatic head carcinoma in group A was 3.0±0.94 cm, while that in group B was 3.0±0.92 cm ( Table 1).

Common bile duct and main pancreatic duct dilatation
The two groups also showed no signi cant difference in the proportions of cases showing dilatation of the common bile duct and main pancreatic duct. The number of patients showing common bile duct dilatation in groups A and B was 24/43 (55.8%) and 29/50 (58%) respectively (P = 0.832). The number of patients showing main pancreatic duct dilatation in groups A and B was 33/43 (76.7%) and 37/50 (74%), respectively (P = 0.76) ( Table 1).

Duodenal invasion
Normal enhancement of the duodenal wall was interrupted, with uniform/nodular thickening of the duodenal wall, when the low-density pancreatic head tumor spread from the pancreas to the adjacent duodenal wall. Although the duodenal invasion in group A (26/43, 60.47%) showed an increasing trend in comparison with that in group B (24/50, 48%), the difference was not statistically signi cant (P = 0.229). (Table 1)

Discussion
Pancreatic cancer is one of the most aggressive human tumors. Complete surgical resection is the only known curable way for pancreatic cancer, but even patients who have undergone successful surgery often die of the disease due to metastases or relapses, which can be partly attributed to nerve invasion.
Peripheral nerve invasion can be attributed to different mechanisms, and to date, three theories have been proposed to explain this nding. (i) Jin et al. identi ed lymphatic vessels in the superior mesenteric artery peripheral nerve plexus by immunohistochemical staining, providing factual evidence for the theory that tumor cell in ltration in neurogenic PNI was transmitted through lymphatic channels [11]. (ii) The "low resistance theory" proposed in the 1930s identi ed three weak points in the fascialis-the nerve endings, the blood vessels entering the nerve, and the reticular bers entering the nerve-and suggested that cancer cells could easily invade and grow along the interfascialis space [12]. (iii) In 1994, Dale Bockman et al. observed high expression levels of transforming growth factor α (TGF-α) in nerve cells and high expression levels of epidermal growth factor receptor (EGFR) in cancer cells by electron microscopy, indicating the possibility of interactions between tumor cells and nerves [13]. Subsequently, numerous basic experiments gradually led to the theory of the neural microenvironment. This theory proposed that cells stimulated by cancer cells in the microenvironment could in uence the occurrence of PNI-related processes through autocrine or paracrine mechanisms [14][15][16]. For example, pancreatic cancer releases colony-stimulating factor (CSF-1) to recruit macrophages, and cancer cells stimulate macrophages to secrete glial cell-derived neurotrophic factor (GDNF) to enhance the invasive ability of cancerous cells [20].
Pancreatic cancer with nerve invasion has been reported to signi cantly reduce progression-free survival and overall survival. Recent imaging studies focusing on preoperative determination of nerve invasion in pancreatic head cancer have con rmed the accuracy of MSCT in the diagnosis of EPNI and its effect on prognosis [18][19][20][21][22]. However, even in pancreatic head cancer with nerve invasion, individual patients show distinct differences in prognoses. Therefore, the present study attempted to further investigate the invasion of different pathways of the nerve plexus by pancreatic head cancer.
Although the surgical methods for Whipple surgery have continued to improve over decades of development, screening patients who are eligible for resection is more important to ensure greater patient bene ts in the long run. Some scholars have proposed that surgical treatment is not recommended for patients who have been preoperatively diagnosed with EPNI/duodenal invasion of pancreatic cancer. However, since the peripheral nerve plexus invasion rate of pancreatic cancer can be as high as 53%-100% [23,24], categorization of all patients with peripheral nerve invasion as unresectable may deprive some patients of the opportunity to undergo essential surgery. Therefore, we tried to group pancreatic head cancers involving different nerve plexuses on the basis of the ndings of preoperative MSCT examinations, and preliminarily investigated the differences in prognosis among these groups to provide a theoretical basis for the subsequent exploration of treatment options for different groups.
From the embryological point of view, the pancreas originates from the ventral and dorsal pancreatic primordia, and a single pancreas is formed through the development and fusion of both primordia. The ventral pancreatic primordium forms the dorsal part of the head of the pancreas and the uncinate process of the pancreas, while the dorsal pancreatic primordium gradually develops into the ventral part of the head of the pancreas as well as the neck, body, and tail of the pancreas. The boundary between the ventral and dorsal pancreas, i.e., the line connecting the portal/superior mesenteric vein and the anterior edge of the intrapancreatic bile duct, can be roughly de ned on CT images (Fig. 4). The ventral pancreas lies on the dorsal side of this boundary, while the dorsal pancreas occupies the ventral region [25,26]. Pancreatic head cancer originating from the dorsal pancreas often invades the anterior nerve pathway. In contrast, pancreatic head cancer originating from the ventral pancreas often shows cancer cells invading PLX1, PLX2, and ROM. In addition, since the three pathways of nerve plexuses cross at the uncinate process of the pancreas and are located close to each other, pancreatic head cancers are prone to multipathway invasion of these nerve plexuses. On the basis of the differences in the anatomical locations of the four pathways of nerve plexus and the growth locations of pancreatic head cancer, this study divided pancreatic head cancers invading different pathways of the nerve plexus into two groups: pancreatic head cancers that invaded AN (group A), and those that invaded PLX1, PLX2, and ROM (group B).
Wang et al [27] found that the clinicopathological features of pancreatic head cancer arising from dorsal or ventral pancreas were different. The frequencies of some regional lymph node metastasis were different too. Tumors arising from dorsal pancreas were more likely to invade the common bile duct and duodenum than the others. Finally, their ndings indicated that the pancreatic head cancer arising from ventral pancreas had a better survival outcome, which is consistent with our results. The weakness of their study [27] is that it is often di cult to accurately classify the primary tumor location on preoperative CT imaging, and some patients could not be grouped. Differently, we speculate that it is feasible to divide the pancreatic head cancer into anterior and posterior groups according to the preoperative CT diagnosis of nerve invasion. Our study preliminarily con rmed the difference in prognosis between the two groups: group A, pancreatic head cancer anteriorly invading AN, showed worse prognosis than group B, pancreatic head cancer posteriorly invading PLX1/ PLX2/ROM.
The limitations of this study need to be acknowledged. First, this was a retrospective study, and the prognosis of patients with pancreatic head cancer with nerve plexus invasion in different groups still needs to be con rmed by further prospective studies. Second, the small sample size precluded adequate strati ed studies. Third, traditional observation indicators such as lymph node invasion and surgical margin were not included in the evaluation criteria, because these indicators need to be con rmed by postoperative pathology and cannot be accurately diagnosed by preoperative CT. Fourthly, multivariate analysis of factors affecting postoperative survival of pancreatic head cancer was not conducted in this study, because EPNI is an independent prognostic factor affecting postoperative survival time of pancreatic head cancer [7].

Conclusions
In summary, our study showed radiographic ndings other than those pertaining to nerve invasion were not signi cantly different between pancreatic head cancer invading AN (group A) and PLX1/ PLX2/ROM (group B), but the prognosis of group A was evidently worse than that of group B. The clinical signi cance of preoperative MSCT diagnosis of pancreatic cancer groups A/B in the selection of treatment methods will be prospectively studied in the follow-up.

Availability of data and materials
The datasets used and analysed during the current study are available from the corresponding author on reasonable request.  The radiographic manifestations of pancreatic head cancer involving the peripheral nerve plexus. 2a.
Axial multislice spiral CT image. Pancreatic head cancer (T) in ltrating the inferior pancreaticoduodenal artery (arrow) along pancreatic plexus 2. Mass-and-strand pattern (arrowhead). S, superior mesenteric artery. 2b. Small eld of view volume rendered constructed in 20° right coronal oblique and axial planes perpendicular to the plane of the superior mesenteric artery. Pancreatic head cancer (T) involves both pancreatic plexus 2 and the root of the mesenteric pathway. The mass completely surrounds (white arrow) the posterior inferior pancreaticoduodenal artery and in ltrates (white arrowhead) upward along the superior mesenteric artery (S). The mass also grows toward the root of the mesenteric pathway (black arrowhead). Mass and strand pattern (black arrow). 2c. Small eld of view constructed the same as that in gure 2b. Pancreatic head cancer (T) in ltrates the gastroduodenal artery (GDA) along the anterior nerve plexus and extends to the common hepatic artery (CHA) (white arrow). Coarse reticular pattern (white arrowhead) and mass and strand pattern (black arrowhead). 2d. Axial multislice spiral CT image. Pancreatic head cancer extends along pancreatic plexus 1. A small tumor of the head of the pancreas (T) extends posteriorly to the portal vein (PV) in a coarse reticulation pattern (arrowhead). PD, pancreatic duct; IVC, Inferior vena cava; S, superior mesenteric artery.

Figure 3
The Kaplan-Meier survival curve showed that the prognosis was signi cantly poorer when pancreatic head cancer involved the anterior nerve pathway (group A) in comparison with the invading pancreatic plexus 1/ pancreatic plexus 2/root of the mesenteric pathway (group B) (P = 0.001).