Well-differentiated pancreatic neuroendocrine tumors (PNETs) are a rare form of epithelial neoplasm that has been increasingly diagnosed worldwide over the past several decades.1,2 Complete resection is the cornerstone of clinical management and the sole curative option for patients with PNETs.3 The clinical courses of PNETs are diverse and difficult to predict.4–8 The most widely accepted prognostic factor is the World Health Organization (WHO) histological grading scheme, in which PNETs are classified into G1, G2, and G3 based on mitotic counts and Ki-67 labeling indices.9,10 However, because this classification system requires tissue specimens, a preoperative assessment for the risk of malignancy is generally difficult in patients with PNET.11

Previous studies have shown that tumor size is one of the most reliable, noninvasive predictors for nodal and distant metastases.1,2,4,12 However, a larger tumor size alone is far less reliable to determine aggressive clinical behavior in clinical practice.3,13 Therefore, new preoperative factors that can be assessed in the clinical setting have been needed. We recently observed that some patients with PNETs who presented with stricture of the main pancreatic duct (MPD) showed a malignant clinical course. A review of the literature found similar case reports, which prompted us to speculate that MPD involvement in PNETs may indicate a poor prognosis.14–17

Based on our hypothesis, this multicenter, retrospective study was conducted to examine the clinical significance of MPD involvement in PNETs in a systematic way.

Methods

Case Selection

The study protocol was approved by the Ethics Committee at Kobe University Graduate School of Medicine. Between January 2004 and April 2016, 110 patients with well-differentiated PNETs were surgically treated at our institutes [Kobe University Hospital (n = 55), Kindai University Hospital (n = 30), and Kagawa University Hospital (n = 25)]. Eight incidental PNETs found in surgical specimens resected for other purposes were excluded. Another patient who had both PNET and intraductal papillary mucinous neoplasm (IPMN) also was excluded from this study, because it was difficult to determine whether the duct dilatation was due to PNET or IPMN.

Study Cohort

The study cohort consisted of 49 men and 52 women with a median age of 64 (range 23–87) years at the time of resection. Twenty-five cases (25%) were functioning (21 insulinomas, 3 gastrinomas, and 1 VIPoma). Three patients had multiple endocrine neoplasia type 1, whereas two others had a history of von Hippel–Lindau disease. Standard surgical procedures with regional lymph node dissection were performed in the following 66 of 101 cases (65%): pancreatoduodenectomy (n = 28), distal pancreatectomy (n = 37), and total pancreatectomy (n = 1). Organ/parenchymal-sparing resections [spleen-preserving distal pancreatectomy (n = 10), central pancreatectomy (n = 7), or enucleation (n = 18)] were considered for small tumors (<20 mm) without any signs of metastasis. Complete resection was achieved in 98 of 101 patients (97%). Seven patients with G1 PNETs (5–12 mm at the greatest diameter) were followed at an outpatient clinic and subjected to repeat contrast-enhanced computed tomography (CT) every 12 months. The remaining patients followed were subjected to repeat CT every 3–6 months during the first 2–3 years and yearly thereafter. Additional magnetic resonance imaging, endoscopic ultrasonography, and positron emission tomography studies were performed if recurrence was suspected. The median follow-up time was 43 (range 2–133) months.

Data Collection

Medical records, imaging data, and histology slides were reviewed retrospectively in terms of age, sex, tumor size, number of tumors, tumor location, and tumor functionality. Tumor sizes were measured based on the last enhanced CT scans before to surgery. For patients with multiple PNETs, the largest tumor was evaluated. Tumors were graded based on Ki-67 indices and mitotic counts, according to the WHO classification.18

Definition of MPD Involvement

The presence or absence of MPD involvement was evaluated retrospectively on magnetic resonance cholangiopancreatography (MRCP) images by at least two experienced pancreatic surgeons at each institution. In the preliminary study performed in the Kobe University cohort (n = 51), the accuracy of MRCP and that of enhanced CT for diagnosing MPD involvement were in 100% agreement. Therefore, enhanced CT data were used alternatively for cases with no available MRCP data (36/101 cases, 36%). MPD involvement was defined as stenosis of the MPD in the tumor mass associated with up-stream dilatation of the MPD more than double the diameter of the proximal part of the MPD. MPD dilatation from other causes was excluded.

Immunohistochemical Study

In the Kobe University cohort (n = 51), original hematoxylin and eosin-stained sections were further reviewed for histological changes representing MPD stenosis. Microscopic venous and lymphatic invasion and perineural infiltration were evaluated using immunostaining for CD31 and D2-40 as described previously.19 Preoperative serum pancreatic amylase and lipase levels also were measured to determine parenchymal injury due to ductal obstruction.

Statistics

To identify predictors of a progressive clinical course of PNET, clinicopathological factors were correlated with nodal metastases or recurrence. Tumor recurrence was evaluated instead of overall survival, as PNETs are slow-growing neoplasms requiring longer follow-up periods for analyzing patient survival. Analyses were performed with the Fisher’s exact test for categorical variables and the Wilcoxon test for a continuous variable. A two-sided P value <0.05 was considered statistically significant. Univariate analyses were performed with logistic regression models, and variables with P values <0.05 were further evaluated in a multivariate logistic regression model to calculate an adjusted odds ratio (OR) with 95% confidence intervals (CI). Survival curves were estimated with the Kaplan–Meier method and compared by the log-rank test. All analyses were performed with JMP 11.0 for Macintosh (SAS Institute, Inc., Cary, NC).

Results

Patient Characteristics

MPD involvement was observed in 13 of 101 patients (13%). Figure 1a shows a representative MRCP image of stenosis and upstream dilatation of the MPD. Table 1 compares multiple clinicopathological parameters between patients with and without ductal involvement. PNETs with MPD involvement were all nonfunctioning and were significantly larger in size and more common in the pancreatic head than those without ductal involvement. The most significant difference was noted in the rate of nodal metastasis [10/13 (77%) with involvement vs. 11/88 (13%) without involvement; P < 0.001]. During follow-up, 6 of 12 PNET cases (50%) with MPD involvement and 6 of 86 PNET cases (7%) without MPD involvement had recurrence (P < 0.001). Recurrent tumors developed in the liver (n = 9) or multiple organs (n = 3).

Fig. 1
figure 1

Representative radiological and histopathological findings for MPD involvement in PNETs. a Representative magnetic resonance cholangiopancreatography showing stenosis (arrowhead) and upstream dilatation of the MPD. b The MPD (D) entrapped in the tumor (T) appears to be narrow due to periductal tumor infiltration (original magnification 40×). Unlike classic PNETs, the tumor is associated with fibrotic stroma. Arrows indicate the border between the tumor and the background parenchyma. c The tumor (T) shows extensive intraductal infiltration (original magnification 40×). Arrows indicate the MPD. MPD main pancreatic duct, PNET pancreatic neuroendocrine tumor

Full size image
Table 1 Patient characteristics
Full size table

Evaluation of the Clinical Impact of Stenotic Involvement of the MPD

The Kaplan–Meier curve (Fig. 2) illustrated that patients with PNETs with MPD involvement had a shorter recurrence-free period than did those without MPD involvement. The 5 years recurrence-free rate of patients with PNETs with MPD involvement was 41%, which was lower than that of patients with PNETs ≥ 15 mm (75%) or G2 PNETs (74%).

Fig. 2
figure 2

Recurrence-free survival of patients with PNETs. Patients with MPD involvement (a P < 0.001 vs. no MPD involvement), tumor size ≥15 mm (b P = 0.004 vs. tumor size <15 mm), or G2 histological grade (c P = 0.003 vs. G1 histological grade) showed significantly worse recurrence-free survival. PNET pancreatic neuroendocrine tumor, MPD main pancreatic duct

Full size image

Multivariate analysis also was conducted to determine whether MPD involvement is an independent prognostic factor. Seven potential prognostic features (age, sex, tumor size, tumor location, functionality, MPD involvement, and histological grade) of nodal metastasis and tumor recurrence were evaluated by univariate analysis (Table 2). Those factors significantly associated with nodal metastasis on univariate analysis, tumor size (P < 0.001), MPD involvement (P < 0.001), and histological grade (P = 0.002) were subjected to multivariate analysis, which revealed MPD involvement (P < 0.001) as an independent factor significantly correlated with nodal metastasis. For tumor recurrence, tumor size (P < 0.001), MPD involvement (P < 0.001), and histological grade (P < 0.001) were identified as significant factors on univariate analysis, and all three factors (P = 0.023, P = 0.009, P = 0.018, respectively) appeared to be independent predictors for tumor recurrence on multivariate analysis.

Table 2 Uni- and multi-variate analyses for risk factors for nodal metastasis and recurrence
Full size table

Clinicopathological Correlation Study

To elucidate the radiology–pathology correlation, histology slides from the Kobe University cohort (n = 51) were reviewed. In this cohort, ten cases had MPD involvement. The pathological examination revealed that all ten tumors with MPD involvement exhibited an invasive growth pattern with fibrotic stroma. Entrapped MPDs were directly infiltrated by PNETs (Fig. 1b). Three cases (30%) also showed additional intraductal polypoid infiltration in the MPD (Fig. 1c). Microscopic venous invasion and perineural infiltration were observed more commonly in PNETs with MPD involvement than in those without MPD involvement (Table 3). Although the Ki-67 index was higher in patients with MPD involvement than in those without MPD involvement [7/10 (70%) vs. 12/41 (30%), P = 0.028], the histological grade did not differ between the two groups (P = 0.079).

Table 3 Pathological and clinical details of the Kobe University cohort
Full size table

In terms of clinical and serological features, two patients with MPD involvement presented with abdominal pain, indicating pancreatitis. MPD involvement also was correlated significantly with higher preoperative serum concentrations of pancreatic amylase (P = 0.018) and lipase (P = 0.002; Table 3). However, no imaging features suggestive of severe acute pancreatitis, such as peripancreatic fluid correction, were observed in any of the patients.

Discussion

Previous studies have not reached reliable conclusions regarding the clinical predictive factors for a progressive clinical course in patients with PNETs.20–22 The present study revealed that MPD involvement in this uncommon pancreatic neoplasm is associated with multiple unfavorable clinical and pathological features. Most importantly, stricture of the MPD appeared to be an independent factor associated with nodal metastasis at the time of surgery and subsequent tumor recurrence during follow-up. Kaplan–Meier analysis also suggested that MPD involvement may be a stronger indicator of aggressive clinical behavior than large size (≥ 15 mm) or G2 PNETs, given that the 5 years recurrence-free rate of PNETs with MPD involvement was just 41%.

Although the biological nature of MPD involvement in PNETs remains elusive, MPD involvement in PNETs may relate to a unique growth pattern.15,16,23–26 Unlike classic PNETs, which proliferate in a medullary pattern with scant stroma, PNETs with MPD involvement appeared to be more fibrotic and infiltrative into adjacent structures, which is likely the reason that the MPD is entrapped inside the tumor masses. Therefore, a sign of MPD involvement on imaging may be a good indicator of an infiltrative histological growth pattern. Previous studies have shown that a subset of PNETs produce multiple tumor-related growth factors (e.g., platelet-derived growth factor, transforming growth factor-beta, and basic fibroblast growth factor), and they are suspected to stimulate a fibroblastic reaction and vascularization.25,26 These factors may be involved in the infiltrative growth and development of the fibrotic stroma in PNETs with MPD involvement.

Several studies have suggested that nodal metastasis of PNETs is related to shorter disease-related survival and that regional lymph node dissection may decrease the risk of locoregional tumor recurrence.2,4,27,28 However, it remains difficult to determine whether nodal dissection is truly necessary on a case-by-case basis in clinical practice. In this study, more than half of the cases with MPD involvement had nodal metastases, suggesting that regional lymph node dissection should be considered in PNET cases with MPD involvement. Similarly, given the high risk of tumor recurrence in patients with PNETs involving the MPD, close follow-up also is required for such patients. To the best of our knowledge, no imaging features that reliably predict the biological behavior of PNETs have been determined. MPD involvement in PNETs will be an easy-to-assess, reproducible imaging finding suggestive of aggressive clinical behaviors.

This study has several limitations. First, only surgical cases were examined. Second, standard lymph node dissection was not completed in all cases; regional lymph nodes were removed in 72% of the nonfunctioning tumors and 44% of the functioning tumors. Third, the follow-up period (median 43 months; range 2–133 months) may not be long enough for slow-growing tumors like PNETs. Finally, the MPD involvement in cases of pancreatic body-tail tumors might be underestimated, because distal duct dilatation is suspected to be less apparent.

Conclusions

The results of the current study suggest that MPD involvement in PNETs may be a useful imaging sign, indicating an aggressive clinical course, such as nodal metastasis and tumor recurrence. This is probably because MPD involvement is linked to a histological infiltrative growth pattern. Regional lymph node dissection and long-time follow-up will be required in patients with resectable PNETs with MPD involvement.