Lymphoepithelial Cyst of the Pancreas: Can Common Imaging Features Help to Avoid Resesction?

Background: Differentiation of cystic pancreatic neoplasms remains a challenging task for radiologists with main aim of characterizing malignant and premalignant conditions. Purpose: The study aimed to compare radiological features of lymphoepithelial cysts (LEC) with other cystic pancreatic lesions, which could help to differentiate them in order to avoid unnecessary resection and optimize surveillance. Material and Methods: We retrospectively reviewed 12 cases of resected and histopathological conrmed LECs in last 12 years, for 10 patients imaging studies were available. 20 patients with mucinous cystic neoplasms (MCN) and 20 patients with branch-duct intraductal papillary mucinous neoplasms (BD-IPMN) were selected consecutively to serve as control groups. Imaging ndings as well as clinical data were analyzed. Results: Three imaging subtypes of LEC were identied: simple cystic morphology (20%) and mixed cystic-solid lesions (80%) with either diffuse subsolid component (30%) or mural nodule (50%). All lesions revealed exophytic location with strong male predominance (9:1). MCNs were presented exclusively in middle-aged woman and IPMN in both sexes showing slight male predominance (13:7). Mean patient age in IPMN (70.5+7.7 years) was signicantly higher compared to other groups (p<0.001 for LEC, p=0.005 for MCN). Unenhanced CT-attenuation of LEC was higher than MCNs (p=0.025) and IPMNs (p=0.021). Conclusion: The present study provides three imaging subtypes of LEC with key features for the differentiation from other cystic pancreatic lesions such as increased native attenuation, absence of connection to main pancreatic duct (MPD) and exophytic location. Clinical data, such as male predominance in LEC, is crucial in differentiating cystic pancreatic lesions.


Introduction
Cystic pancreatic lesions are divided in neoplastic and non-neoplastic lesions, with the bigger part being neoplastic lesions. Most common include intraductal papillary mucinous neoplasms (IPMN), mucinous cystic neoplasms, serous cystic neoplasms (SCN), and solid pseudopapillary tumor, which account approximately 70% of cystic pancreatic lesions. Non-neoplastic lesions are divided in the epithelial and nonepithelial lesions, with most common epithelial lesions being congenital cysts, alimentary duplication cysts, endometrial cysts, and lymphoepithelial cysts, and most common nonepithelial lesions being pseudocysts or walled-off necrosis (1).
LEC of the pancreas are rare benign cystic lesions. They were rst described in 1985 by Luchtrat (2) and the majority of available data on radiological features of LECs are case reports. Over the past 30 years, since the rst description, over 100 cases were reported (3). As they are rare, their clinical and pathologic features have not yet been fully characterized, especially regarding the differentiation from other cystic lesions (4,5). In the clinicopathologic analysis of 12 patients, LECs were reported to constitute approximately 0.5% of pancreatic cysts. They were seen in middle-aged patients, predominantly, but not exclusively in men (M/F = 4/1) (6). Recent reviews documenting the demographic features of LECs indicate a strong male preponderance as well (7).
LECs are true pancreatic cysts that are lined by squamous, non-dysplastic epithelium, and surrounded by mature lymphoid tissue (8). Etiological theories include formation from squamous metaplasia of pancreatic ducts, derivation from epithelial remnants in lymph nodes, and displacement of ultimo-branchial remnants that go on to fuse with the pancreas during embryogenesis. There is also a possibility that LECs are a form of teratoma (6).
Because the imaging features of the LECs vary and sometimes are very similar to other pancreatic lesions, it is di cult to differentiate LECs from other pancreatic lesions, however, a correct differentiation and classi cation is crucial for the clinical decision making and treatment planning. LECs are benign and do not possess malignant potential and, thus, an accurate identi cation of these lesions is important to avoid unnecessary intervention (8). A correct preoperative diagnosis of a LEC could help prevent surgical interventions, though, today ne-needle aspiration is the only tool that can achieve a diagnosis without resection (3,(9)(10)(11).
Both MCNs and IPMNs are characterized by a neoplastic mucin-producing epithelial cell lining with the potential to malignant transformation. Since the rare data about imaging features of LEC, that are available from case series, show an overlapping with MCNs and IPMNs, differentiation of LEC from these potentially malignant pancreatic lesions remains di cult and an important radiological problem. The present study aimed to compare LECs with other cystic lesions of the pancreas and to identify common radiological features which could help to differentiate them from other cystic neoplasms in order to avoid unnecessary resections.

Surgical procedure and histopathology
The patients were considered symptomatic if they had a left upper quadrant (LUQ) or epigastric pain, postprandial complaints (i.e. cramps) and did not have other speci c cause for these symptoms. Clinically, abdominal pain was the most common clinical feature in all three groups. Clinical symptoms served as the main indication for surgical treatment, all 50 patients in the present study underwent a surgical resection.
LEC-patients underwent cyst resection or left (distal) pancreatic resection with or without splenectomy. After surgery, 60% of the patients had minor local complications: four had uid collections that had to be drained, one had a pancreatic stula, and one developed a pseudocyst. All complications were treated with radiological interventions without the need of the additional surgical procedures (Fig. 1).
Histopathological reports con rmed LECs of the pancreas in all 10 patients without evidence of malignant transformation. An example of histopathological image is shown in Fig. 2.
Among MCN there were 14 lesions with low grade MCN, 4 lesions with high grade MCN and 2 lesions with invasive carcinoma associated with MCN. Among IPMN there were 18 BD-IPMN and 2 mixed-type IPMN with 13 low grade IPMN, 6 IPMN with carcinoma in situ and 1 IPMN with associated invasive carcinoma.

Baseline results
LECs were more common in middle-aged males (nine males, one female), ranging between 43 to 63 years (mean age: 50.9 ± 6.4 years). Summary of patient demographics, initial LEC ndings and surgical procedures are presented in Table 1. In contrast, MCNs were presented exclusively in the middle-aged woman with the age range of 37 to 67 years (mean age 51 ± 13 years). There was no statistically signi cant difference in age between LEC-group and MCN-group (p = 0,983).
Branch duct-IPMNs were found in both sexes showing a slight male predominance (13 male and 7 female). The mean patient age was higher compared to the other two groups, with mean value of 70.5 ± 7.7 years (rage 55 to 78 years), the difference was statistically signi cant (LEC compared to IPMN p < 0.001 and MCN compared to IPMN p = 0.005).
Imaging ndings: All patients with LEC revealed a well-de ned exophytic lesion within the tail (70%) or body (30%) of the pancreatic parenchyma, without communication with the main pancreatic duct. The mean size was 6.1 cm (2.7-22 cm). There were no calci cations within the pancreatic parenchyma or within the cystic lesions. On unenhanced CT scans nine LECs had high attenuating uid or material, which strongly indicated the presence of keratinous material with unenhanced CT attenuation mean value of 22 HU (ranged 16 to 34 HU). Cystic liver lesions were incidentally found in one patient, another incidental nding was cholecystolithiasis in another patient.
MCN lesions had a solitary / macrocystic appearance in the distal pancreas with a mean size of 5.9 cm. In all cases a pancreatic tail was involved, with 5 out of 20 occupying both the body and the tail of the pancreas. There was also no connection to the pancreatic duct.
In 13 out of 20 cases patients with IPMN the head of the pancreas was affected (with one lesion involving the entire pancreas and the other -the pancreatic head and tail), seven were found in the pancreatic tail. The mean size was signi cantly smaller than in previous two groups with the mean size of 1.8 cm (p < 0.001). The key differential feature was evidence of communication to the main pancreatic duct, which was easily con rmed using MRCP images (when available).
When comparing LEC with other cystic lesions there was no statistically signi cant difference in size between LEC and MCN (p = 0.912). On the contrary, branch duct IPMNs were signi cantly smaller that LEC (p < 0.001). Unenhanced CT attenuation of LEC was signi cantly higher than MCN (p = 0.025) and IPMNs (p = 0.021). Summary of comparison between three groups is presented in Table 2. Taken the imaging features together, we were able to identify three imaging subtypes of LECs. In 20% the LECs where simple cystic lesions. In 80% LECs where mixed solid-cystic lesions with different morphology: rst group (30%) had diffuse subsolid component (Fig. 3) and the second group (50%) had mural nodules ( Fig. 1 and Fig. 4).
Subsolid component was de ned as homogeneous or heterogenous area of higher attenuation than simple uid on noncontrast CT, with unchanged attenuation on arterial/venous phase. The mean size of the mural nodule was 2.1 cm, ranged from 0.8 to 5.6 cm. Four LECs were unilocular, six multilocular (≥ 4) and ve had septations. Both septations and solid components showed contrast enhancement in the multiphase studies. One LEC had areas of sponge-like appearance due to septations with some degree of enhancement, similar to those in serous cystic neoplasms.
None of the lesions caused pancreatic duct or bile duct dilation and no atrophy of the pancreatic parenchyma. We found this feature bene cial in differentiating LECs to IPMNs, with IPMNs showing a slight dilatation of MPD in 50% of cases with the mean diameter of 4 mm (ranged 3 to 6 mm).
Overlapping imaging features of LEC and branch-duct IPMN are shown in Fig. 5.
The most likely differential diagnosis in the provided case of a small cystic lesion in the pancreatic tail is IPMN due to lesion morphology. Nevertheless, there was no clear connection with the main pancreatic duct on the MRCP images, so the diagnosis was not certain. Due to the presence of a second bigger lesion both lesions were resected.
Overlapping imaging features of LEC and MCN are shown in Fig. 6. Due to the lesion morphology, similar to MCN and lesion size so it was resected.

Discussion
In the present study, which is one of the largest monocentric radiological studies on LEC, we identi ed three First typical clinical feature of LEC is a clear male predominance with a ratio of 9 to 1 (M to F). This nding was the main differential feature to MCNs and is in concordance with two LEC-cohorts assembling a male to female ratio of 4 to 1 (6) and 7 to 1 (11). It is reported that the majority of patients with LECs were middle aged men with an average age of 54 years (range 26-82 years in different studies), which corresponds with our results with the mean age of LEC patients being 50 years. This nding served as the main differential feature to IPMN.
Kim et al. (16) as well as Adsay et al. (6) reported that the most characteristic morphological and imaging nding of LEC is its exophytic location with contour deformation; we were able to verify this nding. This feature can be a result of development of LEC from epithelial remnants in the lymphatic node that normally has an extrapancreatic location. Adsay (6) und Sewkani (7) reported that LEC could be found in any location of the pancreas including head, body, and tail. In the following studies (17)(18)(19)(20)(21), the LECs were localized in the body and tail, which is in line with our ndings with distribution ratio of tail to body of 7:3.
LECs also showed a higher HU attenuation on the native CT scans when compared to both MCNs and branch duct IPMNs. This feature can be a result of the granular keratinized material in the cyst uid intraluminal, one of the pathologic characteristics of the LECs. We found it bene cial in differentiation LECs to MCNs and IPMNs. Kudo et al. (22) reported slight signal reduction in out-of-phase MRI compared to that of in-phase, indicating co-existence of fat and water. Fukukura et al. (23) stated that LECs could have a lipid component which has negative CT attenuation values. We were not able to verify these nding in our patient collective.
A malignant transformation of LEC has not been reported, epithelial changes are often reactive due to secondary in ammation of the cyst. The elevation of tumor markers is not a distinctive clinical feature of malignancy as around 30-40% of LEC showed elevated CEA and CA 19 − 9 levels being benign lesions (24).
With imaging it is possible to differentiate pseudocyst when a clinical history of acute pancreatitis and follow-up imaging are given. Pseudocysts are most commonly a result of pancreatic in ammation with a rupture of the pancreatic duct due to acute pancreatitis. They can also form as a result of pancreatic trauma with parenchymal rupture. They have no malignant potential and normally are followed-up or treated with minimal invasive intervention, therefor we didn't include these patients as a control group in our study.
CT and MRI remain the main radiological modalities for the differentiation of cystic pancreatic lesions; however, radiological features of LEC could be indistinguishable from other cystic pancreatic neoplasms. The major limitation of our study its limited number of patients, explanation to this is the rarity of the LECs.
Secondly, there was no standardized imaging protocol of CT and MRI within 12 years because of the retrospective study design; some patients already had imaging studies, performed outside our institution and were referred to our European pancreatic center for treatment. Although some imaging differences between patients were present, the image quality and provided diagnostic information was su cient for decision making. Such imaging heterogeneity also re ects realistic clinical work ow.

Conclusion
In conclusion, clinical data is crucial in the differential diagnosis of the cystic lesions of the pancreas. The present study showed a strong male predominance of LEC, which is bene cial in the differential diagnosis with MCN.
Our study provides key radiological features, such as connection to MPD, dilatation of MPD, and location, for the imaging differentiation of LEC and other cystic pancreatic lesions. Firstly, compared to branch-duct IPMN, LEC showed no communication and no dilatation of the main pancreatic duct. Secondly, the location can also serve as a differential feature with lesions located in the pancreatic head unlikely to be LEC with most typical location in pancreatic tail. However, under some circumstances LECs are indistinguishable from other pancreatic neoplasms and a correct preoperative diagnosis cannot be done using imaging only. In such cases, a multidisciplinary approach should be strongly considered in order to correlate the imaging nding with clinical data and the possibility of watch and wait approach.

Methods
Ethics approval and consent

Radiological evaluation
Two radiologists (EK, 10 years of experience in abdominal imaging and EEG with 2,5 years of experience), blinded to the histopathological reports, analyzed the images independently, using a Picture Archiving and Communication System (PACS) workstation. Imaging analysis included the following parameters: location (head, body, tail), lesion appearance (cystic, subsolid), septation (present, absent), enhancing mural nodules (present, absent), and dilatation of main pancreatic duct (MPD) > 3 mm and common bile duct (CBD) > 8 mm upstream of the lesion (present, absent; for lesions without proximity to the MPD or CBD and for lesions located at the tip of the pancreatic tail the latter two parameters were classi ed as not applicable).
Size of lesions (mm) and attenuation values (Houns eld units, HU) on unenhanced images, as well as in the arterial and venous phases were measured. To assess the attenuation on unenhanced, arterial and venous CT scans, HU were measured using circular regions of interest (ROI) located in the most homogeneous part of the lesion. Afterwards, discrepancies in image interpretation were resolved by consensus between the two radiologists.
The following parameters were assessed for CT images only: calci cation (present, absent) and CT density on unenhanced phase (hypodense, isodense, hyperdense).
The following parameters were assessed for MR images only: signal intensity (hypointense, isointense, hyperintense) and signal uniformity (homogeneous, heterogeneous) of the lesion on unenhanced T1-and T2-weighted imaging with and without fat saturation, presence of T1-hyperintense spots with corresponding T2-hypointensity as evidence for haemorrhage (present, absent), signal intensity on DWI with b ≥ 500 s/mm2 compared to the surrounding pancreas (hypointense, isointense, hyperintense).

Statistical analysis
Data management was carried out by SAS software release 9.4 (SAS Institute, Cary, North Carolina, USA) and statistical analysis was made using IMB SPSS software, version 24 (IBM Corp.). Data is reported as mean ± SD. Agreement between radiologists was quantitated using Cohen's kappa for nominal categorical variables (12) and weighted kappa (linear weights) for ordinal categorical variables (13). As proposed by Landis and Koch (14), kappa values were interpreted as poor (< 0.00), slight (0.00-0.20), fair (0.21-0.40), moderate (0.41-0.60), substantial (0.61-0.80), and almost perfect agreement (0.81-1.00). Mann-Whitney U test was performed to compare continuous parameters between groups. For categorial parameters absolute numbers are shown. Two-sided p-values were computed, the differences were considered statistically signi cant at P-value of 0.05 or less.