Multicenter Study on the Diagnostic of Imaging Tests and EUS-guided ne needle aspiration of Solid Pseudopapillary Neoplasms of the Pancreas

Purpose Imaging diagnosis of SPN is dicult. Preoperative diagnosis by EUS-FNA) is possible, safe, and has been reported in the literature. However, its usefulness is still controversial. The aim of this study was to determine the accuracy of the EUS-FNA and imaging ndings in CT and MRI/MRCP exams in the diagnosis of patients with SPN. Methods We retrospectively reviewed the medical records of patients undergoing EUS-FNA with suspected diagnosis of SPN on imaging studies in 5 high-volume hospitals. The nal diagnosis was obtained after the histological examination of the surgical specimen. Demographic data, CT, MRI and EUS ndings, anatomopathological specimen and McH results obtained by EUS were analyzed.


Introduction
Pancreatic solid-pseudopapillary neoplasms (SPN) are rare and 90% occur in women in their 3rd and 4th decades of life [1,2]. The symptoms are nonspeci c or absent and such lesions are mostly incidentally diagnosed during imaging exams such as abdominal ultrasound (US), computerized tomography (CT) or magnetic resonance cholangiopancreatography (MRI/MRCP) [3]. SPN appears as a circumscribed, encapsulated, solid mass, but eventually with cystic areas [4]. Most patients have localized disease and up to 15% have metastasis or vascular invasion [5]. Conventional treatment is surgical resection, and unlike ductal pancreatic adenocarcinoma (PAC), the result is favorable with a ve-year survival rate of 97% [1].
Accurate diagnosis of SPN is di cult based solely on imaging, especially in solid lesions, and can commonly mimic a non-functioning pancreatic neuroendocrine tumor (NF-NET), serous cystadenoma (SCA) and less commonly PAC [6]. Based on diagnostic imaging of a pancreatic nodule, SPN may be surgically removed without a preoperative diagnosis and previous anatomopathological con rmation.
However, since such lesions are commonly found at the head of the pancreas, a duodenopancreatectomy (DP) procedure is associated to considerable morbidity and mortality rates.
Endosonography (EUS) is increasingly being used in the evaluation of pancreatic tumors. Recently, some studies evaluated the role in EUS-FNA in SPN [7][8][9][10][11][12], as well as comparing it with CT [13] and MRI for diagnosis, but its accuracy in SPN diagnosis remains unclear.
The aim of this study is to determine the diagnostic yield of imaging ndings and EUS-FNA in the preoperative diagnosis of pancreatic solid cystic lesions.

Methods
This was a retrospective and multicentric case series, carried out in three reference University hospitals and two private tertiary referral hospitals in Brazil. The study was approved by the Research Ethics Committee of each institution and complied with the regulations of the Health Insurance Portability and Accountability Act (HI-PAA).
The study was approved by the Research Ethics Committee on 07/22/2019 under number: 3.518.999. A retrospective review was carried out at medical records of patients with suspected diagnosis of SPN at CT, MRI and EUS or post-operative pathological evaluation. All selected patients signed an informed consent form before undergoing EUS-FNA. Age, sex, symptoms and CT, MRI and EUS-FNA ndings were noted and the nal diagnosis was obtained after the histological examination of the surgical specimen.
EUS was performed with conscious sedation or anesthesia monitored by an anesthesiologist. All procedures were performed by experienced doctors, always using a sectoral echoendoscope. All neneedle biopsies were made under the same procedure. Fine-needle gauge speci cations (19G, 22G, 25G and ProCore 20G), number of needles passes and micro histology (McH) results, in addition to the surgical pathology results were recorded. The occurrence of adverse events (AE) was also documented.

Sample handling
Samples obtained by FNA were placed in 10% formalin (6 to 24 hours) and followed the routine of the pathological anatomy sector of each center participating in the study. The blocks and biopsies embedded in para n were subjected to semi-series histological sections of 3µm thick, at three different depth levels, stained by the Hematoxylin-Eosin method. CellSens Micro Imaging Software (Olympus America Inc, Center Valley, PA) was used to dimension samples in millimeters. The diagnosis of SPN was con rmed through McH with hematoxylin-eosin and, when possible, immunohistochemistry (IHC). The sections stained with hematoxylin-Eosin (H / E) from both FNA and surgical samples were reviewed by specialized pathologists to con rm the diagnosis. The IHC markers (primary antibodies) used in the analysis included β-catenin, CD10, progesterone receptor (PR), CD99 and chromogranin A [14].

Statistical analysis
The collected and recorded data were stored in a database. The statistical analysis of continuous variables was described as mean and standard deviation and dichotomous variables expressed as simple proportions. The de nitive diagnosis of SPN was based on histological analysis, and the patients who did not have it were excluded from the study. True positive cases were the ones that after undergoing CT, MRI, EUS and EUS-FNA, had the SPN diagnosis histologically con rmed. To assess the diagnostic gain acquired with the combination of imaging tests, the numbers of truly positive cases from each method were added. The comparison between the results obtained in the imaging exams was performed using the Mc Nemmar test for each pair of groups (CT + EUS; CT + EUS / FNA; MRI + EUS; MRI + EUS-FNA). The association between the size of the needle used and the diagnosis obtained in the McH was tested using Fisher's exact test. The level of statistical signi cance adopted was 95%. The analyzes were performed using the STATA 14 software.

Results
All 54 patients had suspected SPN pancreatic tumors. Three patients had a non-functioning pancreatic neuroendocrine (NF-NET) tumor type G1 as a nal diagnosis after surgical resection. The others (n = 51) had the diagnosis of SPN con rmed by McH, anatomopathological report of the surgical specimen and IHC analysis. Of the 51 patients with con rmed SPN, 90.2% (n = 46) were female and 9.8% (n = 5) were male, and the mean age was 33.66 years (SD 14.46) ( Table 1).
Twenty-one patients (41.2%) were symptomatic, being the most common symptoms abdominal pain and weight loss associated with nausea or vomiting. Jaundice was present in 4 (7.8%) patients. SPN was detected incidentally in 32 (59.3%) of the patients. The average size of the cyst was 3.8 cm (SD 2.26) and the lesions were distributed between the body (20), head/uncinate (22), and tail (9) portions of the pancreas. One patient had two synchronous lesions identi ed on the body and tail of the pancreas. The solid, solid-cystic (microcystic) and cystic appearance was present in 27 (52.9%), 21 (41.1%) and 4 (7.8%) cases, respectively ( Figure 1). The lesion margin was well de ned in 42 (82.3%) of the cases. Calci cation was present in 3 (5.8%) patients. In 4 (7.8%) patients there was vascular involvement detected by the EUS and con rmed by surgery, as well as dilation of the main pancreatic duct (MP) in 5 (9.8%) patients (Table   1).
EUS-FNA was performed on all patients, without the pathologist presence of a rapid on site evaluation (ROSE) ( Table 4). Conventional needles of 22G in 34 (66.66%), 19G in 9 (17.64%), ProCore 20G in 5 (9.8%) and 25G in 3 (5.8%) were used. A median of 2.5 needle passes (range 1 to 5) was performed for each lesion. The diagnostic performance of the EUS-FNA was not affected by the size of the needle (p = 0.175). No adverse events were documented.
Surgical resection was performed in all patients who had no evidence of lymph nodes or distant metastases.
EUS alone suspected the diagnosis of SPN in 33 (64.71%). Two cases that were not included in this analysis because they received a nal diagnosis of NF-NET, had a previous result of SPN through EUS. EUS suspected PAC in 9, NF-NET in 4 (7.8%), chronic pancreatitis in 1, pancreatic cystic lesion in 1, serous cystadenoma in 2, and lymph node in 1. The McH obtained by the EUS-FNA diagnosed SPN in 45 (88.24%), NF-NET in 2 (3.92%) and was negative in 4 cases (7.84%), all con rmed as SPN after surgery.

Discussion
SPN are rare, with approximately 2744 reported cases [15]. US and CT are the most used imaging methods for diagnosis and represent approximately 80% of imaging studies for the diagnosis of SPN. In the past few decades, the use of MRI has increased substantially, as has EUS, but it still accounts for only 5% of the imaging methods used for the diagnosis of SPN. In addition, there are few studies that determine the additional bene t offered by the EUS-FNA to imaging tests such as CT and MRI [16]. The bene t in the use of EUS in patients with solid and cystic pancreatic tumors with a yield approaching 78.8% and 71.4%, respectively, has been proven [17,18]. Therefore, the inclusion of EUS-FNA increases the diagnostic accuracy of this type of tumor.
The results of this study highlight the di culty of CT, MRI and EUS in correctly classifying the SPN using only the image resources, with a diagnostic performance of 21.95%, 28.88%, 64.71% respectively. These results are similar to another recent study, which was 23.5% and 41.2% for CT and EUS, respectively [13]. When we include the results of EUS-FNA, there is an increase in the diagnostic performance of CT (72.16%) and MRI (65.23%), rising up to 94.11% in both methods.
Classically, SPN are described as circumscribed, hypoechoic, solid, heterogeneous masses and eventually with a cystic component. However, this appearance presents in just 60% of patients, with a predominantly solid lesion found in 32% of cases, suggesting a more classic appearance associated with a neuroendocrine tumor [19]. In our series, the solid, solid/cystic, and cystic morphology was 52.9%, 41.1% and 7.8% respectively. SPN can mimic other pancreatic tumors, which can lead to diagnostic challenges [20].
This fact can be observed in our study, as CT and EUS suspected the presence of NF-NET in (9.75%) and (7.8%), respectively, since the injuries identi ed by these exams had a predominantly solid component. When microcystic, SPN are similar to serous cystadenoma and can cause diagnostic confusion, which occurred in 4% of the cases examined by CT, MRI and EUS. The microcystic pattern must be recognized as part of the morphological spectrum of the SPN, which can lead to confusion regarding the presence of a serous cystadenoma [6]. Previous CT and EUS studies, in which SPN was identi ed as other cystic lesions in the pancreas in up to 50% of cases, including benign lesions such as serous cystadenoma, con rm the ndings of our study [13,21,22]. Some cases were interpreted as pseudocysts, highlighting the di culty in determining the etiology of a cystic lesion based only on morphological characteristics of the image. This occurred in 7.31% and 15.55% of CTs and MRIs, respectively. Therefore, the importance of correctly classifying SPN is emphasized in this study, where 8 (33.3%) cysts were erroneously classi ed by CT, as pseudocysts (2), unspeci c pancreatic cystic lesion (2), unspeci c pancreatic cystic lesion with calci cation (2), hematoma (1) and serous cystadenoma (1), all which have prognosis and management completely different from those adopted in SPN. The same happened with MRI / MRCP which was normal in 3 patients and identi ed unspeci c pancreatic cystic lesion in 4 and pseudocyst in 3 patients.
In the literature, rupture of the SPN is associated with abdominal trauma or may be spontaneous, which is rare. To date, there have been only 3 cases of SPN with spontaneous rupture reported worldwide [23]. In our series, we had two young patients with blunt abdominal trauma who underwent imaging exams that suspected hematoma (1) and pseudocyst (1). In these two cases, the EUS-FNA was crucial for the diagnosis of SPN, changing the management of these patients.
The aim of this study was to con rm the bene t of performing the EUS-FNA to increase the diagnostic performance of other conventional imaging tests such as CT and MRI. The EUS-FNA association increased diagnostic yield by almost 72.16% for CT and 65.23% for MRI. The results are similar to previous studies, which report a low sensitivity rate and absence of adverse events as in the current series [13]. Regarding the adverse events resulting from the EUS-FNA, the rst case of neoplastic cell implantation in the stomach was recently described [24]. In our series, even after prolonged follow-up and interviews with patients and attending physicians, we did not observe this type of complication.
In the present series we found 2 cases of interest, as they had a strong suspicion of SPN on imaging exams. In the rst, CT and MRI con rmed the diagnosis of pseudocyst, the EUS suspected SPN, and the EUS-FNA revealed NF-NET, which was con rmed by uncinate process resection. The second had CT, MRI and EUS image exams with SPN morphology (solid-cystic), the McH obtained by EUS-FNA con rmed the presence of SPN, but the operative piece con rmed NF-NET. In this case, the diagnosis by McH con rming SPN was performed only by HE, as there was not enough material to perform the IHC. These ndings corroborate those found in the literature where the EUS-FNA performed on suspected SPN has excellent positive and negative predictive value, however the most common classi cation errors were with NF-NET which presents no clinical impact, just as it was observed in our series [8]. In addition, in 177 patients studied by the EUS-FNA with suspected NF-NET, discrepancies were reported in 14 patients. In 4 of them the erroneous diagnosis was SPN. Accordingly, it is concluded that when an adequate sample is obtained, the most signi cant error is the incorrect classi cation, which is more often associated with SPN, but the damage associated with this diagnostic error is minimal [25].
The authors conclude that the isolated CT, MRI and EUS have low rates of correct diagnosis, but the association with EUS-FNA increases preoperative diagnosis and should be considered to rule out other pancreatic diseases such as NF-NET and SCA. These results suggest that patients who do not have a clear diagnosis of SPN, as is usually the case, should undergo EUS-FNA.

Declarations
Funding: This study did not obtain any funding.
Con icts of interest / competing interests: All authors declare that there is no con ict of interest.
Availability of data and materials: Data from each of the patients included in this study are and will be available at any time for further investigation.      CT images with enlarged pancreatic head, dilation of the main pancreatic and choledochal duct, in addition to atrophy of the pancreatic gland in a 23-year-old woman with a history of chronic alcoholism (same case in gure 3).