Tumors of the pancreas associated with extracellular mucin production include mucin-producing ductal adenocarcinoma, mucinous cystic neoplasm (MCN), and intraductal papillary mucinous tumor (IPMT). Fine-needle aspiration (FNA) is used as an adjunct to radiologic analysis for the preoperative categorization of these tumors. The current study was designed to identify distinctive cytomorphologic features that would be useful for the categorization of mucinous tumors of the pancreas.
The authors evaluated Papanicolaou smears and Diff-Quik-stained smears of specimens obtained by computed tomography-guided and endoscopic ultrasound-guided FNA of the pancreas in 51 cases of mucinous tumors. In the 19 cases in which the patients underwent surgical excision after FNA, the cytologic features were compared with the histopathologic findings. The remaining cases were categorized as one of the three above-mentioned types of mucinous tumors on the basis of cytomorphologic features identified as indicative of subtype among the cases in which a cytologic-histologic correlation was performed.
Among the 19 cases with cytologic-histologic correlation, 2 cases of serous cystadenoma and 2 cases of gastrointestinal duplication cyst were misdiagnosed on cytologic analysis as low-grade MCN. Among the remaining 15 cases, cytologic features by histologic diagnosis were as follows: ductal adenocarcinoma (n = 4): moderate to high cellularity, mild to moderate background mucin, three-dimensional clusters, high nuclear cytoplasmic ratios, and mild to moderate nuclear membrane irregularities; low-grade MCN (n = 5): mild to moderate cellularity, abundant background mucin, small clusters, and flat sheets of relatively bland glandular cells; mucinous cystadenocarcinoma (n = 1): similar to ductal adenocarcinoma but more abundant background mucin; and IPMT (n = 5): moderate to high cellularity, abundant background mucin, and prominent papillary arrangement of tall columnar cells with mild to moderate nuclear atypia. The remaining 32 cases were categorized based on cytology alone as adenocarcinoma with mucin production (n = 24) and low-grade MCN (n = 8).
Pancreatic neoplasms associated with significant amounts of extracellular mucin production include mucin-producing ductal adenocarcinoma, mucinous cystic neoplasm (MCN), and intraductal papillary mucinous tumor (IPMT) (World Health Organization classification of tumors).1
Ductal adenocarcinoma is the most common of these tumor types. It most often develops in older individuals and presents as a firm, ill-defined mass usually involving the head of the pancreas. Ductal adenocarcinoma associated with significant mucin production differs from conventional ductal adenocarcinoma not otherwise specified with regard to the amount of extracellular mucin that is present in the background of the tumor.
MCNs are uncommon tumors, representing 2–5% of all exocrine pancreatic tumors.1 MCNs occur predominantly in middle-aged women, with a peak incidence in the fourth to fifth decades.1–3 Clinical symptoms are usually nonspecific. MCNs most often are found in the body and tail of the pancreas as unilocular or multilocular cystic masses filled with thick mucoid material and surrounded by an ovarian type of stroma.1, 2 The cystic mass does not communicate with the pancreatic duct system. These tumors are classified as mucinous cystadenomas, borderline mucinous tumors, or mucinous cystadenocarcinomas according to the degree of atypia of the mucinous epithelium lining the cystic spaces.
IPMTs are characterized by dilatation of the main pancreatic duct or its branches, which are filled with mucus and lined by a papillary proliferation of the lining epithelium encompassing varying degrees of atypia. Unlike MCNs, IPMTs are more common in men and occur mainly in the sixth and seventh decades of life. IPMTs present with abdominal discomfort and are associated with mild elevation of pancreatic enzyme levels, mimicking the clinical symptoms of chronic or recurrent episodes of pancreatitis.
Fine-needle aspiration (FNA) biopsy often is used in conjunction with radiologic studies for the initial investigation of pancreatic masses. Because the appropriate management of pancreatic neoplasms associated with mucin production may differ by tumor type, it is important to categorize these tumors as accurately as possible on the basis of FNA smears. The current study was designed to identify distinctive cytomorphologic features that would be useful for the categorization of pancreatic mucinous tumors of the pancreas.
MATERIALS AND METHODS
The cytopathology files of The University of Texas M. D. Anderson Cancer Center were searched for cases in which FNA of the pancreas was performed between January 1995 and December 2002. Approval was obtained from the Institutional Review Board for conducting the study. A total of 830 cases were identified. Of these, 51 corresponded to some type of pancreatic mucinous tumor and were the subject of this study.
Nine of the 51 pancreatic tumors were aspirated by an interventional radiologist under computed tomography guidance with a coaxial technique. Patients initially were scanned in the prone position. Once a suitable aspiration site was selected, the overlying skin was marked and prepared using standard sterile technique, and a local anesthetic then was administered. FNA then was performed with the posterior approach using either an 18-gauge or a 22-gauge Chiba needle.
In the remaining 43 cases, endoscopic ultrasound guidance was used for localization of the tumors and FNA was performed by a gastroenterologist. Patients first were evaluated with endoscopy of the upper gastrointestinal tract; this was followed by endoscopic ultrasound examination. Under direct visualization using a Pentax FG-36UX echoendoscope, (Pentax Precision Instrument Corp., Orangeburg, NY) transgastric or transduodenal FNA was performed using either a Pentax 23-gauge, 4-cm needle or a 22-gauge, 10-cm Wilson-Cook needle (Wilson-Cook Medical GI Endoscopy, Winston Salem, NC).
An average of three passes were made per case, and the specimens were assessed immediately by cytopathologists for specimen adequacy. Direct smears were prepared and were either air-dried for Diff-Quik staining (American Scientific Products, McGraw Park, IL) or fixed in alcohol (modified Carnoy solution) for Papanicolaou staining. The syringe was rinsed in RPMI medium and subjected to centrifugation. Depending on the size of the pellet in the sediment, either cytospin or cell block preparations were made. Cytospins were stained with the Papanicolaou stain. The cell blocks were fixed in formalin, embedded in paraffin, and stained with hematoxylin and eosin. Histochemical staining for mucicarmine, with adequate control, was performed in selected cases.
In 19 of the 51 cases, the patients underwent subsequent surgical excision. The tissue specimens for these cases were retrieved from the files and the specimens were reviewed to confirm the histologic diagnosis. In each of these cases the corresponding FNA material was reviewed retrospectively to identify salient cytomorphologic features that might aid in the categorization of pancreatic mucinous tumors.
The FNA specimens were analyzed for cellularity, background (mucinous or necrotic), architecture (honeycombed flat sheets, three-dimensional clusters, or papillary clusters), nuclear characteristics (membrane, chromatin, pleomorphism, presence of nucleoli, and mitosis), nuclear-cytoplasmic ratio, and intracytoplasmic mucin. These features were scored semiquantitatively on a scale of 1+ to 3+.
On the basis of the cytomorphologic characteristics of the 19 cases in which cytologic histologic correlation was performed, the remaining 32 cases of pancreatic mucinous tumors were reclassified cytologically.
Nine women and 10 men underwent surgical excision after FNA of the pancreas. Histopathologically, these cases were classified as mucin-producing ductal adenocarcinoma in four cases, mucinous cystadenomas in five cases, mucinous cystadenocarcinoma in one case, IPMT in five cases, serous cystadenoma in two cases, and gastrointestinal duplication cyst in two cases. The clinical features, radiologic findings, and cytologic and histologic diagnoses for the 19 cases in which cytologic-histologic correlation was possible are listed in Table 1.
Table 1. Clinical Features, Radiologic Findings, and Cytologic-Histologic Diagnoses in 19 Patients with Pancreatic Mucinous Tumors
Mucinous papillary neoplasm suspicious for adecarcinoma
2.5-cm cystic mass, body
3.5-cm cystic mass head
2.6-cm, septated cystic mass, head
4.5-cm multicystic mass, head
GI duplication cyst
4.0-cm cystic mass, body
GI duplication cyst
3.5-cm septated cystic mass, head
Consistent with MCN
2.3-cm cystic mass, body
Consistent with MCN
The four cases of mucin-producing ductal adenocarcinoma and one case of mucinous cystadenocarcinoma displayed cytologic features of malignancy on FNA smears. These features included high cellularity, with cells arranged in three-dimensional clusters or loosely cohesive groups; a high nuclear-cytoplasmic ratio; irregular nuclear membranes; coarse chromatin; the presence of nucleoli; sparse mitosis; and background necrosis. All the cases demonstrated abundant mucin in the background (Fig. 1).
The aspirates of the five cases of mucinous cystadenoma exhibited moderate to high cellularity and were characterized by small clusters and honeycomb sheets of relatively bland mucin-containing columnar cells present in a background of abundant mucin. The nuclei had regular nuclear membranes, fine chromatin, and inconspicuous nucleoli (Fig. 2).
The aspirates of the five cases of IPMT were highly cellular, with cells arranged predominantly in tall papillary groups with a few flat sheets and small clusters distributed in a background of abundant mucin (Fig. 3). The papillae were lined by mucin-containing columnar cells with basally located nuclei. The nuclear features of the IPMTs were more variable than those of the low-grade MCNs; some cases had relatively bland nuclei and some had mild nuclear atypia. Only one case was found to demonstrate moderate atypia characterized by enlarged nuclei with irregular nuclear membranes and chromatin distribution.
Two cases each of gastrointestinal duplication cysts and serous cystadenomas initially were diagnosed as MCNs on cytology. The aspirates of the two cases of gastrointestinal duplication cysts (Fig. 4) had features similar to those of low-grade MCN, including a few sheets and clusters of relatively bland columnar cells in a background of abundant mucin. These cystic lesions were interpreted radiologically as being localized in the pancreas. The aspirates of the two cases of serous cystadenoma were paucicellular specimens. In the first case, the specimen was comprised of very few clusters of benign, mucin-containing cells with minimal amounts of extracellular mucin focally; in the second case, the specimen was comprised of few clusters of bland, columnar, mucin-containing cells with a moderate amount of extracellular mucin in the background. In retrospect, it was apparent that these two cases were nondiagnostic specimens on the basis of which a definitive diagnosis should not have been rendered. In neither of the two cases did the FNA smears contain diagnostic material representing a serous cystadenoma.
The relation between the cytomorphologic findings and histologic findings in the 19 cases in which a cytologic-histologic correlation was possible are shown in Table 2. The remaining 32 cases were categorized cytologically as follows: adenocarcinoma with mucin production (n = 24) and low-grade MCN, (n = 8) (Table 3).
Table 2. Cytomorphologic and Histologic Features of 19 Cases of Cytologically-Histologically Correlated Pancreatic Mucinous Tumors
Serous cystadenoma with ductal mucinous hyperplasia (1)
N/C ratio: low
Table 3. Cytologic Cateogorization of 32 Cases of Pancreatic Mucinous Tumors in Which a Cytologic-Histologic Correlation Was Not Possible
Cytologic diagnosis (no. cases)
3-D: three-dimensional; ICM: intracytoplasmic mucin; N/C: nuclear-cytoplasmic. Differential diagnosis includes mucin-producing ductal adenocarcinoma and mucinous cystadenocarcinoma, either in situ or invasive types; P: present; A: absent.
2+ to 3+
3-D clusters, loose cohesive groups, and honeycomb sheets
Mucin: 2+ to 3+
Shape: columnar/cuboidal, signet ring
Mucin-producing adenocarcinoma (24)
N/C ratio: high
1+ to 3+
Small clusters and honeycomb sheets
Mucin: 2+ to 3+
Mucinous cystic neoplasm, low-grade (8)
N/C ratio: low
The management of pancreatic mucinous tumors depends on the specific type of neoplasm. For patients with MCNs (including mucinous cystadenomas, borderline mucinous cystic neoplasms, and mucinous cystadenocarcinomas in situ), partial pancreatectomy is the surgery of choice. In contrast, IPMTs may be associated with neoplastic epithelium extending along the pancreatic duct beyond the point of obvious tumor. The importance of frozen section evaluation of the pancreatic transection margin and the occasional need for total pancreatectomy is discussed with the patient prior to the operation. Therein lies the importance of defining preoperatively which patients may have IPMTs.4–6 For invasive mucinous cystadenocarcinoma and mucin-producing ductal adenocarcinoma, protocol based multimodality therapy is the treatment of choice.7, 8 Because management is dependent on the tumor type, preoperative diagnosis of the different types of pancreatic mucinous tumors is extremely important in making the right therapeutic choices.
A few investigators to date have studied and reported the cytologic features of small series of pancreatic cystic neoplasms. The bland cytologic features of low-grade mucinous neoplasms comprised of honeycomb sheets of mucin-containing cells and the overt malignant features of mucinous cystadenocarcinomas are clearly elucidated in these studies.9–13 Although the cytologic features of IPMT per se have been described previously,14 to our knowledge the distinction between IPMT and MCN on the basis of cytologic features has not been very well described to date. In addition, none of the previous studies of the cytologic features of pancreatic cystic neoplasms discussed the overlapping cytologic features of mucinous cystadenocarcinoma and mucin-producing ductal adenocarcinoma.
In the current study, a retrospective review of FNA specimens of pancreatic mucinous neoplasms that subsequently were surgically excised helped us establish the distinctive cytomorphologic features of these tumors and thereby formulate an approach to their cytologic diagnosis. The cytologic features of low-grade MCN and IPMT proved to be distinctive, enabling their recognition on FNA. The current study findings regarding the cytologic features of these two entities largely agree with the findings of previous investigators. However, these findings appear to conflict with those of Dodd et al.,11 who found significant overlap between the features of low-grade MCN and those of IPMT. While the IPMTs in the current study displayed tall papillae lined by mucin-containing columnar cells with minimal atypia, low-grade MCN was found to exhibit honeycomb sheets and clusters of mucin-containing columnar cells with, very rarely, small papillary clusters. Therefore, although low-grade MCN may demonstrate a few papillary clusters, they are not as tall, abundant, and striking as the clusters observed in IPMT. It also should be noted that because of the heterogeneity of the epithelial lining of MCNs, there may be discrepancies between the cytologic typing and subsequent histologic diagnosis of these tumors. Although we did not encounter such discrepancies, it is plausible that after thorough and extensive sampling of the MCN on histology, tumors classified as low-grade MCN on cytology may exhibit a spectrum of changes including not only those characteristic of mucinous cystadenoma but also those characteristic of borderline mucinous tumors and may even demonstrate areas of focal carcinoma. Similarly, with regard to IPMTs, there may be discrepancies because of sampling issues with regard to the degree of atypia in the tumor noted on cytologic and histologic analysis. It is important to remember that IPMTs may be either noninvasive or invasive, and this distinction cannot be made based on cytologic analysis.
The current study also highlighted the pitfalls that may be encountered in making a cytologic diagnosis of low-grade MCN. Two cases of serous cystadenoma were mistakenly diagnosed as low-grade MCN on the basis of limited material comprised of a few clusters of bland columnar cells in a background of focal mucin. These findings most likely reflected the sampling of normal structures adjacent to the serous cystadenomas. Because of the paucicellular material, the FNA samples from these two cases should have been categorized as nondiagnostic, and repeat sampling should have been recommended. This pitfall of mistaking normal epithelium for MCN on limited material also has been alluded to by Jones et al.10 The distinction between serous cystadenoma and MCN is extremely important from the treatment point of view, and FNA can be immensely helpful in this respect. Most patients with serous cystadenoma do not require resection unless they have abdominal pain, jaundice, or recurrent pancreatitis.4, 6 In contrast, virtually all MCNs should be resected. Radiologically, serous cystadenomas have multiple small (< 2 cm) cystic areas, often producing a starburst appearance because of a centrally located scar,4 whereas MCNs invariably contain several cystic areas measuring > 2 cm in greatest dimension. MCNs may be solitary and macrocystic, with papillary excrescences or may have septae within the cystic structure with even a solid component within the wall of the cystic mass.4 Cytologically, serous cystadenomas have a few sheets of cuboidal cells with clear glycogen-containing cytoplasm,12, 15 whereas MCNs have honeycombed sheets and clusters of mucin-containing cells. In addition, MCNs have abundant mucin in their background, which is not a feature of serous cystadenoma.
The other two cases that were falsely identified as low-grade MCN on cystologic analysis were gastrointestinal duplication cysts that demonstrated cytologic features that were indistinguishable from low-grade MCN. These two cases were presumed radiologically to be localized in the pancreas. In the absence of accurate radiologic localization, this pitfall would be unavoidable. Centeno et al.13 and Sperti et al.16 also have cautioned that a variety of nonpancreatic cysts such as adrenal cysts, mesenteric cysts, and mesenteric cystic lymphangioma may be mistakenly localized in the pancreas radiologically and that this may be a source of error in the diagnosis of low-grade MCN.
The clearly malignant nature of mucinous cystadenocarcinoma and mucin-producing ductal adenocarcinoma was well evident on FNA smears. Making a diagnosis of malignancy was relatively straightforward in these cases in the current study. However, these two entities exhibit similar cytologic features and to our knowledge cannot be distinguished from one another on FNA. Therefore, the cytologic findings need to be correlated with the radiologic findings before one diagnosis can be favored over the other. Even when a diagnosis of mucinous cystadenocarcinoma is suggested on FNA because of the presence of a predominantly multicystic pancreatic mass on radiology, it must be remembered that the distinction between in situ and invasive mucinous cystadenocarcinoma is not possible on FNA.
Low-grade MCNs and IPMTs have distinctive radiologic and cytologic features that may be utilized to diagnose them correctly and distinguish them from one another and from other cystic tumors such as serous cystadenomas and pancreatic pseudocysts. In contrast, although mucinous cystadenocarcinoma and mucin-producing ductal adenocarcinoma can be clearly recognized as malignant, to our knowledge they cannot be distinguished from one another cytologically. Familiarity with the salient radiologic features of mucinous pancreatic tumors and a knowledge of the cytologic features of these tumors, as well as the relevant potential pitfalls and limitations in cytologic analysis, may be instrumental in categorizing these tumors more accurately on FNA.
The authors would like to thank Yun Gong, M.D., and Wei Sun, M.D., for their help with the data search for this article.