CANCER
163
CYTOPATHOLOGY
Intraductal Papillary Mucinous Neoplasm of the Pancreas Cytologic Features Predict Histologic Grade
Paul J. Michaels, M.D.1 Elena F. Brachtel, M.D.1 Brenna C. Bounds, M.D.2 William R. Brugge, M.D.2 Martha Bishop Pitman, M.D.1
BACKGROUND. Intraductal papillary mucinous neoplasm (IPMN) is an increasingly recognized cystic neoplasm of the pancreas, histologically classified by the degree of epithelial atypia and by the presence or absence of invasion of the cyst wall. To the authors’ knowledge, the cytologic features of this neoplasm are poorly characterized, especially with respect to tumor grade. METHODS. Thirty-three endoscopic ultrasound (EUS)-guided pancreatic fine-nee-
1
James Homer Wright Pathology Laboratories and Cytopathology Laboratory, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts.
2
Department of Gastroenterology, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts.
The authors thank Joanne Schiavo for excellent secretarial assistance. Address for reprints: Martha Pitman, M.D., Department of Pathology, Warren 105-D, Massachusetts General Hospital, 55 Fruit St., Boston, MA 02114; Fax: (617) 724-6564; E-mail:
[email protected] Received May 10, 2005; revision received July 13, 2005; accepted August 9, 2005.
dle aspiration biopsy (FNAB) samples and 1 pancreatic duct brush specimen from 25 patients with a histologically confirmed IPMN were retrospectively reviewed. Blinded to tumor grade, background mucin, inflammation, necrosis, overall cellularity, the presence of gastrointestinal-contaminating epithelium, architecture of cell clusters, and nuclear and cellular morphology were evaluated. In cases in which special stains for mucin were performed, the diagnostic utility of these stains was assessed. These cytologic features were subsequently correlated with the histologic diagnosis. RESULTS. The 34 cytology samples represented 4 adenomas, 15 IPMN-moderate dysplasias, 7 intraductal carcinomas, and 8 IPMNs with invasive carcinoma. Extracellular mucin was present in 97% of all cases; 53% had thick, viscous, “colloidlike” mucin. Special stains for mucin were positive in 6 of 11 cases (54%), helping to identify thin mucin in only 2 cases. Gastrointestinal contamination did not appear to create diagnostic difficulty due to an apparent dual (dysplastic-nondysplastic) epithelial population, but only 4 adenomas were evaluated in this study. Necrosis distinguished IPMN with carcinoma from IPMN-adenomas and IPMN with moderate dysplasia (P ⬍ .00001), and was more often observed with invasion than IPMN-carcinoma in situ (P ⬍ .05). Tight epithelial cell clusters with hyperchromatic nuclei and a high nuclear to cytoplasmic ratio was more significant in IPMN of at least moderate dysplasia (P ⫽ .03). Pale nuclei with parachromatin clearing was found to be a nuclear feature that was suspicious for at least carcinoma in situ (P ⬍ .001). In addition, significant background inflammation (neutrophils and histiocytes) was found to be more characteristic of IPMN with at least carcinoma in situ (P ⫽ .002). CONCLUSIONS. The presence of thick, “colloid-like” mucin is noted in half of the IPMN cases, but was not found to be specific to grade. The absence of such mucin does not exclude an IPMN. The presence of tight epithelial cell clusters is consistent with a neoplasm of at least moderate dysplasia, and abundant background inflammation and parachromatin clearing correlated with the presence of at least carcinoma in situ. Necrosis was the only feature found to be strongly suggestive of invasion. Cancer (Cancer Cytopathol) 2006;108:163–73. © 2006 American Cancer Society.
KEYWORDS: intraductal papillary mucinous neoplasm, mucinous cysts, fine-needle aspiration biopsy, pancreas, pancreatic neoplasms, endoscopic ultrasound.
© 2006 American Cancer Society DOI 10.1002/cncr.21838 Published online 20 March 2006 in Wiley InterScience (www.interscience.wiley.com).
164
I
CANCER (CANCER CYTOPATHOLOGY) June 25, 2006 / Volume 108 / Number 3
ntraductal papillary mucinous neoplasm (IPMN) is an increasingly recognized mucin-producing cystic neoplasm of the pancreas first distinguished from mucinous cystic neoplasm and ductal adenocarcinoma in 1982.1 These neoplasms have been referred to in the past by a variety of terms, including mucin-producing tumor,1 intraductal mucin-hypersecreting neoplasm,2 mucinous duct ectasia,3 and villous adenoma of the main pancreatic duct.4 With the evolving clinical, radiologic, and histologic characterization of IPMN over the past 20 years, it is estimated that IPMN currently represents 3% to 5% of all pancreatic neoplasms and 20% of cystic neoplasms of the pancreas.5,6 Macroscopically, IPMNs are characterized by mucinous dilatation of the pancreatic ducts, with involvement of either the main duct alone (main duct type), side branch ducts (side branch type), or both (combined type). These neoplasms are classified histologically based on architectural complexity and the severity of epithelial atypia into IPMN-adenoma, IPMN with moderate dysplasia, or IPMN with carcinoma, either in situ or invasive (tubular type or colloid carcinoma).7 Compared with conventional ductal adenocarcinoma, IPMNs with invasive carcinoma have a better prognosis,8 and for IPMNs, the presence or absence of an invasive carcinoma is the most important prognostic factor for survival. The 5-year survival of a patient with a resected noninvasive IPMN is reported to be 90% to 100%, compared with a 40% 5-year survival rate for patients with an IPMN containing a focus of invasive carcinoma.9 –11 An invasive carcinoma can be predicted by the presence of a mass in the cyst wall as detected by endoscopic ultrasound (EUS), endoscopic retrograde cholangiopancreatography, or magnetic resonance cholangiopancreatography; however, such a visible mass is not always present in neoplasms that have invaded. Microinvasive neoplasms have a similarly reduced prognosis as those with endoscopic evidence of invasion, although the size of the invasive component has been shown to correlate with prognosis.12,13 Currently, to our knowledge there is no established or accurate means of preoperatively distinguishing an IPMN with invasive carcinoma from lower-grade neoplasms. Preoperative diagnosis of pancreatic cysts is currently performed using fine-needle aspiration biopsy (FNAB), increasingly performed by EUS. The use of EUS-FNAB has allowed for increased access to small neoplasms, but the introduction of gastric or duodenal epithelium and mucin into the specimen during the procedure (eg, gastrointestinal [GI] contamination) has created diagnostic challenges, particularly in the area of mucinous cysts of the pancreas.14 To our knowledge, the impact of GI contamination on the
ability to make a cytologic diagnosis on EUS-FNAB of pancreatic cysts has received little attention to date. Although the sensitivity and specificity of the diagnosis of pancreatic cysts in general is low,15 and the cytologic diagnosis of IPMN often underestimates the grade of the neoplasm,16,17 cytology is still considered a necessary adjunct to EUS for the preoperative diagnosis of pancreatic cysts.15 Given the common location of IPMNs in the head of the pancreas, a pancreaticoduodenectomy (a Whipple procedure) is the typical surgical treatment. Although the morbidity and mortality associated with a pancreaticoduodenectomy have been significantly decreasing in recent years, many patients with IPMNs are often in their sixth or seventh decade of life, an age at which other comorbid conditions are frequently present, making the procedure more dangerous. Although all suspected IPMNs should ideally be removed given the current limitations of accurate preoperative diagnosis, a subset of patients with IPMNs who are poor surgical candidates may benefit from close clinical follow-up rather than surgical excision if their cystic neoplasms can be identified preoperatively as low-grade neoplasms. To our knowledge, only a few studies to date have evaluated and compared the cytologic and histologic findings in IPMNs,16 –19 but none have attempted to grade IPMNs by cytology and few have addressed the impact of GI contamination on the ability to make a cytology diagnosis on EUS-FNAB.17,18 In the current study (which to our knowledge is the largest series reported to date), we analyzed the cytologic features of 33 specimens obtained preoperatively by EUS-FNAB and 1 main duct brush, all with histologic follow-up, and correlated the various cytologic features with histologic grade. We also assessed the contribution of the special stains for mucin on the ability to diagnose a mucinous cyst and the impact of GI contamination on diagnostic outcome.
MATERIALS AND METHODS A search of the laboratory information system of the Department of Pathology at Massachusetts General Hospital identified 25 patients with histologically proven IPMNs, for whom 34 preoperative cytology specimens were available. The histology slides were reviewed for consistency in grading and classification according to the current accepted nomenclature.7 The final classification rested with the review diagnosis. In cases with a heterogeneous lining, the highest degree of atypia was used. The original diagnosis was compared with the histologic diagnosis. The cytology specimens consisted of direct smears stained using a custom laboratory-made Pa-
Cytologic Grading of IPMN of Pancreas/Michaels et al.
panicolaou (Pap) stain. Air-dried, Romanowskystained smears were not evaluated. A subset of cases were also stained with hematoxylin and eosin for rapid interpretation. Specimens that were received in the laboratory as cyst fluid typically were processed as a cytospin (Thermo-Shandon Instruments, Pittsburgh, PA) or ThinPrep威 (Cytyc Corporation, Marlborough, MA) preparation. If the fluid was sufficient in quantity, 2 cytospins were made for mucicarmine and Alcian blue (pH 2.5) mucin stains. All original cytology reports were reviewed, but the cytology specimens were blindly evaluated without knowledge of the IPMN histologic grade to assess the cytologic features present. The original diagnoses were classified as nondiagnostic (ND) if reported as such or if the report gave a vague descriptive diagnosis without mention of a mucinous neoplastic process, as negative for malignancy, atypical, suspicious for malignancy, or positive for malignancy. These original diagnoses were compared with the histologic diagnosis. The background and cytologic features evaluated included overall cellularity, presence and quality of extracellular mucin, necrosis, and background inflammatory cells. Background mucin was assessed as either thick or thin. Thick mucin was defined as an easily detectable, opaque, “colloid-like” coagulum of mucus diffusely or focally present on the slide, or readily recognized as thick mucinous wisps on cytospins or ThinPrep. Thin mucin was defined as wispy mucus strands diffusely or focally present on the slide of all preparations. The presence or absence of inflammation and necrosis also was assessed. Specimen cellularity was divided into three groups: low, moderate, and marked cellularity. Each specimen was categorized into 1 of these groups based on the percent of the slide that contained epithelium. All cytology specimens were assigned to a group based on the overall density of epithelial cells at low power, excluding obvious GI contamination. Specimens labeled as low cellularity had ⬍10% of the specimen area of the glass slide covered with epithelium, those with 10% to 40% epithelium were graded as moderate cellularity, and those with ⬎ 40% epithelium were deemed high cellularity. GI epithelium was recognized by criteria outlined by Nagle et al.20 Duodenal epithelium forms large, often folded, flat monolayered sheets studded with goblet cells. The enterocytes are dense and nonvacuolated. A luminal edge and brush border are often present. Gastric epithelium more often presents as smaller monolayered sheets without a brush-bordered
165
luminal edge and without goblet cells. The cytoplasm may or may not be visibly vacuolated. The architectural features of the epithelial cells, including papillary fragments, monolayered sheets, single mucinous or dysplastic cells, and tight epithelial cell clusters, were assessed. Three-dimensional epithelial clusters, with or without a fibrovascular core, that were at least 2 to 3 times as long as wide were considered papillary fragments. In contrast, tight epithelial cell clusters were characterized by rounded groups of epithelial cells with hyperchromatic nuclei, a high nuclear-to-cytoplasmic (N:C) ratio, and often prominent cellular wrapping. Single dysplastic cells were intact epithelial cells with a high N:C ratio resembling severe squamous dysplasia on a cervical smear. We also evaluated the specific cellular and nuclear features of the cells, such as the presence or absence of vacuolated cytoplasm, mitoses, parachromatin clearing (open nuclei with peripherally clumped chromatin similar to papillary carcinoma of the thyroid), prominent nucleoli, and nuclear grooves. The cytologic characteristics of individual specimens were compared with one another based on the histologic grade of the resected IPMN. The statistical analysis of the comparison between the cytology specimen features and the histologic grade of the IPMN was performed using the chisquare test.
RESULTS Clinical Features All 25 patients underwent both preoperative procedures and resections at Massachusetts General Hospital. There were 13 men and 12 women with a mean age of 69 years (range, 43-84 yrs). Of the 25 neoplasms, 17 (68%) were present in the head of the pancreas, 6 were identified in the body/tail region, and 2 diffusely involved the pancreatic ductal system. The IPMNs were categorized as follows: 2 main duct type (both with moderate dysplasia), 11 side branch type (4 adenomas, 6 with moderate dysplasia, and 1 carcinoma in situ [CIS]), and 12 combined type (4 with moderate dysplasia, 3 CIS, and 5 invasive [4 with tubular and 1 with colloid carcinoma]). The 34 cytology specimens were samples representing 4 IPMN-adenomas, 15 IPMN-moderate dysplasias, and 15 IPMN-carcinomas (7 CIS and 8 invasive). Given the small number of adenomas, adenomas and neoplasms with moderate dysplasia were combined into a category of “low grade” for statistical analysis. Among the 13 men, 5 had IPMN-carcinomas (2 CIS and 3 invasive) and the remaining 8 had low-grade IPMNs. Similarly, 3 of the 12 women had either CIS (1
166
CANCER (CANCER CYTOPATHOLOGY) June 25, 2006 / Volume 108 / Number 3
TABLE 1 Comparison of Original Cytology and Final Histology Diagnoses
TABLE 2 Cytologic Features in 34 Cytology Specimens of IPMN
Original diagnosis
Adenoma
Moderate dysplasia
Carcinoma in situ
Invasive carcinoma
Total
Nondiagnostic Negative Atypical Suspicious Positive Total
2 1 0 1 0 4
3 5 6 1 0 15
1 0 2 2 2 7
0 1 1 1 5 8
6 7 9 5 7 34
woman) or invasive (2 women) IPMNs, whereas the other 9 women had low-grade IPMNs. There was no statistically significant difference noted in the number of IPMN carcinomas when comparing men and women.
Thick mucin ⱖ Moderate cellularity Inflammatory background Necrosis Papillary fragments Tight epithelial cell clusters Nucleoli Open/even chromatin Mitosis Grooves Vacuolated cytoplasm
Low-grade IPMN (n ⴝ 19) (%)
IPMN-CIS (n ⴝ 7) (%)
IPMN with invasive carcinoma (n ⴝ 8) (%)
10 (53) 2 (11) 8 (42) 0 1 (5) 10 (53) 6 (40)* 1 (7)* 1(5) 5 (33)* 10 (67)*
4 (57) 5 (71) 7 (100) 3 (43) 3 (43) 6 (86) 6 (100)* 5 (83)* 0 4 (67)* 3 (50)*
4 (50) 4 (50) 7 (88) 8 (100) 5 (63) 8 (100) 5 (63) 5 (63) 0 4 (50) 8 (100)
IMPN: intraductal papillary mucinous neoplasm; CIS: carcinoma in situ * Absent epithelial component, therefore n ⫽ 15 for low-grade IPMN and n ⫽ 6 for CIS.
Specimen Type and Preparation The specimens consisted of 25 cyst aspirates, 6 dilated pancreatic duct aspirates, 2 cyst wall mass aspirates, and 1 main duct brush. Nineteen patients had 1 specimen, 4 patients had 2 specimens, 1 patient had 3 specimens, and 1 patient had 4 specimens, all of which were multiple specimens obtained during the same procedure. Direct smears were prepared in 18 cases, 1 of which also had a ThinPrep. Cytospins were prepared in 12 cases, 4 of which had ThinPrep smears, including 1 with a cell block.
Special stains Special stains for mucin were performed in 11 cases, and detected mucin in 6 cases. In all 6 cases, the mucin was visible on routine preparations. Four cases demonstrated thick mucin that was visible on Papstained direct smears (n ⫽ 2 cases) or ThinPrep (n ⫽ 2 cases), and 2 demonstrated thin mucin on Papstained cytospins. Five cases did not demonstrate mucin on special stains, but routine stains showed thin mucin in 4 of 5 Pap-stained cystospins.
Original cytologic diagnoses and comparison with histology The original cytologic diagnoses compared with the final histologic diagnosis are outlined in Table 1. There were 6 ND cytology reports (2 adenomas, 3 moderate dysplasias, and 1 CIS), 7 negative diagnoses (1 adenoma, 5 moderate dysplasias, and 1 invasive carcinoma). 9 atypical diagnoses (6 moderate dysplasias, 2 CIS, and 1 invasive carcinoma), 5 suspicious diagnoses (1 adenoma, 1 moderate dysplasia, 2 CIS, and 1 invasive carcinoma), and 7 positive diagnoses (2 CIS and 5 invasive carcinomas). There were no false-positive cytology diagnoses, but in 1 case the cytology was sus-
picious in 1 adenoma and there was 1 IPMN-moderate dysplasia. The original cytology diagnosis either failed to report the presence of at least a mucinous cyst in general (6 ND cases) or underdiagnosed the grade of the neoplasm in 12 (35%) cases, including 5 IPMNs with moderate dysplasia that were called negative (n ⫽ all 5 cases), 4 IPMN-CIS that were called either atypical (n ⫽ 2 cases) or suspicious (n ⫽ 2 cases), and 3 invasive carcinomas that were called either negative (n ⫽ 1 case), atypical (n ⫽ 1 case), or suspicious (n ⫽ 1 case). This underdiagnosis stemmed from the absence of epithelium (n ⫽ 4 cases) or sampling error (n ⫽ 8 cases).
Cytologic features The cytologic features stratified by histologic grade are outlined in Table 2.
Smear Background Mucin was present in the background in all but 1 of the 34 specimens. In 18 of the 34 (53%) cases, the mucin was thick, viscous, and at least focally “colloidlike” (Fig. 1A,B). The easy identification of mucin as fibrillary material on a ThinPrep slide qualified as thick mucin (Fig. 1C). In the remaining cases, the mucin was thin and watery. Thin mucin was more easily detected on direct smears (Fig. 2A) than on ThinPrep (Fig. 2B) or cytospin (Fig. 2C) preparations. Mucicarmine and Alcian blue stains highlighted the obvious mucin present, and only confirmed the presence of thin mucin in 2 of 6 cases identified on Papstained preparations (Fig. 1D). The mucin quality did not appear to correlate with the histologic grade of the neoplasm.
Cytologic Grading of IPMN of Pancreas/Michaels et al.
167
FIGURE 1. Thick, viscous “colloid-like” mucin was present (A) diffusely or (B) focally on direct smears and thick mucin detected on ThinPrep slides (C) was apparent as clumps of fibrillar material. Special stains for mucin (D) can highlight thin mucin, as shown on this cytospin preparation (A-C, Papanicolaou stain; D, Alcian blue [pH 2.5]). Original magnification ⫻200 (A–D).
Necrosis (Fig. 3) was noted in 11 of the 15 IPMNcarcinomas but was not observed in any of the 19 low-grade IPMNs (73% vs. 0%; P ⬍ .0001). In addition, when solely evaluating IPMN-carcinomas, necrosis was identified in all 8 cases of IPMNs with an associated invasive carcinoma but was found in only 3 of the 7 in situ carcinomas (100% vs. 43%; P ⬍ .01). Similarly, inflammatory cells, typically neutrophils and histiocytes (Fig. 4), were identified in the background of 93% (14 of 15) of the IPMN-carcinomas compared with less than half (8 of 19) of the specimens from low-grade IPMN (93% vs. 42%; P ⫽ .001).
Of the IPMN-carcinomas, 9 of the 15 specimens (60%) were characterized as having either a moderate or marked degree of cellularity at low magnification. Interestingly, the in situ carcinomas displayed an overall higher cellularity (5 of 7 cases, 71%) than the invasive carcinomas (4 of 8 cases, 50%). In contrast, only 2 of the 19 (10%) low-grade IPMNs were identified as having a similar degree of cellularity (60% vs. 10%; P ⬍ .01). The absence of epithelial cells did not exclude carcinoma, as 1 case of IPMN-carcinoma in situ was acellular, although abundant thick “colloid-like” mucin was present, allowing for a diagnosis of a neoplastic mucinous cyst.
Cellularity
GI Contamination
Epithelial cells, either in sheets, clusters, or scattered singly in the background, were noted in 28 of 34 cases (82%). Overall hypercellularity of the cytology specimens was found to be correlated with higher histologic grade.
Prominent GI contamination was identified in only 2 of 34 (6%) of the cases (Fig. 5A). In these 2 cases, on review we, as well as the original diagnostic pathologist, felt that there was a definite distinction between
168
CANCER (CANCER CYTOPATHOLOGY) June 25, 2006 / Volume 108 / Number 3
FIGURE 2. Thin mucin was composed of faint wispy mucoid threads on (A) direct smears and was found to be more attenuated on (B) ThinPrep and (C) cytospin preparations (A, Papanicolaou stain). Original magnification ⫻200 (A–C). the neoplastic mucinous epithelium and the benign contaminating epithelium of the GI tract because the background neoplastic cells were obviously dysplastic. However, the initial cytologic diagnosis either questioned or favored the source of the epithelium present on the slides in 3 other cases as being from the GI tract (all of which were cases of IPMN-moderate dysplasia). In all these cases, the epithelium was atypical enough on blinded review to distinguish between gastric and duodenal epithelium (Fig. 5B). None of the 4 cases of adenoma contained GI-contaminating epithelium. As a result, the presence of GI contamination did not impact our ability to make a diagnosis of IPMN in the current study, although, as stated above, it did on the initial diagnosis in 3 of 34 (9%) cases.
Architectural Arrangement FIGURE 3. Background necrosis presented as a coagulum of cellular debris diffusely or in clumps as on this ThinPrep slide (Papanicolaou stain, original magnification ⫻400).
Papillary fragments (Fig. 6) were noted in 8 of the 15 IPMN-carcinomas, 3 of 7 (43%) in situ carcinomas, and 5 of 8 (62%) invasive carcinomas. In contrast, only 1 of the 19 low-grade IPMNs contained any papillary
Cytologic Grading of IPMN of Pancreas/Michaels et al.
169
FIGURE 4. Background inflammation was significant in the form of abundant neutrophilic inflammation admixed with histiocytes (Papanicolaou stain, original magnification ⫻400). tissue fragments (53% vs. 5%; P ⫽ .001). No statistical difference was noted among the histologic grades of IPMNs with respect to the presence of monolayer sheets of epithelial cells or the identification of single mucinous epithelial cells. Small, tight epithelial budlike clusters (Fig. 7A) and single dysplastic cells (Fig. 7B) were identified in 10 of 19 (53%) low-grade IPMNs, 6 of 7 (86%) IPMN with CIS, and 8 of 8 IPMN with invasive carcinomas. Although this feature did not distinguish CIS from invasive carcinomas, or even carcinomas from neoplasms with moderate dysplasia, the presence of these groups were noted in 23 of 30 IPMNs that were of moderate dysplasia or higher and in only 1 case of an IPMN-adenoma (77% vs. 25%; P ⫽ .03).
Cellular Features Prominent nucleoli and parachromatin clearing (Fig. 8) were nuclear characteristics that were noted more often in IPMN-carcinomas compared with low-grade IPMN. Four cases of low-grade IPMN and 1 case of IPMN-CIS failed to contain sufficient epithelial cells to evaluate nuclear features. Of the 14 specimens from patients with IPMN-carcinomas, 11 (78%) had prominent nucleoli, compared with only 6 (40%) of the 15 low-grade IPMN (78% vs. 40%; P ⫽ .03). Similarly, 10 of 14 cases (71%) of CIS or invasive carcinoma had parachromatin clearing compared with only 1 of the 15 (7%) low-grade IPMNs (71% vs. 7%; P ⬍.001). The majority of cases of low-grade neoplasms had a chromatin pattern best described as granular or finely speckled. Mitoses, nuclear grooves, and cytoplasmic vacuolization were not helpful in distinguishing between the various histologic grades of IPMNs.
FIGURE 5. (A) Gastric epithelium was recognized by the presence of large, orderly, often folded monolayered sheets of glandular epithelium with a luminal edge with or without mucin. The edge is lined by foveolar cells with mucinous cytoplasm associated with thin mucin. (B) Dysplastic epithelium from an intraductal papillary mucinous neoplasm-moderate dysplasia initially favored to be gastrointestinal (GI) contamination on initial evaluation was distinguished from GI contamination by the disorderly nuclei with slight nuclear irregularity (A, Papanicolaou stain). Original magnification ⫻200 (A); ⫻400 (B).
DISCUSSION Intraductal papillary mucinous neoplasms are cystic neoplasms of the pancreatic ductal system. These neoplasms are histologically classified based on the degree of epithelial and architectural atypia of the cyst lining into adenomas, neoplasms with moderate dysplasia, CIS, and IPMN with invasive carcinoma. Despite the fact that most patients with low-grade neoplasms (adenomas and IPMN-moderate dysplasia) will outlive the likely progression of these neoplasms to invasive carcinoma,21 the recommended management of all IPMNs is surgical resection given the current inability to accurately diagnose and grade these neoplasms preoperatively. In this study of various cytologic features of these neoplasms sampled primarily
170
CANCER (CANCER CYTOPATHOLOGY) June 25, 2006 / Volume 108 / Number 3
FIGURE 6. Papillary fragments are characterized by epithelial groups that are at least twice as long as they are wide. These cell clusters may or may not have a fibrovascular core (Papanicolaou stain, original magnification ⫻400).
FIGURE 8. Parachromatin clearing was typically associated with intraductal papillary mucinous neoplasm (IPMN) carcinomas, and rarely was observed in low-grade IPMNs (Papanicolaou stain, original magnification ⫻1000).
FIGURE 7. (A) Tight epithelial cell clusters are small groups of small epithelial cells with a high nuclear-to-cytoplasmic ratio and hyperchromatic nuclei. (B) Single dysplastic epithelial cells resembled severe squamous dysplasia on a cervical smear (Papanicolaou stain, original magnification ⫻1000).
by EUS-guided FNAB, we correlated a number of cytologic features, including smear background characteristics, epithelial architectural features, and cellular details, with the histologic grade of the IPMN in 34 aspirates from 25 neoplasms in an attempt to preoperatively classify them on aspirate cytology. The mucin observed in the background of these specimens varied. In a slight majority of cases (53%), thick, viscous, at least focally “colloid-like” mucin was identified in contrast to thin, watery mucin. The quality of the mucin did not appear to correlate with the grade of the neoplasm. In other words, specimens composed of diffuse, thick “colloid-like” mucin without an atypical epithelial component did not exclude a high-grade IPMN (moderate dysplasia or higher). In previous studies, this thick and viscous quality of the mucin was deemed sufficient to distinguish the source of the mucin as originating from the cyst and not from the GI tract, warranting a diagnosis of a neoplastic mucinous cyst.18 Such mucin was not noted in direct samples from the GI tract.20 As should be expected, mucin stains in such cases will be positive, as was the case in all 4 of the cases in the current series with thick mucin in which stains were performed. The need for special mucin stains is not necessary for the diagnosis of a neoplastic mucinous cyst in these cases because the presence of this “neoplastic” mucin is apparent on the routine stains and no doubt correlates with the macroscopic appearance of thick mucoid fluid aspirated during the biopsy. Given that we only examined histologically confirmed IPMN and not other cystic lesions in and around the pancreas, we cannot say from this study that thick and viscous mucin is only encountered in aspirates of neoplastic mucinous cysts
Cytologic Grading of IPMN of Pancreas/Michaels et al.
(albeit this has been our experience), but we can say that the absence of such thick and viscous mucin does not exclude an IPMN. Thin and watery mucin is nonspecific and is often seen in association with GI contamination (unpublished data), and therefore does not necessarily signify the presence of an IPMN. Identification of degenerated cells, atypical mucinous epithelial cells, and histiocytes within the mucin are also helpful and reassuring that the mucin present is likely associated with a neoplastic process.18,20,22 Special stains for mucin may not highlight thin mucin even if noted on the routinely stained specimen. This was the case in 6 of our cases, in which thin mucin was present on routine stains but mucin stains were positive in only 2 cases. Also, even with a positive mucin stain, the source of thin mucin is ambiguous and could be GI contamination. As such, the clinical utility of performing mucin stains at all in these cases remains questionable. GI contamination is a known diagnostic pitfall encountered when evaluating EUS-guided FNAB.17,18,23 The contaminating epithelium is either duodenal or gastric in origin. Surprisingly, only 2 of the 34 specimens in the current study had significant GI contamination, none of which were adenomas. In each case, distinguishing the contaminant epithelium from the neoplastic epithelium was not difficult, aided by the fact that in both of these cases an obvious 2-cell population was present. Three cases had GI contamination noted as the probable source of the epithelium on the original cytology report but were not identified on blinded review as cases with GI contamination. All three were IPMN-moderate dysplasia, and the epithelium was deemed atypical enough to warrant such an interpretation. The lack of uniform nuclear spacing and focal nuclear crowding and overlap, as well as subtle nuclear membrane abnormalities, are features not described in either duodenal or gastric epithelium.20 High-grade nuclear features and papillary architecture clearly distinguished the neoplastic epithelium from the bland, organized nuclear arrangement of the cells in the epithelium from the GI tract. However, we acknowledge that differentiating adenomatous epithelium in mucinous cysts from GI epithelium may be impossible in certain cases. Background features identified at low magnification are often helpful in suggesting a diagnosis. We found that the presence of necrosis was highly associated with IPMN carcinomas (both CIS and invasive lesions), and was the only feature that was significantly different between CIS and invasive carcinoma (43% vs. 100%, respectively; P ⬍ .01). The distinction between invasive and noninvasive carcinomas is of critical importance for patient prognosis. The 5-year
171
survival rate for surgically resected noninvasive IPMNs is close to 100%.9,12,24,25 In contrast, the 5-year survival rate for IPMNs associated with an invasive carcinoma is ⬍ 50%.9,24,26 Similar to the association with necrosis, we found the presence of background inflammation to be important in distinguishing low-grade IPMNs from CIS and invasive IPMNs (42%, 100%, and 88%, respectively). This finding is particularly important to keep in mind when the clinical and radiologic differential diagnosis includes a pancreatic pseudocyst, given that inflammation is not an uncommon finding in the cyst fluid aspirated from pseudocysts. Therefore, the presence of numerous background inflammatory cells in an FNAB of a cystic pancreatic mass should prompt a thorough search for mucin, as well as epithelial cells, which, by definition, are absent from aspirates of a pseudocyst. Papillary fragments, as the name of the tumor implies, are not uncommonly seen in intraductal papillary mucinous neoplasms, and are, in fact, quite characteristic. However, we found an association between papillary fragments and only IPMN carcinomas. One can think of papillary fragments as a manifestation of marked architectural complexity seen in IPMNcarcinoma. Therefore, it is not surprising that papillary fragments are more likely to be found in those neoplasms that characteristically have the highest degree of cell proliferation and papillary in foldings. Papillary fragments were 1 of 2 features found to distinguish IPMN from other cystic lesions of the pancreas in a recent study,17 the other being intracellular mucin, but the significance of this feature with respect to the grade of IPMN was not explored. Similarly, tight epithelial cell clusters (TECs) were seen in the IPMNs with at least moderate dysplasia. TECs represent the cellular projections or papillary buds that are often found histologically at the tips of papillae and tufted along the lining of IPMN cysts with at least moderate dysplasia. We found the presence of these groups corresponded to an IPMN of at least moderate dysplasia in all but 1 case (23 of 24 neoplasms with TECs present). Hara et al.27 found “small papillary cohesive clusters” only in carcinomas, mostly intraductal carcinomas. Therefore, as a practical issue, the identification of TECs should alert one to the presence of at least moderate dysplasia. Layfield and Cramer17 found that epithelial cells with 2⫹ to 3⫹ nuclear atypia with nuclear molding, prominent nucleoli, nuclear irregularity, and cell crowding were correlated with nuclear grade as well; however, there was no case of an IPMN-adenoma in their study and only 1 case of a mucinous cystadenoma. In addition to architectural features, cellular fea-
172
CANCER (CANCER CYTOPATHOLOGY) June 25, 2006 / Volume 108 / Number 3
tures such as chromatin pattern, nucleoli, nuclear membranes (grooves), and cytoplasmic features are important features to consider when grading IPMNs. Because many IPMN carcinomas are well differentiated, it is imperative to rely on subtle cellular features of malignancy, as has been described in conventional ductal adenocarcinoma.28,29 In this series of cases, nuclear grooves were present in both low-grade IPMN and carcinomas, both CIS and invasive, with no significant differences noted between the groups (P ⫽ .20). Prominent nucleoli were observed more often in IPMN carcinomas (11 of 14 carcinomas [6 of 6 CIS and 5 of 8 invasive]) compared with low-grade IPMN (6 of 15 cases [40%]). The association between malignant IPMNs and prominent nucleoli is not surprising, but this was not a feature that Layfield and Cramer17 found helpful in distinguishing IPMN-moderate dysplasia from carcinomas according to their data table. Epithelial cells with pale nuclei and parachromatin clearing were not features evaluated by Layfield and Cramer, but are previously described features of conventional well-differentiated adenocarcinoma.28,29 We found this feature to be present in 5 of 6 (83%) CIS and 5 of 8 (62%) invasive carcinomas, but in only 1 of 15 (7%) low-grade IPMN (P ⬍.01). Therefore, we find this feature at least suspicious for carcinoma, but it is not a feature that can distinguish in situ from invasive carcinoma. Vacuolated cytoplasm was observed in all grades of IPMN: 10 of 15 (67%) low-grade IPMN, 3 of 6 (50%) CIS, and 8 of 8 (100%) IPMN with invasive carcinoma. Given that most of the low-grade IPMN in the current study were neoplasms with moderate dysplasia, the absence of apparent cytoplasmic mucin correlates with the increasing N:C ratio noted in dysplasia; a similar argument holds for CIS. Large vacuolated malignant cells were a consistent feature in invasive carcinomas. The presence of intracytoplasmic mucin was 1 of the 2 features found by Layfield and Cramer17 to distinguish IPMN from other cysts of the pancreas (P ⫽ .029), and according to their data table, was most prominent in invasive carcinomas. The results of the current study demonstrate that cytologic features can predict the histologic grade in many IPMNs. In the appropriate clinical setting of an IPMN with a markedly dilated pancreatic duct with or without a side branch cyst, the presence of thick “colloid-like” mucin is consistent with an IPMN and is not the quality of mucin from GI contamination, but the mucin quality was not found to be correlated with tumor grade. The presence of epithelial cells is not necessary for the diagnosis of a neoplastic mucinous cyst. In addition, the absence of thick mucin does not exclude a neoplasm. Tight epithelial clusters of cells
with a high N:C ratio are found most often in neoplasms of at least moderate dysplasia. A smear background rich in acute inflammation also supports the presence of a neoplasm of at least moderate dysplasia. The presence of epithelial cells with pale nuclei and parachromatin clearing is suspicious for at least IPMN-CIS. Only the presence of necrosis suggests the possibility of invasion, but the results of the current study did not show this feature as being unique to IPMN with invasive carcinoma compared with IPMNCIS. It will be important to apply these criteria prospectively to assess their accuracy in a large series of all pancreatic cysts. Currently, patients with neoplastic mucinous cysts all undergo resection if the patient is a surgical candidate, but given the expanding research of these neoplasms and the search for cellular and molecular features that will assist in their classification, the ability of cytology to help triage patients with noninvasive carcinoma may have important management implications, potentially precluding the need for surgery.
REFERENCES 1.
Ohhashi K, Murakami Y, Mardyama M, et al. Four cases of “mucin producing” cancer of the pancreas on specific findings of the papilla of Vater [abstract; in Japanese]. Prog Dig Endosc. 1982;20:348-351. 2. Rickaert F, Cremer M, Deviere J, et al. Intraductal mucinhypersecreting neoplasms of the pancreas. Gastroenterology. 1991;101:512-519. 3. Agostini S, Choux R, Payan MJ, et al. Mucinous pancreatic duct ectasia in the body of the pancreas. Radiology. 1989; 170:815-816. 4. Payan MJ, Xerri L, Moncada K, et al. Villous adenoma of the main pancreatic duct: a potentially malignant tumor? Am J Gastroenterol. 1990;85:459-463. 5. Kloppel G, Kosmahl M. Cystic lesions and neoplasms of the pancreas. The features are becoming clearer. Pancreatology. 2001;1:648-655. 6. Tanaka M. Intraductal papillary mucinous neoplasm of the pancreas: diagnosis and treatment. Pancreas. 2004;28:282288. 7. Hruban RH, Takaori K, Klimstra DS, et al. An illustrated consensus on the classification of pancreatic intraepithelial neoplasia and intraductal papillary mucinous neoplasms. Am J Surg Pathol. 2004;28:977-987. 8. Maire F, Hammel P, Terris B, et al. Intraductal papillary and mucinous pancreatic tumour: a new extracolonic tumour in familial adenomatous polyposis. Gut. 2002;51:446-449. 9. Chari ST, Yadav D, Smyrk TC, et al. Study of recurrence after surgical resection of intraductal papillary mucinous neoplasm of the pancreas. Gastroenterology. 2002;123:15001507. 10. Cuillerier E, Cellier C, Palazzo L, et al. Outcome after surgical resection of intraductal papillary and mucinous tumors of the pancreas. Am J Gastroenterol. 2000;95:441-445.
Cytologic Grading of IPMN of Pancreas/Michaels et al. 11. Yamaguchi K, Ohuchida J, Ohtsuka T, et al. Intraductal papillary-mucinous tumor of the pancreas concomitant with ductal carcinoma of the pancreas. Pancreatology. 2002; 2:484-490. 12. D’Angelica M, Brennan MF, Suriawinata AA, et al. Intraductal papillary mucinous neoplasms of the pancreas: an analysis of clinicopathologic features and outcome. Ann Surg. 2004;239:400-408. 13. Maire F, Hammel P, Terris B, et al. Prognosis of malignant intraductal papillary mucinous tumours of the pancreas after surgical resection. Comparison with pancreatic ductal adenocarcinoma. Gut. 2002;51:717-722. 14. Stelow EB, Bardales RH, Stanley MW. Pitfalls in endoscopic ultrasound-guided fine-needle aspiration and how to avoid them. Adv Anat Pathol. 2005;12:62-73. 15. Brugge WR, Lewandrowski K, Lee-Lewandrowski E, et al. Diagnosis of pancreatic cystic neoplasms: a report of the cooperative pancreatic cyst study. Gastroenterology. 2004; 126:1330-1336. 16. Maire F, Couvelard A, Hammel P, et al. Intraductal papillary mucinous tumors of the pancreas: the preoperative value of cytologic and histopathologic diagnosis. Gastrointest Endosc. 2003;58:701-706. 17. Layfield LJ, Cramer H. Fine-needle aspiration cytology of intraductal papillary-mucinous tumors: a retrospective analysis. Diagn Cytopathol. 2005;32:16-20. 18. Stelow EB, Stanley MW, Bardales RH, et al. Intraductal papillary-mucinous neoplasm of the pancreas. The findings and limitations of cytologic samples obtained by endoscopic ultrasound-guided fine needle aspiration. Am J Clin Pathol. 2003;120:398-404. 19. Shimizu M, Hirokawa M, Manabe T, et al. Cytologic findings in noninvasive intraductal papillary-mucinous carcinoma of the pancreas: a report of two cases. Acta Cytol. 1999;43:243246. 20. Nagle J, Wilbur DC, Pitman MD. Cytomorphology of gastric
21.
22.
23.
24.
25.
26.
27.
28.
29.
173
and duodenal epithelium and reactivity to B72.3: a baseline for comparison to pancreatic neoplasms aspirated by EUSFNAB. Diagn Cytopathol. 2005;33:381-386. Nakagohri T, Asano T, Kenmochi T, et al. Long-term surgical outcome of noninvasive and minimally invasive intraductal papillary mucinous adenocarcinoma of the pancreas. World J Surg. 2002;26:1166-1169. Lau SK, Lewandrowski KB, Brugge WR, et al. Diagnostic significance of mucin in fine needle aspiration samples of pancreatic cysts. Mod Pathol. 2000;13:48A. Jhala NC, Jhala DN, Chhieng DC, et al. Endoscopic ultrasound-guided fine-needle aspiration. A cytopathologist’s perspective [review]. Am J Clin Pathol. 2003;120:351-367. Kitagawa Y, Unger TA, Taylor S, et al. Mucus is a predictor of better prognosis and survival in patients with intraductal papillary mucinous tumor of the pancreas. J Gastrointest Surg. 2003;7:12-18; discussion, 18-19. Sugiura H, Kondo S, Islam HK, et al. Clinicopathologic features and outcomes of intraductal papillary-mucinous tumors of the pancreas. Hepatogastroenterology. 2002;49:263267. Gigot JF, Deprez P, Sempoux C, et al. Surgical management of intraductal papillary mucinous tumors of the pancreas: the role of routine frozen section of the surgical margin, intraoperative endoscopic staged biopsies of the Wirsung duct, and pancreaticogastric anastomosis. Arch Surg. 2001; 136:1256-1262. Hara H, Suda K, Oyama T. Cytologic study of noninvasive intraductal papillary-mucinous carcinoma of the pancreas. Acta Cytol. 2002;46:519-526. Robins DB, Katz RL, Evans DB, et al. Fine needle aspiration of the pancreas. In quest of accuracy. Acta Cytol. 1995;39:110. Lin F, Staerkel GA. Cytologic criteria for well differentiated adenocarcinoma of the pancreas in fine-needle aspiration biopsy specimens. Cancer. 2003;99:44-50.
