Intraductal papillary-mucinous neoplasms of the pancreas

An analysis of in situ and invasive carcinomas in 28 patients




Intraductal papillary-mucinous neoplasms (IPMNs) of the pancreas are intraductal tumors with variable amounts of papilla formation, mucin production, and cytoarchitectural atypia. Associated invasive carcinomas, reported to occur in up to 30% of patients, often are mucinous and clinically indolent.


The clinical and pathologic features of 28 IPMNs resected at Memorial Sloan-Kettering Cancer Center between 1983 and 1997 were reviewed.


There were 16 females and 12 males with a mean age of 68 years (range, 44–79 years) and a mean tumor size of 4.5 cm (range, 1.5–11.0 cm). The head of the gland was the predominant tumor site (89%). Abdominal pain, weight loss, and acholic stool were the most common symptoms at presentation. According to histology, two types of papillae were identified: intestinal (22 patients) and pancreatobiliary (6 patients). In the intraductal component, cytologic atypia was minimal (i.e., intraductal papillary-mucinous [IPM] adenoma) in 2 patients and moderate (IPM borderline tumor) in 5 patients, and severe atypia (IPM carcinoma in situ) was seen at least focally in 21 patients. In addition, invasive carcinoma was identified in 15 patients (53%), 4 of whom had only microscopic foci. Invasive carcinoma was of the mucinous type (colloid) in six patients and of the tubular type (conventional ductal adenocarcinoma) in nine patients. At a median follow-up of 35 months, four patients died of disease; two of these patients had only borderline atypia with no identified in situ or invasive carcinoma in the sections submitted. Eighteen patients had no evidence of disease, 1 patient was alive with recurrent disease, and 5 patients died of other causes. The actuarial 5-year disease free survival rate was 78%. Of the 14 patients with invasive carcinoma, 5 of 6 patients with colloid type tumors were free of tumor at a mean of 55 months. Of the patients with tubular type invasive carcinoma, two patients died of their disease (at 4 years and 7 years), three patients died of other causes, and four patients were alive (three were free of disease, and one experienced disease recurrence) at an average follow-up of 7.5 years.


Two distinct patterns of intraductal papillae are seen in patients with IPMNs: intestinal and pancreatobiliary. Both in situ and invasive carcinoma may be encountered more commonly than previously recognized. Tubular type invasive carcinomas occur as well as mucinous type (colloid) carcinomas. Although the neoplasms are less aggressive as a group than conventional pancreatic ductal adenocarcinoma, patients with IPMNs may pursue a deadly course, even in the absence of identifiable invasive carcinoma. Conversely, patients with tubular type invasive carcinoma arising in the background of IPMN may follow a more favorable course than patients with conventional ductal adenocarcinoma without IPMN, emphasizing the importance of recognizing the IPMN component in patients with pancreatic adenocarcinoma. Cancer 2002;94:62–77. © 2002 American Cancer Society.

Intraductal papillary-mucinous (IPM) neoplasm (IPMN) is a recently created category of pancreatic neoplasms,1, 2 comprising an estimated 0.5–9.8% of pancreatic exocrine tumors.3 In the past, tumors that are now included in this category have been reported under a plethora of names, each emphasizing a different aspect of this entity: villous adenoma4–6 for its papillary nature, ductectatic mucinous cystadenoma or mucinous duct ectasia7–9 for the cystic dilatation of the ductal system, and mucin-producing tumor10–12 for the excessive production of mucin exuding from the ampulla of Vater, to such an extent that some authors have referred to these tumors as an endoscopic syndrome.13 Recently, because of their histologic and clinical overlap, these tumors have been collected under the heading of IPMN,1, 2, 14 a term that also is endorsed by the World Health Organization (WHO).15 The distinction of IPMNs from mucinous cystic neoplasms, which are their close kindreds but are characterized by the presence of an ovarian-like stroma, a predilection for afflicting younger females, and a propensity for involvement of the tail of the pancreas, is now well established.2

Although numerous cases of IPMN have been reported, and the clinical and pathologic features are becoming well recognized, several important issues remain unresolved. IPMNs display a spectrum of cytoarchitectural atypia. Similar to mucinous cystic neoplasms of the pancreas, IPMNs have been divided into adenoma, borderline, and carcinoma categories, the last including both in situ and invasive varieties.2, 15 Invasive carcinoma has been found in approximately 31% of patients with IPMN: Invasion was noted in 101 of 321 patients reported in the more comprehensive recent studies.7, 12, 16–25 Mucinous (colloid) carcinoma (also referred to as mucinous noncystic carcinoma) was more common than conventional ductal (tubular) type invasive carcinoma: Twenty-one of 32 reported patients (66%) with the histologic type of invasive carcinoma specified had mucinous carcinomas.12, 17, 21, 23, 24 It has been established recently that patients with colloid carcinomas of the pancreas (with or without an IPMN component) have a significantly better prognosis than patients with conventional ductal adenocarcinomas.26

Although some interesting data have emerged about tumor types and biologic behavior, the features of IPMN-associated carcinoma have yet to be defined in detail. Furthermore, different morphologic patterns may be found in the intraductal papillae. Although most tumors resemble the intestinal and gastric foveolar types of epithelia found in adenomas of the tubular gastrointestinal tract, other types occur. Recently, we reported intraductal pancreatic tumors with oncocytic epithelium, and other patterns remain to be described in detail. In this report, we present the clinical and pathologic features of 28 patients with IPMN who were diagnosed and treated at Memorial Sloan-Kettering Cancer Center, focusing on the histologic patterns of the papillae and the correlation of carcinoma in situ and invasive carcinoma with survival.


The minimum criteria for inclusion of a tumor in this study were defined as either 1) cystic dilatation of the pancreatic ducts detected radiographically or on macroscopic examination, 2) mucin extrusion from the ampulla combined with mucinous lining of cystically dilated ducts on microscopic examination, or (3) marked (macroscopically or radiologically visible) intraductal papilla formation with ductal dilatation. In all patients, the size of the cystic dilatation or papillary focus was > 1 cm.

Twenty-eight resected specimens that fulfilled these criteria were identified in the files of the Departments of Surgery and Pathology, Memorial Sloan-Kettering Cancer Center between 1983 and 1997, which included 780 pancreatic resection specimens. Clinical information was obtained from patient charts and included symptoms at presentation, findings of diagnostic work-up, type of operation performed, and postoperative course. Clinical outcomes were classified as no evidence of disease, dead of disease, alive with disease, or dead of other causes.

The reports of radiologic studies were reviewed. Twenty-three patients had computerized tomography scans, 11 patients had ultrasound tests, and 2 patients had magnetic resonance images of the abdomen. Endoscopic retrograde cholangiopancreatography had been performed in 13 patients.

All pathology materials (slides and reports) were reviewed. The location, size, and extent of the tumor were noted. The presence and extent of macroscopic papillae, mucin, and invasive carcinoma were recorded.

An average of 35 slides (range, 8–82 slides) and 9 tumor slides (range, 3–65 tumor slides) per patient were examined microscopically. The following data were collected: extent of ductal dilatation, amount of luminal mucin, architecture of the papillae, cytoarchitectural features, proportion of any in situ or invasive carcinoma, type of any invasive carcinoma (mucinous or tubular), and extent of tumor (perineural or vascular invasion, extension to adjacent structures or peripancreatic tissues, and lymph node metastases).

The degree of papilla formation was scored as minimal, moderate, or marked. Tumors with only rare microscopic folds or micropapillae comprised of < 30–40 cells in thickness were graded as minimal. Tumors with overt villous architecture or branching papillae were classified as moderate. Tumors with macroscopic nodular papillary tumors that measured > 1 cm were graded as marked. Based on the maximum degree of cytoarchitectural atypia in the intraductal component, each tumor was classified as adenoma (no or minimal atypia), borderline tumor (moderate atypia), or intraductal carcinoma (severe atypia) according to the WHO classification system.15 The correlation of the morphology of the intraductal papillae and the type and extent of any invasive carcinoma with the clinical outcome was studied.


Clinical Findings

The clinical findings are presented in Table 1. The patients included 16 females and 12 males. The average age was 67.8 years (range, 44–79 years). Chief complaints at the time of presentation included abdominal pain (n = 16 patients; 57%), acholic stool (n = 12 patients; 43%), and weight loss (n = 12 patients; 43%). Only four patients had a history of pancreatitis (14%), six patients had jaundice (18%), and two patients were asymptomatic (detected incidentally during a work-up for other diseases). A history of prior malignancy was noted in eight patients (29%): one patient with melanoma, two patients with urothelial carcinomas (kidney and bladder), one patient with squamous cell carcinoma of the oral cavity, one patient with prostatic adenocarcinoma, one patient with squamous cell carcinoma of the lung, one patient with chondrosarcoma, and patient with ampullary carcinoma.

Table 1. Clinical Findings
PatientPathologic diagnosisAge (yrs)aGenderSymptoms at presentationOther malignancyRadiologic studiesResults of radiographic studies
  • IPM: intraductal papillary mucinous, (A, adenoma; B, borderline; C, carcinoma); TB: with tubular (usual ductal-type) invasion; CC: colloid (mucinous noncystic) carcinoma; F: female; M: male; MRI: magnetic resonance imaging; CT: computerized tomography; US: ultrasonography; ERCP: endoscopic retrograde cholangiopancreatography; N/A: not available.

  • a

    The average age was 68 years.

1IPMA69FAbdominal painLung, squamous ca.CT, ERCP, MRI, USCystic dilatation of the duct, pancreatic mass lesion
2IPMA76MAbdominal painCT, ERCPPancreatic mass involving uncinate, narrowing of proximal pancreatic duct
3IPMB76FAbdominal pain, weight lossERCP, CT, USDilated pancreatic duct, mass
4IPMB72FAsymptomaticCT, ERCPHypodense area within the uncinate process, ductal ectasia and mucin
5IPMB70FFlu-like symptomsUS, CT, ERCPSolid and cystic mass in the head of pancreas
6IPMB76FAbdominal painChondrosarcoma, low gradeMRI, CTMass in distal pancreas and in liver
7IPMB64MFatigue, weight lossOral, squamous ca.CTCystic lesion, head of pancreas
8IPMC68FAnorexia, fatigue, jaundice, light stoolNoneN/A
9IPMC78FFatigue, nauseaMelanomaCTPancreatic mass
10IPMC71MAbdominal pain, diarrhea, fatigue, light stool, weight lossKidney, urothelial ca.CTPancreatic mass in the head with distal pancreatic duct dilatation
11IPMC75MAsymptomaticProstate adenoca.CT, ERCPEnlarged pancreatic duct, complex cystic enlargement of the head of the pancreas
12IPMC76MAnorexia, fatigue, jaundice, weight lossCT, USComplicated cystic mass in the head of the pancreas
13IPMC73FAbdominal painUS, ERCPMass in head of pancreas, dilatation of the ducts
14IPMC-CC71MDiarrhea, weight lossCTPancreatic mass
15IPMC-CC60MAbdominal pain, back painCT, ERCP, USPancreatic mass; homogenous and causing a cut off of the duct at the neck
16IPMC-CC78FDiarrhea, light stool, weight lossCTCystic mass at head of pancreas (6 cm)
17IPMC-CC75FDiarrhea, light stoolERCP, US, CTCystic mass, with fistulous (?) connection to duodenum
18IPMC-CC44MBloody stool, diarrhea, weight lossCT, USDuodenal ulcer, mass at the head of the pancreas
19IPMC-CC61FAbdominal pain, constipationCT, US, angiogramPancreatic mass encasing gastroduodenal artery and splenic vein
20IPMC-TB48FAbdominal pain, light stools, nausea, emesisNoneN/A
21IPMC-TB69FAbdominal pain, anorexia, fatigue, fever, jaundiceCT, ERCPGall stones
22IPMC-TB62MAbdominal pain, nausea, emesisBladder urothelial ca.CTPancreatic mass
23IPMC-TB58FAbdominal pain, fatigue, jaundice, light stool, pruritis,Ampullary ca.CT, US, ERCPN/A
24IPMC-TB69MAbdminal pain, anorexia, jaundice, light stool, pruritisNoneN/A
25IPMC-TB67MAbdominal painCT, ERCPPancreatitis, marked defect in junction of body and tail of the pancreas
26IPMC-TB79FLight stool, weight lossCT, ERCPDouble duct sign, irregularity compressing both biliary and pancreatic ducts
27IPMC-TB69MAbdominal pain, dark urine, fatigue, jaundice, light stoolCT, USPancreatic mass (head)
28IPMC-TB61FAbdominal pain, anorexia, diarrehea, weight lossNoneN/A

Radiographic studies documented the presence of a mass in 17 patients and dilatation of the ducts in 5 patients. In seven patients, the lesions were reported radiologically as cystic.

Twenty-three patients were treated with pancreatoduodenectomy, 3 with distal pancreatectomy and 2 with total pancreatectomy. In none of the patients was metastasis detected intraoperatively.

Pathologic Findings

The pathologic findings are summarized in Table 2.

Table 2. Pathologic Findings and Clinical Outcome
PatientDiagnosisSiteSize (cm)aType of papillae (extent)InvasionExtent of diseaseFollow-up (months)b
%Type and size (cm)
  • IPM: intraductal papillary mucinous (A, adenoma; C, carcinoma; B, borderline); H: head; B: body; T: tail; TB: tubular (usual ductal) type invasion; CC: colloid (muconodular) type invasion; PB: pancreaticobiliary; IN: intestinal; +: positive; CIS: carcinoma in situ; LNs: lymph nodes; NED: no evidence of disease; DOC: dead of other causes; DOD: dead of disease; AWD: alive with disease.

  • a

    The overall average tumor size was 4.5 cm.

  • b

    The overall average follow-up was 42 months.

1.IPMAT3.0IN (moderate)0NoneNED (19)
2.IPMAH3.0IN (minimal)0NoneNED (53)
3.IPMBH2.0IN (minimal)0NoneMargin + for IPMBDOD (13)
4.IPMBH4.5IN (moderate)0NoneNED (38)
5.IPMBH2.5IN (minimal)0NoneNED (26)
6.IPMBT1.8IN (minimal)0NoneNED (29)
7.IPMBH3.5IN (minimal)0NoneMargin + for IPMBNED (1)
Average 1–72.9(26)
8.IPMCH6.5IN (moderate)0NoneMargin + for CISDOC (3)
9.IPMCH2.0PB (marked)0NoneDOD (36)
10.IPMCH8.0PB (marked)0NoneNED (30)
11.IPMCH6.5IN (moderate)0NoneNED (48)
12.IPMCH7.0IN (moderate)0NoneNED (33)
13.IPMCH2.0IN (moderate)0NoneNED (25)
Average 8–135.3(29)
14.IPMC-CCH6.0IN (marked)1CC, 0.1NED (84)
15.IPMC-CCH3.5PB (marked)40CC, 1.5NED (55)
16.IPMC-CCH6.0IN (moderate)20CC, 1.5Perineural invasionNED (57)
17.IPMC-CCT, B, H11.0IN (marked)40CC, 4.5Margin + for invasionDOC (2)
18.IPMC-CCH6.2IN (marked)70CC, 4.1NED (97)
19.IPMC-CCH6.8IN (moderate)70CC, 5.0Perineural invasionNED (36)
Average 14–196.6(55)
20IPMC-TBH1.5IN (marked)10TB, 0.2DOC (52)
21IPMC-TBH3.8IN (marked)5TB, 0.2NED (92)
22IPMC-TBH3.3PB (marked)10TB, 0.3Vascular inv., metastasis to liverDOD (44)
23IPMC-TBH2.2IN (moderate)55TB, 1.2Vascular invasion, 1 of 8 LNsAWD (25)
24IPMC-TBH3.8IN (moderate)40TB, 1.5Perineural invasion, 1 of 14 LNsDOD (87)
25IPMC-TBH3.0PB (moderate)60TB, 2.0Perineural and vascular invasionNED (40)
26IPMC-TBH3.5IN (moderate)70TB, 2.5Perineural invasion, 3 of 18 LNsDOC (0)
27IPMC-TBH4.5IN (moderate)70TB, 3.0Perineural invasion, 1 of 17 LNsDOC (9)
28IPMC-TBH5.0PB (marked)70TB, 3.5Perineural invasion, margin + for CISNED (153)
Average 20–283.4(56)

Macroscopic findings

The head of the gland was the predominant tumor site (25 of 28 patients). In two patients, the tumor was in the tail. One patient had involvement of the entire ductal system. The average tumor size was 4.5 cm (range, 1.5–11.0 cm); however, it often was difficult to determine the exact size of the lesion because of the complicated nature of the tumor, with widely distributed papillae and dilated ducts.

Macroscopic ductal dilatation was present in all patients, occasionally presenting as multilocular cystic transformation of the ductal system. In four patients, only the main pancreatic duct showed cystic changes and appeared tortuous, measuring up to 2.5 cm in greatest dimension (Fig. 1A). In four patients, the secondary ducts were involved predominantly and exhibited multilocular cystic changes, especially in the uncinate region. In the remaining 20 patients, both the main and secondary ducts were dilated. In some instances, the ductal dilatation was very localized, and careful dissection of the ductal system was necessary to demonstrate the intraductal nature of the lesion. Intraductal mucin accumulation was noted as a prominent macroscopic finding in 18 patients. Papillary tumor formation was recorded macroscopically in 17 patients (Fig. 1B). Papilla formation ranged in appearance from subtle granulations on the walls of the cystically dilated ducts to exuberant papillary masses filling an 11-cm segment of the duct in one patient. The papillae were usually soft, tan, and friable.

Figure 1.

Photographs of cystic (A) and papillary (B) examples of intraductal papillary-mucinous neoplasms (IPMNs) of the pancreas. (A) The main pancreatic duct (arrows) is markedly dilated and tortuous and is significantly larger than the common bile duct (CBD). The ducts contain mucinous material but no papillae are seen macroscopically. (B) Nodular papillary proliferations are filling and expanding the major pancreatic duct (arrows). The papillae are beige-tan, soft, and friable.

Invasive carcinoma was detected macroscopically in 10 patients. The invasion was described as white-to-gray, firm areas in patients with tubular type carcinoma. Mucinous carcinomas had a more gelatinous quality.

Microscopic findings

The tumors were comprised predominantly of cystically dilated ducts that were represented on histologic sections as apparent separate cysts. The walls of these cystic ducts appeared hyalinized, often containing scattered islets of Langerhans or periductal ductules. No stromal hypercellularity (ovarian-like stroma) was detected in any of the patients. The cystic ducts frequently contained sparse mucin, and calcification in the mucin was seen in two patients. Luminal mucin accumulation was minimal in 7 patients, moderate in 13 patients, and significant in 8 patients.

The lining of the involved ducts exhibited marked variation in cytoarchitectural features, both from patient to patient and also between different regions of individual tumors. In four patients, the ducts were lined exclusively by simple columnar cells with abundant apical mucin, resembling gastric foveolar cells, and basally located, small, and monotonous nuclei. This appearance was present focally in variable amounts in most of the other 24 patients.

Variable degrees of papilla formation were seen (Fig. 2). Morphologically, the papillae could be divided into two types: intestinal (IN; n = 22 patients) and pancreatobiliary (PB; n = 6 patients). The papillae of the IN type (Figs. 3A, 4A,B) were very similar to those of intestinal villous adenomas. They were lined by pseudostratified cells and had columnar (cigar-shaped) nuclei with coarse chromatin and small, if any, nucleoli. Like their intestinal counterparts, some examples (7 patients) had more abundant intracellular mucin (Fig. 4A). Although the intestinal pattern was villous in most patients, a minor tubular component also could be seen in two patients.

Figure 2.

Intraductal papillary-mucinous neoplasms (IPMNs) of the pancreas have a spectrum of cyst and papilla formation in the ducts. This low-power photomicrograph of an IPMN displays the spectrum of cyst and papillae formation. The major pancreatic duct contains a villous proliferation, whereas a branch duct is represented as a cyst with focal micropapillae. Most neoplasms have variable amounts of cysts and papillae.

Figure 3.

There are two distinctive patterns of papillae in intraductal papillary-mucinous neoplasms of the pancreas. (A) Intestinal type papillae are similar morphologically to colonic villous adenomas. They have a villous architecture, and the lining cells are pseudostratified and cigar-shaped. (B) Pancreatobiliary type papillae have more complex, branching papillae with cuboidal cells showing abundant cytoplasm.

Figure 4.

The cytoarchitectural spectrum in intraductal papillary-mucinous neoplasms of the pancreas. (A) An intraductal papillary-mucinous adenoma with the mucinous variant of intraductal (IN) type papillae is seen. The fibrovascular cores of the papillae show edema. The cells are small, monotonous, and well polarized and are devoid of any cytoarchitectural atypia. Abundant apical mucin is present in the cytoplasm. (B) More common appearance of IN type papilla showing mild (borderline) atypia (left) with transition to more severe atypia-carcinoma in situ (right).

In six patients, a pattern distinct from the IN type was predominant. These papillae were arborizing with complex branching (Fig. 3B) and abundant micropapillae. They were lined by one to five layers of cuboidal cells with moderate amounts of acidophilic cytoplasm and round nuclei with single, prominent, and eccentric nucleoli. The cells were largely devoid of mucin. We referred to this pattern as the PB type.

Based on the maximum degree of cytoarchitectural atypia identified in the intraductal component, the intraductal components were classified as IPM adenoma in two tumors (Fig. 4A), IPM borderline tumor in five tumors (Fig. 4B), and IPM carcinoma in situ in 21 tumors. In all tumors that were classified as adenoma or borderline tumor, the papillae were of the IN type; thus, all tumors with PB type papillae were classified as IPM carcinoma in situ. In some instances, small peripheral ducts showed dysplastic changes that were indistinguishable morphologically from the pancreatic intraepithelial neoplasia27 seen in association with conventional ductal carcinoma.

Invasive adenocarcinoma was identified in 15 tumors. In all of the tumors with invasive carcinoma, the intraductal component was at least focally carcinoma in situ. In nine tumors, invasion was of the tubular type (Fig. 5), and these tumors were classified as IPM carcinoma with invasive tubular adenocarcinoma. In the remaining six tumors, the invasive carcinoma was of mucinous (colloid) type (Fig. 6): IPM carcinoma with invasive mucinous (colloid) carcinoma. In four tumors, the invasion was microscopic (< 0.5 cm; three tubular and one colloid type).

Figure 5.

Tubular (usual ductal) type invasion associated with intraductal papillary-mucinous neoplasms (IPMNs) of the pancreas. (A) Low-power photomicrograph of an IPMN (top) associated with tubular type invasion (bottom). This type of invasion was seen in nine neoplasms. (B) High-power photomicrograph of tubular carcinoma. It is characterized by infiltrating tubular units embedded in desmoplastic stroma, the classic appearance of conventional ductal pancreatic adenocarcinoma.

Figure 6.

Intraductal papillary-mucinous neoplasm (IPMN) of the pancreas (thin arrows) with mucinous type (colloid, mucinous noncystic) invasion. IPMN involves the main (thick arrows) and branch (arrows) ducts and is associated with extensive colloid type invasion (A). This type of invasion is characterized by mucin lakes, many of which contain strips, clusters, and individual epithelial cells (B).

Seven of the invasive carcinomas had perineural invasion (two mucinous and five tubular). Invasion of the duodenal mucosa was evident in four tumors. Metastasis to regional lymph nodes was seen in four tumors (all were tubular type). Distant metastasis was documented in only one tumor (tubular type). The adjacent pancreatic tissue was unremarkable in seven tumors. Atrophy and chronic pancreatitis were seen in the remaining 21 tumors (6 mild, 5 moderate, and 10 severe).


The clinical outcome is shown in Tables 2 and 3. At a median follow-up of 35 months (mean, 42 months), 4 of 28 patients had died of disease. Eighteen patients had no evidence of disease at a median follow-up of 39 months (mean, 51 months). One patient was alive with recurrent disease and five had died of other causes. The actuarial 5-year disease free survival rate was 78%.

Table 3. Disease Outcome
Type and extent of papillaeDead of diseaseNo evidence of diseaseDead of other causes
No. of patientsMedian survival (months)No. of patientsMedian follow-up available (months)No. of patientsMedian survival (months)
  1. PB: pancreatobiliary; IN: intestinal; CC: colloid; TB: tubular.

PB type papilla (n = 6)240448
IN type papilla (n = 22)25020373
Noninvasive (n = 13)225103013
Invasive (n = 15)26587046
CC type invasion (n = 6)0055712
TB type invasion (n = 9)26539239
Invasion < 2 cm (n = 8)265470152
Invasion > 2 cm (n = 7)46932

Of the 15 patients with invasive carcinoma, 5 of 6 patients with mucinous (colloid) type tumors were free of disease at a median of 56 months (mean, 55 months). The remaining patient died of other causes. Of the nine patients with tubular type invasion, two patients died of disease (at 4 years and 7 years), three patients died of other causes, and four patients were alive with no evidence of disease at an average follow-up of 7.5 years.

Two of the patients who died of their disease had no invasive carcinoma identified in the sections sampled (25 blocks and 28 blocks, respectively). One of these patients had metastases to the liver after 3 years and died 7 months. later. The other patient, who had a tumor classified as IPM borderline tumor with a positive distal pancreatic margin, developed disseminated tumor in the abdomen and died 13 months after the diagnosis.

Pathologic Findings in Relation to Invasion and Clinical Outcome


Of the five patients who either died of their disease or had a recurrence, one patient had an IPM borderline tumor (died at 13 months), one patient had an IPM carcinoma in situ (died at 36 months), and three patients had invasive tubular type carcinomas, all with local lymph node metastases (one patient died at 44 months after metastasis to the liver at 3 years, another patient died at 97 months, and the third patient was alive with a recurrent tumor at 25 months) (Tables 2 and 3).


The average tumor size of IPM adenomas or IPM borderline tumors was 2.9 cm, whereas, for IPM carcinomas (with or without invasion), the average tumor size was 4.9 cm; the tumors with colloid type invasion especially tended to be larger (average size, 6.6 cm) than the remainder of the group (average, 3.8 cm). The average size of the tumors in patients who died of their disease was 5.0 cm, whereas the average size for the entire group was 4.5 cm. The tumor sizes in the two patients with in situ carcinoma that died of their disease were 2.5 cm and 3.8 cm, respectively. Thus, the size of the entire tumor did not correlate with the clinical outcome.

Extent of papilla formation

All seven patients from the IPM adenoma or IPM borderline tumor groups had either minimal (in five patients) or moderate (in two patients) degrees of papilla formation. Conversely, all of the patients from the IPM carcinoma group had either moderate or marked papillae. Of the four patients who died of disease, two patients had marked papillary proliferations. The other two patients had moderate or minimal papilla formation. Of the patients with invasive carcinoma, seven patients had moderate papillary components and eight patients had marked papillary components within the ducts.

Type of papilla

In all six patients with PB type papillae, the cytoarchitectural atypia was at the level of carcinoma in situ. Furthermore, three patients had tubular type invasion, and one patient had colloid type invasion. Two patients died of disease at 36 months and 44 months, respectively; one patient had a noninvasive tumor, and the other patient had a microinvasive tumor. Four patients were alive with no evidence of disease at an average follow-up of 68 months. In contrast, all seven patients with tumors classified as IPM adenoma or IPM borderline tumor had IN type papillae. Fifteen tumors with IN type papillae had atypia at the level of carcinoma in situ. Both patients with adenoma had the mucinous variant of this pattern (foveolar-like mucin, Fig. 4A). Of the 22 patients with IN type papillae, 6 patients had tubular type invasion, and 5 patients had colloid type invasion. Five patients died of other causes, two patients died of disease at 13 months and 87 months., and 15 patients were alive with no evidence of disease at an average of 46 months.

Invasion and its type

There was no detectable difference in the disease specific survival between patients with invasive carcinoma compared with patients who were without invasion. Of the 15 patients with invasive carcinoma, 9 patients had tubular type adenocarcinoma, and the average follow-up for these nine patients was 56 months. Four patients were alive at an average of 6.5 years (25 months, 40 months, 92 months, and 153 months, respectively), two patients died of their disease at 4 years and 7 years, and three patients died of other causes. Five patients with colloid type invasive carcinoma had no evidence of disease at an average follow-up of 55 months, and one patient died of other causes at 2 months.

Percent and size of invasive component

Of the 15 patients with invasive carcinoma, 4 patients had microinvasion (< 5 mm). In seven patients, invasive carcinoma measured > 2 cm. Invasive carcinoma constituted < 10% of the tumor mass in four patients, 10–50% of the tumor mass in five patients, and > 50% of the tumor mass in seven patients. It is noteworthy that both patients with invasive carcinoma who died of their disease had an invasive component comprising 10–50% of the tumor mass and measuring < 2 cm.

Vascular invasion

Three patients had vascular invasion, all with tubular type invasive carcinoma. One patient was alive with recurrent disease at 25 months, one patient was dead of disease at 44 months, and one patient had no evidence of disease at 40 months.

Margin involvement

Margin involvement with invasive carcinoma was seen in one patient who had colloid type invasive carcinoma and died of other causes at 2 months. Margin involvement with intraductal tumor was seen in four patients. Two patients had carcinoma in situ at the margin: One patient died of other causes at 3 months, and the other patient was alive with no evidence of disease at 153 months. The remaining two patients had borderline atypia at the margin: One patient had no identifiable invasive carcinoma in the specimen and died of disease at 13 months, and the other patient was alive with no evidence of disease at 1 month.

Lymph node metastasis

Metastatic carcinoma to local lymph nodes was identified in only four patients, all of whom had tubular type invasive adenocarcinoma. One patient was alive with recurrent disease at 25 months, one patient died of disease at 87 months, and two patients died of other causes at 9 months and postoperatively, respectively. In two patients with colloid type invasion, direct extension to the lymph nodes was identified: One patient died of other causes at 2 months, and the other patient was alive with no evidence of disease at 97 months.

Distant metastasis

Only one patient had documented distant organ metastasis. This patient had predominantly intraductal papillary carcinoma of the PB type, with only a microscopic (0.3 cm) invasive focus in the primary resection specimen. Vascular invasion also was identified. The patient developed hepatic metastases 3 years after the operation and died within another year.


Before the term IPMN was proposed to encompass the spectrum of these tumors under one heading,1, 2, 17 they had been reported in the literature under a plethora of terms, including villous adenomas or intraductal papillary neoplasms,4–6, 28–37mucin-producing tumors orintraductal mucin-hypersecreting tumors,10–13, 38–50 and mucinous duct ectasia or ductectatic mucinous cystic neoplasms.7–9, 51–56 Our study confirms that there is significant histologic overlap among these clinical entities and that it is justified to categorize them as one group: IPMN.2, 15, 17 This unifying category has found broad acceptance17, 21, 23–25, 57, 58 and is endorsed by the WHO.15 The actual incidence of IPMNs is difficult to determine: IPMNs constituted 3.6% of the tumors resected at the authors' institution, which serves as a referral center; therefore, this number may not reflect a true incidence. Most patients with IPMN presented with nonspecific symptoms, such as abdominal pain, acholic stool, or weight loss. A history of pancreatitis, a common finding in some studies,25 was noted in only 4 of 28 patients in our series. It is noteworthy, however, that eight patients (29%) had a history of another malignancy. This may be incidental, because the incidence of malignancy increases with age, and IPMNs are seen in an older age group. Moreover, it is possible that investigation of the other tumors lead to the diagnosis of IPMN, an indolent tumor that otherwise may have gone undetected. In one of our patients, the IPMN was detected incidentally during a work-up for prostatic adenocarcinoma.

Previous studies have not emphasized the morphologic heterogeneity in the appearance of the papillae in IPMNs. In this study, two distinctive morphologic patterns of papilla formation were found: IN and PB types. All patients with adenoma or borderline tumors in this study had IN type papillae, and atypia was at the level of carcinoma in all six patients with PB type papillae. The PB pattern showed some similarities to the pattern of papillae seen in intraductal oncocytic papillary neoplasms (IOPNs), a new category of intraductal pancreatic tumors.59 However, IOPNs have distinctive features, including intraepithelial lumina formation and oncocytic cells. Histologic overlaps were seen, confirming the close relation of IOPNs to IPMNs. Additional studies are needed to determine whether these two types of papillae differ in their immunohistochemical and molecular phenotypes.

Whereas the type and extent of the papillae did not appear to have any impact on the clinical outcome in this study, the tumors that were predominantly cystic and lacked papilla formation tended to be devoid of cytologic atypia as well and, thus, were classified as adenomas. In addition, tumors that were classified as adenoma or borderline tumors also generally were smaller than those with carcinoma. Recent studies have suggested that IPMNs restricted to the branch ducts are more likely to be smaller and to lack severe atypia or carcinoma compared with the main duct tumors.60

Like mucinous cystic neoplasms of the pancreas,61–66 IPMNs display a spectrum of cytoarchitectural atypia ranging from none to borderline to marked and also can be associated with invasive carcinoma. More than two in three of the previously reported instances of invasive carcinoma associated with IPMNs have been of the mucinous (colloid) type,12, 17, 21, 23, 24 characterized by mucin lakes that contain (or are partially lined by) sparse clusters of malignant cells.

In the current study, the prevalence of invasive carcinoma (53%) was greater than in previous studies, which quote numbers ranging from 20–45%, with an average of 30% in 321 patients.7, 12, 16–25 In addition, our study included more patients with tubular type invasive adenocarcinoma. In part, the greater prevalence of invasive carcinomas in this series may reflect the fact that we included all patients with a macroscopically or clinically recognizable component of IPMN, regardless of the extent of any associated invasive carcinoma. Thus, some tumors were included in which the invasive component constituted as much as 70% of the total tumor mass. Other authors may have classified such tumors as conventional ductal adenocarcinomas. It may seem that the inclusion of these patients with tumors that have a substantial component of invasive tubular adenocarcinoma would impact prognosis adversely, because it would appear that a tumor measuring > 2 cm with a histologic appearance indistinguishable from other ductal adenocarcinomas would pursue the same highly aggressive course as conventional pancreatic carcinoma. In fact, this may not be the case; our data suggest that the tubular type of invasive carcinomas associated with IPMNs may have a more favorable prognosis (median survival, 44 months). Although death from tumor did occur in two patients, the survival was relatively long (44 months and 87 months, respectively), and four additional patients survived more than 2 years, three of whom had no evidence of disease. Obviously, the number of patients in this study was not sufficient to establish whether patients with tubular type adenocarcinomas arising in association with IPMNs have a significantly better prognosis compared patients who have stage-matched, conventional ductal adenocarcinomas that lack an IPMN component. However, we believe that additional study is warranted, and the possibility that IPMN-associated tubular adenocarcinomas may be different biologically and genetically from other pancreatic ductal adenocarcinomas may be explored.

Although insufficient numbers of patients with IPMN-associated tubular adenocarcinomas have been described to established their clinical behavior adequately, it seems clear that patients with mucinous (colloid) carcinomas arising in association with IPMNs have a substantially better prognosis than patients with typical ductal adenocarcinomas. In our study, five of six such patients were alive and free of disease 2–97 months after surgery (median, 56 months), despite having invasive components measuring as much as 4.8 cm; the sixth patient died of postoperative complications. Other authors also have reported a favorable prognosis for patients with the mucinous (colloid) type of IPMN-associated invasive carcinoma,12, 17, 21, 23, 24 and we previously showed a 55% 5-year survival rate for patients with colloid carcinoma of the pancreas with or without an IPMN component.26 Thus, the colloid phenotype appears to reflect a much more indolent biology in pancreatic neoplasms. It will be most interesting to examine any molecular differences between conventional ductal adenocarcinomas and mucinous carcinomas that may help explain the striking differences in behavior.

Taken as a group, IPMNs have a far better clinical behavior than usual ductal type adenocarcinoma. In this study, the actuarial 5-year disease free survival rate was 78%. Although the number of patients was not sufficient to prove survival differences, it seems reasonable that the prognosis for patients with tumors that lack an invasive component should be better compared with the prognosis for patients with invasive carcinoma. However, in two of four patients who died of their disease, no invasive carcinoma was identified. A similar experience has been reported with mucinous cystic neoplasms, and, for this reason, some authors have argued that all mucinous cystic neoplasms should be considered malignant, even in the absence of identifiable invasive carcinoma.61, 62 Others believe, however, that patients with mucinous cystic neoplasms that resulted in death in the absence of identifiable invasive carcinoma were sampled inadequately at the time of initial pathologic evaluation and that the clinical course of mucinous cystic neoplasms differ significantly between invasive and noninvasive tumors.64–66 This possibility also must be considered for the two instances in the current series of IPMNs resulting in the death of patients in the absence of identifiable invasive carcinoma. In neither of these patients was the entire resected pancreas submitted for histology; thus, foci of invasive carcinoma may have been missed. Furthermore, one of these patients had intraductal carcinoma extending to the distal pancreatic margin, raising the possibility that a clinically unapparent focus of invasive carcinoma remained unresected. Given the propensity for IPMNs to spread along the ductal system, the potential for multifocal tumors would appear greater than for circumscribed mucinous cystic neoplasms. The time course for the development of invasive carcinoma from residual or recurrent IPMN is unknown, although most studies in the literature suggest that IPMNs progress slowly over many years before the development of invasive carcinoma. Another issue related to the spread of IPMNs along the pancreatic ducts is the evaluation of the pancreatic resection margins after pancreatoduodenectomy or distal pancreatectomy. It is not uncommon for the margins to show extension of the tumor, although, in most tumors, the bulk of the lesion is localized sufficiently to permit removal without a total pancreatectomy. Because the natural history of each grade of IPMN is essentially unknown, it is not clear how much surgery may be justified to remove a peripheral portion of an IPMN, especially an IPM adenoma or IPM borderline tumor. In our study, four patients had margins positive for IPMN, two of whom showed carcinoma in situ. One of these patients died subsequently of their disease (see above), but no other local recurrences were found. Obviously, it will be important to carefully follow patients with positive ductal margins to determine the true risk for tumor recurrence.

Despite the disconcerting finding of disease-associated death in patients who lacked identified invasive components, we believe that the WHO classification system proposed for IPMNs15 is reasonable and should be employed. The vast majority of previously reported patients with IPMNs without invasive carcinoma have been cured by complete surgical removal, and it appears that the risk of death from an undetected invasive carcinoma is very low. We would not like to suggest that all IPMNs should be considered potentially malignant but would rather encourage thorough sampling of resected tumors and recognize the potential for multicentric or recurrent disease in patients with an intraductal neoplasm.

The new WHO classification system for pancreatic tumors does include a category for invasive mucinous carcinomas associated with IPMNs: Papillary mucinous carcinoma is the term proposed to designate this entity.2, 15 Because the intraductal and invasive components can be recognized separately, the intraductal component likely precedes the development of invasive carcinoma, and some invasive components are not mucinous, we believe that the two components should be designated separately and that the type of invasive carcinoma should be specified. Furthermore, although insufficient data exist to prove the point, it seems likely that the extent of the invasive component may correlate with the prognosis. Thus, we believe that an IPMN with only a microscopic focus of invasive carcinoma should be distinguished from an IPMN with a substantial (or predominant) invasive component.

Because IPMNs often are detected at a predominantly or exclusively intraductal stage, they provide a unique opportunity to study the development of invasive carcinoma in the pancreas. Further studies should focus on the biologic and molecular differences between IPMN-associated carcinomas and conventional ductal adenocarcinomas. Perhaps, by understanding better these clinically detectable, preinvasive pancreatic neoplasms, better techniques can be developed to diagnose patients with ductal adenocarcinoma at an earlier, potentially treatable stage.


The authors thank Karen Lamhaouar for assistance in obtaining clinical information, Benjamin True for assistance in the preparation of the photomicrographs, and Cheryl Lubinski for assistance in the preparation of this article.