Gastrointestinal Imaging
Original Research
Histopathologic Findings of Multifocal Pancreatic Intraductal Papillary Mucinous Neoplasms on CT
Audio Available | 
Share
OBJECTIVE. The criteria for resection of solitary pancreatic side-branch intraductal papillary mucinous neoplasm (IPMN) have been well described by the Sendai consensus statement. However, the management of multiple pancreatic cystic lesions is less certain, with no clear guidelines in the literature to date. The purpose of this study was to evaluate the histopathologic findings in pancreatic IPMNs in patients with multiple (≥ 4) pancreatic cysts.
MATERIALS AND METHODS. The CT scans of all patients with a pathologically proven IPMN at our institution were reviewed, and a total of 52 patients with four or more pancreatic cysts were found. Each case was reviewed for the number of cysts and the presence of signs of invasive malignancy including a coexistent solid pancreatic mass, pancreatic ductal dilatation, and mural nodularity.
RESULTS. A total of 52 patients (19 men, 33 women; mean age, 71.8 years) were found to have multifocal IPMNs, defined as four or more cysts, on CT. Of these 52 patients, nine also had evidence of a solid pancreatic mass on CT. Retrospective review of the pathologic results for the remaining 43 patients (17 men, 26 women; mean age, 71.76 years) showed 18 cases of an IPMN with either high-grade dysplasia or a coexistent invasive carcinoma. Most important, 37% (7/19 patients) had no CT findings of an invasive malignancy according to the Sendai criteria (i.e., cysts ≥ 3 cm in the axial plane, main pancreatic ductal dilatation ≥ 6 mm, or mural nodularity within a cyst) but were found to have an IPMN with either high-grade dysplasia or invasive carcinoma. When the pancreas contained 10 or more cysts, high-grade dysplasia or invasive carcinoma tended to be more likely than low- or intermediate-grade dysplasia (odds ratio, 3.83; 95% CI, 0.87–16.8; p = 0.075).
CONCLUSION. The presence of multiple pancreatic cysts should be looked on with suspicion, particularly when there are a large number of cysts, even when none of the cysts individually meet the imaging criteria for resection according to the Sendai consensus recommendations. At the very least, these patients need to be followed very closely.
Keywords: CT, intraductal papillary mucinous neoplasms, pancreas
As a result of improvements in MDCT over the past decade, as well as the development of 3D reconstruction techniques, it has become a routine part of daily practice to identify small cystic lesions in the pancreas. Moreover, the efficacy of MRI and endoscopic ultra-sound in depicting these small incidental cysts has proven to be even greater than that of CT, and both of these modalities have gradually begun playing larger roles in evaluating the pancreas [1]. Most of these pancreatic cysts are small side-branch intraductal papillary mucinous neoplasms (IPMNs), mucin-producing pancreatic lesions that arise in and can produce dilatation of the main pancreatic duct, side branches, or both [1, 2].
Side-branch IPMNs have a lower risk of associated invasive carcinoma (6–51%) than main duct IPMNs, and as a result, there has been debate regarding the management of these lesions [1]. The Sendai consensus statement issued in 2006 by Tanaka et al. [3] established a number of criteria for the re-section of pancreatic side-branch IPMNs, including an axial diameter of 30 mm or greater, main pancreatic ductal dilatation of 6 mm or greater, and mural nodularity within a cyst. Several subsequent studies have validated these criteria and have suggested that patient symptoms and elevated tumor marker levels (e.g., cancer antigen [CA] 19-9) are also suspicious features that may indicate cyst resection is warranted [2].
However, as imaging technology has improved, it has become increasingly common to visualize multiple side-branch IPMNs and it is also now common for pathologists to identify multifocal IPMNs in histologic specimens [4, 5] (Fig. 1). Although there are many articles in the literature about the management of a single isolated pancreatic side-branch IPMN, there are few about the management of multiple pancreatic cysts particularly when none of the individual cysts meet the Sendai criteria for resection.
 View larger version (172K) | Fig. 1 —66-year-old man with innumerable cysts throughout pancreas on axial contrast-enhanced CT image. No cysts measured more than 3 cm, no solid pancreatic mass was identified, and no individual cyst was found to have mural nodularity. Invasive colloid carcinoma was discovered after total pancreatectomy. |
The aim of this study was to establish the likelihood of lesions with high-grade dysplasia or invasive carcinoma when there are multiple pancreatic IPMNs on MDCT.
| Materials and Methods |   |
|---|
Before commencement of this study, permission was obtained from our institutional review board. Patient informed consent was not required.
A search was then performed of the surgery department computer database for all IPMNs surgically resected from 1998 to 2011. The available CT scans for each of these 279 patients were retrospectively reviewed by an abdominal radiologist, and a total of 52 patients were found to have four or more cystic lesions in the pancreas. This group of patients formed the study population and was composed of 19 men and 33 women. The 52 patients had a mean age of 71.8 years and a median age of 72 years. Thirty patients underwent a Whipple procedure (pancreaticoduodenectomy), 12 patients underwent distal pancreatectomy, and 10 patients underwent total pancreatectomy.
All MDCT examinations were performed on one of four scanners, all of which were manufactured by Siemens Healthcare, in use at our institution during the time course of the study: Somatom Plus 4 Volume Zoom scanner (detector collimation, 4 × 1 mm; reconstruction, 3-mm slice thickness; 120 kVp; 140–180 mAs), Somatom Sensation 16 (detector collimation, 16 × 0.75 mm; reconstruction for diagnostic interpretation, 3-mm slice thickness and 3-mm slice intervals; reconstruction for multiplanar reformation [MPR], 0.75-mm slice thickness and 0.5-mm intervals; 120 kVp; 150–200 mAs), Somatom Sensation 64 (detector collimation, 64 × 0.6 mm; reconstruction for diagnostic interpretation, 3-mm slice thickness and 3-mm slice intervals; reconstruction for MPR, 0.75-mm slice thickness and 0.5-mm intervals; 120 kVp; 150–200 mAs), and Somatom Definition Flash dual-source (detector collimation, 128 × 0.6 mm; reconstruction for diagnostic interpretation, 3-mm slice thickness and 3-mm slice intervals; reconstruction for MPR, 0.75-mm slice thickness and 0.5-mm intervals; 120 kVp; quality reference, 290 mAs for online dose modulation system [CareDose 4D, Siemens Healthcare]).
All patients, regardless of which scanner was being used, ingested 750–1000 mL of water over the 15- to 20-minute period before scanning commenced. In all cases, a scout topogram was acquired, followed by the acquisition of arterial and venous phase images at 25–30 and 50–60 seconds, respectively, after the administration of IV contrast material. Iohexol (Omnipaque 350, GE Health-care), 120 mL, was injected IV rapidly through a peripheral vein at 3–5 mL/s.
All image data were reconstructed using a body soft-tissue algorithm. All studies were interpreted on a PACS workstation (Ultravisual PACS, version 5.35.4.193, Emageon). The studies acquired on the Somatom Sensation 16, Somatom Sensation 64, and Somatom Definition Flash dual-source scanners included automatically reconstructed coronal and sagittal reformations. The oldest studies acquired on the Somatom Plus 4 Volume Zoom scanner did not include automatically reconstructed reformations, and reformats were manually created by the reviewing radiologists on the PACS workstation using the Ultravisual PACS software package.
The CT scans for each of the 52 patients in the study were independently read by two radiologists with subspecialization training in abdominal imaging. The more senior of the two radiologists had 14 years of postfellowship experience, and the second radiologist had 1 year of postfellowship experience. After the cases were independently read by each of the two radiologists, the studies were read by the two readers in consensus to resolve any discrepancies. The abdominal radiologists reviewing the studies were blinded to the pathologic results for each patient when reviewing the CT scans.
The following parameters were evaluated for each CT examination: number of cysts in the pancreas, size of the five (if applicable) largest cysts in the pancreas, presence of mural nodularity within any of the cysts, presence of a separate solid mass within the pancreas, main pancreatic ductal dilatation (defined as present if any portion of the pancreatic duct is ≥ 6 mm) and duct size, presence of either thin septations (≤ 3 mm in thickness) or thick septations (> 3 mm), and presence of biliary dilatation (common bile duct measuring > 1 cm) or presence of a biliary stent.
The postsurgical pathology records were then retrospectively searched for each patient, and the results were placed into one of four categories on the basis of the most recent World Health Organization classification for IPMN [6]: IPMN with low-grade dysplasia, IPMN with intermediate-grade dysplasia, IPMN with high-grade dysplasia, or IPMN with an associated invasive carcinoma. Each of the pathology results was interpreted by a board-certified pathologist with subspecialty training in surgical pathology and a subspecialty focus in gastrointestinal and pancreatic diseases.
For the purposes of analysis, IPMN with low-grade dysplasia and IPMN with intermediate-grade dysplasia were grouped together, as were IPMN with high-grade dysplasia and IPMN with invasive carcinoma. Characteristics of these groups were compared using the Fisher exact test for categoric outcomes and Wilcoxon rank sum test for continuous outcomes. To identify CT features that were associated with having high-grade dysplasia or invasive carcinoma, we used multivariate logistic regression models that adjusted for patient age and sex and we ran the models with high-grade dysplasia or invasive carcinoma as the dependent variable and different CT features as independent variables. Analyses were completed with statistical freeware (R, version 2.14.2, The R Project for Statistical Computing). All p values < 0.05 were considered statistically significant.
| Results | |
|---|
Of the 52 patients in the study population, nine were found to have a distinct solid pancreatic mass on CT in addition to their pancreatic cysts. The characteristics of these patients are summarized in Table 1; all nine patients were found to have invasive carcinoma after surgical resection. The mean maximum axial diameter of these solid masses was 2.5 cm with a median maximum axial diameter of 2.5 cm.
Table 2 summarizes the remaining 43 patients, none of whom had evidence of a solid pancreatic mass on CT. This group of patients had a mean number of pancreatic cysts of 14.1 and a median number of cysts of 9. Eighteen of the 43 patients (41.9%) in this group had evidence of high-grade dysplasia or coexistent invasive carcinoma. Notably, of the patients in this group, 15 patients had pancreatic ductal dilatation of 6 mm or greater, six patients had evidence of mural nodularity within at least one pancreatic cyst, 18 had evidence of a septation within at least one cyst, and 12 had evidence of at least one cyst measuring 3 cm or larger in the axial plane.
Table 3 summarizes the patients with and those without high-grade dysplasia or invasive carcinoma. Notably, no statistically significant differences could be identified between the two groups with regard to mural nodularity within a cyst (p = 0.68), thin septations (p > 0.99), thick septations (p = 0.42), cyst calcification (p = 0.17), or a cyst size of 3 cm or larger (p > 0.99).
Although the proportions of patients with main pancreatic duct sizes of 3 mm or greater, 6 mm or greater, and 7 mm or greater were not significantly different between the two groups, patients with high-grade dysplasia or invasive carcinoma had significantly (p = 0.03) larger duct sizes overall (median, 5.2 mm; range, 1–27 mm) compared with patients with low-grade or intermediate-grade dysplasia (median, 4.5 mm; range, 1–8 mm).
A greater proportion of patients with high-grade dysplasia or invasive carcinoma compared with patients with low- or intermediate-grade dysplasia (67% v 32%) had 10 or more pancreatic cysts (p = 0.03). The difference between the two groups trended toward significance for patients with 20 cysts or more (33% vs 8%, p = 0.05).
Table 4 summarizes the results of regression models examining the associations between high-grade dysplasia and invasive carcinoma and various CT characteristics. Mural nodularity, cyst size of 3 cm or larger in the axial plane, septations (thin or thick), biliary ductal dilatation (for those lesions in the head or uncinate process), or calcifications were not found to be significantly associated with high-grade dysplasia and invasive carcinoma. Patients with a main pancreatic duct of 6 mm or larger (odds ratio [OR], 4.9; 95% CI, 0.95–25.1; p = 0.06), 10 or more cysts (OR = 3.8; 95% CI, 0.87–16.8; p = 0.08), and 20 or more cysts (OR = 13.7; 95% CI, 1.5–126.9; p = 0.02) showed trends toward being more likely to have high-grade dysplasia or invasive carcinoma.
Table 5 summarizes the characteristics of the 19 patients in the study who did not meet any of the Sendai consensus statement's imaging guidelines for cyst resection, including a cyst measuring 3 cm or larger in the axial dimension, mural nodularity, or main pancreatic ductal dilatation of 6 mm or greater. This group of patients was found to have seven cases (37%) of either high-grade dysplasia or invasive carcinoma.
| Discussion | |
|---|
The management of main duct IPMNs has become fairly well established: It is recommended that all main duct IPMNs be resected. Although there has been some debate regarding the management of side-branch IPMNs, the Sendai consensus statement in 2006 [3] established imaging criteria for their resection, including a maximum axial diameter of 3 cm or greater, dilatation of the main pancreatic duct of 6 mm or more, and internal mural nodularity [2, 3, 7–12]. Several additional criteria have been proposed in the literature including patient symptoms, biliary duct obstruction, elevated CA19-9 level, and abnormal cytologic results after endoscopic ultrasound–guided cyst aspiration. These criteria were thought to denote cysts with a greater likelihood of containing foci of high-grade dysplasia or invasive carcinoma [13, 14]. Conversely, although there is some ambiguity in the literature, most small cysts not meeting these criteria (i.e., simple cysts < 3 cm) are thought to be safe to observe [15, 16].
Accordingly, the risk of invasive carcinoma in side-branch IPMNs and small pancreatic cysts has been extensively studied in the literature: Sahani et al. [17, 18] reported 87% of pancreatic cysts smaller than 3 cm were benign and that 97% of small unilocular cysts were benign. Lee et al. [19] reported that only 31 of 166 pancreatic cysts (19%)—regardless of morphology—were malignant. In terms of studies looking specifically at side-branch IPMNs, most do not specifically distinguish between unifocal and multifocal IPMNs in their series, but many have shown fairly low rates of malignancy for side-branch IPMNs. Schmidt et al. [14] found an 18% malignancy rate with regard to unifocal IPMN. In their series of 43 patients with side-branch IPMN, Terris et al. [20] found no cases of invasive carcinoma and only 15% had evidence of high-grade dysplasia. Matsumoto et al. [21] found no cases of invasive carcinoma or high-grade dysplasia in their series of patients with side-branch IPMN measuring less than 30 mm without mural nodularity. In a series by Sugiyama et al. [22], none of the patients with side-branch IPMN was found to have invasive carcinoma.
However, as MDCT technology has improved, it has become increasingly common to visualize multiple side-branch IPMNs in the pancreas. This is concordant with the results of histopathologic studies, which have noted that IPMN can often be multifocal, with rates perhaps ranging as high as 25% (depending on the method used) [4]. Moreover, even when additional IPMNs are not initially visualized on CT, Sohn et al. [7] noted the development of either invasive or noninvasive IPMN in the pancreatic remnant of several patients after resection of a noninvasive IPMN, a result confirmed in at least one other study. The presence of completely negative margins in the pathologic samples of that study suggests that these recurrences were not secondary to the presence of residual dysplastic tissue after surgery but, rather, were due to multifocal disease [7, 23]. Interestingly, a study by Chari et al. [24] noted no recurrences after total pancreatectomy.
It is unclear if these multifocal pancreatic IPMNs are clonally related tumors arising from a single progenitor neoplasm that grows throughout the ductal system, if they reflect the development of multiple unrelated lesions within a ductal epithelium that may be genetically predisposed to the development of neoplasms, or if they reflect the development of multiple neoplasms due to genetic chance [4]. At least one genetic study of histopathologic specimens in patients with known multifocal side-branch IPMNs has shown distinct genetic alterations within different cysts in the same patient, suggesting that these multiple lesions are the result of several distinct genetic events [4]. Investigators have suggested that the multicentric model of IPMN development may reflect hyperplasia and dysplasia in different branch ducts due to different KRAS mutations, a genetic marker thought to be highly specific for detecting a mucinous neoplasm [25, 26].
Nevertheless, the proper management of multifocal IPMNs, when visualized on CT, has not been extensively studied in the literature, with only a few series detailing the association between IPMN multifocality and malignancy. Salvia et al. [4] examined 131 patients with multifocal side-branch IPMNs, roughly 8% of whom underwent surgery. Forty months after surgery, disease in every conservatively managed patient was radio-graphically stable without evidence of progressive disease and only one of the surgically managed patients had evidence of invasive carcinoma [4]. In their study looking at 145 patients with side-branch IPMN, Rodriguez et al. [27] reported that roughly 25% were found to have some evidence of multifocality on histologic specimens, with no statistically significant difference between the benign and malignant groups. In a study by Schmidt et al. [14], multifocality was identified in 41% of pathologic specimens, but there was no statistically significant difference in malignancy between the unifocal (18%) and multifocal (7%) groups. A pathologic study of specimens with multifocal side-branch IPMNs has found that most of lesions are of the gastric-foveolar histologic subtypes and generally contain low- to intermediate-grade dysplasia with a lesser risk of high-grade dysplasia or invasive carcinoma [5].
Although no consensus protocol has emerged regarding the management of these multifocal lesions, the aforementioned studies have suggested that no distinct management algorithm is needed for multifocal disease and that the Sendai criteria should be applied individually to each cyst in the pancreas [2, 4]. However, most studies advocate close radiologic surveillance of these patients especially given the practical difficulties in the surgical management of multiple pancreatic lesions that may involve all portions of the pancreas [4].
However, the results of this study suggest the possibility that the presence of multiple side-branch IPMNs should be considered as a suspicious feature. Notably, of the 43 patients in our study with multifocal disease (≥ 4 pancreatic cysts), 18 of the 43 (41.9%) had evidence of high-grade dysplasia (19%) or invasive carcinoma (23%) despite the failure to visualize a discrete soft-tissue mass in any of the patients. This result represents a higher percentage of patients compared with other studies that have looked solely at unifocal side-branch IPMN. Moreover, logistic regression models with high-grade dysplasia or invasive carcinoma as the outcome found 20 cysts or more to be significantly associated with the outcome (OR, 13.7; p = 0.02), and 10 cysts or more trended toward significance (OR, 3.8; p = 0.07) (Table 4). In addition, a significantly larger percentage of patients with high-grade dysplasia or invasive carcinoma were found to have 10 cysts or more, and the median number of cysts in patients with high-grade dysplasia or invasive carcinoma was significantly higher than in those with low- or intermediate-grade dysplasia (Table 3). These analyses are limited by the small sample size, however, and should be validated in larger populations.
Perhaps most important is that a total of 19 patients did not meet any of the Sendai consensus statement's imaging parameters for resection (i.e., cyst size ≥ 3 cm in the axial plane, mural nodularity, main pancreatic duct dilatation ≥ 6 mm) and seven (37%) of these patients had evidence of high-grade dysplasia or invasive carcinoma (Table 5). Clearly, the strict application of the Sendai imaging criteria to many of these cases may have inappropriately led to surveillance of lesions that deserved surgical resection.
Several confounding elements must be considered. First, although the number of cases of high-grade dysplasia or invasive carcinoma in our study is certainly concerning, we did not perform a comparison with patients with a solitary pancreatic cyst. Second, only eight of 43 patients in this study underwent total pancreatectomy, with the remaining undergoing either a Whipple procedure or distal pancreatectomy. Given the presence of multiple cysts in all parts of the pancreas in many of these cases, it is unclear whether the pathologic results for the resected portions of the pancreas are reflective of the unresected portions of the pancreas [7]. It is conceivable that this study could be underestimating the degree of significant dysplasia within a pancreas with multifocal disease. Moreover, we did not include analysis of those cases with a remnant pancreas after surgery in this study, and it is possible that some of these patients developed progressive disease in their pancreatic remnant after surgery. Third, every case in this study underwent surgery and we did not look at patients merely undergoing surveillance. It is likely that this exclusion biases the results toward significant dysplasia or carcinoma. Moreover, we only studied the Sendai imaging criteria for resection and did not look at patient symptoms, endoscopic ultrasound results, or tumor markers, all of which are undoubtedly also important factors in the decision to operate.
Ultimately, however, the presence of multiple pancreatic cysts may need to be viewed with added suspicion even when none of the cysts individually meet size or resection criteria according to the Sendai consensus. The presence of high-grade dysplasia or invasive carcinoma cannot be excluded in these specimens particularly when confronted with a large number of cysts. It is important not to be too cavalier about the risks associated with surgery: Even in the highest-volume centers, there is significant morbidity and mortality associated with the Whipple procedure or total pancreatectomy. Nevertheless, surgery may be an important option in many of these patients, although total pancreatectomy is currently reserved for the most concerning cases as a result of its significant physiologic impact [7]. Moreover, if surveillance is chosen, it is important not to rely solely on CT because many of the cases of invasive carcinoma in this study were completely unsuspected on the basis of CT; surveillance with endoscopic ultrasound, tumor markers, or both may be equally important. From a radiology perspective, given the proven advantages of MRI with regard to evaluation of the morphology and internal architecture of cysts, MRI may be a valuable component of follow-up in these patients for whom CT clearly shows limitations [28].
| References | |
|---|
