| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, Division of Abdominal Imaging and Intervention,
Brigham and Women's Hospital, Harvard Medical School, 75 Francis St., Boston,
MA 02115.
2 Department of Pathology, Brigham and Women's Hospital, Boston, MA.
3 Department of Radiologic Pathology, AFIP, Washington, DC.
4 Department of Radiology, Uniformed Services University of Health Sciences,
Bethesda, MD.
5 Department of Endocrine and Rhino-Oto-Laryngic-Head and Neck Pathology, AFIP,
Washington, DC.
6 Department of Medicine, Division of Gastroenterology, Brigham and Women's
Hospital, Boston, MA.
Received October 14, 2007;
accepted after revision May 13, 2008.
Address correspondence to K. J. Mortelé
(kmortele{at}partners.org).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Three radiologists, blinded to the pathologic IPMN subtype, retrospectively and independently reviewed the preoperative CT scans of 38 patients with surgically resected pancreatic IPMN: 11 intraductal papillary mucinous adenomas, 11 intraductal papillary mucinous carcinomas, and 16 intraductal papillary mucinous carcinomas with invasion. The patients, 16 women and 22 men, ranged in age from 38 to 80 years (mean age, 64.3 years). CT findings were correlated with each pathologic subtype using the chi-square (two-sided) test and analysis of variance. Interobserver agreement of the CT diagnosis of pathologic subtype and agreement between the CT diagnosis and pathologic subtype were also studied (kappa statistic).
RESULTS. Predominant main pancreatic duct (MPD) involvement (p = 0.04) and a wide (> 1 cm) connection of a side-branch lesion with the MPD (p = 0.03) correlated with intraductal papillary mucinous carcinoma with invasion. Tumor size, MPD diameter, number of tumors per patient, number of pseudoseptations per tumor, common bile duct dilatation, enlarged lymph nodes, intraductal calcifications, papillary bulging, and presence and size of a solid mass yielded no statistically significant relationship with pathologic subtype. Both interobserver agreement of CT diagnosis (range, 0.004–0.359) and agreement between CT diagnosis and pathologic subtype (range, 0.046–0.317) ranged from slight to fair.
CONCLUSION. Prediction of the pathologic subtypes of pancreatic IPMNs by CT is limited. Predominant MPD involvement and a wide connection of a side-branch lesion with the MPD are the only CT findings that can be used to predict the pathologic subtype of pancreatic IPMN.
Keywords: CT • cystic pancreatic tumors • intraductal papillary mucinous neoplasms • pancreas • pancreatic cancer
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Intraductal papillary mucinous adenoma is defined as a benign intraductal tumor with columnar mucin-producing epithelium showing papillary growth in the MPD or its side branches; no features of malignancy are present. Intraductal papillary mucinous carcinomas include noninvasive lesions that have histologic features of malignancy; these lesions have also been referred to as carcin oma in situ or IPMN with high-grade dysplasia. Intraductal papillary mucinous carcinomas with invasion include tumors that contain neoplastic epithelium outside the main or side-branch pancreatic duct, indicating invasion. Whether an IPMN is an intraductal papillary mucinous adenoma, carcinoma, or invasive carcinoma is important because patient care decisions and prognosis are affected [4]. For example, benign intraductal papillary mucinous adenomas may be followed conservatively or locally excised, whereas intraductal papillary mucinous carcinomas with or without invasion require radical resection and lymph node sampling. Also, noninvasive intraductal papillary mucinous carcinomas are associated with a better prognosis than invasive tumors [4].
Although previous studies have examined the imaging features that differentiate benign from malignant IPMNs, to the best of our knowledge, no studies have assessed whether CT can be used to distinguish intraductal papillary mucinous adenomas from intraductal papillary mucinous carcinomas and intraductal papillary mucinous carcinomas from intraductal papillary mucinous carci nomas with invasion. Therefore, we sought to evaluate whether CT can be used to predict the pathologic subtypes of pancreatic IPMNs.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The remaining 38 patients formed the study population; their mean age was 64.3 years (range, 38–80 years). Twenty-two patients (58%) were men and 16 (42%) were women. The most common presenting signs and symptoms were abdominal pain in 18 patients (47%) followed by weight loss in 10 (26%); diarrhea in six (16%); anemia in two (5%); and rectal bleeding, hematemesis, and jaundice in one patient (3%) each. Eight patients (21%) were asymptomatic. The mean time interval between the date of the last preoperative CT study and surgical resection was 51.2 days (range, 1–244 days); the time interval between the date of the last preoperative CT study and surgical resection was less than 30 days in 16 patients (42%), between 30 and 60 days in 12 patients (32%), between 60 and 90 days in four patients (10%), and more than 90 days in six patients (16%).
The study was HIPAA compliant. Institutional review board approval for reviewing medical records and images related to this study was obtained from both participating institutions. Informed patient consent was waived.
CT Technique and Analysis
The CT scans of each of the 38 patients in the study group were
independently and retrospectively reviewed by three radiologists, each of whom
was aware of the diagnosis of IPMN but blinded to the pathologic subtype. The
experience of the three readers was 15, 14, and 5 years of interpreting
abdominal and pancreatic CT scans, respectively. Twenty-four of 38 patients
had CT scans that were stored on DVD-ROM (digital video disc read-only memory)
and viewed on computer monitors, 10 patients had CT scans viewed using a PACS,
and the remaining four patients' original CT scans were viewed on hardcopy
film.
Of 38 CT examinations, 20 were performed after IV contrast material administration only, 15 patients underwent both unenhanced and contrast-enhanced CT, and three patients underwent only unenhanced CT. Contrast-enhanced CT scans were obtained during the portal venous phase (60–70 seconds after contrast injection) in all 35 patients; four patients also had delayed images obtained 5 minutes after contrast administration. The CT scan section thickness varied between 2.5 and 10 mm (mean section thickness, 5.2 mm); the section thickness was 3 mm in nine patients (24%), 5 mm in 22 patients (58%), 7 mm in two patients (5%), 8 mm in two patients (5%), and 10 mm in three patients (8%). Thirty-four of the 38 patients received oral contrast material before CT.
Each reader recorded the predominant distribution of tumor (main duct, side-branch duct, or combined). When the distribution was considered combined, the predominant pattern of ductal involvement (main duct or side-branch duct) was determined. Predominant main duct combined tumors were defined as those that had significant main duct abnormalities with less pronounced side-branch dilatation, and predominant side-branch duct combined tumors were defined as those that had significant side-branch duct abnormalities with less pronounced main duct dilatation. For tumors involving the MPD, readers were asked to classify the tumors as diffuse or segmental. Diffuse tumors involved the entire MPD; segmental tumors involved one or more segments of the MPD while sparing the other segments.
Side-branch tumors were classified further by location. The uncinate process was defined as the inferior and posterior aspects of the pancreatic head posterior to the mesenteric vessels. The pancreatic head was defined as the portion of the pancreas to the right of the superior mesenteric vein. The pancreatic neck was defined as the portion of the pancreas anterior to the superior mesenteric vein. The remaining pancreas to the left of the superior mesenteric vein was divided in half to differentiate pancreatic body from tail—the most distal half representing tail and the proximal half representing body.
The readers were also asked to determine whether the tumor was continuous with the MPD; in these cases, readers measured the largest width of the structure that connected the MPD to the tumor. The readers recorded the largest diameter of the tumor if side branch or combined; the MPD diameter if main duct or combined; the presence of a solid mass or calcified content within the tumor; the largest diameter of the solid mass; the number of tumors per patient; the number of pseudoseptations per tumor (none, 1–3, 4 or 5, > 5); and the presence of papillary bulging, common bile duct dilatation (reported in millimeters), invasion of adjacent organs, peripancreatic lymph nodes greater than 5 mm in the short axis, and liver lesions suggestive of metastases. Pseudo septations were defined as septationlike linear soft-tissue structures cours ing through the cystic lesion. Because we know that IPMNs pathologically have no true septations but instead have undulations of a dilated pancre atic duct or side branch, we opted to call them "pseudoseptations" to avoid inaccurately char acter izing IPMN as a tumor that has septations.
Finally, each reader was asked to provide independently an overall assessment of the pathologic subtype of IPMN: intraductal papillary muci nous adenoma, intraductal papillary mucinous carcinoma, or intraductal papillary mucinous carci noma with invasion. For this pur pose, the readers used a combination of CT features that were scored in each patient. For example, features that made readers choose carcinoma or carcinoma with invasion included marked main duct dilatation; side-branch duct dilatation larger than 3 cm; presence of a solid mass; presence of papillary bulging, calcifications, or biliary obstruction; and presence of local invasion and metastatic disease.
Pathologic Analysis
Two pathologists, aware of the reported diagnosis of IPMN, retrospectively
reviewed the pathologic slides from their own institution. Experience of the
two pathologists was 13 years and more than 20 years, respectively. Both
pathologists used a three-tier classification of IPMN subtypes: intraductal
papillary mucinous adenoma, intraductal papillary mucinous carcinoma, or
intraductal papillary mucinous carcinoma with invasion. Benign intraductal
papillary mucinous adenomas were defined as benign tumors, including those
with mild and moderate dysplasia, that were histologically characterized as
intraductal tumors with columnar mucin-producing epithelium showing papillary
growth into the MPD or side-branch pancreatic duct. Intraductal papillary
mucinous carcinoma was defined as a noninvasive tumor in which the intraductal
epithelium showed cytologic and architectural features of malignant cells
(e.g., increased nuclear–cytoplasmic ratio, nuclear pleomorphism,
cellular stratification, and coars ened chromatin). Intraductal papillary muci
nous carcinoma with invasion was defined as a tumor that had neoplastic
epithelium indicat ing invasion outside the MPD or side-branch pancreatic
duct. Tumors that contained foci of more than one subtype were classified as
the most aggressive type. The presence or absence of histologic evidence of
chronic pancreatitis was also recorded for each patient.
|
|
Before correlating (binary) presence or absence of specific CT features with pathology subtype, disagreements among readers were settled by consensus. The correlation of each of the consensus CT findings with pathologic tumor subtype was then evaluated using the Fisher Holton exact test.
For CT findings that had numeric values, the mean for all three readers was computed. Correlations between the means of these continuous variables and histologic tumor subtype were compared using the analysis of variance. Statistical comparisons that assessed the correlation of the number of pseudoseptations per tumor and num ber of tumors per patient, as detected with CT, with pathologic subtype were made using the Kruskal-Wallis test. To evaluate the diameter of the MPD as a discriminating variable of pathologic subtype, receiver operating characteristic (ROC) analyses were applied. First, data were analyzed by grouping tumors as intraductal papillary mucinous adenoma versus others (i.e., intraductal papillary mucinous carcinoma and intraductal papillary mucinous carcinoma with invasion); second, tumors were classified as intraductal papillary mucinous carcinoma with invasion versus others (i.e., intraductal papillary mucinous adenoma and intraductal papillary mucinous carcinoma).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Radiologic–Pathologic Correlation
Type of ductal involvement—On CT, 15 (39%) of 38 patients
were classified as having side-branch duct tumors, seven (18%) were classified
as having main duct tumors, and 16 (42%) were classified as having combined
tumors. Of 15 patients with side-branch duct tumors, two (13%) had involvement
of the uncinate process, 10 (67%) had involvement of the head, one (7%) had
body involvement, and two (13%) had tail involvement
(Fig. 1). Four (57%) of the
seven patients with solely main duct tumors had diffuse involvement of the
duct, whereas two (29%) had segmental involvement of the tail and one (14%)
had segmental involvement of the head (Fig.
2). Of the 16 patients classified as having combined tumors, five
had predominantly side-branch duct involvement and 11 had predominantly main
duct involvement (Fig. 3A,
3B). Although a strong tendency
was noted, there was no statistically significant relationship (p =
0.052) between the CT classification of type of duct involvement and
pathologic subtype of IPMN (Table
1).
|
|
|
A total of 18 patients had predominantly (n = 11) or exclusively (n = 7) MPD involvement with either diffuse or segmental MPD involvement. Of those 18 patients, 12 (67%) had intraductal papillary mucinous carcinomas with invasion, seven (58%) of whom had diffuse MPD involvement. All seven patients with diffuse involvement had intraductal papillary mucinous carcinoma with invasion. Of 11 patients (61%) with segmental involvement, two (18%) had intraductal papillary mucinous adenoma, four (36%) had intraductal papillary mucinous carcinoma, and five (45%) had intraductal papillary mucinous carcinoma with invasion. Overall, involvement of the MPD was predictive of intraductal papillary mucinous carcinoma with invasion (p = 0.04) (Table 2).
|
The remaining 20 patients had predominantly (n = 5) or exclusively (n = 15) side-branch duct tumors. At pathology, nine (45%) of these patients had intraductal papillary mucinous adenoma, seven (35%) had intraductal papillary mucinous carcinoma, and four (20%) had intraductal papillary mucinous carcinoma with invasion. No statistically significant relationship (p = 0.192) was found between side-branch duct tumor location and pathologic subtype (Table 2).
Continuity with MPD—Ten (50%) of the 20 tumors that either exclusively or predominantly involved side branches of the MPD were continuous with the MPD on CT (Figs. 3A, 3B and 4A, 4B, 4C). Of these, five (50%) were intraductal papillary mucinous adenoma, two (20%) were intraductal papillary mucinous carcinoma, and three (30%) were intraductal papillary mucinous carcinoma with invasion (p = 0.39). The median diameter of the connection was significantly wider among patients with intraductal papillary mucinous carcinoma with invasion (10.2 mm) than among patients with intraductal papillary mucinous carcinoma (4.8 mm) or intraductal papillary mucinous adenoma (3.9 mm) (p = 0.03) (Table 3).
|
|
|
|
Other CT features—No statistically significant correlations were found between IPMN pathologic subtype and the following qualitative CT features: the presence of a solid mass or calcified content within the tumor, papillary bulging, common bile duct dilatation, invasion of adjacent organs, or enlarged peripancreatic lymph nodes (Table 1).
Similarly, no statistically significant correlations were found between IPMN pathologic subtype and the following numeric CT features: the largest diameter of the tumor (side-branch duct or combined), the MPD diameter (main duct or combined), the largest diameter of solid mass, the number of tumors per patient, and the number of pseudoseptations per tumor (Table 3).
Regarding the distinction between intraductal papillary mucinous adenoma versus the others (i.e., intraductal papillary mucinous carcinoma and intraductal papillary mucinous carcinoma with invasion) when using the MPD diameter as a discriminating value, ROC analysis yielded a cutoff value of 8 mm as being most accurate. Tumors associated with an MPD diameter smaller than 8 mm were intraductal papillary mucinous adenomas, with a sensitivity and specificity of 57.1% and 92.9%, respectively (Fig. 3A, 3B).
Regarding the classification of tumors as intraductal papillary mucinous carcinoma with invasion versus others (i.e., intraductal papillary mucinous adenoma and intraductal papillary mucinous carcinoma), when using the MPD diameter as a discriminating value, ROC analysis yielded a cutoff value of 9.8 mm as being most accurate. Tumors associated with an MPD diameter equal to or larger than 9.8 mm were intraductal papillary mucinous carcinoma with invasion with a sensitivity and specificity of 64.7% and 80%, respectively (Fig. 4A, 4B, 4C).
CT-Based Prediction of Pathologic Subtype
The agreement between each reader's overall CT prediction of tumor subtype
and final pathologic diagnosis revealed slight to fair agreement (reader 1 =
0.285, reader 2 = 0.317, reader 3 = 0.046). The interobserver agreement among
all readers' overall CT prediction of pathologic tumor subtype also showed
slight to fair agreement (readers 1 and 2 = 0.359, readers 1 and 3 = 0.140,
readers 2 and 3 = 0.004).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Although established CT criteria can help differentiate benign from
malignant IPMN, to our knowledge, no studies have evaluated the potential role
of using CT to distinguish intraductal papillary mucinous adenomas from
intraductal papillary mucinous carcinomas and intraductal papillary mucinous
carcinomas from intraductal papillary mucinous carcinomas with invasion. In
prior stud ies, investigators have evaluated the predic tive value of main
versus side-branch pancreatic duct involvement; for example, tumors that
involve the MPD are more likely to be malignant than those that involve only
the side-branch ducts [6]. The
MPD has also been suggested to be wider in patients with malignant IPMN
compared with those with benign IPMN
[7–10].
For example, in their study of 36 patients, 26 of whom had malignant IPMN,
Taouli et al. [8] found that an
MPD diameter of more than 10 mm was associated with a specificity for
malignancy of 92%. Also, Irie et al.
[11] suggested, in their study
of 31 patients, that MPD dilatation greater than 10–15 mm for tumors
that involve the MPD portends a statistically significantly (p <
0.05) poorer prognosis than those that cause less dilatation. Similarly, the
results of prior research have shown that large side-branch duct tumors
[7,
10,
12,
13], papillary bulging
[7], common bile duct
dilatation [14], a large
number of tumors [7,
8,
11,
15], solid components
[7,
8,
12,
16] or calcified ductal
content [7,
8], invasion of adjacent organs
[17], enlarged peripancreatic
lymph nodes [18], and presence
of liver lesions suggestive of metastases
[18] are each indicative of a
malignant tumor. In their study of 36 patients, 26 of whom had malignant IPMN,
Taouli et al. [8] found the
specificity of the presence of a solid mass, multi-focal involvement, and
intraluminal calcified content was 96%, 77%, and 77%, respectively. In their
study of 36 patients, Kawamoto et al.
[10] found that the size of
the tumor in the branch duct type and combined type and the caliber of the MPD
were significantly larger in patients with invasive carcinoma when compared
with the lesions without invasive carcinoma (4.7 ± 1.7 vs 2.6 ±
1.4 cm [p = 0.0007] and 9.3 ± 5.5 vs 4.6 ± 4.1 mm
[p = 0.006], respectively). A solid mass (p < 0.001),
dilatation of the common bile duct or common hepatic duct (
15 mm), and
the presence of a stent (p = 0.0004) also correlated with the
presence of associated invasive carcinoma. Baba et al.
[12] reported that the
accuracy of detecting malignancy in side-branch duct lesions was highest using
endoscopic sonography to measure the cyst diameter, with an accuracy of 68.2%
at a cutoff value of 33.9 mm, in their study of 121 patients with IPMNs. For
main duct type tumors, accuracy was highest, 71.1%, at a cut-off value of 8.4
mm for main duct diameter using MR cholangiopancreatography
[11].
In our study, the type of ductal involvement seen on CT scans did not yield a statistically significant relationship with the pathologic subtype of IPMN. Nevertheless, tumors involving mainly the side-branch ducts were more likely benign and lesions involving mainly the MPD were more likely intraductal papillary mucinous carcinomas with invasion; combined lesions were also most likely intraductal papillary mucinous carcinomas with invasion. These results support those of prior studies that suggest that side-branch duct tumors are more commonly noninvasive [1, 6]. Sohn et al. [1], for example, found in 136 patients that 70% of branch duct tumors and 60% of combined type tumors were noninvasive compared with only 50% of main duct IPMN variants being non-invasive. Our results differ from those of Sohn et al. in that our data revealed a relative minority of patients with combined type lesions as having noninvasive tumors. Our study showed that 87% of side-branch duct tumors and only 38% of combined type tumors were noninvasive compared with 43% of the main duct tumors being noninvasive. All patients with diffuse MPD involvement had intraductal papillary mucinous carcinomas with invasion; of the patients with segmental MPD involvement, those with invasive carcinoma (45%) accounted for the greatest percentage. Thus, if the MPD is predominantly involved, regardless of whether involvement is diffuse or focal, IPMN is most likely carcinoma with invasion.
Among the CT findings, we found that the diameter of the connection of a
side-branch duct tumor with the MPD was significantly less for benign IPMN
compared with intraductal papillary mucinous carcinomas with invasion. We
found no statistically significant relationship between IPMN pathologic
subtype and any other individual CT finding, including the presence of a solid
mass in the MPD, bulging papilla of Vater, ductal calcifications, enlarged
lymph nodes, number of pseudoseptations per lesion, median diameter of the
MPD, and number of lesions per patient. Thus, our findings illustrate the
difficulty in using CT to predict where a particular tumor lies along the
adenoma–invasive carcinoma continuum. Other researchers have also
reported the lack of accuracy of predictive criteria of IPMN subtype
[19]. Bernard et al.
[19] retrospectively evaluated
the pathology findings in 53 IPMNs and found that, in contradistinction to
other researchers
[7–10],
neither the diameter of the tumors (branch duct lesions 30 mm) nor the
MPD diameter ( 15 mm in either combined or MPD involvement) correlated
with malignancy [19].
Nevertheless, data derived from our ROC analyses suggest that the MPD diameter
can be used in a practical way to discriminate between IPMN subtypes; for
example, an MPD diameter less than 8 mm appears highly specific for
intraductal papillary mucinous adenoma. Therefore, this information could be
used to support observing IPMN associated with an MPD less than 8 mm in
elderly patients or patients who are poor surgical candidates.
Our study had a number of limitations. First, there was significant variability in the techniques used to obtain the CT scans and how the data were evaluated. Not all CT scans were obtained with a section thickness of 5 mm or less; four CT scans were evaluated from hard copies; because of unavailability, multiplanar imaging was not used in the data analysis; and more than half of the patients were examined with contrast-enhanced CT only. The lack of unenhanced CT scans may have limited radiologists' ability to differentiate high-attenuation mucin globules from solid masses. Solid masses enhance and often appear on the nondependent surface of ducts, unlike mucin globules that may also change in location with positional changes.
A second limitation is that two pathologists were used, each reviewing cases from his or her respective institution. Theoretically, more uniformity in histologic interpretation would have been obtained if a single pathologist had reviewed all cases. However, both pathologists used agreed-on definitions of each subtype in reviewing the slides and assigning diagnoses.
A third limitation is the time interval between preoperative CT and surgical excision (mean, 51.2 days). Some tumors conceivably could have progressed in histologic subtype during the interval; however, these tumors are typically slow-growing and would not be expected to change in that short period of time. Another limitation is slight to fair interobserver agreement between the overall CT prediction of pathologic subtype by the radiologists and the slight to fair interobserver agreement between the overall CT prediction of pathologic subtype and the true pathologic subtype. We believe that these relatively poor interobserver agreements underscore the inability of radiologists to use CT to accurately predict or agree on specific pathologic subtypes of IPMN.
A final limitation is the relatively small number of patients in each subtype and the relatively low prevalence of some CT features in our patients; this may explain the lack of significant correlation between imaging and pathology for some of the CT features evaluated.
In summary, unlike prior reports, our study suggests that prediction of pathologic subtype of pancreatic IPMN using CT is limited. In our study, only the presence of MPD involvement and the diameter of the connection of a side-branch duct tumor with the MPD, if present, can be used to predict IPMN pathologic subtype. Other heretofore proposed predictors of malignancy—that is, a solid mass in the MPD, MPD diameter, bulging papilla of Vater, ductal calcifications, number of pseudoseptations, and number of lesions—did not correlate with pathologic subtypes of IPMN. However, the relatively small number patients in each subtype may explain the lack of significant correlation between imaging and pathology for some of the CT features evaluated. Whether other, yet undefined, CT features can be used to predict IPMN tumor subtype remains to be determined.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
