- Email: [email protected]
Outcomes after pancreatectomy for intraductal papillary mucinous neoplasms of the pancreas: An institutional experience
Outcomes after pancreatectomy for intraductal papillary mucinous neoplasms of the pancreas: An institutional experience A. D. Yang, MD, L. G. Melstrom, MD, D. J. Bentrem, MD, M. B. Ujiki, MD, J. D. Wayne, MD, M. Strouch, MD, R. H. Bell, MD, S. M. Rao, MD, and M. S. Talamonti, MD, Chicago, Ill
Purpose. To evaluate the experience with pancreatectomy for intraductal papillary mucinous neoplasm (IPMN) at a single academic institution. Methods. A prospective pancreatic database was reviewed and identified 43 patients with IPMN who were managed operatively. Clinicopathologic features and predictors of outcome were examined. The World Health Organization pathologic classification of IPMN was utilized. Results. IPMN was diagnosed in 21% of patients who underwent pancreatic resection for solid or cystic mass lesions. Ninety-five percent of patients were symptomatic. Patients were managed with total pancreatectomy, pancreaticoduodenectomy, distal pancreatectomy, central pancreatectomy, or enucleation. Nine patients had adenomas, 14 had borderline neoplasms, 10 had carcinoma in situ, and 9 had invasive carcinoma. Overall, 23 patients (53%) had lesions with main duct involvement. Frozen section transection margins were positive for malignancy in 2 patients. With a mean follow-up of 17 months, the 5-year disease-specific survival for patients with main duct involvement was 67%. The 5-year disease-specific survival for patients with benign lesions was 100%, and 61% for patients with malignant lesions (P ⫽ .02). The presence of symptoms, increased CA 19-9, and tumor size were not predictive of malignancy. Increased serum bilirubin concentrations were predictive of malignancy (P ⫽ .03). Main duct involvement was also associated with malignancy (P ⬍ .02). Conclusions. Cancer is found in 65% of patients with IMPN involving the main duct. Based on our data, patients with symptomatic, main duct involvement, especially those with an increased serum bilirubin, should be offered resection. Alternatively, patients with side branch IPMN may be managed conservatively. (Surgery 2007;142:529-37.) From the Department of Surgery, Feinberg School of Medicine, Northwestern University, Chicago, Ill
Intraductal papillary mucinous neoplasms of the pancreas (IPMN) represent a recently described pancreatic lesion, the natural history of which remains incompletely understood. The World Health Organization (WHO) has defined IPMN as papillary mucin-producing neoplasms with tall, columnar, mucin-containing epithelium with or without papillary projections that arise in the main pancreatic duct or its major branches
Both A. D. Yang, MD, and L. G. Melstrom, MD, contributed equally to this manuscript. Accepted for publication July 21, 2007. Reprint requests: Mark S. Talamonti, MD, 201 E. Huron St. Galter 10-105, Chicago, IL 60611. E-mail: [email protected]. 0039-6060/$ - see front matter © 2007 Mosby, Inc. All rights reserved. doi:10.1016/j.surg.2007.07.007
(commonly referred to as “main duct IPMN” and “branch duct IPMN,” respectively).1 IPMNs lack the ovarian stroma characteristic of mucinous cystic neoplasms of the pancreas, which also do not exhibit involvement of the major pancreatic ducts. With the clarified definition of IPMN, as well as the widespread use of advanced imaging, the identification of IPMN found during investigation of related or unrelated medical conditions has increased dramatically in the last decade.2-7 IPMN now constitutes a substantial proportion of resected pancreatic lesions in the United States. Whether the diagnosis of IPMN mandates resection remains in question. Despite decreased morbidity and mortality in experienced hands, pancreatectomy remains a formidable operation that confers considerable risk for patients. As a consequence, it has become paramount to further examine predictors of outcome to enable appropriate selection of patients SURGERY 529
530 Yang et al
with IPMN for operative intervention. Recent reviews and consensus conferences have attempted to standardize the management of patients with IPMNs8-10; however, no single test exists that unequivocally differentiates between benign and malignant neoplasms. Therefore, controversy exists regarding predictors of malignancy in patients with IPMN who are managed by resection. The aims of this study were 3-fold: To determine the incidence of IPMN in our patients with pancreatic mass lesions managed by resection, to examine clinical and pathologic factors in this patient population, and to determine outcome measures and risk factors associated with malignancy to make recommendations regarding management for patients with IPMN. METHODS A prospective pancreatic database from Northwestern Memorial Hospital was reviewed to identify patients with IPMN who were managed operatively. All patients who underwent resection had pathologic evaluation by a single gastrointestinal pathologist (S.M.R.), and complete records were included in this study. To avoid overlap with mucinous cystic neoplasms, only patients with neoplasms of the main pancreatic duct or side branch involvement fulfilling the WHO classification1 were included. Forty-three patients who underwent surgical resection for IPMN between January 1997 and August 2006 were identified from the database. The complete medical record of each patient was examined for demographic and clinicopathologic features. Operative reports were reviewed for type of resection, completeness of resection, and findings of intraoperative frozen section biopsies. Pathology reports were reviewed to confirm the diagnosis and to obtain pathologic variables. These pathologic variables included tumor location and size, surgical margin status, and determination of histologic subtypes. Tumors were classified as either benign (adenoma and borderline tumors) or malignant (carcinoma in situ or invasive cancer) to compare demographic features and/or outcome measures that could predict the presence of malignancy in patients with IPMN. 2 analysis was used to identify factors predictive of malignancy and survival. Survival was estimated by means of the Kaplan-Meier method and compared using log rank.11 Statistical significance was defined as P ⬍ .05. RESULTS Demographics and presentation. IPMN was diagnosed in 21% (43/204) of patients who underwent pancreatic resection for solid or cystic masses of the
Surgery October 2007
pancreas between January 1997 and August 2006. The mean age of the cohort was 62 ⫾ 11 years. There were 28 men (65%) and 15 women (35%). Forty-two patients (98%) were white and 1 was black (2%). A history of smoking was present in 30 patients (70%), and 15 patients (35%) admitted to a history of significant alcohol consumption. The incidence of acute pancreatitis at presentation was 35% (15 patients), and 6 patients (14%) had a previously established diagnosis of chronic pancreatitis. Family history was relevant for some form of cancer in ⱖ1 family members in 29 patients (67%), whereas 10 patients (23%) had a history of a prior personal malignancy. Diabetes mellitus was present in 8 patients (19%). Symptoms were present in 41 of the 43 patients (95%). The majority of patients presented with abdominal pain (67%) and/or weight loss (60%). A minority of patients presented with nausea/vomiting (37%) and/or jaundice (23%). Random serum glucose concentrations of ⬎200 mg/dL were found in 7 patients (16%); 9 patients had random serum glucose concentrations of 100 to 200 mg/dL (21%). Serum total bilirubin concentration was increased in 7 patients (16%). Serum alkaline phosphatase was increased in 8 patients (19%) and serum CA 19-9 was increased in 6 patients (14%). Operative management. Preoperative decision making regarding choice of pancreaticoduodenectomy, distal pancreatectomy, or total pancreatectomy was predicated on a preoperative evaluation that included some combination of endoscopic retrograde cholangiopancreatography, endoscopic ultrasonography, and computed tomography or magnetic resonance imaging. Preoperative findings favoring pancreaticoduodenectomy were main duct involvement of primarily the head of the pancreas and sparing of the body and tail from involvement with IPMN. Distal pancreatectomy or extended distal pancreatectomy was favored in patients without any obvious main duct involvement in the head of the pancreas but involvement of disease to the left of the portal vein. Pancreaticoduodenectomy was the most common procedure, performed in 25 patients (58%), followed by distal pancreatectomy in 10 patients (23%), and total pancreatectomy in 6 (14%) patients. One patient (2.5%) underwent central pancreatectomy with pacreatostrostomy and 1 patient (2.5%) had “enucleation” of the IPMN. Review of pathology. The distribution of neoplasms by location was as follows: head/uncinate process (n ⫽ 25, 58%), neck/body (n ⫽ 5, 12%), tail (n ⫽ 7, 16%), and entire gland (n ⫽ 6, 14%). Thirty-five patients had discrete, measurable le-
Yang et al 531
Surgery Volume 142, Number 4 Table I. Pathologic characteristics (n ⫽ 43) Characteristic Tumor location Head/uncinate process Neck/body Tail Entire gland Duct involvement Main duct Branch duct Histologic class Adenoma Borderline Carcinoma in situ Invasive cancer Lymph node involvement Present Present with malignancy Final transection margin Positive for IPMN Positive for IPMN and malignancy
% (n) 58 (25) 12 (5) 16 (7) 14 (6) 53 (23) 47 (20) 23 (10) 33 (14) 23 (10) 21 (9) 17 (7) 21 (4) 35 (14) 14 (6)
IPMN, intraductal papillary mucinous neoplasm. Median tumor size is 25 mm (range, 3 to 100).
sions with a median tumor size of 25 mm (range, 3 to 100 mm). The remaining 8 patients had multifocal and/or multicentric disease in which no specific size was assigned to their neoplasm. Overall, 23 patients (53%) had lesions with main duct involvement (solely or combined main and sidebranch ducts). Malignant IPMN (carcinoma in situ/invasive cancer) was found in 44% of lesions. Histologic classification identified adenomas in 9 patients (21%), 14 (35%) with borderline neoplasms, 10 (23%) with carcinoma in situ, and 9 (21%) with invasive carcinoma. Of the patients with invasive carcinoma, 4 (21%) had lymph node involvement (Table I). The pancreatic transection margin was evaluated with intraoperative frozen section in 36 of 43 patients. Frozen section transection margins were positive for malignancy in 2 patients, 1 who underwent completion total pancreatectomy. The other patient underwent extended pancreaticoduodenectomy with a final frozen transection margin of IPMN with “dysplasia.” Final pathology of the transection margins showed IPMN involvement in 14 patients. Benign IPMN adenoma was observed at the transection margin in 3 patients, borderline or atypical IPMN in 5 patients, and carcinoma in situ in 6 patients, but invasive carcinoma in no patients. Outcome and predictors of malignant disease. With a mean follow-up of 17 months, the 5-year disease-specific survival for all patients was 76% (Figure) and 64% for patients with main duct in-
volvement. The 5-year disease-specific survival for patients with benign lesions was 100% and 49% for patients with malignant lesions (median survival not reached vs 30 months, P ⫽ .02). No demographic factor was associated with the presence of malignancy in the final resected specimens. The presence of symptoms, increased serum CA 19-9, and lesion size were not predictive of malignancy; in contrast, increased serum bilirubin level was predictive of malignancy (P ⫽ .03). Main duct involvement was associated with malignancy (P ⬍ .02). Malignant lesions were found in 15 of 23 patients with main duct involvement. Of the remaining 20 patients with only branch duct involvement, no patients had invasive cancer and only 4 patients had carcinoma in situ (Table II). DISCUSSION Despite the multitude of recent reports addressing the diagnosis and management of IPMN of the pancreas, it is difficult to assess accurately the true incidence and medical impact of these relatively rare neoplasms. With increased utilization of imaging technology, there has been an increased detection of asymptomatic cystic lesions of the pancreas. Not surprisingly, there has also been an increase in the number of operations for pancreatic cystic lesions, many of which are IPMNs. Notwithstanding this increase in pancreatic resections and subsequent categorization of certain pancreatic tumors as IPMNs, there remains an imperative to increase our understanding of the biologic behavior and the spectrum of diseases represented by these neoplasms. Experience with IPMN continues to be limited to retrospective reports from high-volume centers, and many attempts have been made to identify prognostic factors that might influence the management of these patients. Even as larger series and longer follow-up times evolve, questions remain as to which factors are predictive of outcome. Therefore, we sought to determine the incidence of IPMN in our patients with pancreatic mass lesions managed by resection; to examine clinical and pathologic factors in this patient population; and to determine outcome measures and identify factors associated with malignancy to make recommendations regarding management of patients with IPMN. In some centers, the proportion of pancreatic resections for cystic neoplasms of the pancreas has doubled within the last 10 years. Raut et al4 found the incidence of IPMN among the cystic lesions of pancreas treated by resection was 26% (35/137). We have found a much higher proportion of IPMN
532 Yang et al
Surgery October 2007
Figure. Kaplan-Meier actuarial survival curves for patients with IPMN who were operatively managed at a mean follow-up of 17 months. The overall 5-year disease-specific survival for all patients was 76% (upper graph). The 5-year disease-specific survival was significantly decreased in patients with malignant lesions (49%) compared with patients with benign lesions (100%; P ⫽ .02).
in this group (51%, 43/85), which constituted ⬎20% of our pancreatic resections during the study period. The demographic characteristics of our patients were comparable to those found in most series.2-7,12 Similar to the series reported by D’Angelica et al,3
from Memorial Sloan-Kettering Cancer Center, we found that 23% of our patients had a history of a prior extrapancreatic malignancy. Fernandez-del Castillo et al13 found that 27% of IPMNs were discovered incidentally, and the rest had some complaint referable to the pancreas. Similarly,
Yang et al 533
Surgery Volume 142, Number 4
Table II. Predictors of malignant disease for patients with IPMN
Tumor size (mm) ⬍20 ⬎20 Presence of symptoms Increased total bilirubin Increased CA 19-9 Duct involvement Main duct involvement (Main/Combined) Branch duct IPMN
Benign IPMN, % (n)
Malignant IPMN, % (n)
P
62 (15) 38 (9) 96 (23) 10 (2) 18 (4)
42 (8) 58 (11) 95 (18) 27 (5) 10 (2)
NS NS NS .03 NS
35 (8)
65 (15)
⬍.02
80 (16)
20 (4)
⬍.02
IPMN, intraductal papillary mucinous neoplasm.
we found the majority of patients with IPMNs at our institution presented with symptoms. A majority of our patients underwent pancreaticoduodenectomy. Extent of resection was similar to that of other published studies.3,6 Our proportion of patients with benign disease (adenoma or borderline IPMN) compared with patients with malignant disease (carcinoma in situ or invasive IPMN) reflects current management practices in large referral centers. Although early reports showed increased rates of invasive disease (37% to 48%)3,4,6 with stricter criteria for operative intervention, more liberal selection criteria are now being applied with a resulting lower proportion of invasive disease (21%). Our survival estimates differ somewhat from most published reports because we chose to classify our patients as either benign or malignant, whereas many case series classify their patients as either noninvasive (adenoma, borderline, carcinoma in situ) or invasive. We chose to divide our patients into benign and malignant categories because we agree with the theory put forward by Salvia et al,5 namely, that IPMN progresses from benign disease to eventual invasive cancer. Thus, we conclude that carcinoma in situ is the immediate precursor to invasive cancer and should be considered as malignant disease. This approach explains the increased survival in our benign and malignant IPMN groups, because patients with carcinoma in situ have shifted from the benign to the malignant category. Intraoperative frozen section is not utilized uniformly at other high-volume centers.3,4,6 The decision to perform intraoperative frozen section analysis of the margin at our center was at the operating surgeon’s discretion during the time frame of this study. Currently, we perform intraop-
erative frozen section analysis of the margin routinely with the purpose of determining whether carcinoma in situ or invasive cancer is present at the final transection margin. The influence of invasive or noninvasive disease at the margin of transection on survival is controversial and pertinent to multifocal disease. Although we were unable to assess the association of margin status with malignancy, recurrence, or survival, data from Raut et al4 and others3 suggest that the presence of noninvasive IPMN (adenoma, borderline, or carcinoma in situ) does not affect recurrence and/or survival adversely. Data from White et al14 suggest that, although negative margins should be the operative goal, when achievable with partial pancreatectomy in patients with noninvasive IPMN, the risk of local recurrence is not high enough to mandate total pancreatectomy for all patients with positive margins. As a result, what remains controversial is the importance of the frozen section margin versus the gross character of the duct proximal to the resection margin. Although this controversy is not addressed specifically by our data, we believe that if intraoperative frozen section analysis of the resection margin has any role, it is likely in the identification of remaining invasive cancer in the remnant pancreas. Other investigators have also an increased serum bilirubin concentration to be the only laboratory value predictive of outcome.3,4,6,7,9,12 The presence of invasive disease has correlated consistently with poor outcome. The pathologic factors that correlate with the presence of malignancy are less consistent. D’Angelica et al,3 Sohn et al,6 and others5 have shown main duct involvement to be associated with malignancy. Our results confirm this as a pathologic variable predictive of malignancy. The data in our series regarding relation of tumor size to malignant potential are similar to those published by Fernandez-del Castillo et al,13 which show a large percentage of patients (39%) with malignant potential in symptomatic lesions ⬍2 cm. We found a 42% incidence of malignant IPMN in lesions that measured ⬍2 cm. Thus, small, main duct IPMN may harbor an invasive cancer, whereas a larger, branch duct lesion may contain only benign IPMN. We conclude that tumor size should not be an independent factor in selecting patients with IPMN who require operative intervention. Cancer was found in 65% of patients with IPMN involving the main duct. We did not find invasive cancer in any patients without main duct involvement. Based on our data, patients with symptomatic, main duct involvement, especially those with an increased serum bilirubin concentration, should
534 Yang et al
be offered resection. Alternatively, patients with side branch IPMN may be managed conservatively. Drs Yang and Melstrom contributed equally to the writing of this manuscript. The authors thank Ms. Jane Stocker, LPN, for her invaluable contribution to this paper through her management of the pancreatic resection database. REFERENCES 1. Longnecker DS, Adler G, Hruban RH, Kloppel G. Intraductal papillary-mucinous neoplasms of the pancreas. In: Hamilton SR, Aaltonen LA, eds. World Health Organization Classification of Tumours: Tumours of the Digestive System. Lyon, France: IARC Press; 2000:237-240. 2. Chari ST, Yadav D, Smyrk TC, et al. Study of recurrence after surgical resection of intraductal papillary mucinous neoplasm of the pancreas. Gastroenterology 2002;123: 1500-7. 3. D’Angelica M, Brennan MF, Suriawinata AA, Klimstra D, Conlon KC. Intraductal papillary mucinous neoplasms of the pancreas: an analysis of clinicopathologic features and outcome. Ann Surg 2004;239:400-8. 4. Raut CP, Cleary KR, Staerkel GA, et al. Intraductal papillary mucinous neoplasms of the pancreas: effect of invasion and pancreatic margin status on recurrence and survival. Ann Surg Oncol 2006;13:582-94. 5. Salvia R, Fernández-del Castillo C, Bassi C, et al. Main-duct intraductal papillary mucinous neoplasms of the pancreas: clinical predictors of malignancy and long-term survival following resection. Ann Surg 2004;239:678-85. 6. Sohn TA, Yeo CJ, Cameron JL, et al. Intraductal papillary mucinous neoplasms of the pancreas: an updated experience. Ann Surg 2004;239:788-97. 7. Wada K, Kozarek RA, Traverso LW. Outcomes following resection of invasive and noninvasive intraductal papillary mucinous neoplasms of the pancreas. Am J Surg 2005; 189:632-6. 8. Doi R, Fujimoto K, Wada M, Imamura M. Surgical management of intraductal papillary mucinous tumor of the pancreas. Surgery 2002;132:80-5. 9. Salvia R, Bassi C, Falconi M, et al. Intraductal papillary mucinous tumors of the pancreas. Surgical treatment: at what point should we stop? JOP 2005;6:112-7. 10. Tanaka M, Chari S, Adsay V, et al. International consensus guidelines for management of intraductal papillary mucinous neoplasms and mucinous cystic neoplasms of the pancreas. Pancreatology 2006;6:17-32. 11. Kaplan E, Meier P. Non-parametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457-81. 12. Wiesenauer CA, Schmidt CM, Cummings OW, et al. Preoperative predictors of malignancy in pancreatic intraductal papillary mucinous neoplasms. Arch Surg 2003;138:610-7. 13. Fernández-del Castillo C, Targarona J, Thayer SP, Rattner DW, Brugge WR, Warshaw AL. Incidental pancreatic cysts: clinicopathologic characteristics and comparison with symptomatic patients. Arch Surg 2003;138:427-23. 14. White R, D’Angelica M, Katabi N, et al. Fate of the remnant pancreas after resection of noninvasive intraductal papillary mucinous neoplasm. J Am Coll Surg 2007;204:987-93.
DISCUSSION Dr Keith D. Lillemoe (Indianapolis, Indiana): Your report adds to the body of literature that has
Surgery October 2007
already been collected on this topic and continues to help us clarify this somewhat of a mystery of a disease for which we really do not know all the answers. Although you have answered some questions, you have raised some as well. So if you don’t mind, I would like to just ask a few. You describe in this series that 95% of your patients are symptomatic, which I think is a higher percentage than some of the other series. This may be sort of a self-fulfilling prophecy because these are patients who you chose to operate on. Therefore, I would like to ask you about your management of those patients who you choose not to operate on. How do you decide who to operate and who not to operate on? And specifically, because size did not seem to be a predictor of malignancy in your series and symptoms did not seem to be a predictor of malignancy, how do we deal with the asymptomatic patient who is found on a routine screening CT scan for some other purpose to have a 1.5-cm IPMN, keeping in mind that the consensus conference that has recently come out has suggested that perhaps observation for 3-cm IPMNs might be appropriate? This may have been a typo in the copy of the manuscript that I saw. You referred to a 65% incidence of a positive family history for pancreatic cancer in the manuscript. Here you just said history of cancer. Obviously, that is a big difference. And clearly, some further genetic workup would be necessary if you really had that high a correlation. Finally, you have great survival results on these patients, but you did not answer the question about whether the patients who survived, either with benign or malignant IPMNs, had any recurrent IPMN disease in the remnant gland? How do you follow your patients for surveillance for development of new tumors in the gland? Do you recommend that even people with benign disease be followed with magnetic resonance cholangiopancreatography (MRCP) or endoscopic ultrasound (EUS) to look for new tumors? Dr A. D. Yang (Chicago, Illinois): Thank you, Dr Lillemoe, for your comments and your excellent questions. As to your first question as to how we decide on who to operate or not, we use several criteria. I think one is that you have to consider the patient, their acceptable comorbidities, and the relative safety of the operation. If we have a patient with a 1.5-cm IPMN, size does not necessarily dictate what we are going to do in terms of operative management. If the patient has a main duct type IPMN or if they have an increased serum total bilirubin, that would make us more likely to operate. However, if it was a side
Surgery Volume 142, Number 4
branch IPMN and the patient was asymptomatic, I do not think size is much of a factor in the equation and we could possibly observe these patients. As to your comment on the 65% family history, that was a typo. It is that 65% had a family history of some malignancy. As to your final question on recurrent disease, we have patients with recurrence. We have 1 patient who had a metastatic recurrence and the rest have had local recurrences. None of the patients in our series as of yet had to be operated on for recurrence. How do we follow these patients? Regardless of whether they had malignant or benign IPMN, we follow them very closely after surgery. It is our policy to get an MRCP, which we believe is the best diagnostic study, on these patients every 6 months for the first 1 to 2 years. And when we are satisfied that either there is no evidence that the patient has possibly a cancer or malignant disease, then we go to about 1-year intervals in following these patients with imaging only. Dr E. Christopher Ellison (Columbus, Ohio): I think IPMN is a very interesting disease in that it has the potential of being multicentric, and, as Dr Lillemoe alluded to, this multicentricity may not be necessarily synchronous with the index lesion. A point in fact is that we have observed recurrences in the residual pancreas in 3 patients with malignant main duct IPMN occurring 7, 8, and 10 years after their index operation with negative margins. Indeed, 2 have undergone subsequent resection with documented recurrent disease. The third patient has declined reoperation at this point in time. So in this respect, we tend to follow these patients for a long period of time. And I wanted to know what the policies were at Northwestern in terms of following them carefully. How long do you follow them? Do you stop at 5 years? Or do you continue to follow them out for 10 years after their index operation? A second point is we have observed in some community settings that surgeons may attempt a distal pancreatectomy with splenic preservation for this disease when it involves the tail of the pancreas, and subsequently these patients have had recurrence in the splenic hilum. And I would like to know what your policies are regarding splenic preservation in the presence of IPMN involving the distal pancreas. Dr A. D. Yang (Chicago, Illinois): As to your first question on how long we follow these patients, basically there is no endpoint. We follow them, as stated, with scans every 6 months for the first 1 to 2 years and then follow them clinically on imaging
Yang et al 535
scans. There is no limit at this point. We do not follow them for just 5 years; we follow them perpetually. Your second question regarding distal pancreatectomy: We do not perform splenic preservation in any of our distal pancreatectomies, so that has not been a problem. Dr Christian M. Schmidt (Indianapolis, Indiana): I was intrigued by your finding that size did not matter in IPMNs and was curious to know a few things about that. First, how did you determine size of IPMNs, particularly in cases of multifocal disease? Was your tumor size just the branch-type lesions or main-type lesions and mixed-type lesions? How did you measure these? Second, size did not matter when you compared malignant with nonmalignant lesions. Did you look at adenoma versus borderline versus carcinoma in situ versus invasive? Was there a trend that size mattered as you went up in the pathologic grades? The Hopkins group recently published a paper suggesting it was only between carcinoma in situ and invasive that there was actually a statistically significant size difference. Finally, your whole paper is set up on malignant versus nonmalignant. Did you look at invasive versus noninvasive? And what predictive factors were present in the invasive tumors? Dr A. D. Yang (Chicago, Illinois): The first question was to the size criterion in patients with diffuse or multicentric disease. When we did the calculations for size as related to its predictive value in the malignancy, we excluded those patients who did not have a tumor that could be measured; that is, they had diffusely dilated duct with IPMN within the duct. So we took a cutoff—2 cm—to see if there was an association with malignancy in those patients with measurable tumors, and we did not find a correlation there. And we did look at a trend in size and did not see any between the subtypes that we could find in terms of correlation and size. I did not see that report on carcinoma in situ versus invasive. We did not look at that other than we did not see a trend there. Finally, our decision to classify these patients into malignant versus benign disease was based on our clinical practice at Northwestern. We treat patients with carcinoma in situ basically the same as patients with invasive cancer, especially in terms of operative therapy, so we chose to classify our data in this way. And in the literature, part of the controversy here is that there have been some groups that have classified the tumors as noninvasive versus invasive disease and other groups that have classified it as malignant versus benign disease. And
536 Yang et al
we chose to do it this way based on our clinical practice to correlate better clinically with what we do. Dr Roger G. Keith (Saskatoon, Saskatchewan, Canada): It is interesting over the history of time this disease has become a better and better and better cancer and surgery has become less and less and less invasive. You are suggesting that we ought to consider a total pancreatectomy for people with main duct disease, that we accept that invasive carcinoma and in situ carcinoma are the same, and that we accept that in your group one third of your main duct patients did not have cancer. Have you followed your in situ group to their mortality and do you have any absolute deaths in that group? I think we need to know as these reports come through whether or not in situ carcinoma is lethal. Dr A. D. Yang (Chicago, Illinois): I believe that there were no deaths in the carcinoma in situ group as of yet. Dr Mark S. Talamonti (Chicago, Illinois): I think a lot of the questions have to do with size and invasion versus in situ. And this is a relatively small series compared with other series, so I think to make definitive statements about the implications of size and outcome may be a bit of a limitation to the study. Suffice it to say that we did not find that size alone was a predictor of malignancy. That is different than saying whether or not size within any subgroup is a predictor of outcome. You could have an 8-cm cystic lesion in the head of the pancreas that was all benign IPMN and that may or may not predict whether there is invasion in it. Once you get invasion in it, does size matter if there is an invasive component to it? That is a different but more subtle question. And I do not think we can really address that in the limited number of patients we have. In terms of follow-up for carcinoma in situ and duration of follow-up, I think that point has been made well and emphasized. This is not the type of disease that has an endpoint to it. Just like a woman with carcinoma in situ of the breast would have a yearly mammogram for long-term follow-up, I think when you talk about patients who have a premalignant lesion or a predisposition to developing invasive cancer of the pancreas, those patients need to be followed accordingly. What the exact guidelines are for that are, I think, remain to be written. We follow patients yearly with MRCP of the pancreas indeterminately, as Dr Yang mentioned. We chose to include in situ carcinoma in our survival analysis mainly because we manage in situ
Surgery October 2007
carcinoma similarly to invasive carcinoma. But there is no question that if you put the in situ patients with the invasive patients, the outcomes for those patients will improve. Whether or not that is a statistically valid method to do that, I think if you had a large enough number of patients that you can subcategorize in situ versus invasive cancer with several hundred patients in each group, you might be able to draw some distinctions. But certainly we could not do that in this series. Dr Harry M. Richter (Chicago, Illinois): If I heard you correctly, half a dozen patients on final pathology had a positive transection margin for carcinoma in situ. So the question is, did they fare differently than those who had a more benign or a totally normal transection margin? And did you offer any of them an immediate reoperation in view of your philosophy that it should be treated like an invasive tumor? Dr Mark S. Talamonti (Chicago, Illinois): Those are good questions. Thus far, none of those patients has developed either recurrent invasive cancer or metastatic disease. Many of those patients, though, were elderly, and the willingness to accept a positive margin with in situ, not on a frozen section but on the final one, is very much patient driven. If we have a patient who is well motivated, somebody that we think will accept the implications and the long-term sequelae of insulin and enzymes, then we would offer that patient a completion pancreatectomy. But again, in an elderly patient with a small focus of in situ carcinoma, I am not sure you need to offer that same patient the same option. Dr Syed Ahmad (Cincinnati, Ohio): One of the biggest questions in determining management is the time it takes to go from benign to malignant. And 1 way people have looked at that is looking at a median age of benign patients, invasive patients, and in situ patients. I was wondering whether you can give us some insight into your series and what the median age of patients with benign disease versus in situ versus invasive cancer was. Dr A. D. Yang (Chicago, Illinois): Again, I think a limitation of our study is that our numbers are small. We did look at benign versus malignant patients, but we did not find a significant difference between the ages, although there was a trend, I believe, toward invasive disease being older. Dr Mark S. Talamonti (Chicago, Illinois): This is a very select series of patients who went to the operating room. It does not represent every patient who has been seen with a radiographic finding suggestive of IPMN.
Surgery Volume 142, Number 4
So accepting the fact that this a selected group of patients who had suspicious disease either radiographically or with interventional gastroenterology and who, as pointed out, were symptomatic, it is a very selected group. And I do not know that you can really start to project long-term biologic transformation in a surgical series. That is a limitation of any surgical series that looks at these patients. Dr R. Matthew Walsh (Cleveland, Ohio): I am curious. You gave us nicely the final pathology on the transection margin. What actually were the frozen section diagnoses and how many pathologists do you have reading these frozen sections? How valuable is that frozen section? My last comment is this: The difficulty with the WHO classification is that it does just what you did, it separates benign from malignant and uses that word “malignant.” But I also am not sure that the in
Yang et al 537
situ group is really malignant in the sense of their long-term outcome. Dr Mark S. Talamonti (Chicago, Illinois): One of the major advantages we have is that we have a single pathologist who has looked at every one of these retrospectively. And then during the operation, except when he is out of town, he is the person who looks at these intraoperative frozen sections. The frozen sections did correlate in the sense that if a patient had a completely clear frozen section margin with no abnormal changes, that patient stayed that way on the final margin. Where there was change were in patients who had what they would call intestinal metaplasia or low-grade IPMN. Those patients were sometimes upgraded to a more aggressive borderline or in situ carcinoma. But there is great value in having internal consistency with pathology review, no question about that.
Purpose. To evaluate the experience with pancreatectomy for intraductal papillary mucinous neoplasm (IPMN) at a single academic institution. Methods. A prospective pancreatic database was reviewed and identified 43 patients with IPMN who were managed operatively. Clinicopathologic features and predictors of outcome were examined. The World Health Organization pathologic classification of IPMN was utilized. Results. IPMN was diagnosed in 21% of patients who underwent pancreatic resection for solid or cystic mass lesions. Ninety-five percent of patients were symptomatic. Patients were managed with total pancreatectomy, pancreaticoduodenectomy, distal pancreatectomy, central pancreatectomy, or enucleation. Nine patients had adenomas, 14 had borderline neoplasms, 10 had carcinoma in situ, and 9 had invasive carcinoma. Overall, 23 patients (53%) had lesions with main duct involvement. Frozen section transection margins were positive for malignancy in 2 patients. With a mean follow-up of 17 months, the 5-year disease-specific survival for patients with main duct involvement was 67%. The 5-year disease-specific survival for patients with benign lesions was 100%, and 61% for patients with malignant lesions (P ⫽ .02). The presence of symptoms, increased CA 19-9, and tumor size were not predictive of malignancy. Increased serum bilirubin concentrations were predictive of malignancy (P ⫽ .03). Main duct involvement was also associated with malignancy (P ⬍ .02). Conclusions. Cancer is found in 65% of patients with IMPN involving the main duct. Based on our data, patients with symptomatic, main duct involvement, especially those with an increased serum bilirubin, should be offered resection. Alternatively, patients with side branch IPMN may be managed conservatively. (Surgery 2007;142:529-37.) From the Department of Surgery, Feinberg School of Medicine, Northwestern University, Chicago, Ill
Intraductal papillary mucinous neoplasms of the pancreas (IPMN) represent a recently described pancreatic lesion, the natural history of which remains incompletely understood. The World Health Organization (WHO) has defined IPMN as papillary mucin-producing neoplasms with tall, columnar, mucin-containing epithelium with or without papillary projections that arise in the main pancreatic duct or its major branches
Both A. D. Yang, MD, and L. G. Melstrom, MD, contributed equally to this manuscript. Accepted for publication July 21, 2007. Reprint requests: Mark S. Talamonti, MD, 201 E. Huron St. Galter 10-105, Chicago, IL 60611. E-mail: [email protected]. 0039-6060/$ - see front matter © 2007 Mosby, Inc. All rights reserved. doi:10.1016/j.surg.2007.07.007
(commonly referred to as “main duct IPMN” and “branch duct IPMN,” respectively).1 IPMNs lack the ovarian stroma characteristic of mucinous cystic neoplasms of the pancreas, which also do not exhibit involvement of the major pancreatic ducts. With the clarified definition of IPMN, as well as the widespread use of advanced imaging, the identification of IPMN found during investigation of related or unrelated medical conditions has increased dramatically in the last decade.2-7 IPMN now constitutes a substantial proportion of resected pancreatic lesions in the United States. Whether the diagnosis of IPMN mandates resection remains in question. Despite decreased morbidity and mortality in experienced hands, pancreatectomy remains a formidable operation that confers considerable risk for patients. As a consequence, it has become paramount to further examine predictors of outcome to enable appropriate selection of patients SURGERY 529
530 Yang et al
with IPMN for operative intervention. Recent reviews and consensus conferences have attempted to standardize the management of patients with IPMNs8-10; however, no single test exists that unequivocally differentiates between benign and malignant neoplasms. Therefore, controversy exists regarding predictors of malignancy in patients with IPMN who are managed by resection. The aims of this study were 3-fold: To determine the incidence of IPMN in our patients with pancreatic mass lesions managed by resection, to examine clinical and pathologic factors in this patient population, and to determine outcome measures and risk factors associated with malignancy to make recommendations regarding management for patients with IPMN. METHODS A prospective pancreatic database from Northwestern Memorial Hospital was reviewed to identify patients with IPMN who were managed operatively. All patients who underwent resection had pathologic evaluation by a single gastrointestinal pathologist (S.M.R.), and complete records were included in this study. To avoid overlap with mucinous cystic neoplasms, only patients with neoplasms of the main pancreatic duct or side branch involvement fulfilling the WHO classification1 were included. Forty-three patients who underwent surgical resection for IPMN between January 1997 and August 2006 were identified from the database. The complete medical record of each patient was examined for demographic and clinicopathologic features. Operative reports were reviewed for type of resection, completeness of resection, and findings of intraoperative frozen section biopsies. Pathology reports were reviewed to confirm the diagnosis and to obtain pathologic variables. These pathologic variables included tumor location and size, surgical margin status, and determination of histologic subtypes. Tumors were classified as either benign (adenoma and borderline tumors) or malignant (carcinoma in situ or invasive cancer) to compare demographic features and/or outcome measures that could predict the presence of malignancy in patients with IPMN. 2 analysis was used to identify factors predictive of malignancy and survival. Survival was estimated by means of the Kaplan-Meier method and compared using log rank.11 Statistical significance was defined as P ⬍ .05. RESULTS Demographics and presentation. IPMN was diagnosed in 21% (43/204) of patients who underwent pancreatic resection for solid or cystic masses of the
Surgery October 2007
pancreas between January 1997 and August 2006. The mean age of the cohort was 62 ⫾ 11 years. There were 28 men (65%) and 15 women (35%). Forty-two patients (98%) were white and 1 was black (2%). A history of smoking was present in 30 patients (70%), and 15 patients (35%) admitted to a history of significant alcohol consumption. The incidence of acute pancreatitis at presentation was 35% (15 patients), and 6 patients (14%) had a previously established diagnosis of chronic pancreatitis. Family history was relevant for some form of cancer in ⱖ1 family members in 29 patients (67%), whereas 10 patients (23%) had a history of a prior personal malignancy. Diabetes mellitus was present in 8 patients (19%). Symptoms were present in 41 of the 43 patients (95%). The majority of patients presented with abdominal pain (67%) and/or weight loss (60%). A minority of patients presented with nausea/vomiting (37%) and/or jaundice (23%). Random serum glucose concentrations of ⬎200 mg/dL were found in 7 patients (16%); 9 patients had random serum glucose concentrations of 100 to 200 mg/dL (21%). Serum total bilirubin concentration was increased in 7 patients (16%). Serum alkaline phosphatase was increased in 8 patients (19%) and serum CA 19-9 was increased in 6 patients (14%). Operative management. Preoperative decision making regarding choice of pancreaticoduodenectomy, distal pancreatectomy, or total pancreatectomy was predicated on a preoperative evaluation that included some combination of endoscopic retrograde cholangiopancreatography, endoscopic ultrasonography, and computed tomography or magnetic resonance imaging. Preoperative findings favoring pancreaticoduodenectomy were main duct involvement of primarily the head of the pancreas and sparing of the body and tail from involvement with IPMN. Distal pancreatectomy or extended distal pancreatectomy was favored in patients without any obvious main duct involvement in the head of the pancreas but involvement of disease to the left of the portal vein. Pancreaticoduodenectomy was the most common procedure, performed in 25 patients (58%), followed by distal pancreatectomy in 10 patients (23%), and total pancreatectomy in 6 (14%) patients. One patient (2.5%) underwent central pancreatectomy with pacreatostrostomy and 1 patient (2.5%) had “enucleation” of the IPMN. Review of pathology. The distribution of neoplasms by location was as follows: head/uncinate process (n ⫽ 25, 58%), neck/body (n ⫽ 5, 12%), tail (n ⫽ 7, 16%), and entire gland (n ⫽ 6, 14%). Thirty-five patients had discrete, measurable le-
Yang et al 531
Surgery Volume 142, Number 4 Table I. Pathologic characteristics (n ⫽ 43) Characteristic Tumor location Head/uncinate process Neck/body Tail Entire gland Duct involvement Main duct Branch duct Histologic class Adenoma Borderline Carcinoma in situ Invasive cancer Lymph node involvement Present Present with malignancy Final transection margin Positive for IPMN Positive for IPMN and malignancy
% (n) 58 (25) 12 (5) 16 (7) 14 (6) 53 (23) 47 (20) 23 (10) 33 (14) 23 (10) 21 (9) 17 (7) 21 (4) 35 (14) 14 (6)
IPMN, intraductal papillary mucinous neoplasm. Median tumor size is 25 mm (range, 3 to 100).
sions with a median tumor size of 25 mm (range, 3 to 100 mm). The remaining 8 patients had multifocal and/or multicentric disease in which no specific size was assigned to their neoplasm. Overall, 23 patients (53%) had lesions with main duct involvement (solely or combined main and sidebranch ducts). Malignant IPMN (carcinoma in situ/invasive cancer) was found in 44% of lesions. Histologic classification identified adenomas in 9 patients (21%), 14 (35%) with borderline neoplasms, 10 (23%) with carcinoma in situ, and 9 (21%) with invasive carcinoma. Of the patients with invasive carcinoma, 4 (21%) had lymph node involvement (Table I). The pancreatic transection margin was evaluated with intraoperative frozen section in 36 of 43 patients. Frozen section transection margins were positive for malignancy in 2 patients, 1 who underwent completion total pancreatectomy. The other patient underwent extended pancreaticoduodenectomy with a final frozen transection margin of IPMN with “dysplasia.” Final pathology of the transection margins showed IPMN involvement in 14 patients. Benign IPMN adenoma was observed at the transection margin in 3 patients, borderline or atypical IPMN in 5 patients, and carcinoma in situ in 6 patients, but invasive carcinoma in no patients. Outcome and predictors of malignant disease. With a mean follow-up of 17 months, the 5-year disease-specific survival for all patients was 76% (Figure) and 64% for patients with main duct in-
volvement. The 5-year disease-specific survival for patients with benign lesions was 100% and 49% for patients with malignant lesions (median survival not reached vs 30 months, P ⫽ .02). No demographic factor was associated with the presence of malignancy in the final resected specimens. The presence of symptoms, increased serum CA 19-9, and lesion size were not predictive of malignancy; in contrast, increased serum bilirubin level was predictive of malignancy (P ⫽ .03). Main duct involvement was associated with malignancy (P ⬍ .02). Malignant lesions were found in 15 of 23 patients with main duct involvement. Of the remaining 20 patients with only branch duct involvement, no patients had invasive cancer and only 4 patients had carcinoma in situ (Table II). DISCUSSION Despite the multitude of recent reports addressing the diagnosis and management of IPMN of the pancreas, it is difficult to assess accurately the true incidence and medical impact of these relatively rare neoplasms. With increased utilization of imaging technology, there has been an increased detection of asymptomatic cystic lesions of the pancreas. Not surprisingly, there has also been an increase in the number of operations for pancreatic cystic lesions, many of which are IPMNs. Notwithstanding this increase in pancreatic resections and subsequent categorization of certain pancreatic tumors as IPMNs, there remains an imperative to increase our understanding of the biologic behavior and the spectrum of diseases represented by these neoplasms. Experience with IPMN continues to be limited to retrospective reports from high-volume centers, and many attempts have been made to identify prognostic factors that might influence the management of these patients. Even as larger series and longer follow-up times evolve, questions remain as to which factors are predictive of outcome. Therefore, we sought to determine the incidence of IPMN in our patients with pancreatic mass lesions managed by resection; to examine clinical and pathologic factors in this patient population; and to determine outcome measures and identify factors associated with malignancy to make recommendations regarding management of patients with IPMN. In some centers, the proportion of pancreatic resections for cystic neoplasms of the pancreas has doubled within the last 10 years. Raut et al4 found the incidence of IPMN among the cystic lesions of pancreas treated by resection was 26% (35/137). We have found a much higher proportion of IPMN
532 Yang et al
Surgery October 2007
Figure. Kaplan-Meier actuarial survival curves for patients with IPMN who were operatively managed at a mean follow-up of 17 months. The overall 5-year disease-specific survival for all patients was 76% (upper graph). The 5-year disease-specific survival was significantly decreased in patients with malignant lesions (49%) compared with patients with benign lesions (100%; P ⫽ .02).
in this group (51%, 43/85), which constituted ⬎20% of our pancreatic resections during the study period. The demographic characteristics of our patients were comparable to those found in most series.2-7,12 Similar to the series reported by D’Angelica et al,3
from Memorial Sloan-Kettering Cancer Center, we found that 23% of our patients had a history of a prior extrapancreatic malignancy. Fernandez-del Castillo et al13 found that 27% of IPMNs were discovered incidentally, and the rest had some complaint referable to the pancreas. Similarly,
Yang et al 533
Surgery Volume 142, Number 4
Table II. Predictors of malignant disease for patients with IPMN
Tumor size (mm) ⬍20 ⬎20 Presence of symptoms Increased total bilirubin Increased CA 19-9 Duct involvement Main duct involvement (Main/Combined) Branch duct IPMN
Benign IPMN, % (n)
Malignant IPMN, % (n)
P
62 (15) 38 (9) 96 (23) 10 (2) 18 (4)
42 (8) 58 (11) 95 (18) 27 (5) 10 (2)
NS NS NS .03 NS
35 (8)
65 (15)
⬍.02
80 (16)
20 (4)
⬍.02
IPMN, intraductal papillary mucinous neoplasm.
we found the majority of patients with IPMNs at our institution presented with symptoms. A majority of our patients underwent pancreaticoduodenectomy. Extent of resection was similar to that of other published studies.3,6 Our proportion of patients with benign disease (adenoma or borderline IPMN) compared with patients with malignant disease (carcinoma in situ or invasive IPMN) reflects current management practices in large referral centers. Although early reports showed increased rates of invasive disease (37% to 48%)3,4,6 with stricter criteria for operative intervention, more liberal selection criteria are now being applied with a resulting lower proportion of invasive disease (21%). Our survival estimates differ somewhat from most published reports because we chose to classify our patients as either benign or malignant, whereas many case series classify their patients as either noninvasive (adenoma, borderline, carcinoma in situ) or invasive. We chose to divide our patients into benign and malignant categories because we agree with the theory put forward by Salvia et al,5 namely, that IPMN progresses from benign disease to eventual invasive cancer. Thus, we conclude that carcinoma in situ is the immediate precursor to invasive cancer and should be considered as malignant disease. This approach explains the increased survival in our benign and malignant IPMN groups, because patients with carcinoma in situ have shifted from the benign to the malignant category. Intraoperative frozen section is not utilized uniformly at other high-volume centers.3,4,6 The decision to perform intraoperative frozen section analysis of the margin at our center was at the operating surgeon’s discretion during the time frame of this study. Currently, we perform intraop-
erative frozen section analysis of the margin routinely with the purpose of determining whether carcinoma in situ or invasive cancer is present at the final transection margin. The influence of invasive or noninvasive disease at the margin of transection on survival is controversial and pertinent to multifocal disease. Although we were unable to assess the association of margin status with malignancy, recurrence, or survival, data from Raut et al4 and others3 suggest that the presence of noninvasive IPMN (adenoma, borderline, or carcinoma in situ) does not affect recurrence and/or survival adversely. Data from White et al14 suggest that, although negative margins should be the operative goal, when achievable with partial pancreatectomy in patients with noninvasive IPMN, the risk of local recurrence is not high enough to mandate total pancreatectomy for all patients with positive margins. As a result, what remains controversial is the importance of the frozen section margin versus the gross character of the duct proximal to the resection margin. Although this controversy is not addressed specifically by our data, we believe that if intraoperative frozen section analysis of the resection margin has any role, it is likely in the identification of remaining invasive cancer in the remnant pancreas. Other investigators have also an increased serum bilirubin concentration to be the only laboratory value predictive of outcome.3,4,6,7,9,12 The presence of invasive disease has correlated consistently with poor outcome. The pathologic factors that correlate with the presence of malignancy are less consistent. D’Angelica et al,3 Sohn et al,6 and others5 have shown main duct involvement to be associated with malignancy. Our results confirm this as a pathologic variable predictive of malignancy. The data in our series regarding relation of tumor size to malignant potential are similar to those published by Fernandez-del Castillo et al,13 which show a large percentage of patients (39%) with malignant potential in symptomatic lesions ⬍2 cm. We found a 42% incidence of malignant IPMN in lesions that measured ⬍2 cm. Thus, small, main duct IPMN may harbor an invasive cancer, whereas a larger, branch duct lesion may contain only benign IPMN. We conclude that tumor size should not be an independent factor in selecting patients with IPMN who require operative intervention. Cancer was found in 65% of patients with IPMN involving the main duct. We did not find invasive cancer in any patients without main duct involvement. Based on our data, patients with symptomatic, main duct involvement, especially those with an increased serum bilirubin concentration, should
534 Yang et al
be offered resection. Alternatively, patients with side branch IPMN may be managed conservatively. Drs Yang and Melstrom contributed equally to the writing of this manuscript. The authors thank Ms. Jane Stocker, LPN, for her invaluable contribution to this paper through her management of the pancreatic resection database. REFERENCES 1. Longnecker DS, Adler G, Hruban RH, Kloppel G. Intraductal papillary-mucinous neoplasms of the pancreas. In: Hamilton SR, Aaltonen LA, eds. World Health Organization Classification of Tumours: Tumours of the Digestive System. Lyon, France: IARC Press; 2000:237-240. 2. Chari ST, Yadav D, Smyrk TC, et al. Study of recurrence after surgical resection of intraductal papillary mucinous neoplasm of the pancreas. Gastroenterology 2002;123: 1500-7. 3. D’Angelica M, Brennan MF, Suriawinata AA, Klimstra D, Conlon KC. Intraductal papillary mucinous neoplasms of the pancreas: an analysis of clinicopathologic features and outcome. Ann Surg 2004;239:400-8. 4. Raut CP, Cleary KR, Staerkel GA, et al. Intraductal papillary mucinous neoplasms of the pancreas: effect of invasion and pancreatic margin status on recurrence and survival. Ann Surg Oncol 2006;13:582-94. 5. Salvia R, Fernández-del Castillo C, Bassi C, et al. Main-duct intraductal papillary mucinous neoplasms of the pancreas: clinical predictors of malignancy and long-term survival following resection. Ann Surg 2004;239:678-85. 6. Sohn TA, Yeo CJ, Cameron JL, et al. Intraductal papillary mucinous neoplasms of the pancreas: an updated experience. Ann Surg 2004;239:788-97. 7. Wada K, Kozarek RA, Traverso LW. Outcomes following resection of invasive and noninvasive intraductal papillary mucinous neoplasms of the pancreas. Am J Surg 2005; 189:632-6. 8. Doi R, Fujimoto K, Wada M, Imamura M. Surgical management of intraductal papillary mucinous tumor of the pancreas. Surgery 2002;132:80-5. 9. Salvia R, Bassi C, Falconi M, et al. Intraductal papillary mucinous tumors of the pancreas. Surgical treatment: at what point should we stop? JOP 2005;6:112-7. 10. Tanaka M, Chari S, Adsay V, et al. International consensus guidelines for management of intraductal papillary mucinous neoplasms and mucinous cystic neoplasms of the pancreas. Pancreatology 2006;6:17-32. 11. Kaplan E, Meier P. Non-parametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457-81. 12. Wiesenauer CA, Schmidt CM, Cummings OW, et al. Preoperative predictors of malignancy in pancreatic intraductal papillary mucinous neoplasms. Arch Surg 2003;138:610-7. 13. Fernández-del Castillo C, Targarona J, Thayer SP, Rattner DW, Brugge WR, Warshaw AL. Incidental pancreatic cysts: clinicopathologic characteristics and comparison with symptomatic patients. Arch Surg 2003;138:427-23. 14. White R, D’Angelica M, Katabi N, et al. Fate of the remnant pancreas after resection of noninvasive intraductal papillary mucinous neoplasm. J Am Coll Surg 2007;204:987-93.
DISCUSSION Dr Keith D. Lillemoe (Indianapolis, Indiana): Your report adds to the body of literature that has
Surgery October 2007
already been collected on this topic and continues to help us clarify this somewhat of a mystery of a disease for which we really do not know all the answers. Although you have answered some questions, you have raised some as well. So if you don’t mind, I would like to just ask a few. You describe in this series that 95% of your patients are symptomatic, which I think is a higher percentage than some of the other series. This may be sort of a self-fulfilling prophecy because these are patients who you chose to operate on. Therefore, I would like to ask you about your management of those patients who you choose not to operate on. How do you decide who to operate and who not to operate on? And specifically, because size did not seem to be a predictor of malignancy in your series and symptoms did not seem to be a predictor of malignancy, how do we deal with the asymptomatic patient who is found on a routine screening CT scan for some other purpose to have a 1.5-cm IPMN, keeping in mind that the consensus conference that has recently come out has suggested that perhaps observation for 3-cm IPMNs might be appropriate? This may have been a typo in the copy of the manuscript that I saw. You referred to a 65% incidence of a positive family history for pancreatic cancer in the manuscript. Here you just said history of cancer. Obviously, that is a big difference. And clearly, some further genetic workup would be necessary if you really had that high a correlation. Finally, you have great survival results on these patients, but you did not answer the question about whether the patients who survived, either with benign or malignant IPMNs, had any recurrent IPMN disease in the remnant gland? How do you follow your patients for surveillance for development of new tumors in the gland? Do you recommend that even people with benign disease be followed with magnetic resonance cholangiopancreatography (MRCP) or endoscopic ultrasound (EUS) to look for new tumors? Dr A. D. Yang (Chicago, Illinois): Thank you, Dr Lillemoe, for your comments and your excellent questions. As to your first question as to how we decide on who to operate or not, we use several criteria. I think one is that you have to consider the patient, their acceptable comorbidities, and the relative safety of the operation. If we have a patient with a 1.5-cm IPMN, size does not necessarily dictate what we are going to do in terms of operative management. If the patient has a main duct type IPMN or if they have an increased serum total bilirubin, that would make us more likely to operate. However, if it was a side
Surgery Volume 142, Number 4
branch IPMN and the patient was asymptomatic, I do not think size is much of a factor in the equation and we could possibly observe these patients. As to your comment on the 65% family history, that was a typo. It is that 65% had a family history of some malignancy. As to your final question on recurrent disease, we have patients with recurrence. We have 1 patient who had a metastatic recurrence and the rest have had local recurrences. None of the patients in our series as of yet had to be operated on for recurrence. How do we follow these patients? Regardless of whether they had malignant or benign IPMN, we follow them very closely after surgery. It is our policy to get an MRCP, which we believe is the best diagnostic study, on these patients every 6 months for the first 1 to 2 years. And when we are satisfied that either there is no evidence that the patient has possibly a cancer or malignant disease, then we go to about 1-year intervals in following these patients with imaging only. Dr E. Christopher Ellison (Columbus, Ohio): I think IPMN is a very interesting disease in that it has the potential of being multicentric, and, as Dr Lillemoe alluded to, this multicentricity may not be necessarily synchronous with the index lesion. A point in fact is that we have observed recurrences in the residual pancreas in 3 patients with malignant main duct IPMN occurring 7, 8, and 10 years after their index operation with negative margins. Indeed, 2 have undergone subsequent resection with documented recurrent disease. The third patient has declined reoperation at this point in time. So in this respect, we tend to follow these patients for a long period of time. And I wanted to know what the policies were at Northwestern in terms of following them carefully. How long do you follow them? Do you stop at 5 years? Or do you continue to follow them out for 10 years after their index operation? A second point is we have observed in some community settings that surgeons may attempt a distal pancreatectomy with splenic preservation for this disease when it involves the tail of the pancreas, and subsequently these patients have had recurrence in the splenic hilum. And I would like to know what your policies are regarding splenic preservation in the presence of IPMN involving the distal pancreas. Dr A. D. Yang (Chicago, Illinois): As to your first question on how long we follow these patients, basically there is no endpoint. We follow them, as stated, with scans every 6 months for the first 1 to 2 years and then follow them clinically on imaging
Yang et al 535
scans. There is no limit at this point. We do not follow them for just 5 years; we follow them perpetually. Your second question regarding distal pancreatectomy: We do not perform splenic preservation in any of our distal pancreatectomies, so that has not been a problem. Dr Christian M. Schmidt (Indianapolis, Indiana): I was intrigued by your finding that size did not matter in IPMNs and was curious to know a few things about that. First, how did you determine size of IPMNs, particularly in cases of multifocal disease? Was your tumor size just the branch-type lesions or main-type lesions and mixed-type lesions? How did you measure these? Second, size did not matter when you compared malignant with nonmalignant lesions. Did you look at adenoma versus borderline versus carcinoma in situ versus invasive? Was there a trend that size mattered as you went up in the pathologic grades? The Hopkins group recently published a paper suggesting it was only between carcinoma in situ and invasive that there was actually a statistically significant size difference. Finally, your whole paper is set up on malignant versus nonmalignant. Did you look at invasive versus noninvasive? And what predictive factors were present in the invasive tumors? Dr A. D. Yang (Chicago, Illinois): The first question was to the size criterion in patients with diffuse or multicentric disease. When we did the calculations for size as related to its predictive value in the malignancy, we excluded those patients who did not have a tumor that could be measured; that is, they had diffusely dilated duct with IPMN within the duct. So we took a cutoff—2 cm—to see if there was an association with malignancy in those patients with measurable tumors, and we did not find a correlation there. And we did look at a trend in size and did not see any between the subtypes that we could find in terms of correlation and size. I did not see that report on carcinoma in situ versus invasive. We did not look at that other than we did not see a trend there. Finally, our decision to classify these patients into malignant versus benign disease was based on our clinical practice at Northwestern. We treat patients with carcinoma in situ basically the same as patients with invasive cancer, especially in terms of operative therapy, so we chose to classify our data in this way. And in the literature, part of the controversy here is that there have been some groups that have classified the tumors as noninvasive versus invasive disease and other groups that have classified it as malignant versus benign disease. And
536 Yang et al
we chose to do it this way based on our clinical practice to correlate better clinically with what we do. Dr Roger G. Keith (Saskatoon, Saskatchewan, Canada): It is interesting over the history of time this disease has become a better and better and better cancer and surgery has become less and less and less invasive. You are suggesting that we ought to consider a total pancreatectomy for people with main duct disease, that we accept that invasive carcinoma and in situ carcinoma are the same, and that we accept that in your group one third of your main duct patients did not have cancer. Have you followed your in situ group to their mortality and do you have any absolute deaths in that group? I think we need to know as these reports come through whether or not in situ carcinoma is lethal. Dr A. D. Yang (Chicago, Illinois): I believe that there were no deaths in the carcinoma in situ group as of yet. Dr Mark S. Talamonti (Chicago, Illinois): I think a lot of the questions have to do with size and invasion versus in situ. And this is a relatively small series compared with other series, so I think to make definitive statements about the implications of size and outcome may be a bit of a limitation to the study. Suffice it to say that we did not find that size alone was a predictor of malignancy. That is different than saying whether or not size within any subgroup is a predictor of outcome. You could have an 8-cm cystic lesion in the head of the pancreas that was all benign IPMN and that may or may not predict whether there is invasion in it. Once you get invasion in it, does size matter if there is an invasive component to it? That is a different but more subtle question. And I do not think we can really address that in the limited number of patients we have. In terms of follow-up for carcinoma in situ and duration of follow-up, I think that point has been made well and emphasized. This is not the type of disease that has an endpoint to it. Just like a woman with carcinoma in situ of the breast would have a yearly mammogram for long-term follow-up, I think when you talk about patients who have a premalignant lesion or a predisposition to developing invasive cancer of the pancreas, those patients need to be followed accordingly. What the exact guidelines are for that are, I think, remain to be written. We follow patients yearly with MRCP of the pancreas indeterminately, as Dr Yang mentioned. We chose to include in situ carcinoma in our survival analysis mainly because we manage in situ
Surgery October 2007
carcinoma similarly to invasive carcinoma. But there is no question that if you put the in situ patients with the invasive patients, the outcomes for those patients will improve. Whether or not that is a statistically valid method to do that, I think if you had a large enough number of patients that you can subcategorize in situ versus invasive cancer with several hundred patients in each group, you might be able to draw some distinctions. But certainly we could not do that in this series. Dr Harry M. Richter (Chicago, Illinois): If I heard you correctly, half a dozen patients on final pathology had a positive transection margin for carcinoma in situ. So the question is, did they fare differently than those who had a more benign or a totally normal transection margin? And did you offer any of them an immediate reoperation in view of your philosophy that it should be treated like an invasive tumor? Dr Mark S. Talamonti (Chicago, Illinois): Those are good questions. Thus far, none of those patients has developed either recurrent invasive cancer or metastatic disease. Many of those patients, though, were elderly, and the willingness to accept a positive margin with in situ, not on a frozen section but on the final one, is very much patient driven. If we have a patient who is well motivated, somebody that we think will accept the implications and the long-term sequelae of insulin and enzymes, then we would offer that patient a completion pancreatectomy. But again, in an elderly patient with a small focus of in situ carcinoma, I am not sure you need to offer that same patient the same option. Dr Syed Ahmad (Cincinnati, Ohio): One of the biggest questions in determining management is the time it takes to go from benign to malignant. And 1 way people have looked at that is looking at a median age of benign patients, invasive patients, and in situ patients. I was wondering whether you can give us some insight into your series and what the median age of patients with benign disease versus in situ versus invasive cancer was. Dr A. D. Yang (Chicago, Illinois): Again, I think a limitation of our study is that our numbers are small. We did look at benign versus malignant patients, but we did not find a significant difference between the ages, although there was a trend, I believe, toward invasive disease being older. Dr Mark S. Talamonti (Chicago, Illinois): This is a very select series of patients who went to the operating room. It does not represent every patient who has been seen with a radiographic finding suggestive of IPMN.
Surgery Volume 142, Number 4
So accepting the fact that this a selected group of patients who had suspicious disease either radiographically or with interventional gastroenterology and who, as pointed out, were symptomatic, it is a very selected group. And I do not know that you can really start to project long-term biologic transformation in a surgical series. That is a limitation of any surgical series that looks at these patients. Dr R. Matthew Walsh (Cleveland, Ohio): I am curious. You gave us nicely the final pathology on the transection margin. What actually were the frozen section diagnoses and how many pathologists do you have reading these frozen sections? How valuable is that frozen section? My last comment is this: The difficulty with the WHO classification is that it does just what you did, it separates benign from malignant and uses that word “malignant.” But I also am not sure that the in
Yang et al 537
situ group is really malignant in the sense of their long-term outcome. Dr Mark S. Talamonti (Chicago, Illinois): One of the major advantages we have is that we have a single pathologist who has looked at every one of these retrospectively. And then during the operation, except when he is out of town, he is the person who looks at these intraoperative frozen sections. The frozen sections did correlate in the sense that if a patient had a completely clear frozen section margin with no abnormal changes, that patient stayed that way on the final margin. Where there was change were in patients who had what they would call intestinal metaplasia or low-grade IPMN. Those patients were sometimes upgraded to a more aggressive borderline or in situ carcinoma. But there is great value in having internal consistency with pathology review, no question about that.