Serous Cystadenoma of the Pancreas

Abstract

Objective:

To define the natural history and optimal management of serous cystadenoma of the pancreas.

Summary Background Data:

Serous cystadenoma of the pancreas is the most common benign pancreatic neoplasm. Diagnostic criteria, potential for growth or malignancy, and outcomes are not well defined. As a result, management for patients with serous cystadenomas varies widely in current practice.

Methods:

A total of 106 patients presenting with serous cystadenoma of the pancreas from 1976–2004 were identified. Hospital records were evaluated for patient and tumor characteristics, diagnostic workup, treatment, and outcome. Twenty-four patients with serial radiographic imaging were identified, and tumor growth curves calculated.

Results:

Mean age at presentation was 61.5 years and 75% of patients were female. The most common symptoms were abdominal pain (25%), fullness/mass (10%), and jaundice (7%); 47% were asymptomatic. Mean tumor diameter was 4.9 ± 3.1 cm, which did not vary by location. Tumors <4 cm were less likely to be symptomatic than were tumors ≥4 cm (22% vs. 72%, P < 0.001). The median growth rate in the patients who had serial radiography was 0.60 cm/y. For tumors <4 cm at presentation (n = 15), the rate was 0.12 cm/y, whereas for tumors ≥4 cm (n = 9), the rate was 1.98 cm/y (P = 0.0002). Overall, 86 patients underwent surgery, with one perioperative death.

Conclusions:

Large (>4 cm) serous cystadenomas are more likely to be symptomatic. Although the median growth rate for this neoplasm is only 0.6 cm/y, it is significantly greater in large tumors. Whereas expectant management is reasonable in small asymptomatic tumors, we recommend resection for large serous cystadenomas regardless of the presence or absence of symptoms.

The diagnostic criteria, natural history, and optimal treatment of patients with pancreatic serous cystadenomas remain incompletely delineated. We describe a large, single-center experience with this neoplasm and give recommendations for treatment.

Cystic neoplasms of the pancreas are increasingly diagnosed in the modern era. More frequent use of radiography and advances in imaging techniques have led to larger numbers of cystic lesions being identified.^1–4 For decades, it has been clear that correctly differentiating between a cystic neoplasm and a pseudocyst is vital to avoid incorrect treatment.⁵ More recently, as the divergent natural histories and malignant potentials of different cystic neoplasms have been elucidated, differentiating between mucinous cystic neoplasms (MCNs), intraductal papillary mucinous neoplasms (IPMNs), serous cystadenomas, and other less common tumors has become increasingly important.^6–9

As opposed to pancreatic pseudocysts, serous and mucinous cystic tumors have an epithelial lining. The epithelium of IPMNs and MCNs is made up of columnar mucin-producing epithelium. However, MCNs, which occur almost exclusively in women, are devoid of communication with the ductal system and are supported by ovarian-type stroma, whereas IPMNs arise in the main pancreatic duct or its major branches. In contrast, serous cystic tumors are lined by an inconspicuous single layer of either cuboidal or flattened cells. The cytoplasm of the cells is either clear or eosinophilic and the nuclei are usually centrally located, small, and hyperchromatic. Mitoses are conspicuously absent.

Most serous cystadenomas are microcystic, forming an honeycomb-like appearance, but macrocystic variants have been described.^10–12 The overwhelming majority of these tumors are benign, with only a handful of case reports of serous cystadenocarcinomas.^13–18 Resection is generally carried out for symptoms, large size, or the inability to distinguish a serous cystic neoplasm from a mucinous lesion, which has greater malignant potential. Some authors have recommended resection for all cystic neoplasms of the pancreas,^14,15,19 whereas others (including our group) advocate a more selective approach.^4,20

Our study had three specific aims: to describe a large single-center experience with serous cystadenoma of the pancreas, including presenting signs and symptoms and methods of diagnosis; to analyze serial radiography in available patients to determine the rate of growth; and to make recommendations for the accurate diagnosis and optimal treatment of patients with this neoplasm.

MATERIALS AND METHODS

The Massachusetts General Hospital Institutional Review Board granted protocol approval for retrospective chart review. Patients with the diagnosis of serous cystadenoma of the pancreas were identified by reviewing hospital admission data and outpatient visit logs. In all, 106 patients diagnosed with serous cystadenoma were identified. The inpatient, outpatient, and computerized medical records were reviewed in detail. Information regarding demographics, presentation, comorbid conditions, radiologic and endoscopic evaluation, follow-up after diagnosis, surgical treatment (if any), pathology, postoperative recovery, morbidity, mortality, and length of stay were obtained.

Tumor size was calculated from radiographic images. Maximum diameter was available in all patients and hence was used as the primary metric for tumor size. In addition, tumor volume was estimated were made using the formula π/6 × (d1 × d2 × d3).²¹ If fewer than three diameter measurements were available, the formula was modified accordingly: π/6 × (d1² × d2) or π/6 × (d³). Twenty-four patients were identified who had serial radiographic exams. Tumor growth over time was calculated by plotting the maximum tumor diameter at each time point. A regression line using the least squares method was fitted to the data points. Tumor growth was then derived from the slope of the regression equation. To estimate tumor volume growth, doubling times (DT) were calculated using the Schwartz equation: DT = [(ln2)(t2−t1)]/ln(v2/v1), where t is time and v is volume.^22,23

Normally distributed continuous variables were compared using a two-sample Student t test; nonnormally distributed continuous variables were compared using the Mann-Whitney U test. Categorical variables were compared using a χ² test.

RESULTS

Demographics

Patient characteristics are presented in Table 1. Mean patient age was 61.5 ± 13.1 years and 75% of the patients were women. Women were significantly younger at presentation than were men (59.9 vs. 67.2 years, P = 0.018). Overall, 90% of patients were non-Hispanic whites.

TABLE 1. Patient Characteristics

Presentation

The most common symptoms were abdominal pain (25%), mass or fullness (10%), jaundice (7%), and fatigue and/or malaise (6%). One patient presented with pancreatitis. In all, 47% of patients were asymptomatic and the tumor was identified as part of the workup of a different problem. There was no significant difference in the proportion of males and females that were asymptomatic. Incidentally discovered serous cystadenomas accounted for 53% of all tumors seen from 1995 to 2004, whereas they accounted for only 34% of all tumors seen prior to 1995 (P = 0.074). Thirteen patients (12%) presented in the context of other neoplasms, including three patients with colon adenocarcinoma, two patients with von-Hippel-Lindau disease, and one patient with Zollinger-Ellison syndrome. Of these 13 patients with other malignancies, the pancreatic cystadenoma was asymptomatic in nine (69%) patients.

Tumors

Forty-seven patients (44%) had tumors in the head, neck, or uncinate process of the pancreas, whereas 59 (56%) had tumors in the body or tail of the pancreas. The prevalence of symptoms did not vary significantly by tumor location (any symptoms: head/neck 58% vs. body/tail 48%, P = 0.42). Mean tumor diameter was 4.9 ± 3.1 cm, and similarly did not vary by location (head/neck 5.4 ± 2.8 vs. body/tail 4.5 ± 3.4; P = 0.20). Women had significantly smaller tumors than men (mean 4.5 ± 2.9 vs. 6.3 ± 3.7, P = 0.02). The 46 patients with tumors <4 cm were less likely to be symptomatic than were the 60 patients with tumors ≥4 cm (22% vs. 72%, respectively; P < 0.001).

Endoscopic Ultrasound and/or CT-Guided Fine Needle Aspiration

Endoscopic ultrasound (EUS) or computed tomography (CT)-guided needle aspiration was used in 36 patients. These diagnostic tests were used more frequently in the more recently diagnosed patients (21/39 patients evaluated since the year 2000). In 10 of the 21 patients, EUS or CT-guided FNA was considered diagnostic of serous cystadenoma. In the other 11 patients, specimens were considered inadequate or a diagnosis other then serous cystadenoma was suggested by EUS or CT-FNA, and serous cystadenoma was identified only with operative pathology.

Operations and Perioperative Course

In all, 86 patients underwent surgery, including 36 Whipple procedures, 35 distal pancreatectomies (14 spleen-preserving), nine middle pancreatectomies, and one total pancreatectomy. Figure 1 shows a serous cystadenoma involving the entire pancreas treated with a total pancreatectomy. Two patients underwent enucleation and three underwent exploratory laparotomy with biopsy but no resection. Thirteen patients (15%) had major perioperative complications (Table 1), and there was one perioperative death from pulmonary embolism. Mean length of stay prior to 1998 was 12.7 days (median 9 days), which decreased to 7.3 days (median 7 days) in our more recent experience during 1998–2004 (P = 0.009).

FIGURE 1. Serous cystadenoma involving the entire pancreas. A, CT scan. B, Cut specimen. The patient was 64-years-old and had postprandial fullness and a palpable mass.

Surgical Pathology

All 86 patients who underwent operation had a final pathologic diagnosis of serous cystadenoma. Of these, six patients (7%) were noted to have macrocystic or oligocystic variants of serous cystadenoma. One patient had a simultaneous nonfunctioning neuroendocrine tumor of the pancreas diagnosed on final pathology. There were no cystadenocarcinomas diagnosed by histologic criteria. One patient had “invasion” of the duodenum noted on initial pathology, with benign cellular histology. After secondary review (G.Y.L), the tumor pathology was considered “pushing” rather than invasive, and thus benign.

Nonoperative Patients

Twenty patients were diagnosed with serous cystadenoma and have not undergone operation. All of them were asymptomatic and had CT scans consistent with serous cystadenoma (n = 14) or small indeterminate cystic lesions of the pancreas (n = 6). EUS was performed in 11 of these 20 patients, and gave additional morphologic, cyst fluid, or cytologic information supporting the diagnosis of serous cystadenoma. Five other patients underwent CT-guided biopsy for this same purpose. Four of these nonoperated patients had tumors greater than 4 cm in maximum diameter; one was 80 years of age at presentation and had major medical comorbidities; the second had moderate comorbidities and was 84 years of age; the third had an unresectable hilar cholangiocarcinoma; and the fourth had extensive sequelae of Marfan syndrome.

Tumor Growth

Serial radiography was analyzed for 24 patients: 14 patients had two studies, six patients had three studies, and four patients had more than three studies. The interval between first and last study ranged from 3 to 162 months, with a median of 23.3 months. Growth curves are displayed in Figure 2. Using measurements of maximum diameter, the median growth rate was 0.60 cm/y. There was a significant difference in tumor growth rate depending on the size at first presentation. For tumors <4 cm at presentation (n = 15), the rate was 0.12 cm/y; the upper 75th percentile for tumors <4 cm was 0.48 cm/y. For tumors ≥4 cm (n = 9), the rate was 1.98 cm/y (P = 0.0002) (Fig. 3). Similarly, when examining doubling times for estimated tumor volume, tumors <4 cm in maximum diameter (volume <33.3 cm³) had a doubling time of 2.84 years, whereas tumors ≥4 cm in maximum diameter had a doubling time of 0.64 years (P = 0.017). Figure 4 shows a large serous cystadenoma in which growth was documented over a 6-month period.

FIGURE 2. Serous cystadenoma tumor growth over time. Open circles represent tumors <4 cm at presentation, solid squares represent tumors ≥4 cm at presentation.

FIGURE 3. Comparison of tumor growth for serous cystadenomas <4 and ≥4 cm in maximum diameter.

FIGURE 4. Serial radiographs in an 81-year-old woman with an incidentally discovered serous cystadenoma of the head of the pancreas. The tumor measured 5 × 5.5 cm in September 2003. By April 2004 the patient was complaining of dull epigastric pain, and the tumor measured 5.5 × 6.1 cm.

DISCUSSION

Over 25 years ago, in what is now a landmark publication, Compagno and Oertel described 34 cases of serous cystadenoma of the pancreas⁶ and established the differences between this benign lesion and the mucinous cystic neoplasm with malignant potential. In the Compagno and Oertel series, mean tumor diameter was 10.8 cm and 29% of patients were asymptomatic. In 1992, Pyke et al¹⁹ found a similar percentage (33%) of asymptomatic patients.

The present, more contemporary, series describes the largest single-center experience with pancreatic serous cystadenoma. Our work confirms the benign nature of this tumor and its preponderance in women in the seventh decade of life. Our data also show that the mean age of men with serous cystadenoma is more than 7 years older than that of women. The reasons for this difference are not immediately evident, although the larger size in males (6.3 vs. 4.5 cm) suggests a delay in diagnosis. Our series differs from those of Compagno and Pyke in the proportion of asymptomatic cases, which we found to be 47%. This difference is likely a reflection of increased use of cross-sectional imaging, and explains why our mean tumor diameter was almost 6 cm less than in Compagno and Oertel’s series (4.9 cm vs. 10.8 cm).

This high rate of incidental detection makes management challenging. First, accurate differential diagnosis is crucial to rule out mucin-producing cystic lesions and other pancreatic cystic neoplasms that have malignant potential. Second, even when the diagnosis of serous cystadenoma is firm, up to this point there have been no data that allow one to predict if an asymptomatic tumor will grow sufficiently to cause symptoms during the lifespan of a given patient. This is an important issue, because even though the mortality of pancreatic resection has decreased markedly in experienced hands, the morbidity remains high and the consequences of loss of pancreatic tissue are serious.

Radiographic imaging is a potent tool with which to diagnose serous cystadenoma of the pancreas, but limitations exist. The most widely applicable radiographic test at the current time is helical CT scanning with thin cuts through the pancreas, which often can provide assistance in the differentiation between serous and mucinous neoplasms. Classic CT findings suggestive of serous cystadenoma include a central scar with the “honeycomb” appearance of microcysts, found in the more common microcystic variant of serous cystadenoma. However, the rarer oligocystic or macrocystic variants, which occurred in 7% of our series, may be more difficult to differentiate from mucinous tumors based on CT findings.10–12,20,24 Other modalities such as magnetic resonance imaging and magnetic resonance cholangiopancreatography may be more useful in differentiating mucinous tumors such as IPMT from serous cystadenoma.² In blinded studies, the ability of radiologists to accurately distinguish serous neoplasms has ranged from 23% to 82%, although component cysts smaller than 2 cm have been found to be significantly associated with serous tumors, and peripheral tumor calcification has been found to be significantly associated with mucinous tumors.^25–28 In the future, newer techniques including F-18-fluorodeoxyglucose positron emission tomography (FDG-PET) may help distinguish benign and malignant pancreatic cystic lesions.^28a

EUS has been proposed as an ideal imaging technique for pancreatic cystic lesions.^29–31 Ultrasound can readily characterize cysts and high-resolution imaging of the pancreas can be achieved through endoscopic means. Needle aspiration of pancreatic cystic lesions can be used to obtain fluid for cytology, and cyst fluid tumor markers can be used for diagnostic purposes. Cyst fluid carcinoembryonic antigen (CEA) values are universally low in serous cystadenomas, trend higher in mucinous lesions, and are generally even more elevated in mucinous cystadenocarcinomas.^31,32 Although cytologic samples diagnostic of serous cystadenoma are obtained in less than 50% of cases, when such samples are positive, the specificity is high. Clinical acumen and radiologic testing can often be used to differentiate cystic neoplasms from pseudocysts. However, when the operative or nonoperative plan hinges upon differentiating a serous from a mucinous cystic neoplasm, cyst fluid analysis via endoscopic ultrasound or CT-guided aspiration and biopsy is particularly useful.³³

In the past, serous cystadenomas have been treated in divergent manners, depending on patient characteristics, surgeon preferences, and local practice patterns. Much of this variability can be attributed to lack of information regarding the natural history of these neoplasms. To our knowledge, this is the first study that attempts to define the growth potential of these neoplasms. By reviewing serial radiographs in 24 patients with a median interval of almost 2 years between first and last study, we have plotted growth curves and found a median increase in size of 0.6 cm/y, and a marked difference in the growth of tumors that were 4 cm or larger at presentation (almost 2 cm/y) compared with the smaller lesions (0.12 cm/y). The reasons why larger tumors appear to have a faster rate of growth than smaller tumors remain unclear. Any assessment of tumor growth is limited by the validity of the means or mathematical model used to calculate tumor size. When two-dimensional radiologic images are assessed, maximal tumor diameters are commonly used to assess tumor size. In other situations (for example, in experimental animal tumor models) more complex formulae that approximate tumor volume are often used.^21,34–38 Our calculated tumor doubling times based on volume estimates seem to corroborate the observation that larger tumors grow faster than smaller ones, and correlate with the analogous diameter-based curves, suggesting that tumor diameters, although imperfect, may be used as rough estimates of tumor size and growth. Furthermore, given the cystic natures of these tumors, tumor volumes and doubling time may have less applicability in serous cystadenomas than in relatively solid tumors such as adenocarcinomas. Whether serous cystadenomas that present at and grow to larger dimensions differ biologically, and perhaps bear a greater propensity for malignant degeneration,¹⁴ compared with their smaller counterparts remains an open question.

Based on these findings, we propose the following guidelines. Patients should be diagnosed with serous cystadenoma of the pancreas based upon a compatible clinical presentation and characteristic radiographic evidence. When this concordance is not present, endoscopic ultrasound should be used. Patients with the above criteria, who are asymptomatic and have tumors less than 4 cm in maximal diameter, are candidates for nonoperative management with clinical follow-up and serial imaging (Fig. 5). The interval between serial imaging is unclear, but based on the indolent growth of small tumors observed in this study, 2 years may be reasonable. Patients with symptoms attributable to their tumors, patients in whom a mucinous or other potentially malignant tumor cannot be comfortably excluded, and patients with serous cystadenomas measuring 4 cm or more in maximal diameter who are reasonable surgical candidates should be offered resection. This recommendation to proceed with surgery in asymptomatic patients with larger tumors is based both in their more rapid growth rate as well as in a more than threefold increase in the likelihood of developing symptoms.

FIGURE 5. Diagnostic and management algorithm for suspected serous cystadenoma of the pancreas.

ACKNOWLEDGMENTS

The authors are grateful for the contributions of Marc S. Sabatine, MD, MPH for statistical analysis.

Discussions

Dr. Keith D. Lillemoe (Indianapolis, Indiana): As the authors have noted, this is the largest reported series of these tumors, and certainly one of the few to provide us longitudinal follow-up. Cystic neoplasms are a problem of increased concern for pancreatic surgeons primarily as we are seeing more of these with the widespread use of abdominal imaging. Many of these are being detected in asymptomatic patients who are in their sixth, seventh, or eighth decade, and the decision-making associated with these tumors can be difficult. There certainly must be careful thought given to performing a major pancreatic resection, particularly a pancreaticoduodenectomy for benign asymptomatic incidentalomas. However, the fear of leaving behind a potentially curable malignant or pre-malignant lesion pushes us to be aggressive in many cases. I have a few questions.

First, these tumors usually reside in otherwise normal glands with soft texture and nondilated pancreatic and bile ducts. This generally ups the ante associated with surgical procedures, at least the leak rate associated with pancreaticoduodenectomy. Your overall perioperative mortality of less than 2% and your morbidity of 15% is fabulous and your pancreatic leak rate of 6% is also remarkable. Do you approach these resections differently than you would for resections for malignant disease?

I also note that you performed the classic Whipple procedure in the majority of patients rather than the pylorus-preserving modification. Is this simply surgeon or institutional bias or did the size of these lesions make this necessary?

Considering minimally invasive operations for pancreatic tumors, I note that about 10% of your patients had the midsegment pancreatectomy that Dr. Warshaw has popularized but only 2 patients had tumor enucleation. Again, is this an institutional bias against enucleation, or did you find that the resection of these tumors by enucleation was just not an attractive option? Have you performed any laparoscopic pancreatic resections for these tumors?

Finally, in your algorithm you describe a limited role for endoscopic ultrasound and fine needle aspiration. Despite your institution’s interest in this procedure, you used it in only a minority of patients. Do you really feel comfortable inrecommending observation only without obtaining more information such as cytology – and tumor markers even in the small tumors which lack the classic CT findings?

Dr. Jennifer F. Tseng (Houston, Texas): You are correct, of course, that with a normal pancreas we have more of a risk of leak. Our 5% reported leak rate in our paper is low. And we have looked back through the medical records to ascertain the correctness. We do not do anything differently per se. The surgeons look at the pancreas intraoperatively and make a decision as to the type of anastomosis based on the size of the duct and the pancreatic parenchyma. Of late, our surgeons have been using maneuvers such as using the falciform ligament and using tisseal to decrease further our leak rate.

In terms of pylorus-preserving pancreaticoduodenectomy, I do think there are institutional practices, at least, if not biases. We have found that our patients with standard Whipples have done very well from all perspectives and that we have had a rate of delayed gastric emptying such that we have not been pleased with with pylorus-preserving pancreaticoduodenectomies, so standard Whipple tends to be our institutional practice.

With regard to enucleation, our tumor size was large, almost 5 centimeters. And although this is lower than the 10.8 centimeters originally reported by Compagno and Oertel it is still fairly large, and in most of these cases the tumors could not be removed by enucleation. Also, of our 86 patients, the preoperative diagnosis was called serous or benign in only 20 of those patients and the rest is in cystic tumor NOS. We have not alluded to, for lack of time, the whole spectrum of cystic neoplasms. One of the major issues of course is differentiating a benign from a potentially malignant lesion. So we tend to err on the side of curative operation.

We have not performed any laparoscopic pancreatectomies for these tumors, although diagnostic laparoscopy is increasingly employed. We will see in the future.

In order to safely observe a patient with suspected serous cystadenoma, we must be sufficiently confident in the diagnosis. Of the 20 patients that are being managed nonoperatively in our series, 16 of those patients underwent either endoscopic ultrasound or CT/FNA, 11 of them underwent endoscopic ultrasound FNA, and 5 of them underwent CT/FNA. So it is true in the patients that we observed without operation we have a very high threshold for obtaining additional diagnostic information other than CT in all but the most classic radiographic cases.

Dr. Henry A. Pitt (Indianapolis, Indiana): This paper adds to the growing literature documenting that these tumors grow with time. Last year we reported that in 79 patients who were observed, 19% of the cysts increased in size over a 16-month period. These patients had the spectrum of cystic tumors. The group from Mayo Clinic also published this year a series of patients with an 8-year follow-up with over 40% of the cystic tumors increasing in size. My first question, therefore, is whether you have any data with respect to growth rate of other cystic tumors, the mucin-associated neoplasms or the IPMNs?

The next question I have relates to the recommendation about observation. My question is whether observation is cost effective in most patients. Let’s take, for example, the 50-year-old lady with a 2-centimeter lesion in the body of her pancreas. If you could do a laparoscopic enucleation or a distal pancreatectomy, the cost might be $10,000 or $15,000. If you observe her with serial scans and, perhaps, EUSs, when she is 60, she has a 3-centimeter lesion that is asymptomatic. When she is 70, she has a 4-centimeter lesion that is asymptomatic; however, by your data, when she is 75, she has a 14-centimeter lesion that is symptomatic. Now she requires a distal pancreatectomy with a splenectomy whereas you could have preserved the spleen when she was 50. Now, the cost of following this lesion for 25 years was at least $40,000 or $50,000. Thus, is it really cost effective to observe these patients even with the small benign lesions?

Finally, the other dilemma is that advanced age makes it more likely that these lesions are malignant. So even when you have a paatient over 70 with a small lesion, there is about a 60% chance that it is a malignant cystadenocarcinoma or IPMN. So, is it wise to observe these older patients who are healthy?

Dr. Jennifer F. Tseng (Houston, Texas): Regarding the growth rate of other cystic neoplasms, at the MGH we have a large cohort of patients with cystic tumors and we hope to be able to report back to you about the growth of other tumors.

In terms of the financial costs of observation over time – obviously I think in the future we need to design instruments that way where we can make decisions based on costs, but also more importantly based on patient morbidity and mortality. Because even though we think that a pancreatectomy is a relatively safe operation, there is still considerable morbidity and mortality. We had a 15% perioperative morbidity rate, which is comparable to other major centers, so sadly these are not completely safe operations. Certainly we would like to be able to have the data to be able to make a cost-benefit and risk-benefit analysis for these patients. Clearly we need to look at the observed patient over time and reconsider that patient’s risk/benefit ratio for resection.

Dr. Charles E. Lucas (Detroit, Michigan): I have one question regarding the giant cyst. Your data show that the chances of cancer are no more than 1% for serous cysts and the reasons for operation are for palliation of pain. I wonder, therefore, have you used a partial cystectomy combined with a Roux-en-Y pancreaticojejunostomy for palliation? I have done this in 1 patient who was palliated for about 5 years. It grew back and I did it again at 5 years and repalliated him. What is the role of simple drainage Roux for these large giant cysts which would otherwise require a total pancreatectomy?

Dr. Jennifer F. Tseng (Houston, Texas): That certainly, once you are taking a palliative approach, which is not our practice, would be reasonable for discussion. In any case we have not found in our larger tumors that they are generally entirely 1 macro cyst that would be amenable to drainage. In our institution, our practice is to resect, whether it is an anatomic resection or an enucleation of all cystic tumors of the pancreas.

Dr. Hans G. Beger (Ulm, Germany): Besides the size of the tumor, I would add 2 additional parameters to make the decision from “wait and see” to surgical treatment. These are the wall thickness and the compression of the pancreatic main ducts.

We observed in 20% of 27 patients with serous cystadenoma a carcinoma, 3 of these carcinomas were true serous cystadenoma-carcinomas, confirmed by 2 international pathologists and 2 cysts had additionally in the surrounding a ductal adenocarcinoma. We concluded from this experience the wall thickness apart from the size of the tumor, and even more if the tumor is small, is a criterion to change to surgical treatment. Compression of the pancreatic ducts, particularly narrowing of the duct of the processus uncinatus is an indication to surgical treatment. Compression of the main pancreatic duct is a strong parameter to change to treatment. Would you comment on that?

Dr. Jennifer F. Tseng (Houston, Texas): As you know, we have no cystadenocarcinomas in our series, so we cannot comment on the prevalence of cystadenoma carcinomas in regards to wall thickness. So we defer to your expert judgment in this area. In terms of compression of the main duct and the uncinate process, I think that makes a lot of sense when you see evidence of pancreatic obstruction, including pancreatitis, which occurred in only one patient in our series.

Dr. William W. Turner, Jr. (Jackson, Mississippi): Have you had any experience with either systemic or cyst fluid carbohydrate antigen measurement in predicting the behavior of these tumors?

Dr. Jennifer F. Tseng (Houston, Texas): We find that CEA in general is very low in the serous cystadenoma. And of course it can be difficult to get fluid from these tumors. But if we get fluid back and it has mucin, then we know it is a mucinous tumor. If we get fluid back and it is high in CEA, we suspect mucinous tumor. If it is very high, of course, CEA often trends even higher in mucinous adenocarcinomas. So we find its presence or absence very helpful in the endoscopic ultrasound FNA.

Dr. John L. Cameron (Baltimore, Maryland): At The Johns Hopkins Hospital we have operated upon over 200 serous cystadenomas, and on over 200 intraductal papillary mucinous neoplasms. Many of the IPMNs preoperatively had been diagnoses as serous cystadenomas, but after pathologic examination were switched to the IPMN category. In addition many of the lesions that we thought were IPMNs preoperatively, after histologic examination proved to be serous cystadenomas. Can you really accurately differentiate these 2 lesions well enough to avoid making mistakes in following patients you think have serous cystadenomas, but are actually IPMNs? Many of the IPMNs when first seen are less than 4 cm in diameter, they are 2 or 3 cm in diameter, large enough to be too worrisome to follow. Since all of these lesions, I believe, are premalignant, could you occasionally be making a mistake? How many patients that you followed and/or operated upon for serous cystadenomas, actually proved to be IPMNs, and were dropped out of the study?

Dr. Jennifer F. Tseng (Houston, Texas): Dr. Cameron, you raise a very important point. And the point in my mind is that it is a very careful balancing act. And when you are teetering, you always have to err on the conservative side, which in our minds is additional tests and possibly surgical resection. So when it is possible to confidently predict it is a serous cystadenoma based on CT and perhaps EUS/FNA, then it is reasonable to observe the patient with a small asymptomatic tumor. But if there is any substantial question in one’s mind, you have to go back to the tried and true, which is surgical resection.