|Home | About | Journals | Submit | Contact Us | Français|
Cystic lesions of the pancreas are increasingly identified, and a selective approach to resection is now recommended. The aim of this study was to assess the change in presentation and management of pancreatic cystic lesions evaluated at a single institution over a 15 year time period.
A prospectively maintained registry of patients evaluated between 1995 and 2010 for the ICD-9 diagnosis of pancreatic cyst was reviewed. The 539 patients managed from 1995 to 2005 were compared with the 885 patients managed from 2005 to 2010.
1424 patients were evaluated, including 1141 with follow-up > 6 months. Initial management (within 6 months of first assessment), was operative in 422 patients (37%), and non-operative in 719 patients (63%). Operative mortality in patients initially submitted to resection was 0.7% (n=3). Median radiographic follow-up in patients initially managed non-operatively was 28 months (range: 6–175). Patients followed radiographically were more likely to have cysts that were asymptomatic (72% vs 49%, p<0.001), smaller (1.5 cm vs 3cm, p <0.001), without solid component (94% vs. 68%, p<0.001), and without main pancreatic duct dilation (88% vs 61%, p<0.001). Changes prompting subsequent operative treatment occurred in 47 patients (6.5%), with adenocarcinoma identified in 8 (17%) and pancreatic endocrine neoplasm in 4 (8.5%). Thus of the 719 patients initially managed non-operatively, invasive malignancy was identified in 12 (1.7%), with adenocarcinoma seen in 1.1 %.
Cystic lesions of the pancreas are being increasingly identified, yet are less likely to present with concerning features of malignancy. Carefully selected patients managed non-operatively had a risk of malignancy that was equivalent to the risk of operative mortality in those patients who initially underwent resection.
With widespread use of high quality cross-sectional imaging, cystic lesions of the pancreas are being identified with increasing frequency. The prevalence of incidentally detected cystic lesions of the pancreas is approximately 10%1, and may reach as high as 30% in patients > 70 years of age2. The management of these lesions remains controversial for several reasons: 1. the underlying causal pathology ranges from benign to pre-malignant to malignant; 2. non-operative histologic diagnosis is inaccurate in approximately 25% of patients; 3. pancreatic resection is associated with a substantial risk of morbidity and measurable mortality.
Over the last decade, substantial progress has been made concerning the selection of patients for resection, and routine resection of all cystic lesions of the pancreas is no longer advocated3, 4. A more selective approach to resection is now widely recommended5-7, with resection advocated for symptoms, larger size of the lesion8, the presence of mural nodules or solid component9, the presence of septations7, or in the setting of suspicious fluid cytology10 or serum marker analysis11, 12. Selection criteria have been directed towards identifying groups of patients who have an extremely low risk of malignancy (<2%), where the risk of malignancy approximates the risk of mortality from pancreatic resection, and those patients who have high-risk lesions that are likely to have high-grade dysplasia or invasive disease.
Our group has previously reported (2006) the results of a selective surgical approach to patients with cystic neoplasms of the pancreas9. This study identified the presence of a solid component, cyst size >2.5 cm, and symptoms as criteria associated with the recommendation for resection. Since this report, we have continued to see increasing numbers of patients with cystic lesions of the pancreas. The aim of this study was to update and expand upon this prior reported experience, now over a 15 year time period, including more than 1400 patients, with particular attention to changes in presentation and management.
Memorial Sloan-Kettering Cancer Center’s pancreatic cyst registry was designed to identify all patients (both operative and nonoperative) evaluated at Memorial Sloan-Kettering Cancer Center (MSKCC) by a surgeon or gastroenterologist for a cystic lesion of the pancreas. This registry includes patients evaluated since 1995. Patients are included in this database if they are coded by the evaluating physician for the ICD-9 diagnosis of pancreatic cyst (577.2), and had a cystic lesion of the pancreas on review of imaging studies. Patients within this database who were evaluated between January 1995 and January 2010 were reviewed and included in this study. Approval for this review was obtained from MSKCC’s Institutional Review Board.
A previous report on this registry was published in 20069 and included 539 patients evaluated during the initial 10 year time period (1995-2005). In the current study, patient, radiographic, and treatment-related variables were collected as previously described and were also reviewed. Comparisons were made between those patients presented in our previous publication (Jan 1995 – Jan 2005) and those evaluated in the most recent five years (Jan 2005 – Jan 2010).
Values are expressed as median, range or percentage, as appropriate. Chi-square or Fisher exact tests were used to compare differences in discrete or categorical variables, and the t-test or Wilcoxon rank-sum test was used for continuous variables. For time related probability to undergo operative resection analysis, patients’ follow-up was censored if the patient was still alive at last follow-up without operative intervention. Time related probability to undergo surgery was estimated by the method of Kaplan–Meier and the log-rank test was used to compare survival curves. All tests were two sided. For all tests, statistical significance was defined by p < 0.05. Data were analyzed with the STATA 11 statistical software (StataCorp. 2009. Stata Statistical Software: Release 11. College Station, TX: StataCorp LP). Recursive partitioning was used to determine homogeneous subgroups with respect to the likelihood of initial resection. A decision tree was used to display the results of recursive partitioning. The optimal tree was chosen based on 10-fold cross-validation and cost-complexity pruning. Recursive partitioning was implemented using the “rpart” function in R (www.r-project.org)13.
Between January 1995 and January 2010, 1424 patients were evaluated for a cystic lesion of the pancreas. During this fifteen year period, an increasing number of patients were evaluated each year. During the initial 10 years of the study (1995-2005), 539 patients were evaluated (38%) compared to 885 during the last five years of the study (62%). The percent increase in the number of patients evaluated annually was almost linear (y=12/653x, R2=0.9118), with an approximate 8% annual increase in the number of patients evaluated for a cystic lesion of the pancreas (Figure 1). During the study period the median diameter of the lesions decreased with the median diameter being 2.4 cm (range: 0.3-18 cm) in the 1995-2005 period and 1.6 cm (range: 0.3-14 cm) in the 2005-2010 period (p=<0.0001, Figure 1).
The patient and cyst characteristics of the 1424 patients are presented in Table 1. The median age at the time of diagnosis was 67 years (range: 15-95 years), a third of the patients were male (35%, n=497), and 92% were white (n=1284). Upper gastrointestinal symptoms were the complaint resulting in identification of the lesion in 38% (n=542) and 8% had a documented history of pancreatitis. At the time of presentation the median cyst size was 2 cm (range: 0.3-18), lesions were most often located in the head of the pancreas (62%, n=614), and the majority of patients (72%, n=1025) were found to have a solitary cystic lesion (range: 0, over 10). Septations were identified in 42% (n=491), and 14% had a solid component (n=164).
Comparisons were made between patients initially reported from 1995-2005 and patients from the last five-years of the study period (2005-2010). The median age, race, and gender of the patients did not significantly differ between time periods; however, patients from the more recent time period were less likely to present with symptomatic lesions (1995-2005: 44% vs 2005-2010, 35%, p<0.0001). The location of the lesions did not change significantly over time, (head location: 1995-2005, 47% vs. 2005-2010, vs 48%, p=0.4), however the median diameter of the lesions was smaller during the second part of our experience (2.4 vs 1.6 cm, p<0.0001). The radiographic findings of septations (50% vs 36%, p<0.0001) and solid component (23 vs. 9%, p<0.0001) were also significantly decreased.
During the diagnostic evaluation, CT was the most frequently used imaging modality (97%, n=1384), and 67% (n=963) of patients underwent MRI (Figure 2). Positron Emission Tomography (PET) was used in 11% of patients, however in 2009 only 5% (11/198) of patients underwent PET imaging. Endoscopic evaluation was performed in roughly half of patients with 44% (n=626) undergoing Endoscopic Ultrasonography (EUS) and the majority of these patients undergoing Fine Needle Aspiration (FNA) of the identified lesion (n=615).
Resection was performed in 469 patients (41%), including 422 (37%) within six months of their initial visit (initial resection) and 47 patients (47/719, 6.5%) following an initial (> six month) period of radiographic surveillance (Table 2). In patients who were initially resected (n=422), the most common procedures were pancreaticoduodenectomy (36%, n=150), distal pancreatectomy with splenectomy (34%, n=145), and distal pancreatectomy (14%, n=58). Comparison between patients from 1995-2005 and 2005-2010 revealed no significant difference in operative approach, however significantly fewer patients were selected for initial resection during the more recent time period (initial resection: 1995-2005, 43% vs. 2005-2010, 33%: p=0.001). The overall grade III complication rate was 36% and the 30-day mortality rate was between 0.5% and 1% without significant change over time. The median length of stay was 8 days between 1995-2005 and 7 days between 2005-2010 (p<0.001).
The histopathology of the 422 patients who underwent initial operative resection is presented in Table 3. The most common lesions resected were intraductal papillary mucinous neoplasms (IPMN) (27%, n=114), serous cystadenoma (23%, n=98), adenocarcinoma (14%, n=60), mucinous cystadenoma (11%, n=45), and cystic pancreatic endocrine tumors (7%, n=30). Resection of lesions with carcinoma or high-grade dysplasia represented 23% of the resected lesions (n=94).
During the second part of the study period, the pathology of the resected lesions evolved. The most common resected lesion within the first ten years of the study was serous cystadenoma and this significantly decreased during the last five years (34% vs 13%, p<0.001). The percentage of patients who underwent resection for non-invasive IPMN increased in the last five years of the study (1995-2005, 17% vs 2005 – 2010, 36%, p<0.001). Resection for pseudocyst also decreased in the second period of study (8 vs 1%, p=0.001). Overall, resection for carcinoma or high-grade dysplasia was more frequent in the recent period (17% vs 28%, p=0.01), as well as for lesions known to be at risk of malignant progression (i.e. IPMN, mucinous cystadenoma, pancreatic endocrine tumor, solid pseudopapillary tumor vs. serous cystadenoma, pseudocyst or simple cyst) (40% vs 64%, p<0.001).
Characteristics of the 719 patients who were initially managed non-operatively and had > 6 months of radiographic follow-up, are presented in Table 4. The median radiographic follow-up in this group of patients was 28 months (6 – 176 months), with 39% having more than three years (n=283) of radiographic follow-up, 27% having more than 4 years (n=190), and 17% having more than 5 years of radiographic follow-up (n=125). Compared with patients who were initially resected, patients followed radiographically were older at presentation (69 yrs vs 63 yrs, p<0.001), more likely to have a personal history of malignancy (45% vs 26%, p<0.001), less likely to have symptomatic lesions (symptomatic: 28% vs 51%, p<0.001), had smaller lesions (median diameter: 1.5 cm vs 3 cm, p <0.001), without solid component (solid: 32 vs 7, p<0.001), and without main pancreatic duct dilation (12% vs 40%, p<0.001).
A total of 47 patients (47 of 719 patients, 6.5%) underwent resection after having an initial radiographic surveillance period of > six months. The details of the operative procedure and pathological examination of the resected lesions are summarized in Table 5. The median follow-up between initial visit and resection within this group of patients was 14 months (6 to 121 months). Characteristics of the patients and lesions that underwent delayed resection are presented in Table 6. Resection was performed for increasing cyst size (n=35, 74%), and/or suspicious cytology/fluid (n=14, 30%), and/or the appearance of a solid component (n=16, 35%) and/or main pancreatic duct dilatation (n=4, 8.5%). In 32% (15/47) of patients who underwent resection following initial surveillance, the operation occurred > 24 months after the initial visit, in 23% (11/47) after 36 months, in 17% (8/47) after 48 months and in 10% (5/47) after 60 months. Within the group of 15 patients operated on more than two years after initial visit, two had adenocarcinoma (lost to follow-up before surgery, patients operated on 63 and 102 months), nine had IPMN including three with carcinoma in situ (operated at 30, 39, and 67 months after initial follow-up). Within the group of eight patients who had invasive carcinoma (8/719, 1.1%), six underwent resection within 24 months of their initial visit except for the two patients who were lost to follow-up (operation at 63 and 102 months).
Recursive partitioning was performed on the 885 patients who were evaluated for a pancreatic cyst during the last five years of the study period (2005 -2010). The presence of a cyst under 2 cm was the strongest predictor of initial non-operative management (Figure 3). Previously published recursive partitioning performed on the group of patients between 1995-2005 identified the presence of a solid component as the strongest predictor of initial operative management. The difference between the decision tree analyses over the two periods reflects the changing characteristics of the evaluated lesions. Over the last five years of the study, large numbers of patients were evaluated for very small lesions (2005-2010; <1cm, n=168 and <2cm, n=505), which were very unlikely to have a solid component (solid component: 2009, 16/198; 8%. If cyst <2cm: 6/121, 5%; if cyst <1cm: 1/51, 2%). Because the characteristics which are concerning for malignancy (solid component, dilated duct, symptoms) were distinctly uncommon, size has become the only feature associated with treatment decision for the group as a whole.
Radiographic surveillance was recommended for 89% of all patients with lesions under 2 cm. The likelihood that operative intervention would be recommended for a cyst under 2 cm that had been initially selected for radiographic follow-up was 1.2% at 12 months, 4% at 36 months, and 6.3% at 60 months (Figure 4). The risk of death from causes other than pancreatic cancer within the entire group of patients initially managed non-operatively was approximately 20% at 5 years and 60% at 10 years (Figure 4). The risk of death from pancreatic cancer within the entire group of patients initially managed non-operatively was 2.5% at 5 years and 10% at 10 years. Within this group of patients initially managed non-operatively, 14 died of pancreatic adenocarcinoma. These 14 patients included nine patients who presented with a cystic lesion and pancreatic cancer that did not undergo exploration either because the lesions were unresectable at presentation or because of patient comorbidities. The remaining five deaths were in patients explored after six months of follow-up, two underwent pancreaticoduodenectomy and three underwent by-pass or exploration, (5/719; 0.7%).
The data presented in this study further demonstrate the challenge facing the medical and surgical community in managing patients with cystic lesions of the pancreas. This report documents a dramatic increase in the annual number of patients evaluated at MSKCC for a pancreatic cyst over the 15 year study period (198 patients evaluated in 2009). As the time period progressed, these lesions were more likely to be incidentally discovered (>2/3 incidentally discovered in 2009), were smaller in size (median size 1.6 cm in 2009), and less likely to harbor concerning features for malignancy such as a solid component (<10% with solid component in 2009). Initial management of these lesions also evolved over the study period with fewer patients undergoing initial operative management (1995-2005: 43% vs. 2005-2010: 33%) and fewer benign lesions being resected (1995-2005: 34% of resected lesions SCA vs. 2005-2010: 13%). Despite these trends, patients with high-risk lesions (carcinoma or high-grade dysplasia) comprised 23% of the lesions initially resected and approximately 2% of those initially observed. Thus, patients with cystic lesions of the pancreas should be considered a “high-risk” group for pancreatic cancer and careful selection of operative vs. non-operative management is essential.
Clinical decision making for these patients has been hampered by the inability to obtain accurate histologic diagnosis without operative resection, and many have recommended routine resection because of this limitation. The imaging characteristics of the most common benign (serous) and pre-malignant (mucinous) lesions have been described, however studies have shown that CT is discriminatory in only about 40% of cases 14, 15. Even when a cystic lesion can be radiographically defined as mucinous, the ability to discriminate dysplasia within a given lesion is limited to radiographic criteria such a main duct dilation, or cyst size, which are accurate for high-grade dysplasia or invasive disease in approximately 50% of cases16, 17. When a radiographically equivocal lesion is identified, EUS with FNA may be performed. EUS provides the ability to obtain detailed ultrasound images of the pancreas and also allows the ability to evaluate the cyst fluid. EUS imaging alone however has limited accuracy18 and a high degree of interobserver variability19. Cyst fluid analysis and cytology have shown mixed results with the single best test for discriminating between serous and mucinous lesions being cyst fluid CEA which when >200 ng/ml is accurate for a mucinous lesion in approximately 80% of cases20. Cyst fluid CEA however has not been shown to be a reliable marker for dysplasia or carcinoma within the mucinous group21.
Because of these diagnostic limitations there is no consensus as to which tests are necessary to evaluate a cystic lesion of the pancreas. The diagnostic guidelines of the American College of Gastroenterology22 regarding small cysts (<2cm) or very small cysts (<5mm) remain vague. Multi-detector CT (MDCT), which is widely available, remains the primary imaging modality and was performed in nearly every patient evaluated in the present series. CT imaging should be contrast-enhanced triphasic, with thin cuts through the pancreas (2.5-mm section reconstructed to 1.25 mm). MR imaging may provide improved assessment of communication with the ductal system when compared to MDCT23, 24, but its clinical advantage and cost-effectiveness compared to MDCT remain unknown. MRI does not expose the patient to radiation which may be a concern in a young patient who is to be followed radiographically for a prolonged period of time. Invasive procedures such as EUS and FNA are now associated with a very low morbidity25, and in our series were utilized in approximately half of the patients evaluated. Selective use of EUS has been recommended by others as the most cost-effective strategy in the evaluation of radiographically equivocal and asymptomatic cysts26.
We generally utilize EUS FNA for radiographically equivocal lesions (2-3 cm in diameter, some characteristics of serous) where additional information is desired before recommending a specific treatment approach. We have found the most reliable information from EUS to be for the evaluation of a mass lesion/solid component and the determination of cyst fluid CEA. When a significant solid lesion or mass component is noted on EUS, and the lesion is not consistent with a serous cyst, then resection is typically recommended. A cyst fluid CEA level >200 ng/ml is highly predictive of a mucinous lesion and a level of 0 ng/ml is strongly suggestive of a serous lesion when imaging is also consistent18. Asymptomatic serous lesions are typically followed and mucinous lesions of the IPMN sub-type are typically resected if the main duct is dilated or if the branch duct lesion is >2.5 – 3.0 cm in diameter.
EUS is not necessary in many cases, as a variety of radiographic criteria have been proposed for clinical decision making27-29. Some of these criteria have been proposed for cystic lesions in general, and some have been proposed specifically for the pre-malignant group of mucinous cysts. Regardless, most studies have found that the presence of main duct dilation, a solid component within or adjacent to the lesion, biliary dilation, symptoms, and large size of the cyst (>2.5 - 3cm) to be factors associated with high-grade dysplasia or invasive disease and operative resection should be generally recommended9,30-33. When these factors are absent, most reports have suggested the risk of an invasive lesion to be <1%9.
Between 2005 and 2010, the majority of patients evaluated at our institution for a cystic lesion of the pancreas presented with an asymptomatic lesion that was <2cm in size, without solid component, or main pancreatic ductal dilation. When decision tree analysis was performed, cyst size <2cm was the single factor most associated with treatment decision, and 89% of patients with a lesion <2cm were followed radiographically. It should be emphasized, however, that size alone was not the only factor associated with treatment recommendations, and that 11% of patients with lesions <2cm were resected. Factors such as dilation of the pancreatic or bile duct, or the presence of a mass component must also be considered, and when present would generally favor resection. These latter findings are becoming less common as lesions are identified at a smaller size and in the last five years of the study period a solid component or dilation of the main pancreatic duct was identified in fewer than 10% of all patients evaluated.
In the current study there were 719 patients who were initially placed into radiographic surveillance and had > six months of radiographic follow-up. With a median follow-up of 28 months, 47 (6.5%) patients underwent resection, and eight of these patients had pancreatic cancer at the time of resection (8/719, 1%). The risk of developing pancreatic cancer within this group of patients was identical to the risk of operative mortality in the group of 422 patients who underwent initial resection (1%). In addition, within the group of patients initially selected for radiographic follow-up pancreatic cancer was a minor cause of mortality at both 5 and 10 years from the time of identification of the cyst. The risk of non-pancreatic cancer mortality was 20% at five years and 60% at 10 years (risk of death from pancreatic cancer: 2.5% at five years, 10% at 10 years). We are aware that these findings may be more pronounced within our cohort, as they reflect the population of a cancer center. However, the presence of extra-pancreatic malignancy in patients with cystic lesions of the pancreas (mainly IPMN) has been reported by others.34-37 Bose et al.36 reported that 78% of incidental pancreatic cystic lesions were identified during interpretation of a CT scan performed for staging of a non-pancreatic malignancy. Additional patient factors such as age and comorbidity must be considered when one is considering operative resection for an incidentally discovered cystic lesion of the pancreas.
As the evaluated lesions became smaller in size, and more likely to be incidentally detected, the likelihood of resection also decreased. During the last five years of the study period the percentage of patients initially resected decreased from 43% to 33%. The most common histopathologic diagnosis also changed over the 15 year time period. Resection of serous cystadenoma accounted for a minor percentage (1995-2005: 34% vs. 2005-2010: 13%) of cystic lesions resected over the most recent five years, with non-invasive IPMN being the most common histopathology to be resected (1995-2005: 17% vs. 2005-2010: 38%). This change most likely represents an increased ability to identify serous lesions radiographically as well as a general acknowledgement that these lesions are benign and do not warrant resection except in the presence of symptoms, or possibly when arising in a young patient and large and/or marginally resectable. The relative increase in IPMN also likely represents an increased ability to identify these pre-malignant lesions both radiographically and with cyst fluid CEA, as well as our desire to resect these lesions prior to the development of invasive disease.
When radiographic surveillance is recommended, our general approach has been to perform imaging every six months for two years, and then to perform annual imaging thereafter. This approach requires long-term commitment from both the patient and physician, with a willingness to commit to long-term imaging and assessment. When patients live in parts of the country, or world, where routine surveillance is not feasible; or when they are unwilling to commit to this approach, operative resection may be warranted. This is particularly true for patients in which a known mucinous cyst has been documented. Imaging should be pancreatic dedicated contrast enhanced (either CT or MRI). Interpretation of this imaging by dedicated radiologists is essential, and discussion of equivocal lesions within a multi-disciplinary framework is highly recommended. Careful discussion of the risks of both approaches, operative and surveillance should be performed prior to agreement on an informed treatment/surveillance decision.
In summary, the data presented in this study further demonstrate the increasing number of patients being identified with asymptomatic cystic lesions of the pancreas. Over the past 15 years the patients evaluated at MSKCC have presented with lesions that are smaller in size and less likely to be symptomatic. These lesions are less likely to present with concerning radiographic features of malignancy such as a solid component or ductal dilation. Patients selected for radiographic surveillance were generally older patients with greater comorbidities (personal history of malignancy) who presented with small (<2cm) asymptomatic cysts. With a median follow-up of 28 months, patients selected for initial surveillance had a 6.5% likelihood of developing changes that prompted resection, and a 1% chance of developing pancreatic malignancy. The risk of operative mortality within the group of patients who underwent initial resection was 1%. With the dramatic increase in the number of patients identified with incidentally discovered cysts under 2 cm, the clinical challenge will be to further refine these recommendations for patients with pre-malignant mucinous lesions, and in those undergoing surveillance, to determine which imaging should be performed, over what time period, and how frequently it should be obtained.
The authors would like to thank all colleagues who provided clinical data for the database.
Dr Gaujoux received a grant from the Department of Hepato-Pancreato-Biliary Surgery - Pôle des Maladies de l’Appareil Digestif (PMAD), AP-HP, Beaujon Hospital, Clichy, France and the Assistance Publique des Hôpitaux de Paris – AP-HP.
Presented at Southern Surgical Association 122nd Annual Meeting, Palm Beach, FL, December 2010.
Disclosure Information: Nothing to disclose.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.