- Email: [email protected]
Pancreatic Incidentalomas: High Rate of Potentially Malignant Tumors
Pancreatic Incidentalomas: High Rate of Potentially Malignant Tumors Guy Lahat, MD, Mendy Ben Haim, MD, Ido Nachmany, MD, Ronen Sever, MD, Arye Blachar, MD, Richard Nakache, MD, Josef M Klausner, MD Pancreatic incidentaloma (PI) is an increasingly common diagnosis that has received little attention. We characterized these tumors and compared them with symptomatic pancreatic tumors (nonincidentaloma [NI]). STUDY DESIGN: A retrospective database of 475 consecutive pancreatectomies that were performed from January 1995 to June 2007 at our institution was analyzed. Data for PI and NI patient cohorts were compared. RESULTS: Sixty-four PIs (13.5%) and 411 NIs (86.5%) were identified; 21% of pancreatic body and tail tumors versus 9% of tumors located in the pancreatic head were incidentally diagnosed (p ⫽ 0.001). Twenty-two PIs (34%) versus 278 NIs (67%) were malignant (p ⬍ 0.0001), 38 PIs (60%) were premalignant, and the remaining 4 (6%) had little or no risk for malignant progression. Intrapapillary mucinous cystic tumor was the most common diagnosis in the PI group (23.4%, n ⫽ 15). Of these, 13.3% (n ⫽ 2) were invasive versus 40.6% (n ⫽ 15) in the NI group (p ⫽ 0.02). Likewise, pathologic features for ductal adenocarcinomas were more favorable in PI versus NI tumors. Overall, PI patients had prolonged median disease-specific survival: 145 versus 46 months (p ⫽ 0.001). Median disease-specific survival for PI versus NI patients treated for adenocarcinoma were 22 versus 19 months, respectively (p ⫽ 0.4); 5-year disease-specific survival for PI versus NI patients treated for intrapapillary mucinous cystic tumor/mucinous cystadenoma were 94% versus 68%, respectively (p ⫽ 0.07). CONCLUSIONS: Operation for PI is common, and a substantial proportion of these lesions might be malignant or premalignant. Resection of these early tumors in asymptomatic individuals is associated with improved survival, as compared with patients with symptomatic disease. (J Am Coll Surg 2009; 209:313–319. © 2009 by the American College of Surgeons) BACKGROUND:
The term incidentaloma describes an asymptomatic lesion detected incidentally by an imaging or biochemical diagnostic test. Incidentalomas of the adrenal gland and their management are well characterized;1,2 incidental findings of the thyroid,3-5 parathyroid,6,7 pituitary,8 liver,9 heart,10 prostate,11 and kidney12 are also encountered. Although these incidental findings have received particular attention in the literature, data concerning pancreatic incidentaloma (PI), an increasingly common diagnosis, are lacking. The first case of PI was described by Kostiuk in 2001;13 2 series
of incidental pancreatic cysts have been reported subsequently.14,15 Recently, Winter and colleagues16 reported a large series of pancreaticoduodenectomies performed for pancreatic head and periampullary incidentalomas; but these data do not include incidentalomas of body and tail, which are expected to be more common. In the past 2 decades, enormous advances were made in imaging technology. Particularly, improvements in CT and sonographic technology, their widespread use, and increased awareness have made abnormal pancreatic findings in asymptomatic patients much more common. In addition to abdominal CT and ultrasonography, liver function tests, tumor markers, and even upper gastrointestinal endoscopy can also detect asymptomatic findings within the pancreas. In order to expand the current knowledge about this rising diagnosis and its management, we reviewed a relatively large database of pancreatectomies performed at a single institution. We sought to characterize PIs and compare them with symptomatic tumors. Although the majority of pancreatic tumors are malignant or premalignant,
Disclosure Information: Nothing to disclose. Received November 21, 2008; Revised April 26, 2009; Accepted May 4, 2009. From the Departments of Surgery (Lahat, Ben Haim, Nachmany, Nakache, Klausner) and Radiology (Blachar), Tel Aviv Sourasky Medical Center and the Sackler School of Medicine (Lahat, Ben Haim, Nachmany, Sever, Blachar, Nakache, Klausner), Tel Aviv University, Tel Aviv, Israel. Correspondence address: Guy Lahat, MD, Department of Surgical Oncology, MD Anderson Cancer Center, 8515 Fannin St, Unit 1104, Houston, TX 77030. email: [email protected]
© 2009 by the American College of Surgeons Published by Elsevier Inc.
313
ISSN 1072-7515/09/$36.00 doi:10.1016/j.jamcollsurg.2009.05.009
314
Lahat et al
Pancreatic Incidentalomas
J Am Coll Surg
Table 1. Patient Demographics and Medical History Abbreviations and Acronyms
DSS IPMT MCT NI PI
⫽ ⫽ ⫽ ⫽ ⫽
disease-specific survival intrapapillary mucinous cystic tumor mucinous cystic tumor nonincidentaloma pancreatic incidentaloma
our hypothesis was that most incidentalomas are not overtly malignant, and early incidental detection might improve survival.
METHODS The study was approved by our institutional review board, and a waiver of consent was granted for the proposed review of patient records. From January 1995 to June 2007, 475 consecutive patients who had pancreatic resection at our institution were included in the study population. All patients included in the study cohort remained in active clinical or telephone followup. Medical records of all patients were reviewed and clinical, imaging, and pathologic data were evaluated retrospectively. Based on Winter and colleagues’ article,16 patients were categorized into two groups: patients with symptomatic disease attributable to their pancreatic finding (nonincidentaloma group [NI]) and asymptomatic patients diagnosed with a pancreatic incidental lesion (PI). All patients included in the PI group were asymptomatic for their pancreatic lesion at the time of diagnosis. Imaging incidentalomas were identified by CT scan or other high-resolution cross-sectional imaging technique; biochemical incidentalomas were detected by elevated tumor markers and serum liver or pancreatic enzymes. Demographic characteristics, past medical history, intraoperative data, pathologic data, perioperative morbidity and mortality (30-day), and longterm disease-specific survival (DSS) were compared between the PI and the NI groups. Time of DSS was calculated as the elapsed time from operation at our institution to death from disease or last followup date. Patients who died from other or unknown causes were not included in the DSS analysis. Kaplan-Meier curves were used to determine DSS time.17 Log-rank test was used to compare DSS between subgroups of patients. Comparison between patient groups was performed using t-test and Mann-Whitney tests for continuous variables as applicable; chi-square test was used for comparison of categorical variables. All values are expressed as mean ⫾ SD for parametric and median for nonparametric variables. Statistical significance level was set at 0.05, and SPSS soft-
Demographics and medical history
PI (n ⴝ 64) n %
NI (n ⴝ 411) n %
Age (y), mean (range) Female gender Other malignancy Coronary artery disease Acute MI PVD/CVA Hypertension COPD NIDDM
64 (19⫺84) 37 60.7 11 18 9 14.7 4 6.6 1 1.6 24 39.3 2 3.2 11 18
63 (15⫺87) 214 69 40 9.7 68 16.5 17 4.2 6 1.5 132 32.2 17 4.1 87 21.1
There were no statistically significant differences between PI and NI groups. NI, nonincidentaloma; NIDDM, noninsulin-dependent diabetes mellitus; PI, pancreatic incidentaloma; PVD/CVA, peripheral vascular disease/ cerebrovascular accident.
ware for Windows, version 12.0 (SPSS, Inc) was used for the analysis.
RESULTS Patient demographics and tumor location
Four hundred seventy-five pancreatic resections were performed; these included 304 pancreaticoduodenectomies (64%), 155 distal pancreatectomies (32.6%), 12 total pancreatectomies (2.5%), and 4 enucleations (0.9%). Patients were divided into 2 groups: 64 patients (13.5%) had operations for PI; 411 (86.5%) were clustered in the NI group. As presented in Table 1, patient demographics and past medical history of both PI and NI groups did not differ. Thirty-six of 64 PIs (56%) were located in the distal pancreas, compared with 28 (44%) that were located in the head. Looking at the proportion of PIs in each location, 21% (n ⫽ 33) of tumors located in the body or tail were PIs, and only 9% (n ⫽ 28) of pancreatic head tumors were incidentally diagnosed (p ⫽ 0.001). Diagnostic characteristics
PIs were identified by tomography (n ⫽ 27, 42%), sonography (n ⫽ 25, 39%), serum biochemistry (n ⫽ 9, 14%), and endoscopy (n ⫽ 3, 5%). Twenty-nine incidentalomas (45%) were diagnosed during workup for nonpancreatic symptoms, including gynecological symptoms (n ⫽ 6), unspecified complaints (n ⫽ 6), respiratory symptoms (n ⫽ 5), dysphagia (n ⫽ 2), kidney stones (n ⫽ 3), diverticular disease (n ⫽ 2), small bowel obstruction (n ⫽ 1), trauma (n ⫽ 1), peripheral edema (n ⫽ 1), abdominal pain after an enema (n ⫽ 1), and dysuria (n ⫽ 1). Figure 1 illustrates an incidentally diagnosed solid and cystic papillary tumor of the pancreas; the patient was admitted to the emergency room after an assault, and CT was performed
Vol. 209, No. 3, September 2009
Lahat et al
Pancreatic Incidentalomas
315
Figure 1. Axial CT scan of the upper abdomen demonstrates a round, well-defined, slightly heterogeneous mass with a well-defined hyperdense border consistent with a solid and papillary cystic neoplasm.
because of severe abdominal pain shortly after the blunt trauma. Twelve incidentalomas (18%) were detected during a routine/executive evaluation (physical examination, laboratory tests, or sonography). Ten patients (15.5%) were followed for other malignancy or chronic disease: prostate cancer (n ⫽ 3), colorectal cancer (n ⫽ 2), renal cell carcinoma (n ⫽ 1), ovarian cancer (n ⫽ 1), family history of pancreatic cancer (n ⫽ 1), chronic back pain (n ⫽ 1), and familial polyposis (n ⫽ 1). Figure 2 depicts a pancreatic neuroendocrine tumor detected during followup for colorectal cancer. Four incidentalomas (6%) were diagnosed during routine workup before a planned operation: prostatectomy (n ⫽ 2), cataract (n ⫽ 1) and hernia repair (n ⫽ 1). Reasons were not specified in the remaining 10 cases (15.5%). Surgical outcomes
Surgical outcomes of PI versus NI patients who had pancreaticoduodenectomy or distal pancreatectomy were compared separately (19 patients who had total pancreatectomy or enucleation were excluded for the purpose of this analysis). Twenty-eight PI (43.8%) versus 276 NI patients (67.2%) underwent pancreaticoduodenectomy (p ⬍ 0.0001); perioperative mortality rates were 3.6% (n ⫽ 1) and 2.8% (n ⫽ 8) in the PI and NI groups, respectively (p ⫽ 0.58). Major postoperative complication rates were also comparable between PI and NI patients who had pancreaticoduodenectomy, 15 (53.5%) and 131 (47.4%), respectively (p ⫽ 0.53). Pancreatic fistula occurred in 25% (n ⫽ 7) of PI versus 10.5% (n ⫽ 29) of NI patients (p ⫽ 0.03). Twenty-two NI patients (7.9%) were reoperated, and no PI patient required reoperation (p ⫽ 0.04); mean hospitalization time did not
Figure 2. . (A) Axial CT scan image and (B) coronal maximum intensity projection reconstruction during the pancreatic arterial phase shows a hypervascular round and slightly heterogeneous mass demonstrating intense enhancement consistent with an islet cell tumor.
differ significantly between PI versus NI groups, 22.7 days (SD ⫽ 12.1) versus 24.3 days (SD ⫽ 15.6), respectively (p ⫽ 0.14). Thirty-four PI patients (53.1%) versus 118 NI patients (28.7%) underwent distal pancreatectomy; comparing both subcohorts of patients for all surgical outcomes mentioned here, there were no statistically significant differences. Pathologic diagnoses
The main pathologic diagnoses in the PI group were as follows: pancreatic ductal or ampullary adenocarcinoma (25%, n ⫽ 16), intrapapillary mucinous cystic tumor (IPMT; 23.4%, n ⫽ 15), mucinous cystic tumor (MCT; 15.6%, n ⫽ 10), and neuroendocrine tumor (15.6%, n ⫽
316
Lahat et al
Pancreatic Incidentalomas
J Am Coll Surg
Table 2. Pathologic Diagnoses
Variable
Malignant Pancreatic ductal adenocarcinoma Ampullary adenocarcinoma IPMT Mucinous cystic tumors Distal bile duct cancer Neuroendocrine tumor Malignant neuroendocrine tumor Solid and cystic pseudopapillary tumor Pancreatitis Other
PI (n ⴝ 64) n %
NI (n ⴝ 411) n %
Table 3. Staging of Intrapapillary Mucinous Cystic Tumor and Mucinous Cystadenoma PI (n ⴝ 15) n %
p Value
22* 34.3 278 67.6 ⬍ 0.0001
IPMT
11* 5* 15 10 1 10*
Adenoma or borderline Intrapapillary mucinous cystic tumor with invasive cancer Mucinous cystic tumor Mucinous cystadenoma Mucinous cystadenocarcinoma
17.2 183 44.5 7.8 39 9.5 23.4 37 9 15.6 43 10.5 1.6 9 2.2 15.6 23 5.6
3
4.7
6* 0 6
9.4 0 9.4
5
1.2
0.001 0.67 0.001 0.1 1.0 0.04 0.04
4 0.9 ⬍ 0.0001 16 3.9 0.1 57 13.9
*Statistical significance versus NI group (p ⬍ 0.05). IPMT, intrapapillary mucinous cystic tumor; NI, nonincidentaloma; PI, pancreatic incidentaloma.
10). The most common pathologies in the NI group were pancreatic ductal adenocarcinoma (44.5%, n ⫽ 183), ampullary adenocarcinoma (9.5%, n ⫽ 39), and IPMT (9%, n ⫽ 37). Pathologic diagnoses are presented in Table 2. Twenty-two PIs (34.3%) versus 278 NIs (67.6%) were malignant (p ⬍ 0.0001); 11 PIs (17.2%) versus 183 NIs (44.5%) were diagnosed as pancreatic ductal adenocarcinomas (p ⫽ 0.001). IPMTs and mucinous cystadenomas were subcategorized according to their level of invasion, and the results are presented in Table 3. Fifteen PIs (23.4%) versus 37 NIs (9%) were IPMTs (p ⫽ 0.001); 2 IPMTs (13.3%) in the PI group were invasive, compared to 15 (40.6%) in the NI group (p ⫽ 0.02). Ten PIs (15.6%) compared to 43 NIs (10.4%) were MCTs (p ⫽ 0.28); there were 9 cystadenocarcinomas
NI (n ⴝ 37) n %
p Value
13
87
22
59.4
2* 10 10
13
40.6
100
15 43 34
0
9
0
0.02
79 21
0.1
*Statistical significance versus NI group (p ⬍ 0.05). IPMT, intrapapillary mucinous cystic tumor; NI, nonincidentaloma; PI, pancreatic incidentaloma.
(21%) in the NI group, compared with 0 (0%) in the PI cohort (p ⫽ 0.3). Table 4 summarizes the pathologic data of all malignancies; median tumor diameter was 2.5 cm (range, 0.4 to 16 cm) and 3.5 cm (range, 0.7 to 25 cm) in the PI and NI groups, respectively (p ⫽ 0.32). Metastatic lymph nodes were found in 10 PIs (45.4%) and 130 NIs (46.7%; p ⫽ 0.43), and the mean number of metastatic lymph nodes was 0.4, compared with 1.15 (SD ⫽ 1.17) in the PI and NI groups, respectively (p ⫽ 0.001). Vascular invasion was found in 4 PIs (18.2%), compared to 73 NIs (36.2%; p ⫽ 0.002), and perineural invasion was detected in 6 (27%) and 132 (47.4%) of the PI and NI groups, respectively (p ⬍ 0.0001). Analyses of 238 pancreatic ductal or ampullary adenocarcinomas, according to their degree of differentiation, are presented in Table 4. Well-differentiated carcinomas were more common in the PI group, 6 (37.5%) versus 33 (14.8%; p ⫽ 0.04), and poorly differentiated carcinomas were more common in the NI group, 61 (67.4%) versus 2 (12.5%; p ⫽ 0.02).
Table 4. Pathologic Data for Malignant Tumors All malignancies
Tumor diameter (cm), median (range) Positive margins, n (%) Metastatic LN, n (%) Mean No. of positive LN (SD ⫽ 1.17) Vascular invasion, n (%) Perineural invasion, n (%) Differentiation of ductal and ampullary carcinoma Well, n (%) Moderate, n (%) Poor, n (%) *Statistical significance versus nonincidentaloma group (p ⬍ 0.05). LN, lymph nodes; NI, nonincidentaloma; PI, pancreatic incidentaloma.
PI (n ⴝ 22)
NI (n ⴝ 278)
p Value
2.5 (0.4⫺16) 2 (9.1) 10 (45.4) 0.4* 4 (18.2)* 6 (27)* 16 6 (37.5) 8 (50) 2 (12.5)*
3.5 (0.7⫺25) 17 (9.5) 130 (46.7) 1.15 73 (36.2) 132 (47.4) 222 33 (14.8) 128 (58) 61 (27.4)
0.1 0.64 0.43 0.001 0.002 ⬍ 0.0001
0.026
Vol. 209, No. 3, September 2009
Lahat et al
Pancreatic Incidentalomas
317
DSS
Overall, longterm outcomes were better for the cohort of PI patients (n ⫽ 64) as compared with the cohort of NI patients (n ⫽ 411); 1-, 2-, and 5-year DSS rates for PI and NI patients were 91%, 75%, and 64% versus 75%, 59%, and 47%, respectively, with a median survival of 145 months compared to 46 months (p ⫽ 0.001). The 1-, 2-, and 5-year DSS rates for patients with solid and/or cystic malignant PI (n ⫽ 22) versus NI (n ⫽ 278) were 75%, 46%, and 31% versus 67%, 44%, and 28%, respectively, with a median survival of 23 months versus 20 months (p ⫽ 0.3; Fig. 3A). We also evaluated DSS rates when the data were confined to pancreatic solid ductal or ampullary adenocarcinomas; 1-, 2-, and 5-year survival rates for PI (n ⫽ 16) versus NI (n ⫽ 222) patients were 78%, 44%, and 29% versus 67%, 43%, and 26%, respectively, with a median survival of 22 months versus 19 months (p ⫽ 0.4; Fig. 3B). DSS of PI versus NI patients who had operations for IPMT and/or MCT were evaluated separately; a trend toward significantly improved survival was demonstrated in the PI group (n ⫽ 25) in comparison to the NI group (n ⫽ 81); 5-year DSS rates were 94% versus 68%, respectively; median survival was not yet reached in either group (p ⫽ 0.07; Fig. 3C).
DISCUSSION PI is a rapidly increasing phenomenon, and pancreatic operations for asymptomatic patients are likely to become common. The proportion of PIs operated on in our series of 475 pancreatectomies is 13.5%, ⬎ 90% of them after the year 2000. Since that year, the rate of pancreatic operations for PI is 17.3%! Most commonly, PIs are detected on abdominal CT performed for various other clinical manifestations. As cited by Winter and colleagues,16 since the advent of multislice CT in 1998, the number of CT scans performed in the US has increased dramatically and continues to increase by roughly 20% annually.18 Today, ⬎ 7,000 scanners in the US perform ⬎ 50 million CT scans annually.19 In our institution, 3 multidetector CT scanners are installed (2 Philips Brilliance 16 and 1 Philips Brilliance 64 scanners) and perform approximately 40,000 scans per year. The
Figure 3. (A) Kaplan-Meier disease-specific survival curves for patients with malignant tumors who underwent pancreatectomy for pancreatic incidentaloma (PI) versus nonincidentaloma (NI). (B) Kaplan-Meier disease-specific survival curves for patients with adenocarcinoma who underwent pancreatectomy for PI versus NI. (C) Kaplan-Meier disease-specific survival curves for patients with intrapapillary mucinous cystic tumor or mucinous cystic tumor who underwent pancreatectomy for PI versus NI.
318
Lahat et al
Pancreatic Incidentalomas
number of CT studies per 1,000 emergency department visits has increased from 26.2 in 1995 to 70.8 in 2002. The total number of emergency room visits during the decade between 1995 and 2004 increased by 17%, and the number of CT scans increased by 255%. This demonstrates a disproportionate use of CT in the emergency room. In addition, the total number of CT scans performed at our institution has increased by ⬎ 30% since 2004. In our institution, as in other high-volume centers, there is an increased use of CT, and this likely explains the rising rate of pancreatectomies performed for PI by our group. The implications of an incidental detection of a pancreatic mass are yet to be studied. There is a clear paucity of data concerning the outcomes of patients followed, or operated on, for PI. Winter and colleagues16 reported the experience of Johns Hopkins with pancreaticoduodenectomies performed for PI, and Fernandez-del Castillo and colleagues14 reported the Massachusetts General Hospital experience with operations for pancreatic cystic incidentalomas. As mentioned previously, most incidentalomas are usually detected during workup or followup for other medical conditions.20,21 In our series, approximately two-thirds of PIs were diagnosed during workup for nonpancreatic symptoms or during surveillance for other malignancy or chronic disease; the remaining PIs were diagnosed in asymptomatic healthy individuals. The anatomic distribution of PI was not equal. Not surprisingly, there was a considerably higher proportion of PIs in the distal pancreas, where mass-induced symptoms are less likely to occur. Most notable in this series is the high rate of malignancy or malignant potential of PI, ie, 94%. This rate is higher than reported previously14,16 and includes all carcinomas, IPMT, MCT, and neuroendocrine tumor cases. The likely explanation for this high rate is routine use of preoperative multislice CT and endoscopic ultrasound with fine needle aspiration/biopsy, which enables careful patient selection for pancreatic operation.22 All patients in this series had a preoperative abdominal CT, endoscopic ultrasound was performed in 65% of them (n ⫽ 308), and endoscopic ultrasound-guided fine needle aspiration/ biopsy was performed in 33% (n ⫽ 153). Solid tumors were considered potentially malignant, as were cystic lesions with solid areas, mucin, other characteristic features of IPMT, high CEA levels, or atypical cells in cystic fluid. Forty percent of PIs were ultimately diagnosed as IPMT or MCT. These cystic tumors are likely to progress to invasive cancer with time. The point of malignant transformation cannot be anticipated from time of detection, and once invasive cancer occurs, the prognosis is considerably worse. In this series, none of the incidental MCTs, and only 13% of incidental IPMTs, were invasive, considerably less
J Am Coll Surg
than in comparison with MCT and IPMT in the NI group, where cancer was diagnosed in 21% of the symptomatic MCTs and almost half of the symptomatic IPMTs. Higher incidence of invasive cancer among the NI patients treated for cystic tumors explains the worse prognosis demonstrated in our separate survival analysis for this cohort of patients. Apparently, development of invasive cancer is time dependent in both tumors, and it is expected that early incidental detection and resection of these lesions will prevent cancer. Several recent reports recommend a selective approach toward resection of IPMT;23-25 but adequate data and technology are still lacking in order to establish proper patient selection for a conservative approach. Additional investigation can identify molecular biomarkers predictive of in situ or invasive cancer. Until then, we support an aggressive approach if the patient is not a poor surgical candidate. Rates of malignant PI versus malignant NI were 34.3% versus 67.6% respectively, p ⬍ 0.0001; this significant difference correlates with previous data showing that symptomatic tumors are more likely to be malignant.14,16 This higher rate of overt malignancy in the NI group explains the better 5-year DSS rate of all PI as compared with all NI patients, as well as the better prognosis of patients who had operations for incidental cystic tumors versus those who were symptomatic. It should be noted that the malignancy rate for PI is much higher in comparison with resected adrenal incidentalomas, supporting a thorough diagnostic approach toward incidentally diagnosed pancreatic lesions.16,26 Almost half (45.4%) of malignant PIs had metastasized to lymph nodes, similar to the rate of lymph node involvement in the NI group; but the mean number of involved lymph nodes was lower in the PI group. These results are difficult to interpret, because both groups are composed of differently behaving malignancies; future studies concerning the patterns of lymphatic spread, using larger cohorts of patients, might enable similar analysis for specific pathologies. Analyzing the pathologic data of the pancreatic adenocarcinomas only, PIs had significantly favorable features: smaller tumor diameter (2.5 versus 3.5 cm), lower rates of vascular and perineural invasion, and a higher level of differentiation. These characteristics can be explained by less aggressive local behavior of PIs, which makes them more likely to be asymptomatic; but early detection is also a possible explanation for these favorable pathologic characteristics. Interestingly, our data demonstrate higher survival rates for patients treated for malignant PI versus NI, with median survival lengths of 23 months versus 20 months, respectively. A similar trend was observed comparing patients who had operations for pancreatic ductal and ampullary adenocarcinomas: 22 months versus 19 months, respec-
Vol. 209, No. 3, September 2009
tively. These differences were not statistically significant. Better pathologic features observed in the malignant PI group might explain their superior outcomes; but our data need to be validated using a larger and more homogenous cohort of patients, which might also enable evaluation of incidental detection as an independent predictor of outcomes in pancreatic cancer. Our study has two main limitations: First, it includes surgically resected PIs only, although most pancreatic incidentalomas are followed by gastroenterologists as benign lesions. The second limitation is the retrospective nature of our data collection. In spite of these limitations, the large number of patients in our study enabled us to achieve statistical power in most analyses, thereby advancing the current knowledge concerning PIs and their management. Results of this study show that surgical management of PIs is an increasing phenomenon. Although PIs are asymptomatic by definition, most patients who had operations for such lesions were subsequently found to have malignant or premalignant tumors. The indolent nature of PIs should not mislead clinicians in favor of a conservative approach. Our data demonstrate that overt malignancy, as well as nodal involvement, are not uncommon findings in asymptomatic pancreatic tumors; thus, symptomatic presentation as a marker for malignancy may be wrong. Accidental early detection and resection of tumors with malignant potential can prevent cancer. Interestingly, although the incidence of malignancy in PIs was high, these lesions displayed more favorable pathologic features and longterm outcomes. Future prospective clinicopathologic data collection of all incidentally detected pancreatic tumors is needed to allow a thorough understanding of their nature and to establish an algorithm for their management. Author Contributions Study conception and design: Lahat, Nakache, Klausner Acquisition of data: Lahat, Nachmany, Sever, Blachar Analysis and interpretation of data: Lahat, Ben Haim, Blachar Drafting of manuscript: Lahat, Blachar, Klausner Critical revision: Klausner
REFERENCES 1. Udelsman R, Fishman EK. Radiology of the adrenal. Endocrinol Metab Clin North Am 2000;29:27–42. 2. Thompson GB, Young WF Jr. Adrenal incidentaloma. Curr Opin Oncol 2003;15:84–90. 3. Whineray Kelly EL, Braatvedt G, Harman R. A parathyroid incidentaloma. Aust N Z J Surg 2005;75:367. 4. Aron DC, Howlett TA. Pituitary incidentalomas. Endocrinol Metab Clin North Am 2000;29:205–221.
Lahat et al
Pancreatic Incidentalomas
319
5. Mitchell J, Parangi S. The thyroid incidentaloma: an increasingly frequent consequence of radiologic imaging. Semin Ultrasound CT MR 2005;26:37–46. 6. Silver RJ, Parangi S. Management of thyroid incidentalomas. Surg Clin North Am 2004;84:907–919. 7. Chen YK, Ding HJ, Chen KT, et al. Prevalence and risk of cancer of focal thyroid incidentaloma identified by 18F-fluorodeoxyglucose positron emission tomography for cancer screening in healthy subjects. Anticancer Res 2005;25:1421–1426. 8. Schneider HJ, Stalla GK. 10-minute consultation: incidentaloma of the hypophysis. MMW Fortschr Med 2004;146:67–68. 9. Liu CL, Fan ST, Lo CM, et al. Hepatic resection for incidentaloma. J Gastrointest Surg 2004;8:785–793. 10. Vancollie O, Rombaut E, Donckier J. Cardiac incidentaloma. Ann Cardiol Angeiol (Paris) 2001;50:316–318. 11. Picurelli Oltra L, Sendra Torres A, Fernandez Rodriguez A, et al. Incidental prostatic adenocarcinoma in the era of the PSA. Actas Urol Esp 1997;21:354–356. 12. Pobil Moreno JL, Martinez Rodriguez J, Maestro Duran JL, et al. Renal incidentaloma and pregnancy. Arch Esp Urol 1996;49: 755–757. 13. Kostiuk TS. Observation of pancreatic incidentaloma. Klin Khir 2001;9:62–63. 14. Fernandez-del Castillo C, Targarona J, Thayer SP, et al. Incidental pancreatic cysts: clinicopathologic characteristics and comparison with symptomatic patients. Arch Surg 2003;138:427– 423; discussion 433-434. 15. Handrich SJ, Hough DM, Fletcher JG, et al. The natural history of the incidentally discovered small simple pancreatic cyst: longterm follow-up and clinical implications. AJR Am J Roentgenol 2005;184:20–23. 16. Winter JM, Cameron JL, Lillemoe KD, et al. Periampullary and pancreatic incidentaloma: a single institution’s experience with an increasingly common diagnosis. Ann Surg 2006;243:673–683. 17. Kaplan EL, Meier P. Nonparametric estimator from incomplete observations. J Am Stat Assoc 1958;53:457–481. 18. Davis W. Multi-slice CT: 64 and counting. Med Imaging 2005; 5:1267–1278. 19. Linton OW, Mettler FA Jr. National conference on dose reduction in CT, with an emphasis on pediatric patients. AJR Am J Roentgenol 2003;181:321–329. 20. Brunt LM, Moley JF. Adrenal incidentaloma. W J Surg 2001; 25:905–913. 21. Frilling A, Tecklenborg K, Weber F, et al. Importance of adrenal incidentaloma in patients with a history of malignancy. Surgery 2004;136:1289–1296. 22. Maguchi H, Osanai M, Yanagawa N, et al. Endoscopic ultrasonography diagnosis of pancreatic cystic disease. Endoscopy 1998;30:A108–A110. 23. Walsh RM, Vogt DP, Henderson JM, et al. Natural history of indeterminate pancreatic cysts. Surgery 2005;138:665–670. 24. Allen PJ, Jaques DP, D’Angelica M, et al. Cystic lesions of the pancreas: selection criteria for operative and non operative management in 209 patients. J Gastrointest Surg 2003;7:970–977. 25. Allen PJ, D’Angelica M, Gonen M. A selective approach to the resection of cystic lesions of the pancreas: results from 539 consecutive patients. Ann Surg 2006;244:572–579. 26. Guerrieri M, De Sanctis A, Crosta F, et al. Adrenal incidentaloma: surgical update. J Endocrinol Invest 2007;30:200–204.
The term incidentaloma describes an asymptomatic lesion detected incidentally by an imaging or biochemical diagnostic test. Incidentalomas of the adrenal gland and their management are well characterized;1,2 incidental findings of the thyroid,3-5 parathyroid,6,7 pituitary,8 liver,9 heart,10 prostate,11 and kidney12 are also encountered. Although these incidental findings have received particular attention in the literature, data concerning pancreatic incidentaloma (PI), an increasingly common diagnosis, are lacking. The first case of PI was described by Kostiuk in 2001;13 2 series
of incidental pancreatic cysts have been reported subsequently.14,15 Recently, Winter and colleagues16 reported a large series of pancreaticoduodenectomies performed for pancreatic head and periampullary incidentalomas; but these data do not include incidentalomas of body and tail, which are expected to be more common. In the past 2 decades, enormous advances were made in imaging technology. Particularly, improvements in CT and sonographic technology, their widespread use, and increased awareness have made abnormal pancreatic findings in asymptomatic patients much more common. In addition to abdominal CT and ultrasonography, liver function tests, tumor markers, and even upper gastrointestinal endoscopy can also detect asymptomatic findings within the pancreas. In order to expand the current knowledge about this rising diagnosis and its management, we reviewed a relatively large database of pancreatectomies performed at a single institution. We sought to characterize PIs and compare them with symptomatic tumors. Although the majority of pancreatic tumors are malignant or premalignant,
Disclosure Information: Nothing to disclose. Received November 21, 2008; Revised April 26, 2009; Accepted May 4, 2009. From the Departments of Surgery (Lahat, Ben Haim, Nachmany, Nakache, Klausner) and Radiology (Blachar), Tel Aviv Sourasky Medical Center and the Sackler School of Medicine (Lahat, Ben Haim, Nachmany, Sever, Blachar, Nakache, Klausner), Tel Aviv University, Tel Aviv, Israel. Correspondence address: Guy Lahat, MD, Department of Surgical Oncology, MD Anderson Cancer Center, 8515 Fannin St, Unit 1104, Houston, TX 77030. email: [email protected]
© 2009 by the American College of Surgeons Published by Elsevier Inc.
313
ISSN 1072-7515/09/$36.00 doi:10.1016/j.jamcollsurg.2009.05.009
314
Lahat et al
Pancreatic Incidentalomas
J Am Coll Surg
Table 1. Patient Demographics and Medical History Abbreviations and Acronyms
DSS IPMT MCT NI PI
⫽ ⫽ ⫽ ⫽ ⫽
disease-specific survival intrapapillary mucinous cystic tumor mucinous cystic tumor nonincidentaloma pancreatic incidentaloma
our hypothesis was that most incidentalomas are not overtly malignant, and early incidental detection might improve survival.
METHODS The study was approved by our institutional review board, and a waiver of consent was granted for the proposed review of patient records. From January 1995 to June 2007, 475 consecutive patients who had pancreatic resection at our institution were included in the study population. All patients included in the study cohort remained in active clinical or telephone followup. Medical records of all patients were reviewed and clinical, imaging, and pathologic data were evaluated retrospectively. Based on Winter and colleagues’ article,16 patients were categorized into two groups: patients with symptomatic disease attributable to their pancreatic finding (nonincidentaloma group [NI]) and asymptomatic patients diagnosed with a pancreatic incidental lesion (PI). All patients included in the PI group were asymptomatic for their pancreatic lesion at the time of diagnosis. Imaging incidentalomas were identified by CT scan or other high-resolution cross-sectional imaging technique; biochemical incidentalomas were detected by elevated tumor markers and serum liver or pancreatic enzymes. Demographic characteristics, past medical history, intraoperative data, pathologic data, perioperative morbidity and mortality (30-day), and longterm disease-specific survival (DSS) were compared between the PI and the NI groups. Time of DSS was calculated as the elapsed time from operation at our institution to death from disease or last followup date. Patients who died from other or unknown causes were not included in the DSS analysis. Kaplan-Meier curves were used to determine DSS time.17 Log-rank test was used to compare DSS between subgroups of patients. Comparison between patient groups was performed using t-test and Mann-Whitney tests for continuous variables as applicable; chi-square test was used for comparison of categorical variables. All values are expressed as mean ⫾ SD for parametric and median for nonparametric variables. Statistical significance level was set at 0.05, and SPSS soft-
Demographics and medical history
PI (n ⴝ 64) n %
NI (n ⴝ 411) n %
Age (y), mean (range) Female gender Other malignancy Coronary artery disease Acute MI PVD/CVA Hypertension COPD NIDDM
64 (19⫺84) 37 60.7 11 18 9 14.7 4 6.6 1 1.6 24 39.3 2 3.2 11 18
63 (15⫺87) 214 69 40 9.7 68 16.5 17 4.2 6 1.5 132 32.2 17 4.1 87 21.1
There were no statistically significant differences between PI and NI groups. NI, nonincidentaloma; NIDDM, noninsulin-dependent diabetes mellitus; PI, pancreatic incidentaloma; PVD/CVA, peripheral vascular disease/ cerebrovascular accident.
ware for Windows, version 12.0 (SPSS, Inc) was used for the analysis.
RESULTS Patient demographics and tumor location
Four hundred seventy-five pancreatic resections were performed; these included 304 pancreaticoduodenectomies (64%), 155 distal pancreatectomies (32.6%), 12 total pancreatectomies (2.5%), and 4 enucleations (0.9%). Patients were divided into 2 groups: 64 patients (13.5%) had operations for PI; 411 (86.5%) were clustered in the NI group. As presented in Table 1, patient demographics and past medical history of both PI and NI groups did not differ. Thirty-six of 64 PIs (56%) were located in the distal pancreas, compared with 28 (44%) that were located in the head. Looking at the proportion of PIs in each location, 21% (n ⫽ 33) of tumors located in the body or tail were PIs, and only 9% (n ⫽ 28) of pancreatic head tumors were incidentally diagnosed (p ⫽ 0.001). Diagnostic characteristics
PIs were identified by tomography (n ⫽ 27, 42%), sonography (n ⫽ 25, 39%), serum biochemistry (n ⫽ 9, 14%), and endoscopy (n ⫽ 3, 5%). Twenty-nine incidentalomas (45%) were diagnosed during workup for nonpancreatic symptoms, including gynecological symptoms (n ⫽ 6), unspecified complaints (n ⫽ 6), respiratory symptoms (n ⫽ 5), dysphagia (n ⫽ 2), kidney stones (n ⫽ 3), diverticular disease (n ⫽ 2), small bowel obstruction (n ⫽ 1), trauma (n ⫽ 1), peripheral edema (n ⫽ 1), abdominal pain after an enema (n ⫽ 1), and dysuria (n ⫽ 1). Figure 1 illustrates an incidentally diagnosed solid and cystic papillary tumor of the pancreas; the patient was admitted to the emergency room after an assault, and CT was performed
Vol. 209, No. 3, September 2009
Lahat et al
Pancreatic Incidentalomas
315
Figure 1. Axial CT scan of the upper abdomen demonstrates a round, well-defined, slightly heterogeneous mass with a well-defined hyperdense border consistent with a solid and papillary cystic neoplasm.
because of severe abdominal pain shortly after the blunt trauma. Twelve incidentalomas (18%) were detected during a routine/executive evaluation (physical examination, laboratory tests, or sonography). Ten patients (15.5%) were followed for other malignancy or chronic disease: prostate cancer (n ⫽ 3), colorectal cancer (n ⫽ 2), renal cell carcinoma (n ⫽ 1), ovarian cancer (n ⫽ 1), family history of pancreatic cancer (n ⫽ 1), chronic back pain (n ⫽ 1), and familial polyposis (n ⫽ 1). Figure 2 depicts a pancreatic neuroendocrine tumor detected during followup for colorectal cancer. Four incidentalomas (6%) were diagnosed during routine workup before a planned operation: prostatectomy (n ⫽ 2), cataract (n ⫽ 1) and hernia repair (n ⫽ 1). Reasons were not specified in the remaining 10 cases (15.5%). Surgical outcomes
Surgical outcomes of PI versus NI patients who had pancreaticoduodenectomy or distal pancreatectomy were compared separately (19 patients who had total pancreatectomy or enucleation were excluded for the purpose of this analysis). Twenty-eight PI (43.8%) versus 276 NI patients (67.2%) underwent pancreaticoduodenectomy (p ⬍ 0.0001); perioperative mortality rates were 3.6% (n ⫽ 1) and 2.8% (n ⫽ 8) in the PI and NI groups, respectively (p ⫽ 0.58). Major postoperative complication rates were also comparable between PI and NI patients who had pancreaticoduodenectomy, 15 (53.5%) and 131 (47.4%), respectively (p ⫽ 0.53). Pancreatic fistula occurred in 25% (n ⫽ 7) of PI versus 10.5% (n ⫽ 29) of NI patients (p ⫽ 0.03). Twenty-two NI patients (7.9%) were reoperated, and no PI patient required reoperation (p ⫽ 0.04); mean hospitalization time did not
Figure 2. . (A) Axial CT scan image and (B) coronal maximum intensity projection reconstruction during the pancreatic arterial phase shows a hypervascular round and slightly heterogeneous mass demonstrating intense enhancement consistent with an islet cell tumor.
differ significantly between PI versus NI groups, 22.7 days (SD ⫽ 12.1) versus 24.3 days (SD ⫽ 15.6), respectively (p ⫽ 0.14). Thirty-four PI patients (53.1%) versus 118 NI patients (28.7%) underwent distal pancreatectomy; comparing both subcohorts of patients for all surgical outcomes mentioned here, there were no statistically significant differences. Pathologic diagnoses
The main pathologic diagnoses in the PI group were as follows: pancreatic ductal or ampullary adenocarcinoma (25%, n ⫽ 16), intrapapillary mucinous cystic tumor (IPMT; 23.4%, n ⫽ 15), mucinous cystic tumor (MCT; 15.6%, n ⫽ 10), and neuroendocrine tumor (15.6%, n ⫽
316
Lahat et al
Pancreatic Incidentalomas
J Am Coll Surg
Table 2. Pathologic Diagnoses
Variable
Malignant Pancreatic ductal adenocarcinoma Ampullary adenocarcinoma IPMT Mucinous cystic tumors Distal bile duct cancer Neuroendocrine tumor Malignant neuroendocrine tumor Solid and cystic pseudopapillary tumor Pancreatitis Other
PI (n ⴝ 64) n %
NI (n ⴝ 411) n %
Table 3. Staging of Intrapapillary Mucinous Cystic Tumor and Mucinous Cystadenoma PI (n ⴝ 15) n %
p Value
22* 34.3 278 67.6 ⬍ 0.0001
IPMT
11* 5* 15 10 1 10*
Adenoma or borderline Intrapapillary mucinous cystic tumor with invasive cancer Mucinous cystic tumor Mucinous cystadenoma Mucinous cystadenocarcinoma
17.2 183 44.5 7.8 39 9.5 23.4 37 9 15.6 43 10.5 1.6 9 2.2 15.6 23 5.6
3
4.7
6* 0 6
9.4 0 9.4
5
1.2
0.001 0.67 0.001 0.1 1.0 0.04 0.04
4 0.9 ⬍ 0.0001 16 3.9 0.1 57 13.9
*Statistical significance versus NI group (p ⬍ 0.05). IPMT, intrapapillary mucinous cystic tumor; NI, nonincidentaloma; PI, pancreatic incidentaloma.
10). The most common pathologies in the NI group were pancreatic ductal adenocarcinoma (44.5%, n ⫽ 183), ampullary adenocarcinoma (9.5%, n ⫽ 39), and IPMT (9%, n ⫽ 37). Pathologic diagnoses are presented in Table 2. Twenty-two PIs (34.3%) versus 278 NIs (67.6%) were malignant (p ⬍ 0.0001); 11 PIs (17.2%) versus 183 NIs (44.5%) were diagnosed as pancreatic ductal adenocarcinomas (p ⫽ 0.001). IPMTs and mucinous cystadenomas were subcategorized according to their level of invasion, and the results are presented in Table 3. Fifteen PIs (23.4%) versus 37 NIs (9%) were IPMTs (p ⫽ 0.001); 2 IPMTs (13.3%) in the PI group were invasive, compared to 15 (40.6%) in the NI group (p ⫽ 0.02). Ten PIs (15.6%) compared to 43 NIs (10.4%) were MCTs (p ⫽ 0.28); there were 9 cystadenocarcinomas
NI (n ⴝ 37) n %
p Value
13
87
22
59.4
2* 10 10
13
40.6
100
15 43 34
0
9
0
0.02
79 21
0.1
*Statistical significance versus NI group (p ⬍ 0.05). IPMT, intrapapillary mucinous cystic tumor; NI, nonincidentaloma; PI, pancreatic incidentaloma.
(21%) in the NI group, compared with 0 (0%) in the PI cohort (p ⫽ 0.3). Table 4 summarizes the pathologic data of all malignancies; median tumor diameter was 2.5 cm (range, 0.4 to 16 cm) and 3.5 cm (range, 0.7 to 25 cm) in the PI and NI groups, respectively (p ⫽ 0.32). Metastatic lymph nodes were found in 10 PIs (45.4%) and 130 NIs (46.7%; p ⫽ 0.43), and the mean number of metastatic lymph nodes was 0.4, compared with 1.15 (SD ⫽ 1.17) in the PI and NI groups, respectively (p ⫽ 0.001). Vascular invasion was found in 4 PIs (18.2%), compared to 73 NIs (36.2%; p ⫽ 0.002), and perineural invasion was detected in 6 (27%) and 132 (47.4%) of the PI and NI groups, respectively (p ⬍ 0.0001). Analyses of 238 pancreatic ductal or ampullary adenocarcinomas, according to their degree of differentiation, are presented in Table 4. Well-differentiated carcinomas were more common in the PI group, 6 (37.5%) versus 33 (14.8%; p ⫽ 0.04), and poorly differentiated carcinomas were more common in the NI group, 61 (67.4%) versus 2 (12.5%; p ⫽ 0.02).
Table 4. Pathologic Data for Malignant Tumors All malignancies
Tumor diameter (cm), median (range) Positive margins, n (%) Metastatic LN, n (%) Mean No. of positive LN (SD ⫽ 1.17) Vascular invasion, n (%) Perineural invasion, n (%) Differentiation of ductal and ampullary carcinoma Well, n (%) Moderate, n (%) Poor, n (%) *Statistical significance versus nonincidentaloma group (p ⬍ 0.05). LN, lymph nodes; NI, nonincidentaloma; PI, pancreatic incidentaloma.
PI (n ⴝ 22)
NI (n ⴝ 278)
p Value
2.5 (0.4⫺16) 2 (9.1) 10 (45.4) 0.4* 4 (18.2)* 6 (27)* 16 6 (37.5) 8 (50) 2 (12.5)*
3.5 (0.7⫺25) 17 (9.5) 130 (46.7) 1.15 73 (36.2) 132 (47.4) 222 33 (14.8) 128 (58) 61 (27.4)
0.1 0.64 0.43 0.001 0.002 ⬍ 0.0001
0.026
Vol. 209, No. 3, September 2009
Lahat et al
Pancreatic Incidentalomas
317
DSS
Overall, longterm outcomes were better for the cohort of PI patients (n ⫽ 64) as compared with the cohort of NI patients (n ⫽ 411); 1-, 2-, and 5-year DSS rates for PI and NI patients were 91%, 75%, and 64% versus 75%, 59%, and 47%, respectively, with a median survival of 145 months compared to 46 months (p ⫽ 0.001). The 1-, 2-, and 5-year DSS rates for patients with solid and/or cystic malignant PI (n ⫽ 22) versus NI (n ⫽ 278) were 75%, 46%, and 31% versus 67%, 44%, and 28%, respectively, with a median survival of 23 months versus 20 months (p ⫽ 0.3; Fig. 3A). We also evaluated DSS rates when the data were confined to pancreatic solid ductal or ampullary adenocarcinomas; 1-, 2-, and 5-year survival rates for PI (n ⫽ 16) versus NI (n ⫽ 222) patients were 78%, 44%, and 29% versus 67%, 43%, and 26%, respectively, with a median survival of 22 months versus 19 months (p ⫽ 0.4; Fig. 3B). DSS of PI versus NI patients who had operations for IPMT and/or MCT were evaluated separately; a trend toward significantly improved survival was demonstrated in the PI group (n ⫽ 25) in comparison to the NI group (n ⫽ 81); 5-year DSS rates were 94% versus 68%, respectively; median survival was not yet reached in either group (p ⫽ 0.07; Fig. 3C).
DISCUSSION PI is a rapidly increasing phenomenon, and pancreatic operations for asymptomatic patients are likely to become common. The proportion of PIs operated on in our series of 475 pancreatectomies is 13.5%, ⬎ 90% of them after the year 2000. Since that year, the rate of pancreatic operations for PI is 17.3%! Most commonly, PIs are detected on abdominal CT performed for various other clinical manifestations. As cited by Winter and colleagues,16 since the advent of multislice CT in 1998, the number of CT scans performed in the US has increased dramatically and continues to increase by roughly 20% annually.18 Today, ⬎ 7,000 scanners in the US perform ⬎ 50 million CT scans annually.19 In our institution, 3 multidetector CT scanners are installed (2 Philips Brilliance 16 and 1 Philips Brilliance 64 scanners) and perform approximately 40,000 scans per year. The
Figure 3. (A) Kaplan-Meier disease-specific survival curves for patients with malignant tumors who underwent pancreatectomy for pancreatic incidentaloma (PI) versus nonincidentaloma (NI). (B) Kaplan-Meier disease-specific survival curves for patients with adenocarcinoma who underwent pancreatectomy for PI versus NI. (C) Kaplan-Meier disease-specific survival curves for patients with intrapapillary mucinous cystic tumor or mucinous cystic tumor who underwent pancreatectomy for PI versus NI.
318
Lahat et al
Pancreatic Incidentalomas
number of CT studies per 1,000 emergency department visits has increased from 26.2 in 1995 to 70.8 in 2002. The total number of emergency room visits during the decade between 1995 and 2004 increased by 17%, and the number of CT scans increased by 255%. This demonstrates a disproportionate use of CT in the emergency room. In addition, the total number of CT scans performed at our institution has increased by ⬎ 30% since 2004. In our institution, as in other high-volume centers, there is an increased use of CT, and this likely explains the rising rate of pancreatectomies performed for PI by our group. The implications of an incidental detection of a pancreatic mass are yet to be studied. There is a clear paucity of data concerning the outcomes of patients followed, or operated on, for PI. Winter and colleagues16 reported the experience of Johns Hopkins with pancreaticoduodenectomies performed for PI, and Fernandez-del Castillo and colleagues14 reported the Massachusetts General Hospital experience with operations for pancreatic cystic incidentalomas. As mentioned previously, most incidentalomas are usually detected during workup or followup for other medical conditions.20,21 In our series, approximately two-thirds of PIs were diagnosed during workup for nonpancreatic symptoms or during surveillance for other malignancy or chronic disease; the remaining PIs were diagnosed in asymptomatic healthy individuals. The anatomic distribution of PI was not equal. Not surprisingly, there was a considerably higher proportion of PIs in the distal pancreas, where mass-induced symptoms are less likely to occur. Most notable in this series is the high rate of malignancy or malignant potential of PI, ie, 94%. This rate is higher than reported previously14,16 and includes all carcinomas, IPMT, MCT, and neuroendocrine tumor cases. The likely explanation for this high rate is routine use of preoperative multislice CT and endoscopic ultrasound with fine needle aspiration/biopsy, which enables careful patient selection for pancreatic operation.22 All patients in this series had a preoperative abdominal CT, endoscopic ultrasound was performed in 65% of them (n ⫽ 308), and endoscopic ultrasound-guided fine needle aspiration/ biopsy was performed in 33% (n ⫽ 153). Solid tumors were considered potentially malignant, as were cystic lesions with solid areas, mucin, other characteristic features of IPMT, high CEA levels, or atypical cells in cystic fluid. Forty percent of PIs were ultimately diagnosed as IPMT or MCT. These cystic tumors are likely to progress to invasive cancer with time. The point of malignant transformation cannot be anticipated from time of detection, and once invasive cancer occurs, the prognosis is considerably worse. In this series, none of the incidental MCTs, and only 13% of incidental IPMTs, were invasive, considerably less
J Am Coll Surg
than in comparison with MCT and IPMT in the NI group, where cancer was diagnosed in 21% of the symptomatic MCTs and almost half of the symptomatic IPMTs. Higher incidence of invasive cancer among the NI patients treated for cystic tumors explains the worse prognosis demonstrated in our separate survival analysis for this cohort of patients. Apparently, development of invasive cancer is time dependent in both tumors, and it is expected that early incidental detection and resection of these lesions will prevent cancer. Several recent reports recommend a selective approach toward resection of IPMT;23-25 but adequate data and technology are still lacking in order to establish proper patient selection for a conservative approach. Additional investigation can identify molecular biomarkers predictive of in situ or invasive cancer. Until then, we support an aggressive approach if the patient is not a poor surgical candidate. Rates of malignant PI versus malignant NI were 34.3% versus 67.6% respectively, p ⬍ 0.0001; this significant difference correlates with previous data showing that symptomatic tumors are more likely to be malignant.14,16 This higher rate of overt malignancy in the NI group explains the better 5-year DSS rate of all PI as compared with all NI patients, as well as the better prognosis of patients who had operations for incidental cystic tumors versus those who were symptomatic. It should be noted that the malignancy rate for PI is much higher in comparison with resected adrenal incidentalomas, supporting a thorough diagnostic approach toward incidentally diagnosed pancreatic lesions.16,26 Almost half (45.4%) of malignant PIs had metastasized to lymph nodes, similar to the rate of lymph node involvement in the NI group; but the mean number of involved lymph nodes was lower in the PI group. These results are difficult to interpret, because both groups are composed of differently behaving malignancies; future studies concerning the patterns of lymphatic spread, using larger cohorts of patients, might enable similar analysis for specific pathologies. Analyzing the pathologic data of the pancreatic adenocarcinomas only, PIs had significantly favorable features: smaller tumor diameter (2.5 versus 3.5 cm), lower rates of vascular and perineural invasion, and a higher level of differentiation. These characteristics can be explained by less aggressive local behavior of PIs, which makes them more likely to be asymptomatic; but early detection is also a possible explanation for these favorable pathologic characteristics. Interestingly, our data demonstrate higher survival rates for patients treated for malignant PI versus NI, with median survival lengths of 23 months versus 20 months, respectively. A similar trend was observed comparing patients who had operations for pancreatic ductal and ampullary adenocarcinomas: 22 months versus 19 months, respec-
Vol. 209, No. 3, September 2009
tively. These differences were not statistically significant. Better pathologic features observed in the malignant PI group might explain their superior outcomes; but our data need to be validated using a larger and more homogenous cohort of patients, which might also enable evaluation of incidental detection as an independent predictor of outcomes in pancreatic cancer. Our study has two main limitations: First, it includes surgically resected PIs only, although most pancreatic incidentalomas are followed by gastroenterologists as benign lesions. The second limitation is the retrospective nature of our data collection. In spite of these limitations, the large number of patients in our study enabled us to achieve statistical power in most analyses, thereby advancing the current knowledge concerning PIs and their management. Results of this study show that surgical management of PIs is an increasing phenomenon. Although PIs are asymptomatic by definition, most patients who had operations for such lesions were subsequently found to have malignant or premalignant tumors. The indolent nature of PIs should not mislead clinicians in favor of a conservative approach. Our data demonstrate that overt malignancy, as well as nodal involvement, are not uncommon findings in asymptomatic pancreatic tumors; thus, symptomatic presentation as a marker for malignancy may be wrong. Accidental early detection and resection of tumors with malignant potential can prevent cancer. Interestingly, although the incidence of malignancy in PIs was high, these lesions displayed more favorable pathologic features and longterm outcomes. Future prospective clinicopathologic data collection of all incidentally detected pancreatic tumors is needed to allow a thorough understanding of their nature and to establish an algorithm for their management. Author Contributions Study conception and design: Lahat, Nakache, Klausner Acquisition of data: Lahat, Nachmany, Sever, Blachar Analysis and interpretation of data: Lahat, Ben Haim, Blachar Drafting of manuscript: Lahat, Blachar, Klausner Critical revision: Klausner
REFERENCES 1. Udelsman R, Fishman EK. Radiology of the adrenal. Endocrinol Metab Clin North Am 2000;29:27–42. 2. Thompson GB, Young WF Jr. Adrenal incidentaloma. Curr Opin Oncol 2003;15:84–90. 3. Whineray Kelly EL, Braatvedt G, Harman R. A parathyroid incidentaloma. Aust N Z J Surg 2005;75:367. 4. Aron DC, Howlett TA. Pituitary incidentalomas. Endocrinol Metab Clin North Am 2000;29:205–221.
Lahat et al
Pancreatic Incidentalomas
319
5. Mitchell J, Parangi S. The thyroid incidentaloma: an increasingly frequent consequence of radiologic imaging. Semin Ultrasound CT MR 2005;26:37–46. 6. Silver RJ, Parangi S. Management of thyroid incidentalomas. Surg Clin North Am 2004;84:907–919. 7. Chen YK, Ding HJ, Chen KT, et al. Prevalence and risk of cancer of focal thyroid incidentaloma identified by 18F-fluorodeoxyglucose positron emission tomography for cancer screening in healthy subjects. Anticancer Res 2005;25:1421–1426. 8. Schneider HJ, Stalla GK. 10-minute consultation: incidentaloma of the hypophysis. MMW Fortschr Med 2004;146:67–68. 9. Liu CL, Fan ST, Lo CM, et al. Hepatic resection for incidentaloma. J Gastrointest Surg 2004;8:785–793. 10. Vancollie O, Rombaut E, Donckier J. Cardiac incidentaloma. Ann Cardiol Angeiol (Paris) 2001;50:316–318. 11. Picurelli Oltra L, Sendra Torres A, Fernandez Rodriguez A, et al. Incidental prostatic adenocarcinoma in the era of the PSA. Actas Urol Esp 1997;21:354–356. 12. Pobil Moreno JL, Martinez Rodriguez J, Maestro Duran JL, et al. Renal incidentaloma and pregnancy. Arch Esp Urol 1996;49: 755–757. 13. Kostiuk TS. Observation of pancreatic incidentaloma. Klin Khir 2001;9:62–63. 14. Fernandez-del Castillo C, Targarona J, Thayer SP, et al. Incidental pancreatic cysts: clinicopathologic characteristics and comparison with symptomatic patients. Arch Surg 2003;138:427– 423; discussion 433-434. 15. Handrich SJ, Hough DM, Fletcher JG, et al. The natural history of the incidentally discovered small simple pancreatic cyst: longterm follow-up and clinical implications. AJR Am J Roentgenol 2005;184:20–23. 16. Winter JM, Cameron JL, Lillemoe KD, et al. Periampullary and pancreatic incidentaloma: a single institution’s experience with an increasingly common diagnosis. Ann Surg 2006;243:673–683. 17. Kaplan EL, Meier P. Nonparametric estimator from incomplete observations. J Am Stat Assoc 1958;53:457–481. 18. Davis W. Multi-slice CT: 64 and counting. Med Imaging 2005; 5:1267–1278. 19. Linton OW, Mettler FA Jr. National conference on dose reduction in CT, with an emphasis on pediatric patients. AJR Am J Roentgenol 2003;181:321–329. 20. Brunt LM, Moley JF. Adrenal incidentaloma. W J Surg 2001; 25:905–913. 21. Frilling A, Tecklenborg K, Weber F, et al. Importance of adrenal incidentaloma in patients with a history of malignancy. Surgery 2004;136:1289–1296. 22. Maguchi H, Osanai M, Yanagawa N, et al. Endoscopic ultrasonography diagnosis of pancreatic cystic disease. Endoscopy 1998;30:A108–A110. 23. Walsh RM, Vogt DP, Henderson JM, et al. Natural history of indeterminate pancreatic cysts. Surgery 2005;138:665–670. 24. Allen PJ, Jaques DP, D’Angelica M, et al. Cystic lesions of the pancreas: selection criteria for operative and non operative management in 209 patients. J Gastrointest Surg 2003;7:970–977. 25. Allen PJ, D’Angelica M, Gonen M. A selective approach to the resection of cystic lesions of the pancreas: results from 539 consecutive patients. Ann Surg 2006;244:572–579. 26. Guerrieri M, De Sanctis A, Crosta F, et al. Adrenal incidentaloma: surgical update. J Endocrinol Invest 2007;30:200–204.