ORIGINAL ARTICLE. Identification of Unknown Primary Tumors in Patients With Neuroendocrine Liver Metastases

ORIGINAL ARTICLE Identification of Unknown Primary Tumors in Patients With Neuroendocrine Liver Metastases Sam C. Wang, MD; Justin R. Parekh, MD; Marlene B. Zuraek, MD; Alan P. Venook, MD; Emily K. Bergsland, MD; Robert S. Warren, MD; Eric K. Nakakura, MD, PhD Background: For patients with neuroendocrine tumor (NET) liver metastases, resection of the primary tumor may prevent local complications (obstruction, ischemia, and bleeding) and improve survival. Despite preoperative evaluation, the primary tumor location may remain unknown. Design: Retrospective cohort analysis of pathology database from January 1, 1993, to August 15, 2008. Setting: Academic medical center. Patients: One hundred twenty-three patients with NET liver metastases. Main Outcome Measures: Successful identification and resection of the primary tumor. Results: Fifteen patients underwent surgical exploration. The primary tumor was located in 13 patients (86.7%) in the small intestine and resected in 12 patients. Primary tumors in the small intestine found during surgical exploration were significantly smaller than those identified preoperatively (1.38 vs 1.91 cm, P=.03) and were often multifocal (54.2% [n=15]). Computed tomography (34.6% [n=78]) and somatostatin receptor scintigraphy (26.2% [n=42]) were not sensitive in locating a primary NET in the gastrointestinal tract. Colonoscopy was sensitive in detecting colonic NETs (86.7% [n=15]). Conclusion: For patients with NET liver metastases and unknown primary tumor, surgical exploration effectively identifies and resects occult primary tumors that are often located in the small intestine. Primary tumors are usually small and multifocal, so careful palpation of the small intestine is essential. Before patients are considered for surgery, a multidisciplinary team assessment and evaluation consisting of computed tomography, somatostatin receptor scintigraphy, and upper and lower endoscopy should be done. Arch Surg. 2010;145(3):276-280 Author Affiliations: Department of Surgery, Division of Surgical Oncology (Drs Wang, Parekh, Zuraek, Warren, and Nakakura) and Department of Medicine, Division of Hematology/ Oncology (Drs Venook and Bergsland), University of California; UCSF Helen Diller Family Comprehensive Cancer Center (Drs Wang, Parekh, Zuraek, Venook, Bergsland, Warren, and Nakakura), San Francisco. NEUROENDOCRINE TUmors (NETs), previously known as carcinoid tumors, arise most commonly in the gastrointestinal (GI) tract and bronchopulmonary tree. 1 Although rare, their incidence has increased over the past 30 years, and they now occur in 2.5 to 5 individuals per 100 000 persons per year in the United States. 2,3 Unfortunately, most patients with GI NETs are not candidates for potentially curative surgery because they usually have extensive metastatic disease, and survival rates have not improved over 30 years. 2,4 The small intestine is the most common site; approximately 67% to 75% of patients with NETs of the small intestine present with regional or distant spread, and less than 50% are alive at 5 years. 4 Because GI NETs are rare and most are advanced when diagnosed, they present unique clinical challenges that few physicians have experience in handling. Although NETs of the small intestine are often small, they have the propensity to cause a characteristic fibrosis of the mesentery, resulting in bowel obstruction, ischemia, or perforation in approximately one-third of patients. 5,6 Therefore, resection of the primary tumor and regional lymph node metastases is accepted as standard treatment for locoregional disease. In addition, for carefully selected patients with NET liver metastases, hepatic resection seems to improve survival, although most patients are not candidates for liver resection because of extensive disease. 7,8 Recent studies 9,10 suggest that resection of the primary tumor when unresectable NET liver metastases are present may be associated with improved survival. Thus, some 276

advocate that the primary tumor be localized and resected, even in asymptomatic patients with unresectable NET liver metastases. 6,9 A recent epidemiological study of NETs that included almost 20 000 patients showed that approximately 11% to 14% had metastatic disease with unknown primary tumor. 1 Nevertheless, there are no clear recommendations on how best to identify unknown primary tumors in patients with NET liver metastases. Previous studies 11-14 have suggested a limited role for computed tomography (CT), somatostatin receptor scintigraphy, enteroclysis, or capsule endoscopy for detecting the primary NET. An evolving technique, magnetic resonance imaging enteroclysis, identified small intestinal NETs in 8 of 12 patients. 15 Recently, surgical exploration was used to successfully localize and resect occult primary NETs in 17 of 22 patients (77%), but the use of preoperative evaluations in identifying the primary tumor was not described. 6 The 2-fold purpose of our study was to determine the use of preoperative evaluation for detecting the primary tumor in a larger number of patients with well-differentiated NET liver metastases and whether laparoscopic or open surgical exploration is effective for identifying and resecting an unknown primary tumor in patients with NET liver metastases. METHODS We performed a retrospective review of the University of California, San Francisco (UCSF) pathology database to identify all patients with a tissue diagnosis of well-differentiated NET liver metastases from January 1, 1993, to August 15, 2008. Demographic, clinical, radiographic, procedural, and pathological data were obtained from electronic medical records. The UCSF Committee on Human Research approved the study. Statistical comparisons were calculated using the t test. RESULTS We identified 123 patients with NET liver metastases confirmed by biopsy (Table 1 and Figure 1). Nearly twothirds (64.2%) presented to UCSF with NET liver metastases, and 27.6% presented with localized disease that subsequently metastasized to the liver. The most common primary tumor sites were the pancreas (35.0%), small intestine (26.8%), and colon (12.2%). Of the 79 patients who presented with NET liver metastases, 8 had no further workup because of extensive disease or comorbid conditions. The 71 remaining patients underwent evaluations to localize the primary tumors at the discretion of the treating physicians. Each patient underwent at least 1 CT scan, and most (59.2%) underwent somatostatin receptor scintigraphy. Many patients underwent upper and lower endoscopy (29.6% and 59.2%, respectively) to identify the primary tumor and to assess for synchronous neoplasms. Rarely, ultrasonography, positron emission tomography, magnetic resonance imaging, small-bowel series, and capsule endoscopy were done. All patients with NETs arising from the GI tract and pancreas had their diagnosis confirmed by tissue biopsy. This meant we could retrospectively evaluate the sensitivity of Table 1. Characteristics of 123 Patients With Metastatic Neuroendocrine Tumor of the Liver Characteristic Finding, No. (%) Age, mean (range) 57.0 (18-96) Sex Male 66 (53.7) Female 57 (46.3) Race/ethnicity White 91 (74.0) Black 9 (7.3) Asian 8 (6.5) Hispanic 3 (2.4) Other/unknown 12 (9.8) Primary tumor location Pancreas 43 (35.0) Small intestine a 33 (26.8) Colon/rectum 15 (12.2) Pulmonary 5 (4.1) Thyroid 2 (1.6) Stomach 1 (0.8) Kidney 1 (0.8) Presacral 1 (0.8) Unknown b 22 (17.9) a Includes patients who initially had occult primary tumors that were later identified on surgical exploration. b Unknown owing to inadequate records, or lack of workup secondary to the extent of disease or other comorbidities, or failed workup. different diagnostic methods for detecting the primary tumor, although our data did not permit us to determine specificity. We found that for patients with NET liver metastases and a primary tumor of the pancreas, CT successfully localized the primary tumor in every case, probably because the average size of the primary tumors of the pancreas (7.98 cm) was much larger than that of the small bowel (1.70 cm, P.01) and large bowel (3.75 cm, P.01). In contrast, for patients with NET liver metastases and primary tumors in the GI tract (ie, stomach, small intestine, colon, or rectum), CT successfully identified the primary tumor just 34.6% of the time, and somatostatin receptor scintigraphy localized it even less frequently (Table 2). It was not always possible to determine from the records whether every ileal primary tumor was located in the terminal ileum and potentially amenable to colonoscopic detection. Thus, we likely underestimated the ability of colonoscopy to identify primary tumors in the ileum. Tumors of the small intestine detected on CT were significantly larger than tumors that were not (2.09 vs 1.60 cm, P=.05) (Table 3). In contrast, the ability to detect primary tumors in the colon on CT did not correlate with size. As for endoscopy, the primary tumors in the colon that were not seen actually tended to be larger, although there were only 2 such patients (Table 2). Finally, the ability of somatostatin receptor scintigraphy to locate the primary tumor did not seem to depend on size. Despite extensive evaluation, 17 patients (13.8%) had occult primary tumors, 2 of whom underwent surgery for reasons other than to locate the primary tumor. One patient underwent a hepatectomy to resect NET liver metastases. The other had laparoscopic radiofrequency ablation of NET liver metastases. In the operative reports, there was no description of a careful surgical exploration for the pri- 277

123 Patients with NET liver disease seen at UCSF from January 1, 1993, to August 15, 2008 34 Patients presented without liver 79 Patients presented with metastatic 10 Patients with inadequate records metastases, then developed metastases disease 56 Primary tumors were 15 Primary tumors not 8 Patients had no workup located located despite workup because of comorbidities, disease extent, or death 15 Patients underwent surgical exploration 2 Patients underwent a surgical procedure with goal of locating primary tumor to manage liver disease only 7 Patients underwent 8 Patients underwent laparoscopic exploration open exploration 6 Operations were 1 Operation failed 7 Operations were 1 Operation failed successful successful Figure 1. Patient cohort. *Tumor location of patients who presented without neuroendocrine tumor (NET) liver disease: pancreas, 16; small intestine, 3; colon, 8; lung, 3; thyroid, 2; and mesentery, 2 (primary unknown). Within this cohort, 2 patients who presented with mesenteric masses and without NET liver disease proceeded to surgical exploration. Tumor location as determined by workup at the University of California, San Francisco (UCSF): pancreas, 27; small intestine, 17; colon, 7; lung, 2; stomach, 1; kidney, 1; and presacral, 1. Table 2. Sensitivity of Diagnostic Methods for Locating Neuroendocrine Tumors in the Gastrointestinal Tract Primary Identified Diagnostic Method No. of Studies (Sensitivity), No. (%) Computed tomography 78 27 (34.6) Magnetic resonance imaging 3 0 Somatostatin receptor 42 11 (26.2) scintigraphy PET scan 8 2 (25.0) Upper endoscopy Total 21 1 (4.8) Stomach NET 1 1 (100.0) Lower endoscopy Total 42 20 (47.6) All colon/ileum NET 33 20 (60.6) Colon/rectum NET 15 13 (86.7) Ileum NET 18 7 (38.9) Capsule endoscopy 1 0 Small-bowel series 1 1 (100.0) Ultrasonography 3 0 Abbreviations: NET, neuroendocrine tumor; PET, positron emission tomography. mary tumor. The 15 remaining patients underwent surgical exploration to locate the occult primary tumor and to achieve other operative goals, such as cholecystectomy in anticipation of long-term octreotide therapy and/or resection of NET liver metastases. Seven of 8 laparotomies and 6 of 7 laparoscopic explorations were successful in locating the primary tumor (total, 86.7% success). In 1 patient, resection was not attempted owing to extensive mesenteric fibrosis. For 2 patients in the laparoscopic group, conversion to an open surgical procedure was required to resect the mesenteric disease safely. All primary tumors found on surgical exploration were in the small intestine; 12 were in the ileum, and 1 in the jejunum. When compared with small-intestine tumors located on preoperative evaluation, the 13 occult lesions found on surgical exploration were significantly smaller (1.91 vs 1.38 cm, P=.03). No differences were noted in demographics, symptoms, multifocality, or biochemical levels between tumors identified during surgical exploration and those found preoperatively. For patients with small-intestinal primary tumors, most (13 of 24 [54.2%]) had multifocal disease, and most (20 of 24 [83.3%]) had documented mesenteric fibrosis. COMMENT This study sought to determine the use of preoperative evaluation for detecting the primary tumor in a large number of patients with well-differentiated NET liver metastases and whether laparoscopic or open surgical exploration is effective for identifying an unknown primary tumor in patients with NET liver metastases. To our knowledge, our series of 71 patients is the largest to date. We found that preoperative evaluation fails to identify the primary tumor in 21.1% of patients, which agrees with the findings of another study. 1 Moreover, we show that laparoscopic or open surgical exploration is effective for identifying and resecting an unknown primary tumor in patients with NET liver metastases. Our high success rate (86.7%) in identifying the occult primary tumor in patients with NET liver metastases is the highest reported (22%-77%). 6,12-16 We believe our experience with this rare condition will help other surgeons manage patients with NET liver metastases. Among 123 patients evaluated at UCSF with NET liver metastases, the pancreas was the most common primary site identified (35.0%). In every case, CT detected the primary 278

Table 3. Success of Diagnostic Methods for Locating Primary Small-Intestine and Colonic Neuroendocrine Tumors Relative to Tumor Size Modality Result Size of Tumor in Small Intestine (SD), cm [No. Found] Computed tomography Positive 2.09 (0.62) [12] Negative 1.60 (0.75) [29] Colonoscopy Positive 2.13 (0.70) [6] Negative 1.53 (0.64) [12] Somatostatin receptor scintigraphy Positive 1.53 (0.48) [6] Negative 1.76 (0.79) [22] a t Test was used for statistical analysis. P.05 was statistically significant. P Value a Size of Tumor in Colon (SD), cm [No. Found].05 3.39 (1.56) [7] 3.00 (2.50) [3].08 3.03 (1.77) [8] 5.75 (0.35) [2].51 1.75 (1.77) [2] 2.17 (1.44) [3] P Value a.77.07.79 tumorofthepancreas,whichwaslarge(meansize,7.98cm). Our findings may not be surprising, given the retroperitoneal location of the pancreas and unlikelihood for nonfunctioning pancreas NETs to cause symptoms until reaching a threshold size. However, we believe our results suggest an important point: it is highly unlikely that an occult primary tumor of the pancreas is present in patients with NET liver metastases. Therefore, our findings also suggest that endoscopic ultrasonography to look for a subcentimeter tumor in the pancreas is unwarranted in a patient with NET liver metastases. In our series, CT and somatostatin receptor scintigraphy were the most commonly used methods for detecting tumors. Both had low sensitivity for detecting the primary tumor in the stomach, small intestine, and colon/rectum (CT, 34.6%; somatostatin receptor scintigraphy, 26.2%). The small size of primary NETs in the GI tract makes preoperative detection challenging. Indeed, reported sensitivities for detection of primary NETs are low: CT, 0%- 22% 12 ; capsule endoscopy, 38%-45% 13 ; CT enteroclysis, 50% 14 ; and double balloon enteroscopy, 33%. 16 In our series, the average size of primary NETs in the small and large intestine was only 1.84 cm. Not unexpectedly, the smallintestine tumors discovered on CT were significantly larger than ones that were not detected. Although somatostatin receptor scintigraphy had a low sensitivity for detecting primary NETs, in conjunction with CT, somatostatin receptor scintigraphy can help to determine the extent of disease (eg, in lymph nodes, liver, and bone), which may provide useful prognostic information. Computed tomography may be useful in detecting mesenteric masses that may result from extension of the primary NET or lymph node metastases with associated fibrosis, suggesting a small-intestine primary tumor. In our database search, we identified 5 patients who presented with a mesenteric mass seen on CT (2 of whom have NET liver metastases and are included in our study cohort). Four of these patients had primary NETs in the small intestine; 1 had NET in the right colon. Our patient cohort underwent a total of 42 colonoscopies, with 20 successfully detecting the primary tumor. However, only 33 were done in patients who were later confirmed to have a primary NET in the colon or ileum. In the 9 other patients, the primary NETs were in locations not amenable to colonoscopic detection. Thus, the true sensitivity of colonoscopy in this patient group is at least 60.6% (20 of 33 patients). We were unable to determine if all patients marked as having ileal NETs in our database actually had terminal ileal disease that would have been amenable to colonoscopic detection. Colonoscopy was especially effective in locating NETs in the colon (86.7%). In the 15 patients in our series who had NET liver metastases and an unknown primary tumor, we successfully identified the primary NET in 13 (86.7%), all of which were located in the small intestine. We found that laparoscopy readily detected the primary NET in the small intestine since it typically causes a dimpling of the serosa (Figure 2). Because most patients had multifocal primary tumors (54.2%), which are often pea-sized, we believe careful palpation of the entire small intestine is essential. During laparoscopic surgery, we used the soft-tissue wound retractor (Alexis Wound Retractor; Applied Medical, Rancho Santa Margarita, California) or the hand-assisted laparoscopic device (GelPort; Applied Medical) to aid in the exteriorization of the bowel to permit palpation. If laparoscopic exploration fails to identify the primary tumor, we do not believe open surgical exploration is indicated since careful palpation of the small bowel and preoperative upper or lower endoscopy should detect most primary tumors in the GI tract that are likely to cause symptoms. As a retrospective analysis of a cohort from a single institution, our study had no control subjects and is subject to selection and referral bias. Therefore, although ours is the largest series to date, the following recommendations must be offered cautiously. For patients with welldifferentiated NET liver metastases, localization and resection of the primary tumor should be considered to treat and prevent complications (ie, pain, obstruction, ischemia, or perforation) and, perhaps, improve survival. 9,10 Computed tomography and somatostatin receptor scintigraphy should be obtained to identify the primary tumor and assess the extent of disease. If no pancreatic tumor is seen on CT, our study suggests that an occult primary tumor of the pancreas is unlikely and that up- 279

Figure 2. Primary ileal neuroendocrine tumor. Note the tumor is characteristically white and dimples the serosa. per and lower endoscopy should be done to search for the primary tumor and to rule out synchronous neoplasms. Synchronous or metachronous neoplasms can occur in an estimated 20% of patients with NET. 2 In our cohort, we identified 4 patients with synchronous and 5 with metachronous disease. If the patient is a suitable candidate for surgery, even if he or she has unresectable NET liver metastases, laparoscopy with careful palpation of the entire small intestine should be done to identify the primary NET. Frequently, there are multifocal primary tumors that are often very small. Laparoscopic resection is preferred but may be infeasible if the regional adenopathy, fibrosis, or both extend into the root of the mesentery. Importantly, even in the absence of a desmoplastic mesenteric mass, a wide mesenteric dissection encompassing regional lymph nodes is required; if the surgeon cannot achieve this laparoscopically, then an open approach should be taken. Perioperative octreotide infusion should be given to prevent carcinoid crisis. At the time of surgery, a cholecystectomy should be done to prevent gallstone-related complications associated with longterm somatostatin analog therapy and to prevent gallbladder necrosis from hepatic artery embolization. In conclusion, for patients with NET liver metastases and an unknown primary tumor, surgical exploration effectively identifies and resects occult primary tumors that are often located in the small intestine. Primary tumors are usually small and multifocal, so careful palpation of the small intestine is essential. Before a patient is considered for surgery, a multidisciplinary team assessment and evaluation consisting of CT scan, somatostatin receptor scintigraphy, and upper/lower endoscopy should be done. Accepted for Publication: April 3, 2009. Correspondence: Eric K. Nakakura, MD, PhD, 1600 Divisadero St, Room A-724, San Francisco, CA 94143-1932 (eric.nakakura@ucsfmedctr.org). Author Contributions: Drs Wang and Parekh contributed equally to this article. Drs Wang, Parekh, and Nakakura had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Wang, Parekh, and Nakakura. Acquisition of data: Wang and Parekh. Analysis and interpretation of data: Wang, Parekh, and Nakakura. Drafting of the manuscript: Wang, Parekh, and Nakakura. Critical revision of the manuscript for important intellectual content: Wang, Parekh, Zuraek, Venook, Bergsland, Warren, and Nakakura. Statistical analysis: Parekh. Study supervision: Nakakura. Financial Disclosure: None reported. Previous Presentations: Presented as a poster at the 80th Annual Meeting of the Pacific Coast Surgical Association; February 14, 2009; San Francisco, California. Additional Contributions: Pamela A. Derish, MA, Department of Surgery, University of California, San Francisco, provided excellent editorial assistance. REFERENCES 1. Hauso O, Gustafsson BI, Kidd M, et al. Neuroendocrine tumor epidemiology: contrasting Norway and North America. Cancer. 2008;113(10):2655-2664. 2. Modlin IM, Lye KD, Kidd M. A 5-decade analysis of 13,715 carcinoid tumors. Cancer. 2003;97(4):934-959. 3. Modlin IM, Moss SF, Chung DC, Jensen RT, Snyderwine E. Priorities for improving the management of gastroenteropancreatic neuroendocrine tumors. J Natl Cancer Inst. 2008;100(18):1282-1289. 4. Modlin IM, Champaneria MC, Chan AK, Kidd M. A three-decade analysis of 3,911 small intestinal neuroendocrine tumors: the rapid pace of no progress. Am J Gastroenterol. 2007;102(7):1464-1473. 5. Makridis C, Oberg K, Juhlin C, et al. Surgical treatment of mid-gut carcinoid tumors. World J Surg. 1990;14(3):377-384. 6. Boudreaux JP, Putty B, Frey DJ, et al. Surgical treatment of advanced-stage carcinoid tumors: lessons learned. Ann Surg. 2005;241(6):839-845, discussion 845-846. 7. Norton JA, Fraker DL, Alexander HR, et al. Surgery increases survival in patients with gastrinoma. Ann Surg. 2006;244(3):410-419. 8. Chen H, Hardacre JM, Uzar A, Cameron JL, Choti MA. Isolated liver metastases from neuroendocrine tumors: does resection prolong survival? J Am Coll Surg. 1998;187(1):88-92, discussion 92-93. 9. Givi B, Pommier SJ, Thompson AK, Diggs BS, Pommier RF. Operative resection of primary carcinoid neoplasms in patients with liver metastases yields significantly better survival. Surgery. 2006;140(6):891-898. 10. Hellman P, Lundström T, Ohrvall U, et al. Effect of surgery on the outcome of midgut carcinoid disease with lymph node and liver metastases. World J Surg. 2002;26(8):991-997. 11. Sugimoto E, Lörelius LE, Eriksson B, Oberg K. Midgut carcinoid tumours. CT appearance. Acta Radiol. 1995;36(4):367-371. 12. Picus D, Glazer HS, Levitt RG, Husband JE. Computed tomography of abdominal carcinoid tumors. AJR Am J Roentgenol. 1984;143(3):581-584. 13. van Tuyl SA, Kuipers EJ, Timmer R, Stolk MF. Video capsule endoscopy: procedure, indications and diagnostic yield. Neth J Med. 2004;62(7):225-228. 14. Johanssen S, Boivin M, Lochs H, Voderholzer W. The yield of wireless capsule endoscopy in the detection of neuroendocrine tumors in comparison with CT enteroclysis. Gastrointest Endosc. 2006;63(4):660-665. 15. Bader TR, Semelka RC, Chiu VC, Armao DM, Woosley JT. MRI of carcinoid tumors: spectrum of appearances in the gastrointestinal tract and liver. J Magn Reson Imaging. 2001;14(3):261-269. 16. Bellutti M, Fry LC, Schmitt J, et al. Detection of neuroendocrine tumors of the small bowel by double balloon enteroscopy. Dig Dis Sci. 2009;54(5):1050-1058. 280