- Email: [email protected]
Predictors of cancer in patients with suspected pancreatic malignancy without a tissue diagnosis
The American Journal of Surgery 191 (2006) 191–197
Scientific papers–American
Predictors of cancer in patients with suspected pancreatic malignancy without a tissue diagnosis David A. Tessler, D.O.a,*, Andrew Catanzaro, M.D.a, Vic Velanovich, M.D.b, Suzanne Havstad, M.A.c, Sachin Goel, M.D.a a
Department of Internal Medicine, Division of Gastroenterology, Henry Ford Hospital, Gastroenterology, K-7, 2799 West Grand Blvd., Detroit, MI 48202, USA b Department of Surgery, Henry Ford Hospital, Detroit, MI, USA c Department of Biostatistics and Research Epidemiology, Henry Ford Hospital, Detroit, MI, USA Manuscript received March 2, 2005; revised manuscript August 2, 2005
Abstract Background: The aim of this study was to identify predictive factors for malignancy in patients undergoing surgery for suspected pancreatic cancer without a preoperative tissue diagnosis. Methods: Patients were identified by International Classification of Diseases Ninth Revision and current procedural terminology codes, respectively, for pancreatic cancer and pancreaticoduodenectomy at a single tertiary referral center between January 1998 and May 2004. Data were collected retrospectively by chart review. Multivariate analysis of potential predictive factors was performed. Results: A total of 150 patients underwent surgery for documented or suspected pancreatic malignancy; 102 did not have a preoperative tissue diagnosis of cancer. Of these, 75 had neoplastic disease at surgery. Average weight loss was greater for those with malignancy (13.5 vs. 4.8 lbs; P ⫽ .014) as was mean bilirubin (6.1 vs. 3.3 mg/dL; P ⫽ .006). In multivariate analysis, a combination of weight loss ⬎20 lbs, bilirubin ⬎3 mg/dL, and CA 19-9 ⬎37 U/mL had both a specificity and positive predictive value of 100% for predicting malignancy regardless of bile duct abnormalities or mass lesions on endoscopic retrograde cholangiopancreatography or endoscopic ultrasound, respectively. The positive predictive value decreased to 89.5% when any 2 of these findings were present. The presence of a mass on CT or EUS alone had a sensitivity of 84%; however, no other single finding had a sensitivity ⬎65%. Conclusions: In patients suspected of having a pancreatic malignancy, weight loss, hyperbilirubinemia, and increased CA 19-9 level may be predictive of a final cancer diagnosis. Surgical exploration should be considered in these patients even in the absence of a preoperative tissue diagnosis. © 2006 Excerpta Medica Inc. All rights reserved. Keywords: Endoscopic retrograde cholangiopancreatography; Endoscopic ultrasound with fine-needle aspiration; Pancreaticoduodenectomy; Pancreatic malignancy
Although pancreatic cancer accounts for only 2% of newly diagnosed malignancies, it is the fifth most common cause of cancer-related death in the United States [1]. This is attributed to the advanced nature of the disease at diagnosis. Although 10% of patients have localized disease at diagnosis, ⬎40% are found to be locally advanced. Furthermore, 80% to 85% of patients are deemed unresectable at presentation [2]. The overall 5-year survival rate for all patients with pancreatic adenocarcinoma is ⬍5% [3]. Resection offers the only chance for long-term survival for patients with * Corresponding author. Tel.: ⫹1-313-916-2408; fax: ⫹1-313-9169128. E-mail address: [email protected]
pancreatic malignancy. Early surgical intervention is critical to improve the chances of prolonged disease-free survival. The insidious nature and vague symptomatology of the early stages of this disease may make establishing the diagnosis difficult. Imaging studies often fail to demonstrate an obvious mass. Obtaining a tissue diagnosis can be problematic because biopsy samples and brushings may only show a suspicion of neoplasia and be unable to definitively document malignancy. For patients clinically suspected to harbor a malignancy without accompanying definitive documentation of disease, the decision to undergo a major operation may be difficult because of the inherent morbidity and mortality of this type of surgery, especially given the possibility of benign pathology. This study sought to iden-
0002-9610/06/$ – see front matter © 2006 Excerpta Medica Inc. All rights reserved. doi:10.1016/j.amjsurg.2005.08.029
192
D.A. Tessler et al. / The American Journal of Surgery 191 (2006) 191–197
Table 1 Demographics Demographics
Group 1 (n ⫽ 48)
Group 2 (n ⫽ 75)
Group 3 (n ⫽ 27)
P
Age in y (mean ⫾ SD) No. male (%) Race (%) White African American Asian Diagnostic procedures (%) EUS Mass ERCP Biliary stricture PTC CAT Mass Surgical procedures (%) Pancreaticoduodenectomy Distal pancreatectomy Other surgery
64.1 ⫾ 11.3 y 24.0 (50.0)
62.1 ⫾ 12.4 y 40 (53.3)
59.8 ⫾ 12.6 y 12 (44.4)
.34 .73 .19
35 (72.9) 13 (27.1) 0 (0)
40 (54.1) 33 (44.6) 1 (1.3)
17 (63.0) 10 (37.0) 0 (0)
41 (85.4) 34 (82.9) 39 (81.2) 25 (64.1) 2 (4.2) 46 (97.9) 28 (58.3)
55 (73.3) 43 (78.2) 50 (66.7) 32 (65.3) 1 (1.3) 74 (98.7) 45 (60.0)
25 (92.6) 11 (44.0) 19 (70.4) 10 (52.6) 1 (3.7) 26 (96.3) 15 (55.6)
.046 .002 .26 .61 .51 .99 .92
29 (60.4) 3 (6.3) 16 (33.3)
49 (65.3) 9 (12.0) 17 (22.7)
22 (81.5) 2 (7.4) 3 (11.1)
.15 .61 .078
CAT ⫽ computerized axial tomography; ERCP ⫽ endoscopic retrograde cholangiopancreatography; EUS ⫽ endoscopic ultrasound; PTC ⫽ percutaneous transhepatic cholangiography.
tify predictive factors for malignancy in patients undergoing surgery for suspected cancer of the pancreaticobiliary system without a preoperative tissue diagnosis.
Finally, follow-up information was reviewed to determine 30-day mortality and cause of death. Definitions
Methods and Materials Study design This is a retrospective cohort study. Admitting and discharge diagnosis and procedure codes were queried from the Corporate Data Store at Henry Ford Hospital. Current procedural terminology (CPT) and International Classification of Diseases Ninth Revision (ICD-9) codes for “pancreaticoduodenectomy” and “pancreatic cancer” were used to identify patients with proven or suspected pancreas neoplasia undergoing surgery. Manual review of patient records was performed to obtain relevant data points. The Institutional Review Board (IRB) approved the study. Patient characteristics, history, and demographic data were collected for each patient. Where available, reports were obtained for all pertinent endoscopic procedures performed either at our institution or at outside institutions. Reports of imaging studies—including computed axial tomography (CAT) scans, magnetic resonance imaging (MRI), and magnetic resonance cholangiopancreatography (MRCP)— were evaluated. Laboratory studies, including serum bilirubin and CA 19-9, were collected. Weight loss was estimated by patient recollection or from previous office notes where available. Data were also collected about the surgical procedure performed, pathologic diagnosis, complications, and further treatments (eg, chemotherapy and radiation therapy).
Proven pancreatic malignancy was defined as definitive preoperative tissue diagnosis of neoplasm. Suspected pancreatic malignancy was defined as preoperative tissue diagnosis of “suspicious for carcinoma,” “atypical cells,” “benign tissue” or “nondiagnostic.” Statistical analysis Simple descriptive statistics were used. Means and frequencies were calculated based on the numerous data
Fig. 1. Findings on evaluation.
D.A. Tessler et al. / The American Journal of Surgery 191 (2006) 191–197
193
Table 2 Variables significantly correlated with finding neoplasia in patients with suspected pancreas malignancy Variable
Group 1 (n ⫽ 48)
Group 2 (n ⫽ 75)
Group 3 (n ⫽ 27)
P
Median weight loss in lb (range) Median bilirubin in mg/dL (range) Median CA 19-9 in U/mL (range)
15.0 (0–70) 3.7 (0.2–24.2) 22.6 (0–6972)
10.0 (0–60) 2.9 (0.2–29.7) 58.2 (0–5800)
0.0 (0–35) 0.5 (0.2–17.2) 24.7 (1–324)
.006 .007 .15
points. The P values were provided to indicate statistical significance. Chi-square tests were used to compare categoric variables such as sex, and Student t test was used to compare mean age between the groups. Weight loss, bilirubin, and CA19-9 were compared between the 2 groups using Wilcoxon rank sum and Kruskal-Wallis tests. Weight loss, bilirubin, and CA19-9 were also evaluated for predictive purposes of a final cancer diagnosis. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated using cut-offs of 37 U/mL for CA 19-9 (upper limit of normal in our laboratory) and 3 mg/dL for bilirubin (representing clinical jaundice). Furthermore, values were recalculated with the addition of either biliary structuring or obstruction by ERCP or the finding of a mass on CAT or EUS. All statistical analyses were performed using SAS software (Cary, North Carolina).
Results A total of 150 patients underwent surgery for proven or suspected pancreatic malignancy at Henry Ford Hospital between January 1998 and May 2004; 48 patients had a proven malignancy before surgery (group 1), and 102 patients underwent surgery for suspected pancreas malignancy. Of these 102 patients, 75 were found to have neoplasia at surgery (group 2), whereas 27 had benign disease (group 3). The characteristics of the 3 groups can be seen in Table 1. EUS was performed in 80 of the 102 patients with suspected pancreatic malignancy (Fig. 1). The total number
of EUS-fine needle aspiration (FNA) performed was 41. Masses were seen in 54 patients, and 33 of these underwent EUS-FNA. EUS-FNA was not performed in 21 patients with a mass lesion for the following reasons: previous decision to take the patient to surgery regardless of the pathologic findings (n ⫽ 3), technical reasons (n ⫽ 8), forceps biopsy (n ⫽ 1), pre-existing suspicion for pneumoperitoneum (n ⫽ 1), perforation during EUS (n ⫽ 1), and unknown reasons (n ⫽ 7). The remaining 8 EUS-FNAs were performed in patients suspected to have cancer where no obvious mass was seen. ERCP was performed in 68 of the 102 patients with suspected pancreas malignancy (Fig. 1). Forty-two patients had either biliary stricture or obstruction, whereas an additional 13 had dilated extrahepatic bile ducts without an obvious stricture. Similarly, 16 patients had either stenosis or obstruction of the pancreatic duct, and an additional 7 demonstrated a dilated duct without obvious stricture. Of the 102 patients, 100 underwent CAT scan. The remaining 2 patients were seen for a second opinion from an outside institution, and no records were available regarding the findings from the CAT scans performed there. Surgical procedures performed in patients with suspected pancreas malignancy (n ⫽ 102) were pancreaticoduodenectomy (n ⫽ 71), palliative bypass (n ⫽ 19), distal pancreatectomy (n ⫽ 11), and pancreaticojejunostomy (n ⫽ 1). One surgeon performed 79 (77%) of the procedures, and another performed 7 (7%) of them. The remaining 16 surgeries were performed by 10 different surgeons; no surgeon performed ⬎3 of procedures.
Table 3 Predictive factors for finding neoplasia in patients with suspected pancreas malignancy Factors
Sensitivity (%)
Wt loss ⬎20 lb or bili ⬎3 mg/dL or CA 19-9 ⬎37 U/mL Wt loss ⬎20 lb & bili ⬎3 mg/dL & CA 19-9 ⬎37 U/mL Wt loss ⬎20 lb & bili ⬎3 mg/dL & CA 19-9 ⬎37 U/mL & biliary stricture Wt loss ⬎20 lb & bili ⬎3 mg/dL & CA 19-9 ⬎37 U/mL & mass on CAT/EUS
90.5
Specificity (%) 40.0
PPV (%) 82.6
NPV (%) 57.1
45.4
100
100
57.1
66.7
100
100
66.7
83.3
100
100
66.7
bili ⫽ bilirubin; CAT ⫽ computerized axial tomography; EUS ⫽ endoscopic ultrasound; NPV ⫽ negative predictive value; PPV ⫽ positive predictive value; Wt ⫽ weight.
194
D.A. Tessler et al. / The American Journal of Surgery 191 (2006) 191–197
Table 4 Predictive factors for finding neoplasia in patients with suspected pancreas malignancy after exclusion of those with cytology read as “suspicious for carcinoma” (n ⫽ 24) Factors Wt loss ⬎20 lb or bili ⬎3 mg/dL or CA 19-9 ⬎37 U/mL Wt loss ⬎20 lb & bili ⬎3 mg/dL & CA 19-9 ⬎37 U/mL Wt loss ⬎20 lb & bili ⬎3 mg/dL & CA 19-9 ⬎37 U/mL & biliary stricture Wt loss ⬎20 lb & bili ⬎3 mg/dL & CA 19-9 ⬎37 U/mL & mass on CAT/EUS
Sensitivity (%) 93.2
Specificity (%)
PPV (%)
29.4
NPV (%)
77.4
62.5
50.0.0
100
100
62.5
75.0.0
100
100
66.7
100
100
100
100
CAT ⫽ computerized axial tomography; ERCP ⫽ endoscopic retrograde cholangiopancreatography; EUS ⫽ endoscopic ultrasound; PTC ⫽ percutaneous transhepatic cholangiography.
In group 2 (n ⫽ 75), surgical pathology demonstrated pancreatic adenocarcinoma in 59 patients. TNM staging was available for 49 patients (stage I ⫽ 7, stage II ⫽ 15, stage III ⫽ 16, and stage IV ⫽ 11). A resected specimen for final staging was not available in 10 patients because of metastatic disease (n ⫽ 6), vascular encasement (n ⫽ 3), and limited resectability (n ⫽ 1). The other 16 neoplastic lesions were made up of mucinous neoplasms (n ⫽ 10), tubulovillous adenoma of the ampulla (n ⫽ 2), neuroendocrine tumor (n ⫽ 2), carcinoid (n ⫽ 1) and 1 unresectable cancer for which the biopsy specimens showed only the suspicion of a malignancy at surgery, but the diagnosis was later confirmed on further biopsy. Surgical pathology in group 3 (n ⫽ 27) revealed chronic pancreatitis (n ⫽ 19) and inflammatory changes of the pancreas (n ⫽ 8). The difference for weight loss and bilirubin between groups 1, 2, and 3 was statistically significant (Table 2). Specific P values for pairwise testing of the groups showed no statistical significance between groups 1 and 2 for either bilirubin (P ⫽ .76) or weight loss (P ⫽ .39). The differences, however, comparing either group 1 or group 2 with group 3 did show statistical significance for both variables. The median CA 19-9 was different, but this did not meet statistical significance. Evaluation of other variables, including age, sex, and race, failed to yield statistically significant results. Multivariate analysis demonstrates that weight loss (⬎20 lbs.) combined with bilirubin ⬎3 mg/dL and an increased CA 19-9 (⬎100 U/mL) had a 100% specificity and PPV for malignancy regardless of the presence of biliary stricture or obstruction on ERCP or the finding of a mass on either EUS or CAT (Table 3). The sensitivity and NPV were lower but improved when combined with either biliary stricture or mass. Of the 102 patients with suspected pancreas malignancy, 24 patients had a preoperative tissue diagnosis that was read as “suspicious for carcinoma.” Multivariate analysis was repeated after excluding these 24 patients with similar results (Table 4). Surgical morbidity was similar for groups 1, 2, and 3 (Table 5). There was no statistically significant difference
for TNM status between groups 1 and 2 (Table 6). The 30-day mortality for group 1 was 6.2% compared with 4% for group 2 and 3.7% for group 3. Additionally, there was no statistical difference in 30-day mortality between the 3 groups (P ⫽ .83). The reasons for early (⬍30-day) postoperative mortality were postoperative hemorrhage (n ⫽ 2), myocardial infarction (n ⫽ 2), septic shock (n ⫽ 1), and multisystem organ failure (n ⫽ 1). The cause of death for 1 other patient was unknown because he was found dead at home.
Comments Pancreatic cancer has a dismal prognosis without treatment. Surgical resection offers the only hope for long-term survival and even then there is only a 10% to 15% chance of surviving 5 years [4,5]. Survival, however, is increased in those patients undergoing resection compared with those not undergoing surgery regardless of stage [6,7]. Although Table 5 Surgical Morbidity Complication
Group 1 % (n ⫽ 48)
Group 2 % (n ⫽ 75)
Group 3 % (n ⫽ 27)
P
Leak Hernia Dehiscence Pancreatic insufficiency Bleeding Abscess DM Ileus Thrombosis CVA MI SBO Fistula Gastroparesis
6 (12.5) 7 (14.6) 1 (2.1) 17 (35.4) 10 (20.8) 8 (16.7) 1 (2.1) 3 (6.2) 7 (14.6) 2 (4.2) 5 (10.6) 2 (4.2) 0 (0) 3 (6.2)
4 (5.3) 11 (14.7) 5 (6.7) 22 (29.3) 9 (12.0) 14 (18.7) 2 (2.7) 1 (1.3) 11 (14.7) 2 (2.7) 1 (1.3) 3 (4.0) 2 (2.7) 1 (1.3)
2 (7.4) 1 (3.7) 1 (3.7) 5 (18.5) 2 (7.4) 1 (3.7) 0 (0) 0 (0) 1 (3.7) 0 (0) 0 (0) 1 (3.7) 1 (3.7) 0 (0)
.34 .22 .68 .29 .22 .10 .90 .31 .22 .66 .032 .99 .43 .31
CVA ⫽ cerebrovascular accident; DM ; MI ⫽ myocardial infarction; SBO.
D.A. Tessler et al. / The American Journal of Surgery 191 (2006) 191–197 Table 6 TNM status Status
Group 1 % (n ⫽ 48)
Group 2 % (n ⫽ 75)
P
T1 T2 T3 T4 N0 N1 N2 M0 M1
3 (8.3) 10 (27.8) 23 (63.9) 0 (0) 24 (68.6) 10 (28.6) 1 (2.9) 34 (85.0) 6 (15.0)
4 (8.0) 10 (20.0) 33 (66.0) 3 (6.0) 25 (49.0) 26 (51.0) 0 (0) 48 (84.2) 9 (15.8)
.53
.056
.92
resection offers the only chance for long-term survival, median survival after surgery is still only 15 to 19 months [1]. This is unfortunate because as many as 88% of patients with pancreatic cancer are unresectable at diagnosis because of local extension or metastases. The 1995 National Cancer Data Base Report on Pancreatic Cancer reported that 52% of patients with newly diagnosed pancreatic cancer had stage IV disease [6]. Because of the need for an early diagnosis, resection may be considered justifiable in the absence of a preoperative tissue diagnosis [3]. Our study attempted to define whether any factors may aid in differentiating those with malignancy from those without malignancy in patients for whom a preoperative tissue diagnosis is elusive. At our institution, those with a proven pancreatic malignancy are offered enrollment in clinical trials using preoperative chemoradiation in an effort to improve surgical outcomes. Therefore, an aggressive attempt is made to obtain a tissue diagnosis. Failure to document a tissue diagnosis makes a patient ineligible for preoperative chemoradiation. Because preoperative chemoradiation means an inherent delay in surgical intervention, an attempt is often made to decompress the biliary tree during the initial ERCP, especially if the patient is symptomatic, whether or not a diagnosis is eventually obtained; this explains the high number of ERCPs performed in this study. In the absence of a tissue-documented diagnosis, clinical parameters may help guide appropriate patients to the operating room when necessary. The present study demonstrates that jaundice, weight loss, and increased CA 19-9 are predictive of malignancy. Finding a biliary stricture or pancreatic mass added nothing to the specificity, PPV, or NPV for the diagnosis; however, those findings did increase the sensitivity. It is possible that including patients with cytology interpreted as “suspicious for carcinoma” may have enriched the suspected pancreas neoplasm group with patients more likely to have cancer. The multivariate analysis performed after excluding the patients with suspicious cytology, however, generated similar results for sensitivity, specificity, PPV, and NPV. Exclusion of those with distal pancreatic lesions likewise showed similar results despite the fact that patients with
195
distal lesions would not typically present with obstructive jaundice. Re-evaluating the data and excluding patients with metastatic disease did not appreciably change sensitivity or specificity, nor did it diminish our ability to predict those with and without malignancy, thus showing that the advanced nature of the disease does not always correlate with the clinical presentation (data not shown). The previously mentioned criteria were subsequently applied to group 1. Statistical analysis showed a sensitivity of 90.9% when any 1 of the 3 criteria was present. Unfortunately, few patients in this cohort had all 3 criteria, thus limiting adequate statistical evaluation. Thirty-day operative mortality rates for pancreaticoduodenectomy are ⬍5% in high-volume centers, and the majority of deaths are caused by hemorrhage, infection, myocardial infarction, and multisystem organ failure [8]. In our study, the early (⬍30-day) mortality was 4.6%, which is comparable with the published literature. Of all surgeries performed in this study (n ⫽ 150), 1 surgeon performed 123 (82%) of the procedures, and another performed 9 (6%). The remaining 18 surgeries were performed by 10 different surgeons; no surgeon performed ⬎3 of the procedures. Considering all 150 procedures performed, 67% of patients underwent a pancreaticoduodenectomy; 9% underwent distal pancreatectomy, 16% underwent palliative bypass, 7% underwent laparoscopy or laparotomy only, and ⬍1% underwent pancreaticojejunostomy. Other series looking at all procedures performed for pancreatic cancer showed a similar breakdown [9]. Schäfer et al showed no apparent difference in postoperative mortality between pancreaticoduodenectomy and distal pancreatectomy [10]; thus, our mortality data is comparable with other studies that included only cases of pancreatic adenocarcinoma. Furthermore, morbidity and mortality data did not differ by procedure volume of the surgeon in this study. Thompson et al looked at 67 patients undergoing pancreaticoduodenectomy in the pre-EUS era and found a 30% incidence of surgery for presumed, but undocumented, malignancy despite an adequate evaluation including intraoperative biopsies [11]. Of those 20 patients, 9 (45%) were shown after surgery to have malignancy, whereas the remainder had benign disease. It is harder to make the recommendation that all suspicious lesions be resected when 55% of patients underwent surgery for benign disease. The present study, using a larger cohort of patients, demonstrated a much higher rate of neoplasia (74%) in the group with suspected pancreatic malignancy. This may be because all neoplastic lesions were considered, whereas Thompson et al limited their study to cases of adenocarcinoma only. Their results also differ in terms of the disease stage because a much lower rate of T3 disease was found. Additionally, they found a higher proportion of stage I lesions in both the proven pancreatic malignancy group (54%) and suspected pancreatic malignancy group (75%) compared with the findings in this study of 30.5% and 14.2%, respectively, for the 2 groups.
196
D.A. Tessler et al. / The American Journal of Surgery 191 (2006) 191–197
Twenty-seven patients with benign disease underwent surgery for presumed malignancy in this study. Of these, 7 had surgery primarily for symptomatic chronic pancreatitis. For those patients, the possibility of malignancy was listed as the secondary indication for surgery on the operative note. Excluding those patients, benign disease was found in 13% of patients undergoing surgery for presumed malignancy. Other studies, however, have documented surgery for clinical misdiagnosis at 3% to 11% [3,12,13]. Barnes et al reviewed 510 pancreaticoduodenectomies performed at The Johns Hopkins Hospital during an 8-year period. One hundred eight patients (21%) underwent surgery for benign disease, 83 (16%) of those for suspicion of malignancy [14]. Although they recommend an aggressive surgical approach for patients with a clinical suspicion of malignancy, no predictive factors were offered to aid in this decision making. Van Gulik et al believed that no one single test or imaging modality adequately differentiated pancreatic cancer from focal pancreatitis [15]. They described 22 patients with chronic pancreatitis harboring an initially undiagnosed malignancy and 14 inflammatory lesions initially presumed to be malignant at a “center specialized in the diagnosis and treatment of pancreatic cancer” [15]. Twelve of the 14 patients ultimately proven to have benign disease were found to have a mass, whereas biliary ductal or pancreatic ductal stenosis was found in 9 of 13 and 10 of 11 of evaluated patients, respectively. In contrast, an extremely high mortality rate for those with cancer was seen because 20 of the 22 malignancies initially classified as inflammatory were unresectable. This suggests that a more aggressive surgical approach may be warranted if a preoperative diagnosis is uncertain. Other studies have restricted the study protocol to include only pancreatic adenocarcinoma. The present study purposely included all patients with suspected malignancy of the pancreas. This allowed for a larger base of patients on which recommendations could be made because there are other neoplasms of the pancreas, other than adenocarcinoma, for which surgical therapy is appropriate. Subgroup analysis was performed looking only at those with adenocarcinoma and again showed comparable figures for sensitivity (data not shown). Other strengths of this study include the large number of patients, objective end points of the study, and the inclusion only of patients deemed surgical candidates for better comparability. Additionally, this study was able to suggest predictive factors to aid in the differentiation of benign and neoplastic disease. A limitation of the present study was the retrospective design. First and foremost, not all patients had documentation of a preoperative CA 19-9 or adequate evaluation for weight loss even after thorough search through the electronic database. The preoperative workup was not uniform for all patients, thus restricting the data points from which stronger recommendations could be ex-
tracted. Preoperative evaluation differed between medical and surgical services, as well as among individual physicians, often because of personal preference given the lack of a universally accepted algorithm for the workup of patients suspected to have a pancreatic neoplasm at our institution. Additionally, the initial time points of this study coincided with early experience with EUS at our institution, which may explain the lack of diagnostic tissue and failure to perform EUS-FNA in certain patients. Furthermore, another limitation is that as a tertiary referral center, a number of the patients, for whom preoperative as well as follow-up data were not available, came as outside referrals for workup and surgical resection. Mortality for pancreatic surgery has dramatically decreased during the past half century [16]. Concern of procedural morbidity and mortality for disease ultimately proven to be benign should not delay an attempt at surgical resection in cases of suspected pancreatic malignancy. In experienced hands, outcomes for those undergoing surgery without a documented tissue diagnosis are similar to those known to have neoplasia before surgery. In the appropriate clinical setting, a patient with weight loss, jaundice, and increased CA 19-9, often in combination with a biliary stricture or pancreatic mass, should be strongly considered for surgery even if a preoperative tissue diagnosis is not present. Although these criteria were highly suggestive of malignancy in these selected patients, this model needs further study in a prospective fashion before it can be adopted for widespread use. References [1] Hawes RH, Xiong Q, Waxman I, et al. A multispecialty approach to the diagnosis and management of pancreatic cancer. Am J Gastro 2000;95:17–31. [2] Duffy JP, Reber HA. Pancreatic neoplasms. Curr Opin Gastroenterol 2003;19:458 – 66. [3] Abraham SC, Wilentz RE, Yeo CJ, et al. Pancreaticoduodenectomy (Whipple resections) in patients without malignancy: Are they all “chronic pancreatitis”? Am J Surg Pathol 2003;27:110 –20. [4] DiMagno EP. Pancreatic cancer: Clinical presentation, pitfalls and early clues. Ann Oncol 1999;10(suppl):S140 –S142. [5] Yeo CJ, Sohn TA, Cameron JL, et al. Periampullary adenocarcinoma: Analysis of 5-year survivors. Ann Surg 1998;227:821–31. [6] Dimagno EP, Reber HA, Tempero MA. AGA technical review on the epidemiology, diagnosis and treatment of pancreatic ductal adenocarcinoma. Gastroenterology 1999;117:1464 – 84. [7] Carr JA, Ajlouni M, Wollner I, et al. Adenocarcinoma of the head of the pancreas: Effects of surgical and nonsurgical therapy on survival—A ten-year experience. Am Surg 1999;65:1143–9. [8] Ahmad NA, Lewis JD, Ginsberg GG, et al. Long-term survival after pancreatic resection for pancreatic adenocarcinoma. Am J Gastro 2001;96:2609 –15. [9] Wagner M, Dikopoulos N, Kulli C, et al. Standard surgical treatment in pancreatic cancer. Ann Oncol 1999;10(suppl):S247–S251. [10] Schäfer M, Müllhaupt B, Clavien PA. Evidence-based pancreatic head resection for pancreatic cancer and chronic pancreatitis. Ann Surg 2002;236:137– 48.
D.A. Tessler et al. / The American Journal of Surgery 191 (2006) 191–197 [11] Thompson JS, Murayama KM, Edney JA, et al. Pancreaticoduodenectomy for suspected but unproven malignancy. Am J Surg 1994;169:571–5. [12] Van Gulik TM, Reeders JWA, Bosma A, et al. Incidence and clinical findings of benign, inflammatory disease in patients resected for presumed pancreatic head cancer. Gastro Endosc 1997;46:417–23. [13] Smith CD, Behrns KE, van Heerden JA, et al. Radical pancreaticoduodenectomy for misdiagnosed pancreatic mass. Br J Surg 1994; 81:585–9.
197
[14] Barnes SA, Lilliemoe KD, Kaufman HS, et al. Pancreaticoduodenectomy for benign disease. Am J Surg 1996;171:131–5. [15] van Gulik TM, Moojen TM, van Geenen R, et al. Differential diagnosis of focal pancreatitis and pancreatic cancer. Ann Oncol 1999; 10(suppl):S85–S88. [16] Rios G, Conrad A, Cole D, et al. Trends in indications and outcomes in the Whipple procedure over a 40-year period. Am Surg 1999;9: 889 –93.
Scientific papers–American
Predictors of cancer in patients with suspected pancreatic malignancy without a tissue diagnosis David A. Tessler, D.O.a,*, Andrew Catanzaro, M.D.a, Vic Velanovich, M.D.b, Suzanne Havstad, M.A.c, Sachin Goel, M.D.a a
Department of Internal Medicine, Division of Gastroenterology, Henry Ford Hospital, Gastroenterology, K-7, 2799 West Grand Blvd., Detroit, MI 48202, USA b Department of Surgery, Henry Ford Hospital, Detroit, MI, USA c Department of Biostatistics and Research Epidemiology, Henry Ford Hospital, Detroit, MI, USA Manuscript received March 2, 2005; revised manuscript August 2, 2005
Abstract Background: The aim of this study was to identify predictive factors for malignancy in patients undergoing surgery for suspected pancreatic cancer without a preoperative tissue diagnosis. Methods: Patients were identified by International Classification of Diseases Ninth Revision and current procedural terminology codes, respectively, for pancreatic cancer and pancreaticoduodenectomy at a single tertiary referral center between January 1998 and May 2004. Data were collected retrospectively by chart review. Multivariate analysis of potential predictive factors was performed. Results: A total of 150 patients underwent surgery for documented or suspected pancreatic malignancy; 102 did not have a preoperative tissue diagnosis of cancer. Of these, 75 had neoplastic disease at surgery. Average weight loss was greater for those with malignancy (13.5 vs. 4.8 lbs; P ⫽ .014) as was mean bilirubin (6.1 vs. 3.3 mg/dL; P ⫽ .006). In multivariate analysis, a combination of weight loss ⬎20 lbs, bilirubin ⬎3 mg/dL, and CA 19-9 ⬎37 U/mL had both a specificity and positive predictive value of 100% for predicting malignancy regardless of bile duct abnormalities or mass lesions on endoscopic retrograde cholangiopancreatography or endoscopic ultrasound, respectively. The positive predictive value decreased to 89.5% when any 2 of these findings were present. The presence of a mass on CT or EUS alone had a sensitivity of 84%; however, no other single finding had a sensitivity ⬎65%. Conclusions: In patients suspected of having a pancreatic malignancy, weight loss, hyperbilirubinemia, and increased CA 19-9 level may be predictive of a final cancer diagnosis. Surgical exploration should be considered in these patients even in the absence of a preoperative tissue diagnosis. © 2006 Excerpta Medica Inc. All rights reserved. Keywords: Endoscopic retrograde cholangiopancreatography; Endoscopic ultrasound with fine-needle aspiration; Pancreaticoduodenectomy; Pancreatic malignancy
Although pancreatic cancer accounts for only 2% of newly diagnosed malignancies, it is the fifth most common cause of cancer-related death in the United States [1]. This is attributed to the advanced nature of the disease at diagnosis. Although 10% of patients have localized disease at diagnosis, ⬎40% are found to be locally advanced. Furthermore, 80% to 85% of patients are deemed unresectable at presentation [2]. The overall 5-year survival rate for all patients with pancreatic adenocarcinoma is ⬍5% [3]. Resection offers the only chance for long-term survival for patients with * Corresponding author. Tel.: ⫹1-313-916-2408; fax: ⫹1-313-9169128. E-mail address: [email protected]
pancreatic malignancy. Early surgical intervention is critical to improve the chances of prolonged disease-free survival. The insidious nature and vague symptomatology of the early stages of this disease may make establishing the diagnosis difficult. Imaging studies often fail to demonstrate an obvious mass. Obtaining a tissue diagnosis can be problematic because biopsy samples and brushings may only show a suspicion of neoplasia and be unable to definitively document malignancy. For patients clinically suspected to harbor a malignancy without accompanying definitive documentation of disease, the decision to undergo a major operation may be difficult because of the inherent morbidity and mortality of this type of surgery, especially given the possibility of benign pathology. This study sought to iden-
0002-9610/06/$ – see front matter © 2006 Excerpta Medica Inc. All rights reserved. doi:10.1016/j.amjsurg.2005.08.029
192
D.A. Tessler et al. / The American Journal of Surgery 191 (2006) 191–197
Table 1 Demographics Demographics
Group 1 (n ⫽ 48)
Group 2 (n ⫽ 75)
Group 3 (n ⫽ 27)
P
Age in y (mean ⫾ SD) No. male (%) Race (%) White African American Asian Diagnostic procedures (%) EUS Mass ERCP Biliary stricture PTC CAT Mass Surgical procedures (%) Pancreaticoduodenectomy Distal pancreatectomy Other surgery
64.1 ⫾ 11.3 y 24.0 (50.0)
62.1 ⫾ 12.4 y 40 (53.3)
59.8 ⫾ 12.6 y 12 (44.4)
.34 .73 .19
35 (72.9) 13 (27.1) 0 (0)
40 (54.1) 33 (44.6) 1 (1.3)
17 (63.0) 10 (37.0) 0 (0)
41 (85.4) 34 (82.9) 39 (81.2) 25 (64.1) 2 (4.2) 46 (97.9) 28 (58.3)
55 (73.3) 43 (78.2) 50 (66.7) 32 (65.3) 1 (1.3) 74 (98.7) 45 (60.0)
25 (92.6) 11 (44.0) 19 (70.4) 10 (52.6) 1 (3.7) 26 (96.3) 15 (55.6)
.046 .002 .26 .61 .51 .99 .92
29 (60.4) 3 (6.3) 16 (33.3)
49 (65.3) 9 (12.0) 17 (22.7)
22 (81.5) 2 (7.4) 3 (11.1)
.15 .61 .078
CAT ⫽ computerized axial tomography; ERCP ⫽ endoscopic retrograde cholangiopancreatography; EUS ⫽ endoscopic ultrasound; PTC ⫽ percutaneous transhepatic cholangiography.
tify predictive factors for malignancy in patients undergoing surgery for suspected cancer of the pancreaticobiliary system without a preoperative tissue diagnosis.
Finally, follow-up information was reviewed to determine 30-day mortality and cause of death. Definitions
Methods and Materials Study design This is a retrospective cohort study. Admitting and discharge diagnosis and procedure codes were queried from the Corporate Data Store at Henry Ford Hospital. Current procedural terminology (CPT) and International Classification of Diseases Ninth Revision (ICD-9) codes for “pancreaticoduodenectomy” and “pancreatic cancer” were used to identify patients with proven or suspected pancreas neoplasia undergoing surgery. Manual review of patient records was performed to obtain relevant data points. The Institutional Review Board (IRB) approved the study. Patient characteristics, history, and demographic data were collected for each patient. Where available, reports were obtained for all pertinent endoscopic procedures performed either at our institution or at outside institutions. Reports of imaging studies—including computed axial tomography (CAT) scans, magnetic resonance imaging (MRI), and magnetic resonance cholangiopancreatography (MRCP)— were evaluated. Laboratory studies, including serum bilirubin and CA 19-9, were collected. Weight loss was estimated by patient recollection or from previous office notes where available. Data were also collected about the surgical procedure performed, pathologic diagnosis, complications, and further treatments (eg, chemotherapy and radiation therapy).
Proven pancreatic malignancy was defined as definitive preoperative tissue diagnosis of neoplasm. Suspected pancreatic malignancy was defined as preoperative tissue diagnosis of “suspicious for carcinoma,” “atypical cells,” “benign tissue” or “nondiagnostic.” Statistical analysis Simple descriptive statistics were used. Means and frequencies were calculated based on the numerous data
Fig. 1. Findings on evaluation.
D.A. Tessler et al. / The American Journal of Surgery 191 (2006) 191–197
193
Table 2 Variables significantly correlated with finding neoplasia in patients with suspected pancreas malignancy Variable
Group 1 (n ⫽ 48)
Group 2 (n ⫽ 75)
Group 3 (n ⫽ 27)
P
Median weight loss in lb (range) Median bilirubin in mg/dL (range) Median CA 19-9 in U/mL (range)
15.0 (0–70) 3.7 (0.2–24.2) 22.6 (0–6972)
10.0 (0–60) 2.9 (0.2–29.7) 58.2 (0–5800)
0.0 (0–35) 0.5 (0.2–17.2) 24.7 (1–324)
.006 .007 .15
points. The P values were provided to indicate statistical significance. Chi-square tests were used to compare categoric variables such as sex, and Student t test was used to compare mean age between the groups. Weight loss, bilirubin, and CA19-9 were compared between the 2 groups using Wilcoxon rank sum and Kruskal-Wallis tests. Weight loss, bilirubin, and CA19-9 were also evaluated for predictive purposes of a final cancer diagnosis. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated using cut-offs of 37 U/mL for CA 19-9 (upper limit of normal in our laboratory) and 3 mg/dL for bilirubin (representing clinical jaundice). Furthermore, values were recalculated with the addition of either biliary structuring or obstruction by ERCP or the finding of a mass on CAT or EUS. All statistical analyses were performed using SAS software (Cary, North Carolina).
Results A total of 150 patients underwent surgery for proven or suspected pancreatic malignancy at Henry Ford Hospital between January 1998 and May 2004; 48 patients had a proven malignancy before surgery (group 1), and 102 patients underwent surgery for suspected pancreas malignancy. Of these 102 patients, 75 were found to have neoplasia at surgery (group 2), whereas 27 had benign disease (group 3). The characteristics of the 3 groups can be seen in Table 1. EUS was performed in 80 of the 102 patients with suspected pancreatic malignancy (Fig. 1). The total number
of EUS-fine needle aspiration (FNA) performed was 41. Masses were seen in 54 patients, and 33 of these underwent EUS-FNA. EUS-FNA was not performed in 21 patients with a mass lesion for the following reasons: previous decision to take the patient to surgery regardless of the pathologic findings (n ⫽ 3), technical reasons (n ⫽ 8), forceps biopsy (n ⫽ 1), pre-existing suspicion for pneumoperitoneum (n ⫽ 1), perforation during EUS (n ⫽ 1), and unknown reasons (n ⫽ 7). The remaining 8 EUS-FNAs were performed in patients suspected to have cancer where no obvious mass was seen. ERCP was performed in 68 of the 102 patients with suspected pancreas malignancy (Fig. 1). Forty-two patients had either biliary stricture or obstruction, whereas an additional 13 had dilated extrahepatic bile ducts without an obvious stricture. Similarly, 16 patients had either stenosis or obstruction of the pancreatic duct, and an additional 7 demonstrated a dilated duct without obvious stricture. Of the 102 patients, 100 underwent CAT scan. The remaining 2 patients were seen for a second opinion from an outside institution, and no records were available regarding the findings from the CAT scans performed there. Surgical procedures performed in patients with suspected pancreas malignancy (n ⫽ 102) were pancreaticoduodenectomy (n ⫽ 71), palliative bypass (n ⫽ 19), distal pancreatectomy (n ⫽ 11), and pancreaticojejunostomy (n ⫽ 1). One surgeon performed 79 (77%) of the procedures, and another performed 7 (7%) of them. The remaining 16 surgeries were performed by 10 different surgeons; no surgeon performed ⬎3 of procedures.
Table 3 Predictive factors for finding neoplasia in patients with suspected pancreas malignancy Factors
Sensitivity (%)
Wt loss ⬎20 lb or bili ⬎3 mg/dL or CA 19-9 ⬎37 U/mL Wt loss ⬎20 lb & bili ⬎3 mg/dL & CA 19-9 ⬎37 U/mL Wt loss ⬎20 lb & bili ⬎3 mg/dL & CA 19-9 ⬎37 U/mL & biliary stricture Wt loss ⬎20 lb & bili ⬎3 mg/dL & CA 19-9 ⬎37 U/mL & mass on CAT/EUS
90.5
Specificity (%) 40.0
PPV (%) 82.6
NPV (%) 57.1
45.4
100
100
57.1
66.7
100
100
66.7
83.3
100
100
66.7
bili ⫽ bilirubin; CAT ⫽ computerized axial tomography; EUS ⫽ endoscopic ultrasound; NPV ⫽ negative predictive value; PPV ⫽ positive predictive value; Wt ⫽ weight.
194
D.A. Tessler et al. / The American Journal of Surgery 191 (2006) 191–197
Table 4 Predictive factors for finding neoplasia in patients with suspected pancreas malignancy after exclusion of those with cytology read as “suspicious for carcinoma” (n ⫽ 24) Factors Wt loss ⬎20 lb or bili ⬎3 mg/dL or CA 19-9 ⬎37 U/mL Wt loss ⬎20 lb & bili ⬎3 mg/dL & CA 19-9 ⬎37 U/mL Wt loss ⬎20 lb & bili ⬎3 mg/dL & CA 19-9 ⬎37 U/mL & biliary stricture Wt loss ⬎20 lb & bili ⬎3 mg/dL & CA 19-9 ⬎37 U/mL & mass on CAT/EUS
Sensitivity (%) 93.2
Specificity (%)
PPV (%)
29.4
NPV (%)
77.4
62.5
50.0.0
100
100
62.5
75.0.0
100
100
66.7
100
100
100
100
CAT ⫽ computerized axial tomography; ERCP ⫽ endoscopic retrograde cholangiopancreatography; EUS ⫽ endoscopic ultrasound; PTC ⫽ percutaneous transhepatic cholangiography.
In group 2 (n ⫽ 75), surgical pathology demonstrated pancreatic adenocarcinoma in 59 patients. TNM staging was available for 49 patients (stage I ⫽ 7, stage II ⫽ 15, stage III ⫽ 16, and stage IV ⫽ 11). A resected specimen for final staging was not available in 10 patients because of metastatic disease (n ⫽ 6), vascular encasement (n ⫽ 3), and limited resectability (n ⫽ 1). The other 16 neoplastic lesions were made up of mucinous neoplasms (n ⫽ 10), tubulovillous adenoma of the ampulla (n ⫽ 2), neuroendocrine tumor (n ⫽ 2), carcinoid (n ⫽ 1) and 1 unresectable cancer for which the biopsy specimens showed only the suspicion of a malignancy at surgery, but the diagnosis was later confirmed on further biopsy. Surgical pathology in group 3 (n ⫽ 27) revealed chronic pancreatitis (n ⫽ 19) and inflammatory changes of the pancreas (n ⫽ 8). The difference for weight loss and bilirubin between groups 1, 2, and 3 was statistically significant (Table 2). Specific P values for pairwise testing of the groups showed no statistical significance between groups 1 and 2 for either bilirubin (P ⫽ .76) or weight loss (P ⫽ .39). The differences, however, comparing either group 1 or group 2 with group 3 did show statistical significance for both variables. The median CA 19-9 was different, but this did not meet statistical significance. Evaluation of other variables, including age, sex, and race, failed to yield statistically significant results. Multivariate analysis demonstrates that weight loss (⬎20 lbs.) combined with bilirubin ⬎3 mg/dL and an increased CA 19-9 (⬎100 U/mL) had a 100% specificity and PPV for malignancy regardless of the presence of biliary stricture or obstruction on ERCP or the finding of a mass on either EUS or CAT (Table 3). The sensitivity and NPV were lower but improved when combined with either biliary stricture or mass. Of the 102 patients with suspected pancreas malignancy, 24 patients had a preoperative tissue diagnosis that was read as “suspicious for carcinoma.” Multivariate analysis was repeated after excluding these 24 patients with similar results (Table 4). Surgical morbidity was similar for groups 1, 2, and 3 (Table 5). There was no statistically significant difference
for TNM status between groups 1 and 2 (Table 6). The 30-day mortality for group 1 was 6.2% compared with 4% for group 2 and 3.7% for group 3. Additionally, there was no statistical difference in 30-day mortality between the 3 groups (P ⫽ .83). The reasons for early (⬍30-day) postoperative mortality were postoperative hemorrhage (n ⫽ 2), myocardial infarction (n ⫽ 2), septic shock (n ⫽ 1), and multisystem organ failure (n ⫽ 1). The cause of death for 1 other patient was unknown because he was found dead at home.
Comments Pancreatic cancer has a dismal prognosis without treatment. Surgical resection offers the only hope for long-term survival and even then there is only a 10% to 15% chance of surviving 5 years [4,5]. Survival, however, is increased in those patients undergoing resection compared with those not undergoing surgery regardless of stage [6,7]. Although Table 5 Surgical Morbidity Complication
Group 1 % (n ⫽ 48)
Group 2 % (n ⫽ 75)
Group 3 % (n ⫽ 27)
P
Leak Hernia Dehiscence Pancreatic insufficiency Bleeding Abscess DM Ileus Thrombosis CVA MI SBO Fistula Gastroparesis
6 (12.5) 7 (14.6) 1 (2.1) 17 (35.4) 10 (20.8) 8 (16.7) 1 (2.1) 3 (6.2) 7 (14.6) 2 (4.2) 5 (10.6) 2 (4.2) 0 (0) 3 (6.2)
4 (5.3) 11 (14.7) 5 (6.7) 22 (29.3) 9 (12.0) 14 (18.7) 2 (2.7) 1 (1.3) 11 (14.7) 2 (2.7) 1 (1.3) 3 (4.0) 2 (2.7) 1 (1.3)
2 (7.4) 1 (3.7) 1 (3.7) 5 (18.5) 2 (7.4) 1 (3.7) 0 (0) 0 (0) 1 (3.7) 0 (0) 0 (0) 1 (3.7) 1 (3.7) 0 (0)
.34 .22 .68 .29 .22 .10 .90 .31 .22 .66 .032 .99 .43 .31
CVA ⫽ cerebrovascular accident; DM ; MI ⫽ myocardial infarction; SBO.
D.A. Tessler et al. / The American Journal of Surgery 191 (2006) 191–197 Table 6 TNM status Status
Group 1 % (n ⫽ 48)
Group 2 % (n ⫽ 75)
P
T1 T2 T3 T4 N0 N1 N2 M0 M1
3 (8.3) 10 (27.8) 23 (63.9) 0 (0) 24 (68.6) 10 (28.6) 1 (2.9) 34 (85.0) 6 (15.0)
4 (8.0) 10 (20.0) 33 (66.0) 3 (6.0) 25 (49.0) 26 (51.0) 0 (0) 48 (84.2) 9 (15.8)
.53
.056
.92
resection offers the only chance for long-term survival, median survival after surgery is still only 15 to 19 months [1]. This is unfortunate because as many as 88% of patients with pancreatic cancer are unresectable at diagnosis because of local extension or metastases. The 1995 National Cancer Data Base Report on Pancreatic Cancer reported that 52% of patients with newly diagnosed pancreatic cancer had stage IV disease [6]. Because of the need for an early diagnosis, resection may be considered justifiable in the absence of a preoperative tissue diagnosis [3]. Our study attempted to define whether any factors may aid in differentiating those with malignancy from those without malignancy in patients for whom a preoperative tissue diagnosis is elusive. At our institution, those with a proven pancreatic malignancy are offered enrollment in clinical trials using preoperative chemoradiation in an effort to improve surgical outcomes. Therefore, an aggressive attempt is made to obtain a tissue diagnosis. Failure to document a tissue diagnosis makes a patient ineligible for preoperative chemoradiation. Because preoperative chemoradiation means an inherent delay in surgical intervention, an attempt is often made to decompress the biliary tree during the initial ERCP, especially if the patient is symptomatic, whether or not a diagnosis is eventually obtained; this explains the high number of ERCPs performed in this study. In the absence of a tissue-documented diagnosis, clinical parameters may help guide appropriate patients to the operating room when necessary. The present study demonstrates that jaundice, weight loss, and increased CA 19-9 are predictive of malignancy. Finding a biliary stricture or pancreatic mass added nothing to the specificity, PPV, or NPV for the diagnosis; however, those findings did increase the sensitivity. It is possible that including patients with cytology interpreted as “suspicious for carcinoma” may have enriched the suspected pancreas neoplasm group with patients more likely to have cancer. The multivariate analysis performed after excluding the patients with suspicious cytology, however, generated similar results for sensitivity, specificity, PPV, and NPV. Exclusion of those with distal pancreatic lesions likewise showed similar results despite the fact that patients with
195
distal lesions would not typically present with obstructive jaundice. Re-evaluating the data and excluding patients with metastatic disease did not appreciably change sensitivity or specificity, nor did it diminish our ability to predict those with and without malignancy, thus showing that the advanced nature of the disease does not always correlate with the clinical presentation (data not shown). The previously mentioned criteria were subsequently applied to group 1. Statistical analysis showed a sensitivity of 90.9% when any 1 of the 3 criteria was present. Unfortunately, few patients in this cohort had all 3 criteria, thus limiting adequate statistical evaluation. Thirty-day operative mortality rates for pancreaticoduodenectomy are ⬍5% in high-volume centers, and the majority of deaths are caused by hemorrhage, infection, myocardial infarction, and multisystem organ failure [8]. In our study, the early (⬍30-day) mortality was 4.6%, which is comparable with the published literature. Of all surgeries performed in this study (n ⫽ 150), 1 surgeon performed 123 (82%) of the procedures, and another performed 9 (6%). The remaining 18 surgeries were performed by 10 different surgeons; no surgeon performed ⬎3 of the procedures. Considering all 150 procedures performed, 67% of patients underwent a pancreaticoduodenectomy; 9% underwent distal pancreatectomy, 16% underwent palliative bypass, 7% underwent laparoscopy or laparotomy only, and ⬍1% underwent pancreaticojejunostomy. Other series looking at all procedures performed for pancreatic cancer showed a similar breakdown [9]. Schäfer et al showed no apparent difference in postoperative mortality between pancreaticoduodenectomy and distal pancreatectomy [10]; thus, our mortality data is comparable with other studies that included only cases of pancreatic adenocarcinoma. Furthermore, morbidity and mortality data did not differ by procedure volume of the surgeon in this study. Thompson et al looked at 67 patients undergoing pancreaticoduodenectomy in the pre-EUS era and found a 30% incidence of surgery for presumed, but undocumented, malignancy despite an adequate evaluation including intraoperative biopsies [11]. Of those 20 patients, 9 (45%) were shown after surgery to have malignancy, whereas the remainder had benign disease. It is harder to make the recommendation that all suspicious lesions be resected when 55% of patients underwent surgery for benign disease. The present study, using a larger cohort of patients, demonstrated a much higher rate of neoplasia (74%) in the group with suspected pancreatic malignancy. This may be because all neoplastic lesions were considered, whereas Thompson et al limited their study to cases of adenocarcinoma only. Their results also differ in terms of the disease stage because a much lower rate of T3 disease was found. Additionally, they found a higher proportion of stage I lesions in both the proven pancreatic malignancy group (54%) and suspected pancreatic malignancy group (75%) compared with the findings in this study of 30.5% and 14.2%, respectively, for the 2 groups.
196
D.A. Tessler et al. / The American Journal of Surgery 191 (2006) 191–197
Twenty-seven patients with benign disease underwent surgery for presumed malignancy in this study. Of these, 7 had surgery primarily for symptomatic chronic pancreatitis. For those patients, the possibility of malignancy was listed as the secondary indication for surgery on the operative note. Excluding those patients, benign disease was found in 13% of patients undergoing surgery for presumed malignancy. Other studies, however, have documented surgery for clinical misdiagnosis at 3% to 11% [3,12,13]. Barnes et al reviewed 510 pancreaticoduodenectomies performed at The Johns Hopkins Hospital during an 8-year period. One hundred eight patients (21%) underwent surgery for benign disease, 83 (16%) of those for suspicion of malignancy [14]. Although they recommend an aggressive surgical approach for patients with a clinical suspicion of malignancy, no predictive factors were offered to aid in this decision making. Van Gulik et al believed that no one single test or imaging modality adequately differentiated pancreatic cancer from focal pancreatitis [15]. They described 22 patients with chronic pancreatitis harboring an initially undiagnosed malignancy and 14 inflammatory lesions initially presumed to be malignant at a “center specialized in the diagnosis and treatment of pancreatic cancer” [15]. Twelve of the 14 patients ultimately proven to have benign disease were found to have a mass, whereas biliary ductal or pancreatic ductal stenosis was found in 9 of 13 and 10 of 11 of evaluated patients, respectively. In contrast, an extremely high mortality rate for those with cancer was seen because 20 of the 22 malignancies initially classified as inflammatory were unresectable. This suggests that a more aggressive surgical approach may be warranted if a preoperative diagnosis is uncertain. Other studies have restricted the study protocol to include only pancreatic adenocarcinoma. The present study purposely included all patients with suspected malignancy of the pancreas. This allowed for a larger base of patients on which recommendations could be made because there are other neoplasms of the pancreas, other than adenocarcinoma, for which surgical therapy is appropriate. Subgroup analysis was performed looking only at those with adenocarcinoma and again showed comparable figures for sensitivity (data not shown). Other strengths of this study include the large number of patients, objective end points of the study, and the inclusion only of patients deemed surgical candidates for better comparability. Additionally, this study was able to suggest predictive factors to aid in the differentiation of benign and neoplastic disease. A limitation of the present study was the retrospective design. First and foremost, not all patients had documentation of a preoperative CA 19-9 or adequate evaluation for weight loss even after thorough search through the electronic database. The preoperative workup was not uniform for all patients, thus restricting the data points from which stronger recommendations could be ex-
tracted. Preoperative evaluation differed between medical and surgical services, as well as among individual physicians, often because of personal preference given the lack of a universally accepted algorithm for the workup of patients suspected to have a pancreatic neoplasm at our institution. Additionally, the initial time points of this study coincided with early experience with EUS at our institution, which may explain the lack of diagnostic tissue and failure to perform EUS-FNA in certain patients. Furthermore, another limitation is that as a tertiary referral center, a number of the patients, for whom preoperative as well as follow-up data were not available, came as outside referrals for workup and surgical resection. Mortality for pancreatic surgery has dramatically decreased during the past half century [16]. Concern of procedural morbidity and mortality for disease ultimately proven to be benign should not delay an attempt at surgical resection in cases of suspected pancreatic malignancy. In experienced hands, outcomes for those undergoing surgery without a documented tissue diagnosis are similar to those known to have neoplasia before surgery. In the appropriate clinical setting, a patient with weight loss, jaundice, and increased CA 19-9, often in combination with a biliary stricture or pancreatic mass, should be strongly considered for surgery even if a preoperative tissue diagnosis is not present. Although these criteria were highly suggestive of malignancy in these selected patients, this model needs further study in a prospective fashion before it can be adopted for widespread use. References [1] Hawes RH, Xiong Q, Waxman I, et al. A multispecialty approach to the diagnosis and management of pancreatic cancer. Am J Gastro 2000;95:17–31. [2] Duffy JP, Reber HA. Pancreatic neoplasms. Curr Opin Gastroenterol 2003;19:458 – 66. [3] Abraham SC, Wilentz RE, Yeo CJ, et al. Pancreaticoduodenectomy (Whipple resections) in patients without malignancy: Are they all “chronic pancreatitis”? Am J Surg Pathol 2003;27:110 –20. [4] DiMagno EP. Pancreatic cancer: Clinical presentation, pitfalls and early clues. Ann Oncol 1999;10(suppl):S140 –S142. [5] Yeo CJ, Sohn TA, Cameron JL, et al. Periampullary adenocarcinoma: Analysis of 5-year survivors. Ann Surg 1998;227:821–31. [6] Dimagno EP, Reber HA, Tempero MA. AGA technical review on the epidemiology, diagnosis and treatment of pancreatic ductal adenocarcinoma. Gastroenterology 1999;117:1464 – 84. [7] Carr JA, Ajlouni M, Wollner I, et al. Adenocarcinoma of the head of the pancreas: Effects of surgical and nonsurgical therapy on survival—A ten-year experience. Am Surg 1999;65:1143–9. [8] Ahmad NA, Lewis JD, Ginsberg GG, et al. Long-term survival after pancreatic resection for pancreatic adenocarcinoma. Am J Gastro 2001;96:2609 –15. [9] Wagner M, Dikopoulos N, Kulli C, et al. Standard surgical treatment in pancreatic cancer. Ann Oncol 1999;10(suppl):S247–S251. [10] Schäfer M, Müllhaupt B, Clavien PA. Evidence-based pancreatic head resection for pancreatic cancer and chronic pancreatitis. Ann Surg 2002;236:137– 48.
D.A. Tessler et al. / The American Journal of Surgery 191 (2006) 191–197 [11] Thompson JS, Murayama KM, Edney JA, et al. Pancreaticoduodenectomy for suspected but unproven malignancy. Am J Surg 1994;169:571–5. [12] Van Gulik TM, Reeders JWA, Bosma A, et al. Incidence and clinical findings of benign, inflammatory disease in patients resected for presumed pancreatic head cancer. Gastro Endosc 1997;46:417–23. [13] Smith CD, Behrns KE, van Heerden JA, et al. Radical pancreaticoduodenectomy for misdiagnosed pancreatic mass. Br J Surg 1994; 81:585–9.
197
[14] Barnes SA, Lilliemoe KD, Kaufman HS, et al. Pancreaticoduodenectomy for benign disease. Am J Surg 1996;171:131–5. [15] van Gulik TM, Moojen TM, van Geenen R, et al. Differential diagnosis of focal pancreatitis and pancreatic cancer. Ann Oncol 1999; 10(suppl):S85–S88. [16] Rios G, Conrad A, Cole D, et al. Trends in indications and outcomes in the Whipple procedure over a 40-year period. Am Surg 1999;9: 889 –93.