This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Kalady, M. F. Articles by Tyler, D. S. Search for Related Content PubMed PubMed Citation Articles by Kalady, M. F. Articles by Tyler, D. S.

Pancreatic Duct Strictures: Identifying Risk of Malignancy

From the Departments of Surgery (MFK, MWO, OIA-W, BMC, TNP, DST), Biostatistics and Bioinformatics (BP), and Gastroenterology (JB, JKH, PSJ, MSB), Duke University Medical Center, Durham, North Carolina.

Correspondence: Address correspondence and reprint requests to: Douglas S. Tyler, MD, Duke University Medical Center, Box 3118, Durham, NC 27710; Fax: 919-681-8701; e-mail: tyler002{at}acpub.duke.edu

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

 
Background: This study aimed to define PDS characteristics that predict malignancy and would thus invoke further diagnostic evaluation or aggressive treatment.

Methods: 355 cases of PDS were diagnosed by ERCP during a 7-year period at a single institution. A retrospective review identified clinical/demographic patient data and ERCP results.

Results: 218 (61%) patients with a PDS were found to have an isolated PDS. Twelve percent of isolated PDS and 79% of CBD stricture-associated PDS were malignant. The sensitivity and specificity for the double duct sign for malignancy were 77% and 80% respectively, and the positive predictive value was 65%. Predictors of malignancy were statistically similar for both isolated PDS and those associated with a CBD stricture. Univariate predictors of malignancy included stricture location in the pancreatic head/neck, jaundice, and patient age. Predictors of benign disease included a history of pancreatitis, the presence of multiple strictures, pancreatic duct stones, pseudocyst, pancreas divisum anatomy, irregular side branches, and irregular pancreatic duct morphology. Less than 1% of patients with either pancreas divisum anatomy, pancreatic duct stones, or pancreatic pseudocyst had malignancy. Using malignancy as the dependent variable, multivariate factors included in the final prognostic equation were history of pancreatitis (odds ratio 0.009 with history of pancreatitis), stricture location in the head or neck (odds ratio 42) and irregular pancreatic duct side branches (odds ratio 0.05 with irregular branches).

Conclusions: This study demonstrates that certain characteristics of PDS can predict the subset of patients who have an increased risk of cancer.

Key Words: Double duct sign • Endoscopic retrograde cholangiopancreatography • Pancreatic cancer • Pancreatic duct stricture

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

 
Pancreatic duct strictures identified during endoscopic retrograde cholangiopancreatography (ERCP) often present a diagnostic dilemma. Although a pancreatic duct stricture represents underlying pathology, it may result from a variety of etiologies, including chronic pancreatitis, pancreatic neoplasm, pseudocyst, and nonspecific inflammatory processes. Since clinical management of each of these diseases widely differs from observation to resection, it is crucial to determine the underlying diagnosis.

Pancreatic cancer poses the most serious risk and must be considered in any patient with a pancreatic duct stricture. The prognosis of pancreatic cancer is poor, with an overall 5-year survival of less than 5%.¹ Early diagnosis and intervention provide the best chance for increased survival. With resection of early stage disease, median survival approaches 18 months, and 5-year actuarial survival may be as high as 21%.² Therefore, aggressive intervention is warranted to improve outcomes. However, selective intervention is prudent to avoid unnecessary surgical morbidity and mortality in benign disease. Determining which patients should undergo aggressive intervention has been a long-standing clinical challenge.

Clinically suspected pancreatic malignancy is often evaluated with multiple diagnostic techniques, including computed tomography (CT) and ultrasound. The accuracy of these tests, however, relies heavily on the detection of a pancreatic mass. ERCP is an integral part of evaluating patients who have no identifiable mass evident by other diagnostic imaging modalities.³ ERCP delineates the distorted pancreatic ductal system and has been shown to predict adenocarcinoma with a sensitivity and specificity of approximately 70% to 95%.^4–6 In particular, identification of a simultaneous pancreatic duct stricture and a common bile duct stricture, the double duct sign, has been considered a specific sign for ductal carcinoma.^7,8 However, pancreatic cancer occurring in the pancreatic body or tail is frequently indistinguishable from chronic pancreatitis and rarely presents with a double duct sign. Furthermore, not all strictures in the pancreatic head involve the common bile duct. Thus, a double duct sign will not be present in a significant percentage of cancers.

Pancreatic duct strictures without an associated common bile duct stricture, an entity referred to as isolated pancreatic duct strictures, have not been well characterized. This study defines characteristics of all pancreatic duct strictures, including the subgroup of isolated strictures, that assist in predicting malignancy. In addition, we present a mathematical model based on multivariate regression to use as a prognostic clinical tool.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

 
Between June 1993 and November 2000, 7008 ERCPs were performed at a single institution. Results of each procedure were prospectively recorded and maintained in a gastrointestinal procedure database. The database was queried for the diagnosis of pancreatic duct stricture, and the search yielded 421 cases. Patients who had multiple procedures were limited to inclusion of only the first ERCP unless new information was obtained by subsequent examination. In such a situation, cases that provided additional diagnoses or findings were included in the analysis (n = 15). The final population analyzed included 355 cases, of which there were 218 isolated pancreatic duct strictures (Fig. 1).

View larger version (13K): [in this window] [in a new window]   FIG. 1. Distribution of study population by pancreatic duct stricture type (isolated or with an associated common bile duct stricture) and by benign or malignant diagnosis (PDS = pancreatic duct stricture; CBD = common bile duct).

Univariate analyses using the ² test and Student’s t-test were performed to determine characteristics of pancreatic duct strictures associated with benign or malignant disease. A logistic regression model was then fitted to a dependent variable, indicating the diagnosis of pancreatic malignancy. Since only 1 of the 111 patients (0.9%) with pancreatic duct stones, pseudocyst, or pancreas divisum had pancreatic cancer, these patients and those three variables were excluded from further multivariate analysis, leaving a total of 244 patients with which to build a model. Candidate predictor variables were tested with a backward selection process to develop a tentative final model.

With use of this procedure, all variables were initially entered into the model and then were removed individually on the basis of their two-tailed P values. Only those variables with P values less than .01 were permitted to stay in the model. Because of the small sample size, bootstrapping was used to validate this model instead of dividing the data into a test and validation sample. Specifically, the backward selection technique was used to fit logistic regression models to 100 bootstrapped samples. Each bootstrapped sample was created by randomly selecting 244 observations from the dataset, with replacement after each selection. It was decided a priori that only variables occurring in at least 90% of the bootstrapped models would be used in the final prognostic regression equation.

Predicted probabilities of pancreatic cancer were calculated for each patient with use of the final prognostic equation. On the basis of the value of a patient’s predicted probability, regression tree analysis (minimum P value approach) was used to divide patients into three risk groups: low-, intermediate-, and high-risk. This method defines cut-points on the predicted probabilities of cancer that result in the best correlation between risk group and final diagnosis.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

 
Patients and Diagnoses
Three-hundred fifty-five patients were diagnosed with a pancreatic duct stricture on ERCP. This population consisted of 49% males and 51% females, with a mean age of 59 years. Within this group, 218 patients (61%) had an isolated pancreatic duct stricture and 137 patients (39%) had a pancreatic duct stricture associated with a common bile duct stricture (Fig. 1).

Overall, the incidence of cancer associated with a pancreatic duct stricture was 32.7%. Whereas 65% of patients with a double duct sign had pancreatic cancer, only 12% of isolated pancreatic duct strictures were malignant (Fig. 1).; 95% of malignancies included in this study were pancreatic adenocarcinoma. Also included were two cases of cholangiocarcinoma, two cases of pancreatic lymphoma, and two islet cell tumors. The most common diagnoses associated with a benign pancreatic stricture were chronic pancreatitis (65%), pancreatic pseudocyst (23%), pancreatic duct stones (15%), and pancreas divisum (13%). Figures 2 through 4 demonstrate examples of these common findings.

View larger version (120K): [in this window] [in a new window]   FIG. 2. For a 62-year-old woman with epigastric pain, ERCP demonstrated a long stricture in the pancreatic duct head with extravasation of contrast into a complex cyst (arrows) in the pancreatic tail. This patient had pancreatitis with an associated pseudocyst.

View larger version (101K): [in this window] [in a new window]   FIG. 3. For a 57-year-old man with recurrent abdominal pain and jaundice, ERCP showed a stricture in the pancreatic head with dilated main pancreatic duct. There was a single pancreatic duct stone (arrow) in the mid portion of the gland. This patient had pancreatitis.

View larger version (116K): [in this window] [in a new window]   FIG. 4. For a 45-year-old woman with epigastric pain secondary to pancreatitis, ERCP revealed pancreas divisum anatomy with a stricture (arrow) in the ventral duct.

View larger version (112K): [in this window] [in a new window]   FIG. 5. For a 64-year-old woman with jaundice and weight loss, ERCP demonstrated simultaneous strictures in both the distal common bile duct (black open arrow) and the pancreatic duct (white arrow) in the head of the pancreas, i.e., the double duct sign. Pathological diagnosis revealed pancreatic adenocarcinoma.

Multivariate Analysis
There was no significant interaction between the variable defining the presence of a common bile duct stricture and any of the other predictors of malignancy. Therefore, patients with an isolated pancreatic duct stricture and those with a double duct sign were combined to predict malignancy. The final model applies to patients with or without an associated common bile duct stricture. On multivariate analysis, three factors independently statistically predicted disease etiology (Table 2). Stricture location in the pancreatic head or neck correlated with malignancy (odds ratio [OR] = 42), whereas the presence of irregular side branches (OR = 0.05 for not having irregular branches) (Fig. 6) or a history of pancreatitis (OR = 0.009 for no history of pancreatitis) negatively correlated with pancreatic cancer. All three variables had P values < .0001. In the regression analyses on 100 simulated sets of bootstrapped data, the variable history of pancreatitis entered all 100 models, and the variables of irregular branches and location entered 95 and 92 models, respectively. No other variable entered more than 55% of the models. Thus, the final logistic regression model was determined to contain only the above three variables. The model fit extraordinarily well, with a C index of 0.94.

View larger version (115K): [in this window] [in a new window]   FIG. 6. For a 55-year-old man with history of ethanol abuse and recurrent episodes of epigastric pain, ERCP showed multiple pancreatic duct strictures and prominent irregular side branches as well as several pancreatic duct stones. This patient had chronic pancreatitis.

(1)

where XBETA is the sum of the predictive variables multiplied by their respective regression coefficients:

(2)

The variable LOCATION is equal to one if the stricture is located in the pancreatic head or neck and equal to zero if the stricture is in the pancreatic body or tail. The variable BRANCH is assigned a value of one if there are irregular side branches on ERCP and is assigned a zero if side branch morphology is normal. A patient history of pancreatitis is equal to one for the PANCREATITIS variable and equal to zero if there is no history of pancreatitis.

The predicted probabilities were calculated for each patient, and the distribution is shown in Fig. 7. Three distinct groups were readily identified and characterized as low-risk, intermediate-risk, and high-risk. Patients in the low-risk group had a less than 2% chance of malignancy, while patients in the high-risk group had a greater than 93% chance of malignancy. Patients in the intermediate group had probabilities between 12% and 41% (Table 3).

View larger version (19K): [in this window] [in a new window]   FIG. 7. Frequency and distribution of XBETA scores and associated predictive cancer risk for 244 patients used to develop this model. The number next to each bar represents the number of patients in this study with that XBETA value.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Distinguishing malignant from benign pancreatic duct strictures remains a difficult task, and only tissue biopsy provides a definitive diagnosis. Unfortunately, inability to easily access the gland for adequate sampling forces physicians to make clinical decisions based on the best available information. Despite generally accepted norms and intuitive judgments from clinical experience, little has been written to explicitly define the characteristics associated with benign or malignant strictures. This study presents a statistically based approach to pancreatic duct strictures that physicians may use in determining treatment options.

This work set out to evaluate isolated pancreatic duct strictures, with the assumption that isolated strictures might present differently than those associated with a common bile duct stricture and that malignant isolated strictures might present differently than isolated benign strictures. Initial analysis determined that certain characteristics were predictive for malignancy or benign disease, but the predictive factors were statistically similar for both isolated strictures and those associated with a common bile duct stricture. Thus, the final multivariate regression analysis was based on all patients with a pancreatic duct stricture, and the derived prognostic equation may be applied to any patient with a pancreatic duct stricture, isolated or not.

Both stricture characteristics and patient factors were considered for prognostic significance. Stricture location was the strongest predictor of pancreatic cancer by univariate analysis and by multivariate regression analysis. While a stricture in the head or neck of the gland was highly predictive of cancer, a stricture found in the pancreatic body or tail was most likely secondary to benign disease. The presence of an associated common bile duct stricture (the double duct sign) was predictive of cancer on univariate analysis. Since the disease process underlying a double duct sign must anatomically affect the head of the pancreas where the pancreatic and common bile ducts are in proximity, one would expect the variables of location and double duct sign to be statistically correlated. This indeed was the case, but the two variables provided different information. Although all patients with a double duct sign had a stricture in the head or neck of the pancreas, only 47% of patients with a stricture in the pancreatic head or neck (n = 279) had a double duct sign.

Our study determined a sensitivity and specificity of approximately 80% and positive and negative predictive values of 65% and 88%, respectively, for the double duct sign associated with pancreatic malignancy. These data coincide with a recent report from Menges et al., who noted a sensitivity of 85%.⁸ In studying the variables of stricture location and double duct sign, stricture location proved to be the more predictive factor on regression analysis. Stricture location alone, however, cannot account for all cancers. In this series, 96% of cancers were located in the pancreatic head or neck, but 59% of strictures located in the pancreatic head or neck were benign. Despite prior reports of longer stricture length corresponding to malignancy and advanced tumor stage,⁹ stricture length was not predictive of etiology in our series.

Clinical patient factors such as age and the presence of jaundice were also found to be predictive of malignancy on univariate analysis. In general, older patients had an increased likelihood of having malignant disease. The mean age at presentation for patients with cancer was 17 years older than for patients presenting with benign disease. Despite the strong predictive value, age did not enter into the final regression equation as an independent variable because of its high correlation with the variable for pancreatitis. Similarly, jaundice, which was highly predictive of malignancy independently, was strongly correlated with stricture location and thus not included in the final model.

A history of pancreatitis was the strongest predictor of benign disease on multivariate analysis. In our study, 62% of patients had a history of pancreatitis, and 4.1% of these had pancreatic cancer. A patient was determined to have a history of pancreatitis in several ways: documentation of an episode of abdominal pain with an elevated serum amylase level, recorded in the medical record, in the patient history section of the original gastrointestinal database, or in the ERCP report (listed as an indication). We realize that this potentially introduces an imprecise definition of pancreatitis into our model and may skew our data in either direction by omitting undocumented histories or overdiagnosing cases. Although chronic pancreatitis has been reported as a risk factor for pancreatic cancer,^10,11 the exact relationship is not known. The fact that nearly 5% of patients with pancreatitis had an underlying cancer is not surprising and emphasizes the need for close clinical follow-up.

This study also identified several ERCP characteristics whose absence strongly predicted malignancy. In other words, the following findings were predictive of benign disease. Of 57 patients in whom a pseudocyst was identified, 41 had isolated pancreatic duct strictures, of which none were malignant. One patient with a double duct sign and a pseudocyst had pancreatic adenocarcinoma. Pancreatic duct stones were identified in 37 patients, all of whom had benign disease. Both pancreatic duct stones and pseudocyst are known changes associated with chronic pancreatitis.¹² Pancreas divisum anatomy was a third predictor of benign disease. Pancreas divisum anatomy is a congenital anomaly resulting from failed fusion of the major and minor pancreatic ducts during development. This anomaly occurs in approximately 6% of the population and in most cases is asymptomatic.¹³ However, pancreas divisum has been shown to be associated with pancreatitis, and divisum is the etiology of chronic pancreatitis in approximately 10% of cases.^14,15 Epidemiological studies have failed to document risk for cancer in patients with pancreas divisum,¹⁵ and none of the 32 patients in our series with pancreas divisum had malignancy. Divisum anatomy, however, does not preclude cancer and has been reported as the initial diagnosis in cases later diagnosed as pancreatic cancer.¹⁶ In one series of patients with pancreas divisum, the incidence of pancreatic cancer was 1.9%,¹⁷ but no causal relationship was noted. Although the discovery of any or all of the ERCP findings of cyst, stone, or divisum cannot definitively rule out malignancy, the probability is less than 1% in our series.

The accuracy of CT and ultrasonography relies on the ability to detect a mass. ERCP, however, provides anatomic ductal detail and may be particularly helpful in evaluating patients with suspected pancreatic malignancy without a definite mass on CT or ultrasonography. In our series, 281 patients had a record of CT or ultrasonography during the same evaluation period in which ERCP was done. Sixty-eight percent of people evaluated did not have a mass revealed by these techniques. This information helps define the population in which ERCP and our prognostic equation may be most useful. Endoscopic ultrasound (EUS) also plays an important role in the evaluation of patients with a pancreatic duct stricture and suspected pancreatic cancer who have no mass defined by another modality. In particular, EUS guidance may be employed to obtain a tissue biopsy. EUS was not utilized at our institution during the earlier years included in this study, and we cannot comment on its utility for our entire population. We do not employ ERCP at the expense of EUS but view them as complementary tests.

We propose that this model be applied to provide each patient with a general risk category or to specifically assign a numeric probability for that patient. By determining the value of XBETA as defined in Table 3, the risk of cancer may be quickly stratified into low-, intermediate-, or high-risk. To define an exact probability for an individual patient, the XBETA value is inserted into the prognostic equation in Table 3. Since there are three discrete binary variables in the predictive model, there are eight possible probability outcomes. Each potential combination of variables and the associated risk are presented in Table 4. While patients with low prognostic scores may be followed clinically, patients with high prognostic scores warrant aggressive attempts at tissue diagnosis or definitive therapy.

View larger version (29K): [in this window] [in a new window]   FIG. 8. Proposed algorithm and application of pancreatic duct stricture prognostic equation. Clinical and radiographic information is entered into the model as described below. LOCATION = 0 if stricture is located in pancreatic body or tail and 1 if in head or neck; BRANCH = 0 if pancreatic duct side branches are normal and 1 if there are irregular pancreatic duct side branches; PANCREATITIS = 0 if there is no history of pancreatitis and 1 if there is history or an acute episode of pancreatitis.

At the other extreme, take an example of a 55-year-old man with an isolated pancreatic duct stricture in the pancreatic body with irregular pancreatic duct side branches. He has had four episodes of pancreatitis in the past 3 years. There is no mass evident on abdominal CT. Using the XBETA calculation, we would assign a 0 for the variable LOCATION since the stricture is in the body of the pancreas. Because of the irregular side branch morphology, we would assign a 1 for BRANCH. With his history of pancreatitis, his PANCREATITIS variable is 1. His corresponding XBETA value is –8.7681, placing him in the low-risk category. To assign an exact percentage risk, insert the XBETA value into the prognostic equation as shown in Table 3 or apply the appropriate combination of predictive factors into the grid presented in Table 4. This particular patient has a 0.02% probability that his condition is malignant.

A final example illustrates a patient with an intermediate risk for cancer. The patient is a 59-year-old man with vague epigastric pain and no prior episodes of pancreatitis. There is fullness in the head of the pancreas on CT. An ERCP was performed and revealed a stricture in the pancreatic body with regular side branches. There are no stones or cysts and he does not have pancreas divisum. This patient would be scored with a 0 for no prior history of pancreatitis, 0 for the stricture location in the body of the pancreas, and 0 for regular side branches. His XBETA value would thus be –1.0277, which places him in the intermediate risk category. His predicted risk, based on the prognostic equation, would be 26%.

In summary, determining the etiology of pancreatic duct strictures presents a diagnostic challenge. Physicians must assimilate all available information prior to making a treatment decision. This study provides a mathematical model based on patient and ERCP characteristics that helps define the risk of malignancy and may guide clinical judgment.

	ACKNOWLEDGMENTS

 
The acknowledgments are available online in the fulltext version at www.annalssurgicaloncology.org. They are not available in the PDF version.

Presented at the 55th Annual Cancer Symposium of the Society of Surgical Oncology, March 14–17, 2002, Denver, Colorado.

	FOOTNOTES

 
Differentiating pancreatic cancer from benign disease in pancreatic duct strictures remains a diagnostic challenge. This study reviewed 355 endoscopic retrograde cholangiopancreatography cases with pancreatic duct strictures and identified patient and stricture characteristics that help predict malignancy for individual patients.

Received for publication March 15, 2002. Accepted for publication February 25, 2004.

	REFERENCES

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

 

1. Greenlee R, Murray T, Bolden S, Wingo P. Cancer statistics, 2000. CA Cancer J Clin 2000; 50: 7–33.[Abstract]
2. Yeo CJ, Cameron JL, Sohn TA, et al.. Six hundred fifty consecutive pancreaticoduodenectomies in the 1990s: pathology, complications, and outcomes. Ann Surg 1997; 226: 248–57.[CrossRef][Medline]
3. Griffanti-Bartoli F, Arnone GB, Ceppa P, Ravera G, Carrabetta S, Civalleri D. Malignant tumors in the head of the pancreas and the periampullary region: diagnostic and prognostic aspects. Anticancer Res 1994; 14: 657–66.[Medline]
4. Adamek HE, Albert J, Breer H, Weitz M, Schilling D, Riemann JF. Pancreatic cancer detection with magnetic resonance cholangiopancreatography and endoscopic retrograde cholangiopancreatography: a prospective controlled study. Lancet 2000; 356: 190–3.[CrossRef][Medline]
5. Hall TJ, Blackstone MO, Cooper MJ, Hughes RG, Moossa AR. Prospective evaluation of endoscopic retrograde cholangiopancreatography in the diagnosis of periampullary cancers. Ann Surg 1978; 187: 313–7.[Medline]
6. Gilinsky N, Bornman P, Girdwood A, Marks I. Diagnostic yield of endoscopic retrograde cholangiopancreatography in carcinoma of the pancreas. Br J Surg 1986; 73: 539–43.[Medline]
7. Freeny PC, Bilbao MK, Katon RM. ‘Blind‘ evaluation of endoscopic retrograde cholangiopancreatography (ERCP) in the diagnosis of pancreatic carcinoma: the ‘double duct‘ and other signs. Radiology 1976; 119: 271–4.[Abstract]
8. Menges M, Lerch MM, Zeitz M. The double duct sign in patients with malignant and benign pancreatic lesions. Gastrointest Endosc 2000; 52: 74–7.[CrossRef][Medline]
9. Shah SA, Movson J, Ransil BJ, Waxman I. Pancreatic duct stricture length at ERCP predicts tumor size and pathological stage of pancreatic cancer. Am J Gastroenterol 1997; 92: 964–7.[Medline]
10. Lowenfels AB, Maisonneuve P, Cavallini G, et al. Pancreatitis and the risk of pancreatic cancer. International Pancreatitis Study Group. N Engl J Med 1993; 328: 1433–7.[Abstract/Free Full Text]
11. Wakasugi H, Funakoshi A, Iguchi H, et al. Pancreatic carcinoma associated with chronic pancreatitis. Intern Med 1999; 38: 951–6.[Medline]
12. Forsmark C, Toskes P. What does an abnormal pancreatogram mean? Pancreatography 1995; 5: 105–23.
13. Burtin P, Person B, Charneau J, Boyer J. Pancreas divisum and pancreatitis: a coincidental association? Endoscopy 1991; 23: 55–8.[Medline]
14. Kalady MF, Broome AH, Meyers WC, Pappas TN. Immediate and long-term outcomes after lateral pancreaticojejunostomy for chronic pancreatitis. Am Surg 2001; 67: 478–83.[Medline]
15. Bernard JP, Sahel J, Giovannini M, Sarles H. Pancreas divisum is a probable cause of acute pancreatitis: a report of 137 cases. Pancreas 1990; 5: 248–54.[Medline]
16. van Gulik TM, Moojen TM, van Geenen R, Rauws EA, Obertop H, Gouma DJ. Differential diagnosis of focal pancreatitis and pancreatic cancer. Ann Oncol 1999; 10: 85–8.
17. Wang HP, Wu MS, Lin CC, et al. Pancreaticobiliary diseases associated with anomalous pancreaticobiliary ductal union. Gastrointest Endosc 1998; 48: 84–9.

This article has been cited by other articles:

				J. Ahualli The Double Duct Sign Radiology, July 1, 2007; 244(1): 314 - 315. [Full Text] [PDF]

				R. A. Erickson Editorial: ERCP and Pancreatic Cancer Ann. Surg. Oncol., June 1, 2004; 11(6): 555 - 557. [Full Text] [PDF]

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Kalady, M. F. Articles by Tyler, D. S. Search for Related Content PubMed PubMed Citation Articles by Kalady, M. F. Articles by Tyler, D. S.