A prospective pancreatographic study of the prevalence of pancreatic carcinoma in patients with diabetes mellitus




The correlation between diabetes mellitus and pancreatic carcinoma is well documented, but no criteria have been established for the efficient selection of a high-risk group among patients with diabetes mellitus.


Eighty-seven patients were selected prospectively from outpatients with diabetes and underwent endoscopic retrograde pancreatography (ERP) according to the authors' original criteria, including the onset of diabetes after age 55 years, deterioration of diabetes or loss of body weight despite strict medical control, elevation of serum amylase and/or CA19-9 levels, and pancreatobiliary abnormalities on routine ultrasonography. The patients were divided into two groups according to the time from the onset of diabetes to ERP: Patients in Group A had recent-onset diabetes (within 3 years), and Group B patients had diabetes for > 3 years.


A total of 86 patients (excluding 1 patient with unsuccessful ERP who had undergone previous Billroth-2 gastrectomy) were enrolled. There were 33 males and 53 females, age 40–90 years, with a mean age of 65.1 years. ERP demonstrated pancreatic carcinoma, although it was advanced disease in all patients, at an extremely high rate of 7.0% (6 of 86 patients) with no serious complications. The prevalence of pancreatic carcinoma in Group A (13.9%; 5of 36 patients) was significantly greater compared with Group B (2.0%; 1 of 50 patients; P = 0.0442). ERP with an indwelling balloon catheter and subsequent pancreatic juice sampling was performed in 49 patients, yielding positive cytology in 1 patient with pancreatic tail carcinoma, whereas measurements of carcinoembryonic antigen and CA19-9 levels in pancreatic juice were of no use in the diagnosis of pancreatic carcinoma.


Selective ERP in patients with diabetes who were at high risk did not lead to the early diagnosis of pancreatic carcinoma, although this study showed that the 3-year period after the onset of diabetes was critical. A more aggressive diagnostic approach within this period in diabetic patients with the authors' criteria may contribute to the earlier diagnosis of pancreatic carcinoma. Cancer 2002;94:2344–9. © 2002 American Cancer Society.

DOI 10.1002/cncr.10493

Pancreatic carcinoma remains a challenging problem for surgeons and physicians because of its poor prognosis and difficulties in early detection, despite the recent progress of various imaging modalities.1, 2 Many trials of mass screening using tumor markers and/or ultrasonography were not cost effective.3, 4 A few previous articles described the value of screening in selected outpatients with any abdominal complaints4 or with diabetes mellitus, which is known as a correlative factor with pancreatic carcinoma.5–8 Karmody and Kyle7 and Gullo et al.8 reported the association between recent-onset diabetes and the development of pancreatic carcinoma. However, no criteria have been established to efficiently identify high-risk groups among diabetic patients. In this article, we introduce our original indications for endoscopic retrograde pancreatography (ERP) in patients with diabetes mellitus and assess its possible contribution to the process of diagnosing pancreatic carcinoma. In this study, ERP was used as a gold standard for the diagnosis of pancreatic carcinoma despite its invasiveness,9 because no other prospective studies using noninvasive diagnostic modalities have achieved acceptable results.3, 4


Between April 1995 and March 1999, the Unit of Diabetes, Department of Internal Medicine, Fukuoka Red Cross Hospital, annually renewed the registration of approximately 2000 outpatients with diabetes mellitus: There were 1885 patients registered in 1995, 1943 patients registered in 1996, 2161 patients registered in 1997, and 2282 patients registered in 1998. The patients visited the clinic once each month for the purpose of medical control of diabetes. Serum amylase and CA19-9 levels were measured every 2–3 months and ultrasonography was repeated at 6-month intervals, unless the patient was lost to follow-up. A total of 87 patients were selected prospectively from these patients, as indicated for ERP based on one or more of the following criteria: 1) onset of diabetes after age 55 years without obesity, alcoholism, or family history of diabetes; 2) deterioration of preexistent glucose intolerance (fasting and postprandial blood glucose and/or serum concentration of HbA1c); 3) body weight loss (> 5 kg within 3 months) despite strict control of diet and/or dosage of insulin; 4) transient or persistent elevation of serum amylase (> 200 IU/L); 5) CA19-9 level > 300 U/mL; and 6) pancreatobiliary abnormalities on routine ultrasonography (Table 1). These parameters were employed to pick up any signs of the occurrence of malignancy and its consequent adverse effects on the exocrine and endocrine functions of the pancreas. No patients had clinical symptoms, such as biliary colic, that were included in other standard indications for ERP. Computed tomography (CT) scans were performed immediately before or after ERP to supplement and corroborate the pancreatographic diagnosis.

Table 1. Data from Patients with Diabetes Mellitus who Underwent Endoscopic Retrograde Pancreatography and the Diagnostic Rate of Pancreatic Carcinomaa
CharacteristicGroup AGroup BTotal
  • M/F: male/female; ERP: endoscopic retrograde pancreatography; DM: diabetes mellitus.

  • a

    Patients in Group A were evaluated within 3 years from the onset of diabetes, and patients in Group B were evaluated more than 3 years after the onset of diabetes.

  • b

    P = 0.0442.

No. of patients (M/F)36 (16/20)50 (17/33)86 (33/53)
 Mean age in yrs (range)61.1 (40–90)67.9 (43–85)65.1 (40–90)
 Duration from onset of DM in months (range)9.2 (0–36)132.6 (39–312)80.9 (0–312)
Indications for ERP
 1) Onset of DM after age 55 yrs16218
 2) Deterioration of DM182341
 3) Body weight loss527
 4) Elevation of serum amylase5510
 5) Elevation of serum CA19-96511
 6) Abnormal ultrasonogram82028
  Suspicion of solid pancreatic mass459
  Cystic lesion of pancreas044
  Dilatation of biliary or pancreatic ducts3912
  Biliary or pancreatic lithiasis123
Results of ERP
 Conventional ERP162137
 Balloon ERP and pancreatic juice sampling202949
No. of patients with pancreatic carcinoma (%)5 (13.9)b1 (2.0)b6 (7.0)

All 87 patients who were selected provided informed consent for not only the procedure itself but also for participation in the clinical study. On entering the study, when ERP was indicated, each patient was assigned to one of two groups according to the length of time from the onset of diabetes to ERP, according to the report by Gullo et al.8: Patients in Group A had recent-onset diabetes (within 3 years), and patients in Group B had diabetes for > 3 years.

The patient was prepared with pharyngeal anesthesia and intravenous diazepam (5–10 mg). Using a side-viewing duodenoscope (model JF-1T20; Olympus Optical Company, Tokyo, Japan), the main pancreatic duct was entered with a 5-French Teflon catheter and opacified with 60% meglumine sodium diatrizoate (Urografin™; Nihon Schering, Osaka, Japan). A 6-French balloon catheter (model JX-283; Arrow International Inc., Reading, PA) was inserted whenever possible, and the endoscope was removed totally with the catheter left in place, as described previously (balloon ERP).9, 10 After the posture of the patient was changed to the supine position, contrast medium was injected slowly to fill fine pancreatic branches under careful visual control by fluoroscopy. The total amount of contrast medium injected varied depending on drainage through the minor papilla. After obtaining compression spot films of pancreatogram, secretin (50 units of Secrepan™; Eisai, Tokyo, Japan) was administered intravenously. Pancreatic juice was then aspirated manually through the balloon catheter. The first 5–10 mL containing contrast medium was abandoned, and the next 8–15 mL were submitted to cytologic examination and subsequent measurements of carcinoembryonic antigen (CEA) and CA19-9. The total volume of aspirated pancreatic juice was also determined by patency of the minor papilla. Each session of balloon ERP followed by pancreatic juice sampling was completed within 30 minutes. The samples were centrifuged at 1500–3000 revolutions per minute for 5–10 minutes and treated with three kinds of staining on the smears: Giemsa or periodic acid-Schiff staining after fixation in 95% methanol and Papanicolaou staining after preparation in the Y-M solution, which was compounded of 50% ethanol and polyethyleneglycol. The cytologic diagnosis was made with double checking by an expert screener and an authorized pathologist.

After the procedure, the patient was kept fasting and was observed carefully overnight for the appearance of any symptoms. A protease inhibitor was administered intravenously to prevent acute pancreatitis. If the patient was asymptomatic the next morning and the serum amylase level was below 1000 IU/mL, then the patient was discharged after lunch. Hyperamylasemia accompanied by upper abdominal pain that lasted > 24 hours was regarded as acute pancreatitis. Statistical significance was determined using the Fisher exact probability test, and P values < 0.05 were considered significant.


Fine pancreatograms were obtained from 85 of 87 patients. Successive balloon ERPs and pancreatic juice samplings were accomplished in 49 patients. One of two patients with failed ERP had undergone Billroth-2 gastrectomy and was excluded from the analysis, because he was unfit for the assessment of the diagnostic value of ERP. In the other patient, only cholangiography was successful, and irregular stenosis of the intrapancreatic bile duct suggested a diagnosis of pancreatic head carcinoma, which was equivocal on a CT scan. Thus, the study included a total of 86 patients, 33 males and 53 females, age 40–90 years (mean age, 65.1 years). Indications for ERP included the onset of diabetes after age 55 years in 18 patients; deterioration of diabetes in 41 patients; body weight loss in 7 patients; elevations of serum amylase and CA19-9 levels in 10 and 11 patients, respectively; and pancreatobiliary abnormalities on routine ultrasonography in 28 patients (multiple indications in 24 patients). In 13 of 28 patients with abnormal ultrasonograms, a solid or cystic mass of the pancreas was suspected. Dilatation of the biliary or pancreatic duct without a mass was seen in 12 patients, and biliary or pancreatic stones were seen in the remaining 3 patients (Table 1).

ERP was performed in 36 patients from Group A and 50 patients from Group B. The mean time from the onset of diabetes was 9.2 months (range, 0–36 months) in Group A and 132.6 months (range, 39–312 months) in Group B. The average patient age was higher in Group B compared with Group A, and fewer patients had ERP indicated due to the onset of diabetes at age > 55 years in Group B (4.0%) compared with Group A (44.4%). ERP led to a diagnosis of pancreatic carcinoma in 6 of 86 patients (7.0%): 5 of 36 patients (13.9%) in Group A and 1 of 50 patients (2.0%) in Group B. The prevalence of pancreatic carcinoma in Group A was significantly greater compared with Group B (P = 0.0442; one-tailed test) (Table 1).

Pancreatic carcinoma invariably occurred in association with the onset or deterioration of diabetes in all six patients. Five patients were in Group A and had become diabetic after age 55 years. Other indications for ERP in these patients with malignant disease included body weight loss in one patient, elevation of CA19-9 in four patients (380.5–5560.0 U/mL), and abnormal ultrasonography in four patients. All tumors diagnosed exceeded 3 cm in greatest dimension, and three tumors were resected by distal pancreatosplenectomy or pylorus-preserving pancreatoduodenectomy. Local invasion and lymph node metastasis around the superior mesenteric vessels necessitated additional resection of the transverse colon or portal vein in one patient each. One of the two patients with far advanced disease underwent bypass procedures. The remaining patient was unfit for major surgery because of age. The diagnosis of pancreatic carcinoma was confirmed by operative findings and/or clinical course with a poor prognosis (Table 2).

Table 2. Clinical Features of the Patients with Diabetes Mellitus and Pancreatic Carcinoma who were Diagnosed by Endoscopic Retrograde Pancreatography
PatientAge (yrs)GenderDuration from onset of DM (yrs/months)Indications for ERPTreatmentClinical course
  1. DM: diabetes mellitus; ERP: endoscopic retrograde pancreatography; M: male; DP: distal pancreatosplenectomy; US: ultrasound; F: female; PPPD: pylorus-preserving pancreatoduodenectomy; PV: portal vein.

151M7/6Deterioration of DM, elevation of CA19-9Involvement of the stomach liver metastasis → no surgeryDead
278M0Late onset of DM, elevation of CA19-9DPLiver metastasis → dead
390M1Late onset of DM, body weight loss, elevation of CA19-9, abnormal USNo surgeryDead
475F0Late onset of DM, elevation of CA19-9, abnormal USJaundice, duodenal invasion → bypass proceduresDead
568F1/9Late onset of DM, deterioration of DM, abnormal USPPPD and colonic resectionLiver metastasis → dead
665F2/8Late onset of DM, deterioration of DM, abnormal USPPPD and PV resectionDissemination Bone metastasis → dead

Ultrasonography showed dilatation of upstream pancreatic or bile ducts in four of six patients with pancreatic carcinoma but directly demonstrated the pancreatic head mass only in one patient and missed two tumors in the pancreatic tail (Table 3), one of which involved the stomach and metastasized to the liver at the time of the diagnosis according to CT scans and ERP findings (Table 2). Although dynamic CT scans were diagnostic in all six patients, pancreatic tumors were equivocal in only two patients (Table 3).

Table 3. Results of Cross-Sectional Imaging and Endoscopic Retrograde Pancreatography Combined with Pancreatic Juice Sampling in Patients with Diabetes Mellitus and Pancreatic Carcinoma
PatientAge (yrs)GenderLocation of tumorUSCTERPPancreatic juice
  1. US: ultrasound; CT: computed tomography; ERP: endoscopic retrograde pancreatography; CEA: carcinoembryonic antigen; M: male; Pb: pancreatic body; Ph: pancreatic head; Pt: pancreatic tail; MPD: main pancreatic duct; F: female.

151MPb, PtNegativePt mass, liver metastasisMPD obstructionClass II1.82151.9
278MPtNegativePt massMPD obstructionClass IV02820
390MPbMPD dilatationMPD dilatationMPD obstructionClass II0.76.3
475FPhMPD dilatation, biliary dilatation, Ph massMPD dilatation, biliary dilatation, Ph massDouble duct sign
568FPhMPD dilatation, biliary dilatation, gallstonesMPD dilatation, biliary dilatation, Ph swellingDouble duct sign
665FPhMPD dilatationMPD dilatation, Ph massDouble duct sign

ERP delineated irregular stenoses of the Wirsung and bile ducts in the pancreatic head (double duct sign) or obstruction of the main duct in the body or tail of the pancreas (Table 3). Precise analysis of the affected main duct and adjacent branches indicated a diagnosis of pancreatic carcinoma, as described previously.9 Findings regarded as malignant on ERP included abrupt transition to the stenosis or obstruction, deviation of the stenotic ductal lumen, irregular dilatation of peripheral branches or lack of branch filling at the stenosis, and extravasation of contrast medium around the stenosis. These findings supported the definite diagnosis of pancreatic carcinoma, even in two patients with pancreatic tumors that were equivocal on CT scans.

Pancreatic juice sampling yielded positive cytology (Class IV) in only one patient with pancreatic tail carcinoma. There was no significant disagreement between the cytologic diagnoses of the expert screener and the authorized pathologist. In three patients with pancreatic head carcinoma, severe stenosis of the Wirsung duct hampered cannulation of a balloon catheter and subsequent pancreatic juice sampling (Table 3). The CEA and CA19-9 levels in pancreatic juice ranged from 0 ng/mL to 19.1 ng/mL (mean, 1.45 ng/mL) and from 1.0 U/mL to 35,224.0 U/mL (mean, 2657.7 U/mL), respectively. In the majority of patients, the values did not reach the cut-off levels reported in the literature: 15 ng/mL for CEA and 20,000 U/mL for CA19-9.11, 12 The CA19-9 level in pancreatic juice exceeded the cut-off value (35,224.0 U/mL) in only one patient, who was indicated for ERP because of a slight elevation of CA19-9 in the serum (69.4 U/mL). Another patient had an elevation of CEA (19.1 ng/mL) in pancreatic juice but not in the serum. These two patients did not develop pancreatic carcinoma during follow-up of 45 months and 36 months, respectively. Both markers were within normal limits in the patients with pancreatic carcinoma (Table 3).

After the procedure, three patients complained of upper abdominal pain, and two of them required an antispasmodic injection. The three other patients had mild nausea. The serum amylase level ranged from 32 IU/mL to 4493 IU/mL, with a mean of 326 IU/mL, and exceeded 1000 IU/mL in four patients. The symptoms were transient, and the laboratory data returned to normal within 2 days. None of the patients were diagnosed clinically as acute pancreatitis.

During the study period of 4 years, the development of pancreatic carcinoma has been documented in none of the patients with negative ERP findings. Also, none of the other patients in the Unit of Diabetes for whom ERP was not indicated reportedly developed pancreatic carcinoma.


Previous attempts at mass screening for pancreatic carcinoma produced discouraging results with low detection rates.3, 4 In the series by Homma and Tsuchiya,4 mass screening of 10,162 patients age > 40 years yielded only 4 patients (0.04%) with pancreatic carcinoma using serum CA19-9 and elastase-1 in combination with ultrasonography, whereas 85 of 4506 outpatients (1.9%) with gastrointestinal complaints or jaundice had pancreatic carcinoma. Many reports emphasized the correlation between pancreatic carcinoma and diabetes mellitus,5–8 but pancreatic carcinoma is not frequent, even in patients with diabetes. Kessler5 documented 78 deaths (0.4%) from pancreatic carcinoma among 21,447 diabetic patients, a rate that was only approximately twice as much as the statistically standardized incidence rate in nondiabetic population. Thus, the selection of high-risk groups is essential to cost-effective screening for pancreatic carcinoma among diabetic patients. In a survey of 209 patients with pancreatic carcinoma, diabetes was present in 15.3% of patients, whereas 4.3% of patients had diabetes before the diagnosis of pancreatic carcinoma, suggesting a temporal relation from the development of pancreatic carcinoma to that of diabetes.6 The current study identified candidates for ERP using our original criteria and revealed an extremely high prevalence of pancreatic carcinoma (6 of 86 patients; 7.0%). The prevalence was especially high in patients with recent-onset diabetes within 3 years (5 of 36 patients; 13.9%). Karmody and Kyle7 reported that pancreatic carcinoma was diagnosed within 1 year from the onset of diabetes in 40 of 51 patients (78.4%) who had both diseases. Gullo et al.8 also described that 92 of 164 patients (56.1%) had become diabetic either concurrently with pancreatic carcinoma or within 2 years before they were diagnosed as pancreatic carcinoma, but those authors noted no significant association between the two conditions in patients who were diabetic for 3 years or more. Three years after the onset of diabetes seems to be golden time for the early diagnosis of pancreatic carcinoma. In our study, two patients with resectable pancreatic carcinoma had been diabetic and were kept under medical observation for 21 months and 32 months. These patients may have been diagnosed at an earlier stage if ERP had been performed immediately after the onset of diabetes.

Ultrasonography, which was adopted as a screening method in our criteria for ERP, missed advanced pancreatic tail carcinoma in two patients. Although it is well known that dynamic CT scans play an important role in cross-sectional imaging of the pancreas,1 CT scans failed to demonstrate the tumor directly in two of six patients with pancreatic carcinoma. These noninvasive modalities are incompetent for the diagnosis of minute carcinomas (< 1 cm) of the pancreas that carry an acceptable prognosis. ERP maintains its superiority in detecting such small pancreatic tumors2, 9 and provides an opportunity for pancreatic juice sampling as well, which may lead to the detection of in situ carcinoma.10 Ishikawa et al.13 resected occult pancreatic carcinoma (in situ or minimally invasive carcinoma) in 11 patients who were diagnosed by Nakaizumi et al.14 using cytology of pancreatic juice obtained at ERP. Those authors determined the location of the tumors by intraoperative cytology of pancreatic juice. Their results proved the feasibility of early detection of pancreatic carcinoma by cytodiagnosis of pancreatic juice. Successive balloon ERP and pancreatic juice sampling in the current study yielded positive cytology in only one of six patients with advanced pancreatic carcinoma. However, we previously described a patient with in situ carcinoma of the pancreatic body who was diagnosed by the same method.10 Cytologic examination should be positive more frequently at a much earlier disease stage, when malignant cells spill into the pancreatic juice in the process of enthusiastic proliferation before they form a solid mass that is detectable by cross-sectional imaging. Our original criteria for indicating ERP extracted patients with pancreatic carcinoma from diabetics at an extremely high rate, but the timing of pancreatic juice sampling was too late for the early detection of carcinoma by cytology. Our method enables fine pancreatography and sampling of a sufficient amount of pancreatic juice for the cytodiagnosis in one session with the patient in the supine position. Furthermore, this method was completed within 30 minutes and produced no serious complications related to the procedure in the current study. The optimal timing of ERP and pancreatic juice cytology for the early diagnosis of pancreatic carcinoma will be the main issue of our next study.

Measurements of CEA and CA19-9 levels in pancreatic juice were of no use for the diagnosis of pancreatic carcinoma, inconsistent with the results reported by Tatsuta et al.12 These values showed wide variation and did not aid or supplement the cytologic diagnosis. Suehara et al.15 assayed telomerase activity in pancreatic juice, a ribonucleoprotein enzyme recently featured as a marker of various malignancies, and reported much greater sensitivity and almost equal specificity compared with the sensitivity and specificity of cytology for the diagnosis of pancreatic carcinoma. Telomerase assay of pancreatic juice may be a promising diagnostic aid for the early diagnosis of pancreatic carcinoma in combination with cytology.

In conclusion, our original indications for ERP yielded an extremely high diagnostic rate of pancreatic carcinoma in patients within 3 years after the onset of diabetes. Although the patients who were diagnosed with pancreatic carcinoma by ERP invariably had advanced disease, the discouraging results may be attributable to the timing of ERP. A more aggressive diagnostic approach toward the diagnosis of pancreatic carcinoma in diabetic patients with our criteria may contribute to the earlier diagnosis of the disease.


The authors are grateful to Masahiko Yoshida, M.D., and Nobutoshi Imaizumi, M.D., and Fukuoka Red Cross Hospital for their encouragement and support. The valuable contributions of the technical and nursing staffs of Fukuoka Red Cross Hospital also are acknowledged.