Localization of Insulinomas to Regions of the Pancreas by Intra-arterial Stimulation with Calcium

Methods

Diagnosis of insulin-secreting islet cell tumor was based on the development of symptomatic hypoglycemia (blood glucose level, <40 mg/100 mL) with inappropriate plasma insulin levels during prolonged fasting. Ten of the patients were men and 15 were women; their average age was 43 years (range, 24 to 72 years). Five patients had had previous unsuccessful explorations of the pancreas, and 3 had had distal pancreatectomy during these explorations. Two patients had multiple endocrine neoplasia type I.1;0

Most of the 25 patients had had computed tomography (n = 23), magnetic resonance imaging (n = 21), and ultrasonography (n = 22) before having arteriography with calcium stimulation. The first 9 patients had portal venous sampling, but this procedure was not done in the other 16 patients because analysis showed that calcium stimulation provided similar information with less morbidity. This decision was supported by the similar sensitivities of portal venous sampling and intra-arterial secretin stimulation in our patients with the Zollinger-Ellison syndrome [5, 6].

Computed tomography (done using a 9800 HiLite, General Electric, Milwaukee, Wisconsin) was done with 5-mm contiguous sections through the pancreas during the bolus injection of 130 mL of iodinated contrast material (iopamidol [Isovue 300, Bristol-Myers Squibb, Princeton, New Jersey]) at 2 mL per second. Magnetic resonance imaging was done using a 0.5-Tesla scanner (Picker, Highland Heights, Ohio) with 10-mm thick axial T1-weighted (repetition time [TR]/echo time [TE] = 300/10) and short inversion time inversion recovery (STIR) (TR/TI [inversion time]/TE = 1800-2200/100/30) sequences. Gadopentetate dimeglumine (Magnevist, Berlex Lab, Wayne, New Jersey) was not given. Ultrasonography was done using a 3.5- or 5-MHz phased-array sector transducer (Acuson, Mountain View, California).

Pancreatic arteriography was done by selectively injecting nonionic contrast agent (Isovue 300) into the gastroduodenal, splenic, and superior mesenteric arteries. Care was taken to position the catheter at the origin of these vessels so that major pancreatic arteries originating proximally from these vessels, such as the dorsal pancreatic and pancreatic magna arteries, would be perfused. Selective arteriography of the dorsal pancreatic and pancreatic magna arteries was occasionally done, but we did not infuse calcium into these small pancreatic branches because we feared that doing so might increase the risk for pancreatitis. After each selective arteriogram, calcium gluconate 10% (Lyphomed, Rosemont, Illinois), diluted with saline to a volume of 5 mL, was injected into the selectively catheterized artery at a dose of 0.025 mEq Ca^++/kg body weight. Blood samples (5 mL) for insulin determination were obtained from the right (n = 25) and left (n = 17) hepatic veins before and 30, 60, and 120 seconds after calcium infusion. Specimens from the hepatic veins were placed on ice, and plasma was separated in a refrigerated centrifuge and stored at −20°C until insulin levels were measured by radioimmunoassay.

Samples were obtained from the left as well as the right hepatic vein in the first 17 patients because of concern that an insulinoma in the body or tail of the pancreas might be overlooked if splenic venous effluent streamed into the left hepatic lobe. However, it is more difficult to place and maintain a catheter in the left than in the right hepatic vein. To determine whether diagnostic elevations of insulin levels were ever seen only in the left hepatic vein, we compared insulin levels in the right and left hepatic veins in a subset of 10 patients whose insulinomas were in the pancreatic body and tail.

The insulinomas ranged in size from 6 to 25 mm (average, 15 mm). Twelve were located to the right of the superior mesenteric artery (pancreatic head and neck), and 13 were located to the left (pancreatic body and tail). All tumors of the head and neck were enucleated. Tumors of the body and tail were removed by enucleation (n = 5) or distal pancreatectomy (n = 8). Intraoperative ultrasonography (10-MHz transducer, Diasonics, Santa Clara, California) was done in each patient to visualize the tumor, to identify major pancreatic and biliary ducts adjacent to the tumor, and to direct the pancreatic incision for enucleation. All patients were cured.

Data Analysis

The results of sampling from the right (n = 25) and left (n = 17) hepatic veins were plotted for each patient. Graphs were analyzed by selecting the greatest insulin response in a given vessel in the 30- or 60-second sample after injection. Each patient was coded so that, at the time of analysis, the observers were unaware of the results of any other localizing studies or of the location of the tumor at surgery. A response after calcium infusion into the gastroduodenal or superior mesenteric artery localized the adenoma to the head and neck of the pancreas; a response after splenic artery injection localized the adenoma to the body and tail of the pancreas. A response to calcium stimulation usually involved a single artery (Figure 1). When both the gastroduodenal and superior mesenteric arteries showed a response to calcium stimulation, the insulinoma was presumed to be located to the right of the superior mesenteric artery (pancreatic head and neck) (Figure 2). When no vessel was clearly dominant, the response was considered nonlocalizing (Figure 3).

View larger version:

• In this window
• In a new window

Figure 1. Note the eightfold gradient at the 30-second sample from the right hepatic vein ( ) and the rapid decrease in insulin levels at the 60-second sample; this is a frequent, but not invariable, pattern. Lesser gradients are measured in the samples from the left hepatic vein ( ). The conversion factor for insulin levels into pmol/L is 7.175. Typical sampling results from a patient with an insulinoma in the pancreatic tail.topbottom

View larger version:

• In this window
• In a new window

Figure 2. Although the gradients in the left hepatic vein ( ) are substantial, the elevations of insulin levels in the right hepatic vein are higher (sixfold) than those in the left hepatic vein (fourfold). The conversion factor for insulin levels into pmol/L is 7.175. In a patient with an insulinoma of the pancreatic head, greater than twofold gradients were seen after calcium injection into both the gastroduodenal and superior mesenteric arteries, with higher elevations in the gastroduodenal artery (top).bottom

View larger version:

• In this window
• In a new window

Figure 3. Although the 30-second gradient is higher in the splenic vein, this is considered a nonlocalizing study. The tumor was found in the body of the pancreas to the left of the superior mesenteric artery. The conversion factor for insulin levels into pmol/L is 7.175. The only nondiagnostic study in the last 20 cases shows elevated insulin levels in the splenic and gastroduodenal arteries.

The sensitivity of calcium stimulation in all 25 patients was calculated and compared with the sensitivity of the noninvasive imaging studies (computed tomography, magnetic resonance imaging, and ultrasonography) and arteriography. Specificity was irrelevant because all patients in the series had proven insulinomas. In the 9 patients who had portal venous sampling, the sensitivity of calcium stimulation was compared with the sensitivity of portal venous sampling.

To determine whether it was necessary to sample the left hepatic vein, we compared the maximum insulin levels in the right and the left hepatic veins and the ratio of insulin levels in the hepatic vein with those in the peripheral vein in a subset of 10 patients with insulinomas of the body and tail.

Results

The results of all localization studies are summarized in Table 1. A response to calcium stimulation—that is, a greater than twofold elevation of insulin levels in the right or left hepatic vein on the 30- or 60-second samples—occurred in all 25 patients. Calcium stimulation with venous sampling correctly predicted the site of the insulinoma in 22 of 25 patients (sensitivity, 88% [95% CI, 68% to 97%]). In 2 of the 3 patients with false localizations, responses to gastroduodenal and splenic artery injections occurred in the presence of a tumor in the proximal body of the pancreas (Figure 3); in the third patient, a response to a superior mesenteric artery injection occurred in the presence of a tumor in the proximal body. All patients who had a positive response to splenic artery injection only had insulinomas of the body or tail. Two of the three false localizations occurred in our first 5 patients; only one false localization occurred among our last 20 patients.

In the nine patients who had both portal venous sampling and calcium stimulation, portal venous sampling correctly localized six insulinomas (sensitivity, 67%), and calcium stimulation correctly localized seven insulinomas (sensitivity, 78%).

Among 10 patients with surgically proven insulinomas of the body and tail of the pancreas, the maximum insulin levels in response to calcium stimulation were higher in the right than in the left hepatic vein in 8 patients and were equal in the right and left hepatic veins in 1 patient (103 µU/mL compared with 107 µU/mL [739 pmol/L compared with 768 pmol/L]). Only 1 patient with an insulinoma of the pancreatic body had a higher insulin level in the left than in the right hepatic vein (148 µU/L compared with 108 µU/mL [1062 pmol/L compared with 775 pmol/L]), but the ratio of the insulin level in the hepatic vein to the insulin level in the peripheral vein was diagnostic in both the right (6.7) and the left (4.7) hepatic veins. Nine of the 10 patients had higher ratios in the right hepatic vein samples (Table 2). No insulinoma of the body or tail would have been missed by sampling only the right hepatic vein.

Aside from a mild and transient sensation of warmth in the epigastrium during calcium infusion, no complications occurred. Neither measurable hypercalcemia (in the first 5 patients) nor symptomatic hypoglycemia were encountered.

Calcium stimulation with venous sampling was done in two patients not included in this series. In one patient with multiple endocrine neoplasia type I who had had distal pancreatectomy for an insulinoma of the pancreatic tail, recurrent hypoglycemia with a tumor of the pancreatic head was seen on computed tomography, magnetic resonance imaging, and arteriography. The results of a calcium stimulation test and portal venous sampling were also positive for an insulinoma of the head of the pancreas. The patient elected not to have surgery, and the hypoglycemia was controlled medically. In the other patient, the calcium stimulation test was interrupted by equipment failure after only a single vessel (the gastroduodenal artery) was injected (negative response). The results of subsequent portal venous sampling were positive in the splenic vein samples. Arteriography with calcium stimulation was not repeated. A 17-mm insulinoma was resected from the pancreatic tail.

Discussion

In this series of 25 patients with surgically proven insulinomas, calcium stimulation with hepatic venous sampling was the most sensitive of the preoperative localizing studies (22 of 25 [88%]). Arteriographic demonstration of a tumor was not necessary for a positive response to a selective calcium infusion into the same vessel. In 13 of 16 patients with negative arteriograms, the injection of calcium into the same vessel correctly localized an adenoma by the insulin response in hepatic vein samples. Calcium stimulation with hepatic venous sampling was similar to portal venous sampling, which was thought not to be clinically indicated after the results in our first 9 patients were reviewed [4]. This is also true in our larger experience with gastrin-producing islet cell tumors [5, 6]; we have abandoned portal vein sampling in patients with the Zollinger-Ellison syndrome because the procedure is more invasive and painful and is no more sensitive than intra-arterial secretin stimulation [7].

Referral of patients to the National Institutes of Health introduces a bias because a selection toward occult tumors takes place [8]. All patients had had negative or equivocal noninvasive imaging studies before referral to the National Institutes of Health, and many had had nondiagnostic arteriography. This bias accounts for the low sensitivities of computed tomography, ultrasonography, and magnetic resonance imaging in our study compared with sensitivities in other published series [9, 10]. Many patients were referred to our institution specifically because of our extensive experience in portal venous sampling. The calcium stimulation test, done at the same time as standard pancreatic arteriography, provides the same information without requiring special skills or experience.

Samples from the right hepatic vein always contained diagnostic elevations of insulin in the presence of tumors in the body and tail of the pancreas. In fact, maximum insulin levels in the right hepatic vein were higher in 80% and ratios of insulin levels in the right hepatic vein to those in the peripheral vein were higher in 90% of patients with tumors of the body and tail of the pancreas. No preferential streaming of the splenic vein effluent into the left branch of the portal vein was evident. A few observations show that it is not necessary to obtain samples from the left hepatic vein, which is more difficult to catheterize than the right hepatic vein. We recommend that a single catheter be placed in the right hepatic vein for arterial stimulation studies in patients with insulin-producing islet cell tumors.

An intravenous calcium stimulation test has been proposed for use in the diagnosis of insulinoma [11-15], although varying sensitivities have been reported [16-19]. Brunt and colleagues [11] showed that calcium is a more effective secretagogue in patients with insulinomas when it is administered as a rapid intravenous bolus (2 mg Ca^++/kg over 1 minute) than when it is given as a prolonged infusion (12 mg Ca^++/kg over 3 hours). We used one tenth of the “rapid intravenous bolus” dose (0.2 mg Ca^++/kg) injected over 3 to 5 seconds directly into the artery perfusing the islet cell tumor, thus providing a much more intense calcium exposure over a short period of time. One of our concerns was that calcium stimulation might cause a massive release of insulin and severe, even life-threatening, hypoglycemia. However, insulin levels in the hepatic vein usually peaked in the samples taken 60 seconds after calcium stimulation and were often returning toward baseline by 120 seconds after stimulation, suggesting that the release of insulin is confined to the short interval during which β cells are exposed to high serum calcium levels.

This failure of intravenous calcium infusion to stimulate the release of insulin from normal β cells suggests that intra-arterial calcium infusion may provide clues to the diagnosis of factitious hypoglycemia caused by sulfonylurea abuse. One patient not included in this series did not show any insulin gradients in hepatic vein samples after intra-arterial calcium stimulation and was later found to have factitious hypoglycemia caused by sulfonylurea ingestion. Portal venous sampling does not provide specificity; elevations of insulin levels several times above baseline can be found in samples from many sites in the splenic and portal veins of patients without insulinomas (Jensen R and Norton J. Unpublished data).

One patient in this series was receiving diazoxide when the calcium stimulation study was done and showed a prompt elevation of insulin levels after injection of calcium into the gastroduodenal artery. However, we generally prefer that patients not be receiving diazoxide at the time of calcium stimulation. In addition, patients who have substantial cardiac disease or who are receiving cardiac glycoside therapy should be closely monitored during calcium infusion studies. However, the total amounts of calcium administered are insignificant (5.25 mEq in a patient weighing 70 kg over about 30 minutes). None of our first five patients showed any change in serum calcium levels during the study.

Calcium stimulation with venous sampling divides the pancreas into two components, one to the right and one to the left of the superior mesenteric artery. This is similar to the regionalization provided by portal venous sampling. Such a partition has surgical relevance because adenomas to the left of the superior mesenteric artery can be treated by enucleation or by distal pancreatectomy, whereas tumors to the right of the superior mesenteric artery must be localized for enucleation. In these latter tumors, the availability of practitioners experienced in intraoperative ultrasonography is critical because adenomas smaller than 15 mm may be difficult to palpate.

Because of the sensitivity of intraoperative ultrasonography (100% in our study and more than 80% in most reported series [1, 8, 10, 20]), one may ask whether any localization studies are necessary before surgery. At the National Institutes of Health, patients with insulinomas have noninvasive imaging (ultrasonography, computed tomography, magnetic resonance imaging, and octreotide scanning), and, if the results of two or more of these studies are positive at the same site, pancreatic arteriography with calcium stimulation is not done. However, if results of the noninvasive examinations are negative or equivocal, the insulinoma tends to be small and intraoperative localization may be difficult, particularly in hospitals with limited experience in intraoperative ultrasonography. Under such circumstances, we consider intra-arterial stimulation with calcium gluconate—a test that can be competently done at most community hospitals—to be indicated because of its simplicity and high sensitivity.

Our results indicate that calcium stimulation with hepatic venous sampling is as sensitive as and less invasive than portal venous sampling and, thus, should replace portal venous sampling as a technique to localize occult insulinomas. If positive results on a calcium stimulation test indicate an insulinoma of the pancreatic body or tail that cannot be identified intraoperatively, our data suggest that a blind distal pancreatectomy in that setting would be justified.

Addendum

Since this paper was submitted, we studied three more patients with surgically proven insulinomas. In all three patients, the insulinomas were localized correctly (sensitivity for the entire group of 28 patients, 89% [25/28]; CI, 71% to 98%).