CT Features of Pancreatic Fistula After Pancreaticoduodenectomy

04_05_1174_Hashimoto.fm — 3/1/07

Hashimoto et al.

Abdominal Imaging • Clinical Obser vations

Downloaded from www.ajronline.org by 61.11.148.242 on 11/23/13 from IP address 61.11.148.242. Copyright ARRS. For personal use only; all rights reserved

CT of Pancreatic Fistula

CT Features of Pancreatic Fistula After Pancreaticoduodenectomy Manabu Hashimoto1 Makoto Koga1 Koichi Ishiyama1 Jiro Watarai1 Satoshi Shibata2 Tutomu Sato2 Yuzo Yamamoto2 Hashimoto M, Koga M, Ishiyama K, et al.

Keywords: abdominal imaging, CT, pancreatic fistula, pancreaticobiliary imaging DOI:10.2214/AJR.05.1174 Received July 7, 2005; accepted after revision November 3, 2005. 1Department

of Radiology, Akita University School of Medicine, 1-1-1 Hondo, Akita, Akita 010-8543, Japan. Address correspondence to M. Hashimoto ([email protected]).

2Department

of Surgery, Akita University School of Medicine, Akita, Japan. WEB This is a Web exclusive article.

AJR 2007; 188:W323–W327 0361–803X/07/1884–W323 © American Roentgen Ray Society

AJR:188, April 2007

OBJECTIVE. Our objective was to establish the CT features that are indicative of pancreatic fistula after pancreaticoduodenectomy. CONCLUSION. A fluid collection seen on CT around the pancreaticojejunostomy site and in the pancreatic bed may be caused by pancreatic fistula in patients who have undergone pancreaticoduodenectomy. CT depiction of air bubbles in the fluid at these sites may strongly suggest the diagnosis of pancreatic fistula. ancreaticoduodenectomy is often indicated for peripancreatic neoplasms, chronic pancreatitis, and abdominal trauma. It consists of resection of the pancreatic head, duodenum, a short segment of the jejunum, and the gastric antrum followed by hepaticojejunostomy, pancreaticojejunostomy, and gastrojejunostomy. Because it is a complex and invasive procedure, morbidity and mortality rates have been high [1, 2]. The mortality rate has decreased to less than 4% [3, 4], but postprocedural complications leading to a prolonged hospital stay or fatal outcome occur in 30–40% of patients [3–5]. Many surgical reports [6–8] describe pancreatic fistula (leak from pancreaticojejunostomy) as the most frequent complication (10–24%) and one of the causes of delayed hemorrhage after pancreaticoduodenectomy [6, 8]. In most institutions, CT is performed for evaluating tumor recurrence and complications after pancreaticoduodenectomy. Many radiology reports have described postoperative changes and complications after pancreaticoduodenectomy [9–13]. However, those reports did not focus on the CT features of the most frequent complication, pancreatic fistula. Thus, we retrospectively compared CT findings in patients with and without pancreatic fistula and herein describe the features of pancreatic fistula occurring after pancreaticoduodenectomy.

P

Materials and Methods Patients From April 1999 through December 2003, 51 patients underwent pancreaticoduodenectomy at our institution. Indications for surgery included pancreatic

head tumor (n = 18), ampullary carcinoma (n = 8), bile duct carcinoma (n = 22), gallbladder carcinoma (n = 2), and trauma (n = 1). The classic Whipple procedure (involving removal of the gastric antrum) was performed in 46 patients, pylorus-preserving pancreaticoduodenectomy was performed in four, and hepatopancreaticoduodenectomy was performed in one. External drainage catheters were placed around the pancreaticojejunostomy site and hepaticojejunostomy site in all cases. Stent catheters were used for the hepaticojejunostomy and pancreaticojejunostomy; external transhepatic drains were also placed. Eight patients underwent intraoperative radiation therapy, two underwent combined partial hepatic resection of the hepatic hilum, and two underwent combined resection of the portal vein or the superior mesenteric vein. Forty-one of the 51 patients (17 women and 24 men; mean age, 63 years; range, 43–78 years) underwent CT within 30 days after surgery for evaluation of postoperative complications. Complications were suspected in all 41 patients; they had fever or purulent exudate from one of the intraoperatively placed surgical bed drains.

Definition of Postoperative Complications Pancreatic fistula was defined as prolonged or elevated output of amylase-rich fluid through an intraoperatively placed drain (> 3 times normal serum amylase level). Biliary fistula was defined as exudation of fluid containing bile juice through a surgical drain. Enterocolitis, remnant pancreatitis, liver infarction, and wound infection were diagnosed on the basis of clinical, laboratory, and imaging findings. Absence of complications was determined on the basis of the clinical course and laboratory test results. Short-term fever without abdominal pain,

W323

04_05_1174_Hashimoto.fm — 3/1/07

Downloaded from www.ajronline.org by 61.11.148.242 on 11/23/13 from IP address 61.11.148.242. Copyright ARRS. For personal use only; all rights reserved

Hashimoto et al.

A

B

Fig. 1—CT in 49-year-old man 8 days after surgery shows pancreatic fistula that was confirmed by laparotomy and pancreaticojejunostomy and fluid collection around pancreaticojejunostomy. A, CT scan shows pancreaticojejunostomy site (large arrows). Note catheter used to stent pancreaticojejunostomy (small arrow) allows accurate identification of pancreatic anastomosis. B, CT scan obtained just caudal to A shows fluid collection with air bubbles around pancreaticojejunostomy (white arrows). Fluid collection in pancreatic bed is also seen (black arrow).

A

B

Fig. 2—CT in 59-year-old man 16 days after surgery shows fluid collection around pancreaticojejunostomy. Pancreatic fistula was confirmed by pancreatography via stent catheter for pancreaticojejunostomy. A, CT scan shows fluid collection with air bubbles around pancreatic anastomosis (arrows). B, CT scan obtained just cephalad to A shows fluid collection with air bubbles around pancreatic anastomosis (arrows).

abdominal irritability, and increased drainage from a surgical drain were not considered signs of complication. Purulent discharge (< 3 times the normal serum amylase level) was classified as absence of complication if noted only once or twice.

CT Helical CT was performed with a Plus-4 scanner (Siemens Medical Solutions) before and after the administration of iodinated contrast medium. Enhanced CT began approximately 70–90 seconds after the start of mechanical injection of 100 mL of contrast medium (300 mg I/mL) at a rate of 2 mL/s. CT scans were obtained with 7 mm/s cephalocau-

W324

dal table movement and 5-mm collimation, with image reconstruction every 5 mm from the dome of the diaphragm to the iliac crest. When possible, patients were instructed to hold their breath during the helical CT. The CT scans were assessed for the presence of a fluid collection with or without air bubbles. Fluid collection was classified as focal or diffuse. The pancreaticoduodenectomy site was easily detected by the surgically placed pancreaticojejunostomy stent catheter. The sites of focal fluid collection were identified as the area around the pancreaticojejunostomy site (Figs. 1 and 2), the dead space around the pancreatic head (pancreatic

bed) (Figs. 1 and 3), the area around or along the hepaticojejunostomy site (Fig. 4), or other sites. All images were reviewed by two independent radiologists, and any differences in opinion were resolved by consensus.

Statistical Analysis Values are expressed as mean ± SD. Welch’s t test (an adaptation of the Student’s t test) was used to analyze differences in mean values between the two groups. Differences in absolute frequencies were analyzed using Fisher’s exact probability test. A p value of < 0.05 was considered statistically significant.

AJR:188, April 2007

04_05_1174_Hashimoto.fm — 3/1/07

Downloaded from www.ajronline.org by 61.11.148.242 on 11/23/13 from IP address 61.11.148.242. Copyright ARRS. For personal use only; all rights reserved

CT of Pancreatic Fistula

Fig. 3—CT scan of 62-year-old man 7 days after surgery shows fluid collection in pancreatic bed (arrow). This patient had pancreatic fistula.

Results Patients Of the 41 study patients, 23 had a pancreatic fistula and 18 did not (Table 1). Two of 23 patients with pancreatic fistula also had a biliary fistula, two had severe remnant pancreatitis, one had a small liver infarction, and two had a tiny wound infection. The remaining 16 patients had pancreatic fistula alone. In the 23 patients with a pancreatic fistula, it was confirmed by extravasation of contrast medium from the pancreaticojejunostomy via one of the indwelling catheters (n = 11) or by laparotomy (n = 2). Five of the 18 patients without a pancreatic fistula had other complications. Two of these five had severe enterocolitis, two had a biliary fistula, and one had an extensive liver infarction. The remaining 13 patients did not have any complications. We confirmed a biliary fistula by the presence of extravasation of contrast material from the hepaticojejunostomy site by means of direct injection of contrast medium through the hepaticojejunostomy stent catheter in three of the four patients with biliary fistula. CT Findings The time between surgery and CT in patients with pancreatic fistula ranged from 6 to 28 days (mean, 15.1 ± 6.8 days) and in patients without fistula ranged from 4 to 30 days (mean, 15.8 ± 8.4 days, p = 0.79). Fluid around the pancreaticojejunostomy site (Figs. 1 and 2) was found in 20 patients with a pancreatic fistula and in two patients without a fistula (p < 0.0001). Fluid in the pancreatic bed (Figs. 1 and 3) was noted in 18

AJR:188, April 2007

Fig. 4—CT scan of 70-year-old man 8 days after surgery shows fluid collection around hepaticojejunostomy site (arrow). This patient had no complications.

patients with a pancreatic fistula and in six without a fistula (p = 0.005). Patients with a pancreatic fistula showed fluid either around the pancreaticojejunostomy site (n = 5), in the pancreatic bed (n = 3), or both (n = 15). Fluid around the hepaticojejunostomy site (Fig. 4) was found in 16 patients with a pancreatic fistula and in 13 patients without a pancreatic fistula (p = 0.566). A fluid collection other than in the pancreatic bed, around the pancreaticojejunostomy site, or around the hepaticojejunostomy site was found in 18 patients with a pancreatic fistula and in 15 patients without a fistula (p = 0.50). Among patients with a pancreatic fistula, we found air bubbles in the fluid around the pancreaticojejunostomy site in 17 patients, in the pancreatic bed in nine, and around the hepaticojejunostomy site in one (this patient also had a biliary fistula). Among patients without a pancreatic fistula, air bubbles were found in the fluid around the pancreatic bed in one patient (this patient had a biliary fistula). Air bubbles in the fluid at other sites were seen in six patients with a pancreatic fistula and in five without a fistula (these five patients did not have any complications). Fluid collection around the hepaticojejunostomy site was found in all four patients with a biliary fistula. We found fluid collection with air bubbles around the pancreaticojejunostomy site in two patients (these patients also had a pancreatic fistula). Three patients had a fluid collection in the pancreatic bed, and three had a fluid collection in another site. Three patients without any complications did not have a fluid collection at any site on the CT scan obtained

after surgery (14, 16, and 30 days, respectively). No patient had diffuse massive ascites. Collections of fluid at other sites were somewhat large in patients with remnant pancreatitis and enterocolitis and in the patient who underwent hepatopancreaticoduodenectomy. The mean postoperative hospital stay was 54.6 ± 22.7 days (range, 25–115 days) for patients with a pancreatic fistula and 37.8 ± 13.3 days (range, 26–73 days) for patients without a fistula (p = 0.005). All patients except two who underwent laparotomy for a pancreatic fistula were discharged without any intervention. The presence of a pancreatic fistula or other complication was not associated with death. Discussion According to surgical reports, delayed arterial hemorrhage is the most critical complication after pancreaticoduodenectomy; it is associated with a high mortality rate (14–38%) [8, 14–16]. Local sepsis resulting from pancreatic fistula remains the main cause of the delayed hemorrhage [5, 6, 8, 14, 15, 17]. Radiologic studies have suggested that fluid collection in the pancreatic bed is a normal postoperative change and that anastomotic leaks cannot be diagnosed reliably with CT [9, 11–13]. However, those reports were based on a small number of patients, and statistical comparisons were not performed of CT findings between patients with and those without a fistula. If pancreatic fistula can be diagnosed correctly with CT, not only will early diagnosis be possible, but early management and prevention of arterial hemorrhage will also be possible.

W325

04_05_1174_Hashimoto.fm — 3/1/07

Hashimoto et al. TABLE 1: CT Findings in 41 Study Patients With and Without a Pancreatic Fistula

Downloaded from www.ajronline.org by 61.11.148.242 on 11/23/13 from IP address 61.11.148.242. Copyright ARRS. For personal use only; all rights reserved

CT Finding

Pancreatic Fistula (n = 23)

No Pancreatic Fistula (n = 18)

Fluid around pancreaticojejunostomy

20 (17)

2

Fluid in pancreatic bed

18 (9)

6 (1)

Fluid around hepaticojejunostomy

16 (1)

13

Fluid in other sites

18 (6)

15 (5)

54.6 ± 22.7 days

37.8 ± 13.3 days

Hospital stay after surgery (mean ± SD)

p < 0.0001 0.005 0.57 0.5 0.005

Note—Numbers in parentheses are numbers of patients with air in fluid.

We compared CT findings obtained within 30 days after pancreaticoduodenectomy among a relatively large number of patients with and without pancreatic fistula. No significant difference was seen in the time between surgery and CT; the mean interval was 15–16 days in both groups. CT features differed significantly between groups. We found that a fluid collection around the pancreaticojejunostomy site or in the pancreatic bed was significantly associated with a pancreatic fistula. Under postoperative conditions without a pancreatic fistula, a fluid collection does not usually accumulate in these areas. When air bubbles are observed in these areas, the fluid collection is due to a pancreatic fistula in most cases. We speculate that a fluid collection with air bubbles around the hepaticojejunostomy site is a specific CT finding for biliary fistula, although this finding was noted in only one patient with pancreatic and biliary fistulas. We were unable to find CT features of biliary fistulas; examination of many more cases is necessary to determine the significant CT findings. However, we may be able to conclude from our results that a biliary fistula can be a cause of fluid collection (with air bubbles) in the pancreatic bed or around the hepaticojejunostomy. Air bubbles in the pancreatic bed may come from the external drainage catheter, but air bubbles in this site were found only in patients with a pancreatic fistula, a biliary fistula, or both. Air bubbles in the fluid at other sites in patients without a pancreatic fistula (n = 5, 27.8%) were more frequent than that in patients with a fistula (n = 6, 26.1%). The five patients without a fistula did not have any complications. Air bubbles in other sites may come from drainage catheters. Of the 21 patients with a pancreatic fistula who were discharged without any intervention, nine had a postoperative hospital stay longer than 50 days due to persistence of the fistula. CT features in these nine patients and in the remaining 12 patients did not differ. Hence, we

W326

cannot predict the persistence of a pancreatic fistula on the basis of the initial postoperative CT findings. As in our cases, most patients with a low-output pancreatic fistula are managed by maintenance of the peripancreatic drains placed intraoperatively [6, 18]. We performed follow-up CT before discharge in 10 conservatively treated patients. Localized fluid collection at each site was decreased in all patients. In one patient, fluid collection around the anastomotic site and in the pancreatic bed had completely disappeared on the CT scan obtained 21 days after surgery. Follow-up CT should be performed in patients who have a fluid collection with air bubbles in these areas, especially when the patient is symptomatic (leucocytosis, peritoneal irritability, or high fever). If fluids and symptoms persist, surgical or percutaneous intervention may be indicated [15, 18]. In our series, two patients with pancreatic fistula underwent repeated laparotomy. One patient had septic complications and the other had delayed arterial hemorrhage. CT showed fluid collections with air bubbles around the pancreaticojejunostomy site and in the pancreatic bed, although the drainage catheter worked well and the drain output was high in both patients. Another patient suffered delayed arterial hemorrhage; this patient had a highoutput pancreatic fistula. We performed arterial embolization in this patient. CT findings were the same as in the previous two patients. Pancreatic fistula may be severe and lead to life-threatening complications when a fluid collection with air bubbles is seen in these sites in patients who have a relatively large amount of drainage fluid. When fluid caused by a pancreatic fistula locates around major vessels, such as the common hepatic artery or the proper hepatic artery, which are dissected during surgery, we must pay attention to any arterial wall change, such as a pseudoaneurysm [6, 8, 19]. If possible, we should perform arterial dominant phase thin-slice CT in such cases.

Our study has some limitations. First, we reviewed the CT scans of patients clinically suspected of having complications. We could not reliably identify a fluid collection around other sites as a normal part of the postoperative course. However, fluid collections around the pancreaticojejunostomy site and in the pancreatic bed were more important findings suggesting pancreatic fistula. We believe that fluid collections only at other sites around the hepaticojejunostomy site are not a chief CT finding predictive of fistula. Second, 10 of the 23 patients with a pancreatic fistula were diagnosed as having a fistula on the basis of prolonged elevation of the amylase level of the drainage fluid alone. The incidence of pancreatic fistula in this study was quite high, even in comparison with other series in which the incidence was greater than 10% [8, 15, 16]. This is because we included patients just barely meeting the criteria for pancreatic fistula. However, those patients had fever, abdominal pain, or abdominal tenderness. The definition of pancreatic fistula based on the drainage fluid varies from author to author [6, 8, 15, 16, 18]. Pancreatic fistula has been defined as an amylase level in the drainage fluid of 3–10 times the normal blood value [6, 8, 15, 16]. It is set somewhat subjectively by each author. Third, other types of fistula, such as hepaticojejunostomy and gastroenterostomy fistulas, may produce the same CT features that are produced by a pancreatic fistula. However, it is important to know that fluid collection around the pancreaticojejunostomy site or in the pancreatic bed may be due to a pancreatic fistula. In conclusion, fluid collection around the pancreaticojejunostomy site and in the pancreatic bed may be caused by a pancreatic fistula in patients who have undergone pancreaticoduodenectomy. CT depiction of air bubbles in the fluid at these sites may strongly suggest the diagnosis of pancreatic fistula.

References 1. Crile G Jr, Isbister WH, Hawk WA. Carcinoma of the ampulla of Vater and terminal bile and pancreatic duct. Surg Gynecol Obstet 1970; 131:1052–1054 2. Shapiro TM. Adenocarcinoma of the pancreas: a statistical analysis of bypass vs Whipple resection in good risk patients. Ann Surg 1975; 182:715–721 3. Cameron JL, Pitt HA, Yeo CJ, Lillemoe KD, Kaufman HS, Coleman J. One hundred and forty-five consecutive pancreaticoduodenectomies without mortality. Ann Surg 1993; 217:430–438

AJR:188, April 2007

04_05_1174_Hashimoto.fm — 3/1/07

Downloaded from www.ajronline.org by 61.11.148.242 on 11/23/13 from IP address 61.11.148.242. Copyright ARRS. For personal use only; all rights reserved

CT of Pancreatic Fistula 4. Crist DW, Siztmann JV, Cameron JL. Improved hospital morbidity, mortality and survival after Whipple procedure. Ann Surg 1987; 206:358–365 5. Miedema BW, Sarr MG, van Heerden JA, et al. Complications following pancreaticoduodenectomy: current management. Arch Surg 1992; 127:945–949, discussion 949–950 6. Sato N, Yamaguchi K, Shimizu S, et al. Coil embolization of bleeding visceral pseudoaneurysm following pancreatectomy: the importance of early angiography. Arch Surg 1998; 133:1099–1102 7. Grace PA, Pitt HA, Tompkins RK, DenBesten L, Longmire WP. Decreased morbidity and mortality rates after pancreaticoduodenectomy. Am J Surg 1986; 151:141–149 8. Choi SH, Moon HJ, Heo JS, Joh JW, Kim Y. Delayed hemorrhage after pancreaticoduodenectomy. J Am Coll Surg 2004; 199:186–191 9. Coombs RJ, Zeiss J, Howard JM, Thomford NR, Merrick HW. CT of the abdomen after the Whipple procedure: value in depicting postoperative anatomy, surgical complications, and tumor recurrence.

AJR:188, April 2007

AJR 1990; 154:1011–1014 10. Bluemke DA, Fishman EK, Kuhlman J. CT evaluation following Whipple procedure: potential pitfalls in interpretation. J Comput Assist Tomogr 1992; 16:704–708 11. Furukawa H, Kosuge T, Shimada K, Yamamoto J, Ushio K. Helical CT of the abdomen after pancreaticoduodenectomy: usefulness for detecting postoperative complications. Hepatogastroenterology 1997; 44:849–855 12. Mortele KJ, Lemmerling M, de Hemptinne B, De Vos M, De Bock G, Kunnen M. Postoperative findings following the Whipple procedure: determination of prevalence and morphologic abdominal CT features. Eur Radiol 2000; 10:123–128 13. Lepanto L, Gianfelice D, Dery R, Dagenais M, Lapointe R, Roy A. Postoperative changes, complications, and recurrent disease after Whipple’s operation: CT features. AJR 1994; 163:841–846 14. Rumstadt B, Schwab M, Korth P, et al. Hemorrhage after pancreatoduodenectomy. Ann Surg 1998; 227:236–241

15. Bottger TC, Junginger T. Factors influencing morbidity and mortality after pancreaticoduodenectomy: critical analysis of 221 resections. World J Surg 1999; 23:164–172 16. van Berge Henegouwen MI, De Wit LT, Van Guilk TM, Obertop H, Gouma DJ. Incidence, risk factors, and treatment of pancreatic leakage after pancreaticoduodenectomy: drainage versus resection of the pancreatic remnant. J Am Coll Surg 1997; 185:18–24 17. de Castro SMM, Kuhlmann KFD, Busch ORC, et al. Delayed massive hemorrhage after pancreatic and biliary surgery: embolization or surgery? Ann Surg 2005; 241:85–91 18. Buchler MW, Friess H, Wagner M, Kulli C, Wagner V, Z’Graggen K. Pancreatic fistula after pancreatic head resection. Br J Surg 2000; 87:883–889 19. Santoro R, Carlini M, Carboni F, Nicolas C, Santoro E. Delayed massive arterial hemorrhage after pancreaticoduodenectomy for cancer: management of a life-threatening complication. Hepatogastroenterology 2003; 50:2199–2204

W327