Life-threatening postoperative pancreatic fistula (grade C) after pancreaticoduodenectomy: incidence, prognosis, and risk factors

The American Journal of Surgery (2009) 197, 702-709

Clinical Surgery-International

Life-threatening postoperative pancreatic fistula (grade C) after pancreaticoduodenectomy: incidence, prognosis, and risk factors David Fuks, M.D.a, Guillaume Piessen, M.D.b, Emmanuel Huet, M.D.c, Marion Tavernier, M.D.d, Philippe Zerbib, M.D.e, Francis Michot, M.D.c, Michel Scotte, M.D., Ph.D.c, Jean-Pierre Triboulet, M.D.b, Christophe Mariette, M.D., Ph.D.b, Laurence Chiche, M.D.d, Ephraïm Salame, M.D.d, Philippe Segol, M.D.d, François-René Pruvot, M.D., Ph.D.e, François Mauvais, M.D.a, Horace Roman, M.D.c, Pierre Verhaeghe, M.D.a, Jean-Marc Regimbeau, M.D., Ph.D.a,* a

Federation of Digestive Diseases, Amiens North Hospital, University of Picardy Medical Centre, Place Victor Pauchet, F-80054 Amiens cedex 01, France; bDepartment of Digestive and Oncological Surgery, Huriez Hospital, Lille University Medical Centre, Lille, France; cDepartment of Digestive Surgery, Charle-Nicolle Hospital, Rouen University Medical Centre, Rouen, France; dHepatobiliary Surgical Department, Cote de Nacre Hospital, Caen University Medical Centre, Caen, France; eTransplantation and Hepatobiliary Surgery, Huriez Hospital, Lille University Medical Centre, Lille, France KEYWORDS: Pancreatic fistula; Management; Pancreatoduodenectomy

Abstract BACKGROUND: Pancreatic fistula (PF) is one of the most common postoperative complications of pancreatoduodenectomy (PD). A recent International Study Group on Pancreatic Fistula (ISGPF) definition grades the severity of PF according to the clinical impact on the patient’s hospital course. Although PF is generally treated conservatively (grade A), some cases may require interventional procedures (grade B) or may be life-threatening and necessitate emergency reoperation (grade C). The aim of the present study was to evaluate the incidence of postoperative grade C PF after PD and to assess the prognosis and risk factors for this life-threatening condition. STUDY DESIGN: Between January 2000 and December 2006, 680 consecutive patients underwent PD in 5 digestive surgery departments in the northwest region of France (Lille, Amiens, Rouen, and Caen). PF was defined as drain output of any measurable volume of fluid on or after postoperative day 3 with amylase content greater than 3 times the serum amylase activity (ISGPF guidelines). To identify possible risk factors for grade C PF, we reviewed the records of 111 (16.3%) patients with postoperative PF and compared grade C cases with grade A⫹B cases. RESULTS: The median age was 59 years (range 22– 87). The male-to-female ratio was 1.6:1. Fifty-six (50.4%) PDs were performed via pancreaticogastrostomy and 55 via pancreaticojejunostomy. Overall mortality was 2% (n ⫽ 14). Grade C PF was observed in 36 (32%) patients, of whom 17 (47%)

* Corresponding author. Tel.: ⫹33 (0)3 22 66 83 01; fax: ⫹33 (0)3 22 66 86 80. E-mail address: [email protected] Manuscript received February 18, 2008; revised manuscript March 21, 2008

0002-9610/$ - see front matter © 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.amjsurg.2008.03.004

D. Fuks et al.

Life-threatening postoperative pancreatic fistula

703

had sepsis due to an abdominal collection, 16 (44%) had postoperative bleeding, 10 (27.7%) had bleeding associated with abdominal collection, and 3 (9%) had multi-organ failure due to other causes. Of these 36 patients, 35 (97%) underwent reoperation. The mortality rate in grade C PF patients was 38.8%. The major causes of death were sepsis (n ⫽ 6) and recurrent bleeding after reoperation (n ⫽ 5). Grade C PF increased the duration of postoperative hospitalization (46 vs 29 days, P ⬍ .001). Univariate analysis showed that peroperative soft pancreatic parenchyma, peroperative blood transfusion, and postoperative bleeding were significant risk factors for grade C PF, with P values of .011, .003, and .001, respectively. No risk factors for grade C PF were identified in a multivariate analysis. The sensibility, specificity, positive predictive value, and negative predictive value of the presence of the 3 risk factors for grade C PF were 13.89%, 100%, 100%, and 70.75%, respectively. CONCLUSION: Sixteen percent of patients had PF after PD. Among them, 30% had grade C PF, with a mortality rate of about 40%. Achievement of a 100% predictive positive value for grade C PF after PD in individuals with 3 discriminant risk factors (peroperative soft pancreatic parenchyma, peroperative transfusion, and postoperative bleeding) is a first step towards the identification of high-risk patients who should be managed differently from other patients with PF during or after PD. © 2009 Elsevier Inc. All rights reserved.

Despite recent advances in surgical procedures and postoperative management techniques, a significant proportion of patients undergoing pancreaticoduodenectomy (PD) develops a postoperative pancreatic fistula (PF), which is the most frequent cause of postoperative death. Although many different risk factors have been reported in the literature, the definition of PF varies greatly and thus prevents an objective comparison of the various results. A recent publication from the International Study Group on Pancreatic Fistula (ISGPF)1 enables PF to be defined and graded more precisely (grades A, B, and C) according to the clinical procedures and outcome (Table 1). A proportion of PF patients will be asymptomatic (grade A) or poorly symptomatic (grade B), whereas others will develop abscesses, peritonitis, sepsis, and hemorrhage with a high mortality rate (grade C, requiring interventional procedures). Two large series have reported PF in 9% to 29% of cases of PD.2,3 A number of methods for reducing the incidence of PF have been suggested; these include technical adjustments (modification of the pancreaticojejunal anastomosis technique, recon-

Table 1 Grade

Classification of PF (from Bassi et al)1 A

B

Clinical conditions Well Often well Specific treatment* No Yes/no US/CT Negative Negative/ positive Persistent drainage No Usually yes after 3 weeks† Reoperation No No Death related to PF No No Signs of infections No Yes Sepsis No No Readmission No Yes/no

C Ill appearing bad Yes Positive Yes Yes Possibly yes Yes Yes Yes/no

US ⫽ ultrasonography; CT ⫽ computed tomography scan; PF ⫽ pancreatic fistula. * Partial or total parenteral nutrition, antibiotics, enteral nutrition, somatostatin analogue, and/or minimal invasive drainage. † With or without a drain in situ.

struction via pancreaticogastrostomy, and placement of pancreatic duct stents), perioperative infusion of somatostatin analogues, use of adhesive sealants, and early ablation of external drainage.4,5 Despite the implementation of these various techniques and pharmacologic adjuvants, failure of pancreatic anastomosis has a mortality rate of up to 10%.6,7 The aim of the present study was to evaluate the incidence of postoperative grade C PF after PD and to assess the prognosis and risk factors for this life-threatening condition in a large multicenter study.

Methods Patient selection From January 2000 to December 2006, 680 consecutive patients underwent PD in 5 university hospital digestive surgery departments in the northwest region of France (Lille n ⫽ 300, Amiens n ⫽ 80, Rouen n ⫽ 150, Caen n ⫽ 150). Patients’ details were entered into a prospective database in each center. Databases were then merged and retrospectively analyzed. All patients had elective operations after full workup and validation of the requisite preoperative conditions.

Operative technique All operations were performed by experienced surgeons or under their supervision. Technical variations of PD (such as pylorus-preserving techniques, types of pancreatic anastomosis, ductal stenting) were performed according to the surgeons’ preferences and were taken into account in our subsequent risk factor analysis. All patients had abdominal drainage. Injection of octreotide (Sandostatin 100 ␮g, Novartis Basel, Switzerland) was initiated during the operation once resection had been decided and was continued if the surgeon judged the texture of the pancreatic parenchyma to be soft. Adjacent organ resection (colon, gallbladder, liver,

704 stomach) was performed according to malignant extension. All patients received single-shot intravenous antibiotic prophylaxis on induction of anesthesia, in accordance with the French consensus statement.8

Definitions Mortality and morbidity were defined, respectively, as death or complications occurring within 30 days of the operation or during postoperative hospitalization. Complications were graded according to Clavien’s classification9: grade I, any deviation from the normal postoperative course but without any need for pharmacologic treatment or surgical, endoscopic, or radiologic intervention; grade II, complications requiring pharmacologic treatment; grade III, complications requiring surgical, endoscopic, or radiologic intervention; grade IV, life-threatening complications requiring intermediate or intensive care unit (ICU) management; and grade V, death. All abdominal collections were peri-anastomotic (ie, pancreatic anastomosis). PF was defined (in accordance with the ISGPF guidelines) as amylaserich fluid (with an amylase concentration in the drainage fluid more than 3 times the serum concentration) collected by needle aspiration in intra-abdominal collection or from the intraoperatively placed drain from day 3 on.1 PF were graded according to the clinical impact on the patient’s hospital course (grade A, B, or C)1 (Table 1). All data were retrospectively reviewed and graded according to Clavien’s classification and the ISGPF definition. We further analyzed grade C PF patients who underwent invasive procedures in the ICU. These patients were variously treated with total parenteral nutrition (TPN) or continuous enteral nutrition via a nasojejunal tube, nasogastric aspiration, antibiotics, subcutaneous administration of somatostatin analogues, percutaneous drainage, and/or reoperation, as described recently by Bassi et al.1 The parenchyma texture was classified by the operating surgeon as “soft” (normal, friable), “intermediate,” or “hard” (fibrotic, sclerotic). The main pancreatic duct was classified as being less than or greater than 3 mm in diameter. Postoperative bleeding was defined (in accordance with the ISGPF guidelines)10 according to the time of onset (early or late hemorrhage), the location (intraluminal or extraluminal), and the severity (mild or severe).

The American Journal of Surgery, Vol 197, No 6, June 2009 pancreatic anastomosis, ductal stenting); and (3) pancreasrelated factors, including the texture of the pancreatic remnant, the pathologic status of the pancreatic parenchyma, the diameter of the main pancreatic duct on the remnant’s raw surface (⬍ or ⬎3 mm), and the PD indication. The role of neoadjuvant radiochemotherapy on the occurrence of PF was not specifically analyzed.

Aim of the study The aim of the present study was to evaluate the incidence of postoperative grade C PF after PD and to assess the prognosis and risk factors for this life-threatening condition.

Statistical analysis Our univariate analysis used the chi-square test or Fisher exact test for qualitative variables and Student t test for quantitative variables. The Mann-Whitney U test was used for nonparametric variables. A P value of .05 or less was considered as being statistically significant in the univariate analysis. All variables achieving a P value less than .20 in the univariate analysis were included in a multivariate analysis with a stepwise logistic regression model. The independent risk factors of the variables were expressed as odds ratios with their 95% confidence intervals (CIs).

Results Population with PF Of the 680 patients who underwent PD, 111 (16.3%) developed postoperative PF (68 men and 43 women [gender ratio 1.6:1] with a median age of 59 ⫾ 12 years [range 22– 87]). Twenty-five (22.5%) patients had an ASA status greater than 3. The median body mass index was 24.5. Sixteen (14%) patients had diabetes mellitus and 23 (20%) were smokers. Sixty-one (55%) patients displayed preoperative jaundice and preoperative biliary drainage was performed in 26 (42%) of these cases.

Operative details Preoperative and intraoperative data Risk factors for the development of postoperative PF were divided into 3 categories: (1) patient-related factors, such as age, sex, loss of weight, presence of comorbidities, American Society of Anesthesiology (ASA) classification, and preoperative biliary stenting; (2) surgery-related factors, such as operative time, blood loss, intraoperative blood transfusions, the pancreatic anastomosis technique (panceaticojejunostomy or pancreaticogastrostomy), and associated procedures (pylorus-preserving techniques, types of

Fifty-six (50.4%) and 55 PDs were performed using pancreaticogastrostomy and pancreaticojejunostomy, respectively. Eleven (9%) patients underwent pylorus-preserving PD and 2 others underwent portal vein resection with end-to-end anastomosis. The median operative time was 374 ⫾ 120 minutes (range 140 –720). The mean operative blood loss was 951 ⫾ 1,068 mL (range 0 –5,500). Nine (8.1%) patients required operative blood transfusion (mean 8 units [range 4 –15] of packed red blood cells) for major bleeding (⬎2,000 mL).

D. Fuks et al. Table 2

Life-threatening postoperative pancreatic fistula

Indication of PD and grade C PF

Total Malignant disease Pancreatic adenocarcinoma Malignant ampulla Distal cholangiocarcinoma Endocrine tumor Duodenal carcinoma Colonic adenocarcinoma invading pancreas Benign disease Chronic pancreatitis Cystic lesions Intraductal papillary and mucinous neoplasm Others

No. of patients (%)

Grade C PF (%)

111

36

27 20 15 9 8

7 9 4 2 4

(24%) (18%) (14%) (2%) (7%)

(19%) (25%) (11%) (5%) (11%)

4 (3%)

1 (2%)

13 (12%) 3 (3%)

2 (5%) 1 (2%)

4 (3%) 8 (7%)

2 (5%) 4 (11%)

Grade C PF ⫽ postoperative pancreatic fistula with abscess formation, peritonitis, sepsis, and/or hemorrhage with a high rate of mortality requiring interventional procedures.

Postoperative outcomes The indications for PD of patients with PF are summarized in Table 2. Overall morbidity is summarized in Table 3. There were 43 (39%) patients with grade A PF, 32 (29%) with grade B PF, and 36 (32%) with grade C PF. Overall mortality was 2% (n ⫽ 14) and the mortality rate in grade C PF patients was 38.8% (n ⫽ 14). Mortality for grade A and B patients was zero. Patients with grade A PF were asymptomatic and, in all cases, PF was diagnosed by routine assay of drainage fluid amylase levels. The mean duration of drainage output was 13 ⫾ 11 days (range 4 – 45). Patients with grade A PF did not require any specific treatment. Conservative management (TPN and somatostatin) was initiated in all patients with grade B PF. In the latter group, 7 (21%) patients with intra-abdominal collection required percutaneous drainage (success, n ⫽ 7). All 36 patients with grade C PF were symptomatic: 17 (47%) had sepsis due to abdominal collection, 16 (44%) had postoperative bleeding (associated with abdominal collection in 10 cases), and 3 (9%) had multi-organ failure due to other causes. Of the 17 patients with abdominal collection, 4 (23%) developed septic shock. A computed tomography (CT) scan revealed peri-anastomotic abdominal collection in 9 patients (missing data, n ⫽ 8). Of the 16 patients with postoperative bleeding, the location of the latter was externalized via an abdominal drain in 9 patients and was unknown for the other 7 patients. All postoperative bleeding were grade C according to ISGPF classification. Postoperative haemorrhage was detected from day 1 to day 15 after PD. Two of these 16 patients showed haemodynamic instability and a sudden decrease in haemoglobin levels. Three patients had postoperative multi-organ failure (mesenteric ischemia n ⫽ 1, unexplained cause n ⫽ 2). The

705

mean length of stay in hospital for patients with grade C PF was 46 ⫾ 16 days, significantly (P ⬍ .001) longer than the 29 ⫾ 9 days for patients with grade A/B PF.

Management of grade C PF and surgical procedures Abdominal collection. Of the 17 patients with abdominal collection, 7 (41%) underwent percutaneous drainage (failure, n ⫽ 7) before reoperation. In 10 (58.8%) patients with clinical sepsis (septic shock, n ⫽ 4), percutaneous drainage was not performed. All 17 patients required surgical pancreatic drainage: elective surgical drainage of the anastomotic region was performed in 9 (53%) patients and pancreatic anastomosis was reduced (because of necrosis and dehiscence) in 6 patients (missing data, n ⫽ 2). Disconnection of pancreatic anastomosis and resection of the body with conservation of a small pancreatic remnant was performed in 4 patients and total/completion pancreatectomy was performed in 2 cases. Laparotomy revealed a biliary fistula in 2 patients, requiring intubation and drainage. Of the patients reoperated on for abdominal collection, 6 (35%) died due to sepsis: 1 had undergone completion pancreatectomy (mortality rate for the latter procedure: 50%) and the 5 other patients had undergone surgical drainage of the anastomotic region (mortality rate for the latter procedure: 55%). The 2 patients with missing data on reoperation procedures survived.

Table 3 Postoperative complications (Clavien classification, Dindo et al, Ann Surg 2004) Grade of complications Grade I Wound infection Fever Grade II Gastrojejunal fistula Biliary fistula Delayed gastric emptying Central venous catheter infection Urinary tract infection Pneumonia Grade IIIa Intra-abdominal collection Grade IIIb Intra-abdominal collection Grade IV Grade C PF Septic shock Hemorrhage Grade V Grade C PF Hemodynamic instability Hemorrhage recurrence Sepsis Pulmonary embolism Multi-organ failure Mesenteric ischemia

n 15 4 3 7 29 7 9 18 20 17 22 1 16 14 1 4 6 1 1 1

706 Table 4

The American Journal of Surgery, Vol 197, No 6, June 2009 Patient-related risk factors for grade C PF

Factor Total Gender Male Female Median age (y) Body mass index Weight loss (missing data n ⫽ 18) Yes No Weight loss as a % ASA status (missing data n ⫽ 29) ASA⬎III ASA⬍III Diabetes mellitus (missing data n ⫽ 1) Yes No Alcohol abuse (missing data n ⫽ 25) Yes No Tobacco use (missing data n ⫽ 27) Yes No Jaundice Yes No Preoperative biliary stenting Yes No

Grade A/B PF

Grade C PF

75

36

45 30 59.65 24.55

23 13 59.52 25.62

38 23 5.6%

15 17 4.1%

14 48

5 15

13 61

3 33

11 46

9 20

16 40

7 21

42 33

19 17

20 55

6 30

P .69 .96 .35 .15 .26 .82

.19

.22 .72 .74 .24

Postoperative bleeding. Concerning the 16 cases of postoperative bleeding, the 2 (12.5%) patients with hemodynamic instability underwent emergency surgery. In 1 case, the hemorrhage resulted from a pseudoaneurysm of the splenic artery and so splenectomy was performed; the patient subsequently developed recurrent bleeding and died. The second patient required ligation of the stump of a gastroduodenal artery pseudoaneurysm. In both cases, laparotomy revealed a leakage from gastrojejunostomy, which was intubated by drainage. The other 14 patients with bleeding required reoperation with isolated elective hemostasis. Two (12.5%) patients underwent angiography with embolization (failure, n ⫽ 2) prior to reoperation: both underwent iterative ligation of the stump of a gastroduodenal artery pseudoaneurysm and 1 underwent celiac trunk ligation (because of celiac trunk erosion by local sepsis and the pancreatic juice’s proteolytic activity) with an aortohepatic bypass and an aortic endoprosthesis due to an aortic aneurysm. Of the 16 patients with postoperative bleeding, 5 (31%) developed recurrent bleeding (despite extensive transfusion and 3 reoperations) and subsequently died. One (6.2%) patient died of pulmonary embolism during hospitalization (withdrawal of anticoagulants due to postoperative bleeding).

Others. One patient required reoperation for mesenteric ischemia and died immediately after surgery. In another case, a second laparotomy was performed because of unexpected, sudden pain (though in the absence of hemoglobin level changes), and no abdominal explanation was found. Another patient died of multi-organ failure before surgery could be performed.

Risk factors for PF Concerning patient-related factors, 5 patients with grade C PF were rated ASA class III or more, whereas there were 14 ASA class III patients with grade A/B PF (no significant difference in the proportions, 13.8% vs 18.6%, respectively; P ⫽ .82). There was a trend toward a lower prevalence of weight loss in grade C PF relative to patients with grade A/B PF (41.6% vs 50.6%; P ⫽ .15). Additional univariate analyses of patient-related factors are summarized in Table 4. Concerning surgery-related factors (Table 5), there was no difference in PF rates between patients having undergone pancreaticogastrostomy and those having undergone pancreaticojejunostomy (50.4% vs 49.6%; P ⫽ .94). Adjacent organ resection was not identified as a risk factor (2.7% vs 2.6%; P ⫽ .97) between grade A/B and grade C patients. Median operative time was borderline significant (339 vs 389 minutes; P ⫽ .06), with a shorter time in patients with grade C PF. Even though the 2 groups of patients were found not to differ significantly in terms of median blood loss (914 vs 1,013 mL; P ⫽ .68), we determined that peroperative transfusion (51.6% vs 30.8%; P ⫽ .04) and postoperative bleeding (44% vs 2.6%; P ⬍ .001) were indeed risk factors for grade C PF in a univariate analysis.

Table 5 Factor

Surgery-related risk factors for grade C PF Grade A/B PF

Grade C PF

Total 75 36 Pylorus-preserving (missing data n ⫽ 1) Yes 8 3 No 67 32 Pancreas anastomosis Pancreaticogastrostomy 38 18 Pancreaticojejunostomy 37 18 Transfusion (missing data n ⫽ 12) Yes 21 (30.8%) 16 (51.6%) No 47 15 Transfusion ⬎2BU (missing data n ⫽ 16) Yes 4 8 No 62 21 Median length of operation (min) 389 339 Median blood loss (ml) 914 1013

P

.73 .94

.04

.036

.06 .68

D. Fuks et al. Table 6

Life-threatening postoperative pancreatic fistula

Pancreas-related risk factors for grade C PF

Factor Total Peroperative pancreatic parenchyma (missing data n ⫽ 13) Soft Intermediate Hard Histological inflammation (missing data n ⫽ 29) Yes No Histological pancreatitis (missing data n ⫽ 27) Yes No Histological fibrosis (missing data n ⫽ 54) Yes No Wirsung’s duct (missing data n ⫽ 33) ⬎3 ⬍3

Grade Grade A/B PF C PF P 75

36

9 35 21

14 11 8

17 39

10 16

28 29

11 16

15 22

8 12

37 10

8 13

.01

.84

.47

.96 .22

Concerning pancreas-related factors (Table 6), only soft pancreatic parenchyma was a risk factor associated with Grade C PF (38% vs 12%, P ⫽ .01) in a univariate analysis. The 2 patient groups (grade A/B PF and grade C PF) did not significantly differ in terms of prescription of somatostatin analogues (75% vs 73%; P ⫽ .80), disease etiologies, or pathological details (pancreatic inflammation, pancreatitis, and fibrosis). The diameter of the main pancreatic duct (criterion: ⬍3 mm on the remnant raw surface) was not significantly different when comparing the 2 patient groups (36% vs 13%; P ⫽ .22). A multivariate analysis did not reveal any independent predisposing factors for grade C PF (Table 7). Of the 36 patients with PF, 5 had 3 risk factors (discriminated in a univariate analysis), 15 had 2 risk factors, 13 had 1 risk factor, and 3 had no risk factors. When the 3 risk factors (peroperative soft pancreatic parenchyma, peroperative transfusion, postoperative bleeding) for grade C PF were present, the sensitivity, specificity, positive predictive value, and negative predictive value of a grade C PF were 13.9%, 100%, 100%, and 70.8%, respectively.

Comments Management of PF after PD ranges from simple monitoring to emergency interventional procedures (including surgery); it is thus essential to be able to distinguish patients who are likely to develop life-threatening PF (grade C). To the best of our knowledge, this is the first time that a study has reported in detail on grade C PF after PD in a large series of patients. Focusing on patients who are likely to

707

develop grade C PF after PD is pertinent, since despite the extensive range of suggested methods for reducing the incidence of this condition, it still occurs in 10%–30% of patients, with a steady mortality rate of 2%–10%.6,7 It should be noted that the present study did not attempt to analyse the accuracy of embolization or the efficiency of conservative management (including radiological drainage). All 5 digestive surgery departments involved in this series were high- or medium-volume centers, according to certain North American standards,11 and our results in terms of PF rate (16%), mortality rate (2%), peri-anastomic collection rate (2.5%), reoperation rate (5.2%), and length of stay (33 days, in cases of PF) are in agreement with literature data.2,3,6,12–16 After PD, grade C PF occurred in 32% of the patients and led to reoperation in 97% of these individuals: 47% developed sepsis due to an abdominal collection, 44% had postoperative bleeding, and 27.7% had bleeding associated with abdominal collection. The mortality rate in grade C PF patients was 38.8%. Despite the low number of individuals in the grade C PF subgroup, recurrent bleeding (despite reoperation for bleeding in 3 of 5 patients) and surgical drainage of the anastomotic region due to sepsis (6 of 9 patients) were significantly associated with postoperative death. Grade C PF increased the duration of postoperative hospitalization (46 vs 29 days, P ⬍ .001). Our retrospective analysis showed that peroperative soft pancreatic parenchyma, peroperative blood transfusion, and postoperative bleeding were significant risk factors for grade C PF. A multivariate analysis failed to identify any independent risk factors for grade C PF. However, despite the presence of missing data (primarily due to the retrospective, multicentre study design), a 100% predictive positive value for grade C PF after PD in patients with the 3 risk factors is a first step towards identification of high-risk patients. To the best of our knowledge, risk factors for grade C PF have never been specifically assessed, although several series have analysed risk factors for post-PD mortality: Muscari et al12 found 2 independent factors for mortality: an age of 70 or over and extensive pancreatic resection. These factors were not identified as risk factors for grade C PF in our series. The role of neoadjuvant radiochemotherapy was not specifically analyzed in our study. This strategy may Table 7

Multivariate analysis

Grade C PF

95% Odds Standard Confidence ratio Error P value interval

Transfusion Transfusion ⬎2BU Postoperative bleeding Soft parenchyma Diabetes mellitus Loss of weight Length of operation Ampuloma

1.72 1.98 2.21 5.39 1.38 3.08 1.15 .77

2.48 3.06 1.24 4.87 1.28 2.99 1.06 .59

.70 .65 .15 .06 .72 .24 .83 .74

.10 28.75 .09 4.79 .73 6.67 .91 31.71 .22 8.57 .45 2.70 .28 6.31 .17 3.45

708 have a significant impact on occurrence of postoperative bleeding but should be analyzed in a further study. Our univariate analysis showed that peroperative soft pancreatic parenchyma, peroperative blood transfusion and postoperative bleeding were significant risk factors for grade C PF. Peroperatively, the nature of the pancreatic parenchyma has already been reported as a risk factor for PF. In a study of 245 patients after PD, Masson et al reported a correlation between PF, reoperation, and mortality rates and suggested a novel and simple classification for pancreatic parenchyma texture: type 1 (normal), type 2 (fragile, tearing during anastomosis, associated with a small main pancreatic duct), and type 3 (fibrotic parenchyma in the presence or absence of dilatation of the main pancreatic duct). There were no differences between types 1 and 3 in terms of PF occurrence (31% vs 13%) or reoperation (7% vs 3%) and mortality rates (1% vs 0%). In contrast, PF and reoperation and mortality rates were significantly higher in the type 2 subgroup (77%, 40%, and 26%, respectively).17 Mathur et al18 recently reported that fatty pancreatic parenchyma was significantly associated with PF after PD (50% vs 13%, P ⬍ .001, although no data were available concerning the severity of the PF because of the small sample size) and that there is a negative correlation between pancreatic fibrosis and fat. These 2 recent studies raise the question as to whether management of the remnant pancreas after PD (completion pancreatectomy) should be modified according to the nature of the pancreatic parenchyma. Peroperative blood transfusion is a symptom of peroperative difficulties (proximity of vessels, difficulty of retroperitoneal lymphadenectomy, obese patients) which are linked to the postoperative course.3 This factor suggests that PD is a surgery requiring meticulous intraoperative hemostasis. In our multicenter study, all procedures were performed by experimented surgeons. Lastly, postoperative bleeding (especially in the event of recurrence, despite appropriate management) should be considered as a warning sign10,13 because a proportion of patients will require invasive treatment (interventional angiographic embolization, or relaparotomy). However, we believe that identification of these high-risk patients of grade C PF after PD is crucial because they have to be managed differently from other patients with PF after PD. At present, the classification and severity of PF after PD have been clearly defined (notably thanks to Bassi et al and the ISGPF1). What the present study tells us is that in patients with PF after PD, around 30% will have grade C PF and around 40% of the latter will die. Moreover, the 100% predictive positive value for the combination of the 3 identified risk factors in this series for grade C PF may be very useful. Ideally, we would have liked to have identified accurate predictive factors for grade C PF, much as Belghiti et al described for hepatectomy: a combination of the prothrombin time less than 50% and serum bilirubin levels greater than 50 ␮l/L on postoperative day 5 (the “50 –50 criteria”) is a simple, early and accurate predictor of more than 50% of mortalities after hepatectomy. This type of

The American Journal of Surgery, Vol 197, No 6, June 2009 criterion needs to be identified early enough (ie, before clinical evidence of complications appears) for specific interventions to be applied in a timely manner.19 Using our predictive criteria for grade C PF after PD (or once a more accurate, easily applied, early-stage criterion has been identified in prospective studies that examine short-term outcomes and other pathological data), we can offer these high-risk patients a specific, standardized clinical management approach. Intra-abdominal peri-anastomotic collections must be identified and, if possible, fully drained (as judged radiologically). All patients must receive close monitoring and active conservative management (as described by Sauvanet’s group20), including total parenteral nutrition, nasogastric suction, imaging-guided percutaneous drainage of collection when necessary, and administration of somatostatin or its analogues. A helicoidal CT scan with specific vascular reconstructions must be performed to identify possible pseudoaneurysms; secondary systematic arterial embolization must then be offered to these high-risk patients. Early bleeding requires early reoperation,12 whereas endovascular treatment should be used first in delayed bleeding after PD.21-23 In the event of surgery initiated following unsuccessful drainage, septic shock, or hemodynamic instability, the association of hemorrhage and abdominal collection in around a third of all patients7,12,24 –26 (27.7% in the present series) incites us to track these 2 complications and treat them actively. Once surgery has been decided, surgical drainage or complementary pancreatic resection (including total pancreatectomy) could be performed. The role of completion pancreatectomy is subject to debate but it is considered to be a last resort when there is pancreatic necrosis and major dehiscence of the anastomosis.27 Favorable outcomes related to postoperative mortality after completion pancreatectomy have been reported,27–30 although pancreatic endocrine insufficiency is inevitable and thus many surgeons are reluctant to undertake this procedure. An alternative strategy for the treatment of pancreatic leakage has been developed, involving disconnection of the pancreatic anastomosis and resection of the body/tail with conservation of a 5-cm pancreatic remnant. This procedure usually avoids postoperative diabetes mellitus.27,28,31,32 Although the 2 procedures are difficult to compare, de Castro et al suggested that conservation of a pancreatic remnant is not superior to completion pancreatectomy when judged in terms of postoperative mortality. In de Castro et al’s series of 459 patients who underwent PD, 41 developed PF: 9 had completion pancreatectomy (with no mortalities) and a pancreatic remnant was conserved in 10 patients. Three of the latter died and the authors emphasized that completion pancreatectomy still has a role in the management of patients with severe sepsis during PF.27 We wish to stress the fact that conservative management of PF after PD is feasible and successful in over 85% of patients. In the remaining patients (ie, a minority) in whom surgery is mandatory, a single, complete, life-saving procedure must be given high priority (completion of pancreatectomy if

D. Fuks et al.

Life-threatening postoperative pancreatic fistula

major dehiscence of the anastomosis occurs). The knowledge of surgical management of PF raises a question of the role of primary absence of pancreatic anastomosis (with or without completion of pancreatectomy) in prevention of grade C PF in case of soft pancreatic parenchyma found during PD. Lastly, good efficacy for prolonged release lanreotide 30 mg in the treatment of PF has been reported in a randomized, placebo-controlled, double-blind study in terms of decrease in fistula output and time to fistula closure. However, there was no significant impact on mortality (secondary end point: 8.7% in the lanreotide group vs 2.9% in the placebo group).33 It should be noted that these patients at a high risk of grade C PF underwent nonspecific management in the intensive care unit.

Conclusion Despite the success of conservative management, a minority of patients will inevitably develop grade C PF after PD, and the condition is associated with a 40% mortality rate. Accurate and early predictive criteria for grade C PF must be identified because these patients have to be managed differently from others with PF after PD.

References 1. Bassi C, Dervenis C, Butturini G, et al. Postoperative pancreatic fistula: an international study group (ISGPF) definition. Surgery 2005; 138:8 –13. 2. Yeo CJ, Cameron JL, Lillemoe KD, et al. Does prophylactic octreotide decrease the rates of pancreatic fistula and other complications after pancreaticoduodenectomy? Results of a prospective randomized placebo-controlled trial. Ann Surg 2000;232:419 –29. 3. Winter JM, Cameron JL, Campbell KA, et al. 1423 pancreaticoduodenectomies for pancreatic cancer: a single-institution experience. J Gastrointest Surg 2006;10:1199 –210. 4. Poon RT, Fan ST, Lo CM, et al. External drainage of pancreatic duct with a stent to reduce leakage rate of pancreaticojejunostomy after pancreaticoduodenectomy: a prospective randomized trial. Ann Surg 2007;246:425–33. 5. Pratt WB, Maithel SK, Vanounou T, et al. Clinical and economic validation of the International Study Group of Pancreatic Fistula (ISGPF) classification scheme. Ann Surg 2007;245:443–51. 6. Suc B, Msika S, Fingerhut A, et al. Temporary fibrin glue occlusion of the main pancreatic duct in the prevention of intra-abdominal complications after pancreatic resection: prospective randomized trial. Ann Surg 2003;237:57– 65. 7. Cullen JJ, Sarr MG, Ilstrup DM. Pancreatic anastomotic leak after pancreaticoduodenectomy: incidence, significance, and management. Am J Surg 1994;168:295– 8. 8. http://www.sfar.org/antibiofr.html. 9. Dindo D, Demartines N, Clavien PA. Classification of surgical complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann Surg 2004;240:205–13. 10. Wente MN, Veit JA, Bassi C, et al. Postpancreatectomy hemorrhage (PPH): an International Study Group of Pancreatic Surgery (ISGPS) definition. Surgery 2007;142:20 –5. 11. Birkmeyer JD, Warshaw AL, Finlayson SR, et al. Relationship between hospital volume and late survival after pancreaticoduodenectomy. Surgery 1999;126:178 – 83.

709

12. Muscari F, Suc B, Kirzin S, et al. Risk factors for mortality and intra-abdominal complications after pancreatoduodenectomy: multivariate analysis in 300 patients. Surgery 2006;139:591– 8. 13. Blanc T, Cortes A, Goere D, et al. Hemorrhage after pancreaticoduodenectomy: when is surgery still indicated? Am J Surg 2007;194:3–9. 14. Howard TJ, Stonerock CE, Sarkar J, et al. Contemporary treatment strategies for external pancreatic fistulas. Surgery 1998;124:627– 633. 15. Aranha GV, Hodul P, Golts E, et al. A comparison of pancreaticogastrostomy and pancreaticojejunostomy following pancreaticoduodenectomy. J Gastrointest Surg 2003;7:672– 682. 16. Conlon KC, Labow D, Leung D, et al. Prospective randomized clinical trial of the value of intraperitoneal drainage after pancreatic resection. Ann Surg 2001;234:487–93. 17. Masson B, Ewald J, Rault A, et al. Comment prévoir la gravité de la fistule pancréatique après duodénopancréatectomie céphalique. Ann Chir 2006;131:A479. 18. Mathur A, Pitt HA, Marine M, et al. Fatty pancreas: a factor in postoperative pancreatic fistula. Ann Surg 2007;246:1058 – 64. 19. Balzan S, Belghiti J, Farges O, et al. The “50-50 criteria” on postoperative day 5: an accurate predictor of liver failure and death after hepatectomy. Ann Surg 2005;242:824 – 8. 20. Munoz-Bongrand N, Sauvanet A, Denys A, et al. Conservative management of pancreatic fistula after pancreaticoduodenectomy with pancreaticogastrostomy. J Am Coll Surg 2004;199:198 –203. 21. Sato N, Yamaguchi K, Shimizu S, et al. Coil embolization of bleeding visceral pseudoaneurysms following pancreatectomy. Arch Surg 1998; 133:1099 –102. 22. Teramoto K, Kawamura T, Takamatsu S, et al. A case of hepatic artery embolization and partial arterialization of the portal vein for intraperitoneal hemorrhage after a pancreaticoduodenectomy. Hepatogastroenterology 2003;50:1217–9. 23. Otah E, Cushin BJ, Rozenblit GN, et al. Visceral artery pseudoaneurysms following pancreatoduodenectomy. Arch Surg 2002;137: 55–9. 24. De Castro SMM, Kuhlmann KFD, Bush ORC, et al. Delayed massive hemorrhage after pancreatic and biliary surgery: embolization or surgery? Ann Surg 2005;241:85–91. 25. Choi SH, Moon HJ, Heo JS, et al. Delayed hemorrhage after pancreaticoduodenectomy. J Am Coll Surg 2004;199:186 –91. 26. Yoon YS, Kim SW, Her KH, et al. Management of postoperative hemorrhage after pancreatoduodenectomy. Hepatogastroenterology 2003;50:2208 –12. 27. de Castro SM, Busch OR, van Gulik TM, et al. Incidence and management of pancreatic leakage after pancreatoduodenectomy. Br J Surg 2005;92:1117–23. 28. Gueroult S, Parc Y, Duron F, et al. Completion pancreatectomy for postoperative peritonitis after pancreaticoduodenectomy: early and late outcome. Arch Surg 2004;139:16 –9. 29. van Berge Henegouwen MI, van Gulik TM, DeWit LT, et al. Delayed gastric emptying after standard pancreaticoduodenectomy versus pylorus-preserving pancreaticoduodenectomy: an analysis of 200 consecutive patients. J Am Coll Surg 1997;185:373–9. 30. Farley DR, Schwall G, Trede M. Completion pancreatectomy for surgical complications after pancreaticoduodenectomy. Br J Surg 1996;83:176 –9. 31. Ihse I, Anderson H, Andren-Sandberg A. Total pancreatectomy for cancer of the pancreas: Is it appropriate? World J Surg 1996;20: 288 –94. 32. Wu CC, Hwang CR, Yeh DC, et al. Treatment for dehiscence of pancreaticojejunostomy after pancreaticoduodenectomy: is resection of the residual pancreas necessary? Hepatogastroenterology 1996;4: 271– 4. 33. Topart P, Campion JP, Wind P, et al. Intramuscular lanreotide 30 mg LP in the treatment of digestive fistulae: a multicentre, randomized, placebo-controlled, double-blind study. Gastroenterol Clin Biol 2007; 31:A129.