Original article

Diagnosis of postoperative pancreatic fistula

O. Facy1,3, C. Chalumeau1, M. Poussier1, C. Binquet2,P.Rat1,3 and P. Ortega-Deballon1,3

1Department of Digestive Surgical Oncology and 2Clinical Research Unit, Centre d’Investigation Clinique–Epidemiologie Clinique, University Hospital, and 3Institut National de la Sante´ et de la Recherche Medicale´ 866, Locoregional Therapy in Surgical Oncology, Dijon, France Correspondence to: Dr O. Facy, Service de Chirurgie Digestive et Cancerologique,´ Centre Hospitalier Universitaire, 14 Rue Gaffarel, 21079 Dijon Cedex, France (e-mail: [email protected])

Background: Pancreatic fistula (PF) is a major source of morbidity after pancreatectomy. The International Study Group on Pancreatic (ISGPF) defines postoperative fistula by an concentration in the abdominal drain of more than three times the serum value on day 3 or more after surgery. However, this definition fails to identify some clinical fistulas. This study examined the association between lipase measured in abdominal drainage fluid and PF. Methods: Amylase and lipase levels in the abdominal drain were measured 3 days after pancreatic resection. Grade B and C fistulas were classified as clinical fistulas, regardless of whether the measured amylase concentration was considered positive or negative. The PF group included patients with a clinical fistula and/or those with positive amylase according to the ISGPF definition. Results: Sixty-five patients were included. The median level of lipase was higher in patients with positive amylase than in those with negative amylase: 12 176 versus 64 units/l (P < 0·001). The lipase level was 16 500 units/l in patients with a clinical fistula and 224 units/l in those without a clinical fistula (P = 0·001). Patients with a PF had a higher lipase concentration than those without: 7852 versus 64 units/l (P < 0·001). A lipase level higher than 500 units/l yielded a sensitivity of 88 per cent and a specificity of 75 per cent for PF. For clinical fistulas the sensitivity was 93 per cent and specificity 77 per cent when the threshold for lipase was 1000 units/l. Conclusion: Lipase concentration in the abdominal drain correlated with PF. A threshold of 1000 units/l yielded a high sensitivity and specificity for the diagnosis of clinical PF.

Paper accepted 19 March 2012 Published online 27 April 2012 in Wiley Online Library (www.bjs.co.uk). DOI: 10.1002/bjs.8774

Introduction . According to recent guidelines, the diagnosis of relies mainly on lipasaemia7,8.The Postoperative pancreatic fistula (PF) is the main cause of aim of this study was to investigate the value of lipase morbidity and mortality after pancreatic resection, and is concentration in postoperative abdominal drainage fluid in responsible for most septic or haemorrhagic complications the diagnosis of PF defined according to the ISGPF, but following pancreatic surgery. The incidence of PF is also based on clinical and radiological findings suggestive estimated at between 5 and 40 per cent depending on of PF. the definition used1–3. In 2005, the International Study Group on Pancreatic Fistula (ISGPF) defined PF as a level Methods of amylase in the abdominal drain greater than three times 4 the serum value on day 3 or more following operation . Patients who underwent pancreatic resection between A clear definition of PF is essential for evaluating surgical 2008 and 2010 had the concentration of amylase and 5,6 techniques and perioperative care . Nevertheless, the lipase measured in abdominal drains, and data were ISGPF also focused on the shortcomings of its definition, recorded prospectively. Patients who underwent total which may fail to diagnose some clinically obvious fistulas. pancreatectomy were not included. Patients in whom the In recent years, there has been increasing evidence of the lipase concentration was not measured were excluded superiority of lipase over amylase for the diagnosis of acute from analysis. The levels of pancreatic enzymes were

 2012 British Journal of Surgery Society Ltd British Journal of Surgery 2012; 99: 1072–1075 Published by John Wiley & Sons Ltd Lipase and pancreatic fistula 1073

measured systematically in the drains between 3 and 5 days means of the non-parametric Mann–Whitney U test. after surgery. Measurements were repeated after day 5 in To examine the correlation between amylase and lipase, patients with clinical features that suggested the presence Pearson’s correlation coefficient was calculated with of PF (change to any aspect of the fluid in the drain, or its 95 per cent confidence interval (c.i.). P < 0·050 was radiological drainage of perianastomotic abscesses). The considered significant. Areas under the receiver operating  enzyme assay was performed with a Dimension Vista characteristic (ROC) curve were calculated to evaluate analyser (Siemens Healthcare Diagnostics, Erlangen, lipase concentration as a predictor of PF. Triomphe Germany)9. Preoperative data (age, sex, American Society software (designed at Dijon University Hospital) and of Anesthesiologists grade, serum albumin level, body Stata software (Stata Corp, College Station, Texas, USA) mass index (BMI), medical history of diabetes or were used for data analysis. pancreatitis, ongoing medical treatment, indication for surgery, preoperative treatment), intraoperative data (diameter of the Wirsung duct, consistency of the Results pancreatic parenchyma evaluated by the surgeon as Sixty-five patients (34 men and 31 women) with a median fibrous or friable, and blood loss) and postoperative age of 62 (41–81) years were studied. Six patients were clinical data (complications, quality and quantity of drain diabetic and nine had presented with acute pancreatitis. fluid, and findings at relaparotomy) or radiological data The median preoperative BMI was 24 kg/m2 and mean (perianastomotic abscess with or without percutaneous albumin concentration was 28 g/l. Pancreatic resection drainage, and levels of pancreatic enzymes in the was performed for periampullary malignancies in 53 radiological drain) were recorded. patients, endocrine tumours in four, For the diagnosis of PF, an amylase concentration in the in two, intraductal papillary mucinous tumour in four, drain of more than three times the serum value on day 3 or pseudopapillary solid tumour in one and mucinous more after surgery was used to define patients with positive cystadenoma in one patient. drain amylase. PFs were classified as A (subclinical), B or C (clinical and/or radiological) according to the ISGPF 4 definition . Patients with clinical or radiological findings Perioperative data suggestive of PF were classified as having a clinical fistula, regardless of the level of amylase; the fistulas were classified Pancreatic resections consisted of 43 pancreaticoduo- as B or C according to ISGPF criteria (Table 1). denectomies, 21 left pancreatectomies (seven with spleen preservation) and one Frey procedure. The Wirsung duct measured more than 3 mm in 15 patients. The pancreatic Statistical analysis parenchyma was considered soft in 39 patients. Median Values are expressed as median (range) unless indicated blood loss was 500 (100–2300) ml and 15 patients required otherwise. Median levels of lipase were compared by a perioperative blood transfusion.

Table 1 Classification of the International Study Group on Postoperative course Pancreatic Fistula Morbidity occurred in 31 patients (48 per cent) and four Grade A Grade B Grade C patients (6 per cent) died within 30 days of operation. Clinical condition Well Often well Appearing ill/bad According to the Dindo–Clavien classification10,the Specific treatments* No No/yes Yes Ultrasonography/ Negative Negative/positive Positive complications were classified as grade I in 13 patients, computed grade II in seven, grade III in four, grade IV in three and tomography grade V in four patients. Persistent drainage No Usually yes Yes Pearson’s correlation coefficient between amylase and after 3 weeks Reoperation No No Yes lipase concentrations measured in the abdominal drains Death related to PF No No Possibly yes was 0·83 (95 per cent c.i. 0·74 to 0·90; P < 0·001). Signs of infection No Yes Yes Twenty-two patients, including 20 who had a lipase Sepsis No No Yes Readmission No No/yes No/yes concentration of more than 500 units/l in the drain, were classified as amylase-positive (Table 2). The median level *Includes parenteral or enteral nutrition, antibiotics, of lipase in the drain was significantly higher in patients analogue and/or minimally invasive drainage. PF, pancreatic fistula. with a positive amylase concentration than in patients with

 2012 British Journal of Surgery Society Ltd www.bjs.co.uk British Journal of Surgery 2012; 99: 1072–1075 Published by John Wiley & Sons Ltd 1074 O. Facy, C. Chalumeau, M. Poussier, C. Binquet, P. Rat and P. Ortega-Deballon

Table 2 Lipase concentration in abdominal drain 3 days or more P = 0·001). When a threshold of 1000 units/l lipase was after surgery according to each definition of pancreatic fistula, in chosen, the sensitivity of lipase for the diagnosis of clinical 65 patients fistula was 93 per cent (including all 11 patients considered as amylase-positive), with a specificity of 77 per cent and a Fistula No. of Fistula Lipase definition patients grade† (units/l)* P# negative predictive value of 98 per cent. In the same group of patients, the sensitivity and the specificity of amylase ISGPF‡ Amylase-positive 22 A 11 12 176 (0–16 500) < 0·001 were 79 and 78 per cent respectively, with a negative B9 predictive value of 93 per cent. C2 Among the 51 patients without a clinical fistula, 11 Amylase-negative 43 B 1 64 (10–3215) patients were considered as amylase-positive (correspond- C2 Clinical fistula§ ing to grade A fistula) and 13 patients, including nine who Yes 14 B 10 16 500 (1065–20 000) 0·001 were amylase-positive, had a lipase concentration of more C4 than 1000 units/l in the drain. No 51 A 11 224 (0–6500) Pancreatic fistula¶ Yes 25 A 11 7852 (0–20 000) < 0·001 B10 Discussion C4 No 40 64 (10–3215) Lipasaemia is more accurate than amylasaemia for the diagnosis of acute pancreatitis7,8. Based on this, the *Values are median (range). †According to the International Study Group use of lipase for the diagnosis of postoperative PF was on Pancreatic Fistula (ISGPF) classification. ‡Patients were classified as investigated. In the present study, the incidence of clinical amylase-positive when the amylase concentration in the abdominal drain fistula was consistent with the results of other studies, as was more than three times higher than serum value on day 3 or more 11–13 following operation. §Clinical or radiological features suggestive of a were the risk factors . The incidence of fistula in this pancreatic fistula (grade B and C), whatever the amylase level. study is in the upper range of values reported previously, ¶Combination of fistulas diagnosed on the basis of amylase in abdominal possibly because amylase and lipase concentrations were drain and those diagnosed clinically. #Mann–Whitney U test. not measured in eight patients with an unremarkable outcome. A strong correlation was found between amylase a negative amylase concentration (12 176 versus 64 units/l; and lipase concentrations in the abdominal drains. Lipase P < 0·001). measured in the drains more than 3 days after surgery had Twenty-five patients presented with a clinical fistula an area under the ROC curve of 0·89 for the diagnosis and/or were defined as amylase-positive and therefore of PF, and therefore classified 89 per cent of patients included in the PF group. The area under the ROC accurately. Considering fistulas defined according to the curve, which evaluated lipase as a predictor of PF, was ISGPF, the threshold of 500 units/l lipase yielded good 0·89. The median level of lipase in patients with a PF sensitivity and specificity. By definition, in this setting, was significantly higher than that in patients defined as amylase yielded 100 per cent sensitivity and specificity. amylase-negative without a clinical fistula (7852 versus 64 However, when clinical fistulas (grade B or C) were units/l; P < 0·001). When a cut-off of 500 units/l was considered, a threshold lipase level of 1000 units/l in chosen, the sensitivity of lipase for the diagnosis of PF was this study yielded a higher sensitivity and specificity 88 per cent, with a specificity of 75 per cent and a negative than amylase. Measurement of lipase concentration would predictive value of 91 per cent. therefore have diagnosed a number of clinical fistulas that Fourteen patients, including two who died, developed a would have been missed by measuring amylase. These clinical fistula (grade B or C) (Table 2). The duct diameter findings suggest a possible relationship between the level was smaller than 3 mm and the pancreatic parenchyma of lipase and the clinical severity of the fistula, but this was classified as friable in all 14 patients. Three patients remains speculative given the small number of patients. had a clinical fistula but were amylase-negative. One of Finally, the negative predictive value of lipase was greater these patients had a perianastomotic abscess that drained than 90 per cent at both cut-off levels. spontaneously through the midline incision more than The median level of lipase was significantly higher in 5 days after surgery. In two patients, a PF was found all patients considered as amylase-positive, in patients at reoperation but the pancreatic fluid was not analysed. with a clinical fistula and in patients with a PF than in The median concentration of lipase in the drains was patients who were considered negative for fistula based significantly higher in patients with clinical fistulas than in on the three definitions. It is therefore hypothesized that the group without clinical fistulas (16 500 versus 224 units/l; lipase concentration could be accurate for the diagnosis

 2012 British Journal of Surgery Society Ltd www.bjs.co.uk British Journal of Surgery 2012; 99: 1072–1075 Published by John Wiley & Sons Ltd Lipase and pancreatic fistula 1075

of PF, regardless of which definition of fistula is used. 4 Bassi C, Dervenis C, Butturini G, Fingerhut A, Yeo C, In a previous study amylase and lipase concentrations Izbicki J et al. Postoperative pancreatic fistula: an were measured in the abdominal drains of 50 patients international study group (ISGPF) definition. Surgery 2005; undergoing pancreatic surgery, but there were too few 138: 8–13. fistulas to reach a clear conclusion regarding the diagnostic 5 Lee SE, Ahn YJ, Jang JY, Kim SW. Prospective randomized pilot trial comparing closed suction drainage and gravity accuracy of these measurements14. In the present study, drainage of the in pancreaticojejunostomy. thresholds were established that favoured sensitivity while J Hepatobiliary Pancreat Surg 2009; 16: 837–843. achieving acceptable specificity, as the priority is to detect 6 Dong X, Zhang B, Kang MX, Chen Y, Guo QQ, Wu YL. patients at high risk of developing a postoperative PF. Analysis of pancreatic fistula according to the International Technically, amylase and lipase assays depend on Study Group on Pancreatic Fistula classification scheme for colorimetric techniques and the cost of the two methods 294 patients who underwent pancreaticoduodenectomy in a is similar15. The results of both assays may be slightly single center. 2011; 40: 222–228. modified by haemolysis, but to a lesser extent for lipase 7 Yadav D, Agarwal N, Pitchumoni CS. A critical evaluation of than for amylase14. Some authors have suggested that the laboratory tests in acute pancreatitis. Am J Gastroenterol 2002; absolute value of drain amylase, regardless of the serum 97: 1309–1318. value and measured as early as the first day after surgery, 8 Smith RC, Southwell-Keely J, Chesher D. Should serum pancreatic lipase replace serum amylase as a biomarker of could be of interest16,17. In the present study only the acute pancreatitis? ANZ J Surg 2005; 233: 356–367. drain concentration of lipase was measured and this was 9 Winn-Deen ES, David H, Sigler G, Chavez R. Development not correlated with lipasaemia. Lipase and amylase were of a direct assay for alpha-amylase. Clin Chem 1988; 34: not measured on the first 2 days after surgery. 2005–2008. A level of more than 500 units/l lipase in the abdominal 10 Clavien PA, Barkun J, de Oliveira ML, Vauthey JN, drain more than 3 days after operation correlated strongly Dindo D, Schulick RD et al. The Clavien–Dindo with the presence of PF in this study. A lipase concentration classification of surgical complications: five-year experience. higher than 1000 units/l was found to be a good marker of Ann Surg 2009; 250: 187–196. clinical fistulas (grade B or C). 11 Hill JS, Zhou Z, Simons JP, Ng SC, McDade TP, Whalen GF et al. A simple risk score to predict in-hospital mortality after pancreatic resection for cancer. Ann Surg Acknowledgements Oncol 2010; 17: 1802–1807. 12 Kawai M, Kondo S, Yamaue H, Wada K, Sano K, Motoi F O.F. and C.C. contributed equally to this paper. et al. Predictive risk factors for clinically relevant pancreatic The authors thank Dr David Masson for performing fistula analyzed in 1239 patients with the biochemical assays; the staff and nurses from pancreaticoduodenectomy: multicenter data collection as a the Departments of Digestive Surgical Oncology and project study of pancreatic surgery by the Japanese Society of Biochemistry, University Hospital of Dijon, for their Hepato-Biliary-Pancreatic Surgery. J Hepatobiliary Pancreat assistance with data collection; and Philip Bastable for Sci 2011; 18: 601–608. his help in reviewing the manuscript. 13 Pratt WB, Callery MP, Vollmer CM Jr. Risk prediction for Disclosure: The authors declare no conflict of interest. development of pancreatic fistula using the ISGPF classification scheme. World J Surg 2008; 32: 419–428. 14 Kong J, Gananadha S, Hugh TJ, Samra JS. References Pancreatoduodenectomy: role of drain fluid analysis in the management of pancreatic fistula. ANZ J Surg 2008; 78: 1 Frymerman AS, Schuld J, Ziehen P, Kollmar O, Justinger C, 240–204. Merai M et al. Impact of postoperative pancreatic fistula on 15 Foo AY. Amylase measurement – which method? Ann Clin surgical outcome – the need for a classification-driven risks Biochem 1995; 32: 239–243. management. J Gastrointest Surg 2010; 14: 711–718. 16 Molinari E, Bassi C, Salvia R, Butturini G, Crippa S, 2 Barreto G, D’Souza MA, Shukla PJ, Shrikhande SV. The Talamini G et al. Amylase value in drains after pancreatic gray zone between postpancreaticoduodenectomy collections resection as predictive factor of postoperative pancreatic and pancreatic fistula. Pancreas 2008; 37: 422–425. fistula: results of a prospective study in 137 patients. Ann 3 Fuks D, Piessen G, Huet E, Tavernier M, Zerbib P, Surg 2007; 246: 281–287. Michot F et al. Life-threatening postoperative pancreatic 17 Moskovic DJ, Hodges SE, Wu MF, Brunicardi FC, fistula (grade C) after pancreaticoduodenectomy: incidence, Hilsenbeck SG, Fisher WE. Drain data to predict clinically prognosis, and risk factors. Am J Surg 2009; 197: 702–709. relevant pancreatic fistula. HPB (Oxford) 2010; 12: 472–481.

 2012 British Journal of Surgery Society Ltd www.bjs.co.uk British Journal of Surgery 2012; 99: 1072–1075 Published by John Wiley & Sons Ltd