Serum amylase on the night of surgery predicts clinically significant pancreatic fistula after pancreaticoduodenectomy

HPB

DOI:10.1111/hpb.12184

ORIGINAL ARTICLE

Serum amylase on the night of surgery predicts clinically significant pancreatic fistula after pancreaticoduodenectomy Lavanniya K. Palani Velu1,2, Vishnu V. Chandrabalan1,2, Salman Jabbar1, Donald C. McMillan2, Colin J. McKay1, C. Ross Carter1, Nigel B. Jamieson1,2 & Euan J. Dickson1 1

West of Scotland Pancreatic Unit, Glasgow Royal Infirmary, Glasgow, UK and 2Academic Department of Surgery, University of Glasgow, Glasgow, UK

Abstract Objectives: Drainage after pancreaticoduodenectomy (PD) remains controversial because the risk for uncontrolled postoperative pancreatic fistula (POPF) must be balanced against the potential morbidity associated with prolonged and possibly unnecessary drainage. This study investigated the utility of the level of serum amylase on the night of surgery [postoperative day (PoD) 0 serum amylase] to predict POPF. Methods: A total of 185 patients who underwent PD were studied. Occurrences of POPF were graded using the International Study Group on Pancreatic Fistula (ISGPF) classification. Receiver operating characteristic (ROC) analysis identified a threshold value of PoD 0 serum amylase associated with clinically significant POPF (ISGPF Grades B and C) in a test cohort (n = 45). The accuracy of this threshold value was then tested in a validation cohort (n = 140). Results: Overall, 43 (23.2%) patients developed clinically significant POPF. The threshold value of PoD 0 serum amylase for the identification of clinically significant POPF was ≥130 IU/l (P = 0.003). Serum amylase of <130 IU/l had a negative predictive value of 88.8% for clinically significant POPF (P < 0.001). Serum amylase of ≥130 IU/l on PoD 0 and a soft pancreatic parenchyma were independent risk factors for clinically significant POPF. Conclusions: Postoperative day 0 serum amylase of <130 IU/l allows for the early and accurate categorization of patients at least risk for clinically significant POPF and may identify patients suitable for early drain removal. Received 23 July 2013; accepted 31 August 2013

Correspondence Lavanniya K. Palani Velu, Academic Department of Surgery, 4th Floor, Walton Building, Glasgow Royal Infirmary, Alexandra Parade, Glasgow G31 2ER, UK. Tel: + 44 141 211 5440. Fax: + 44 141 211 4991. E-mail: [email protected]

Introduction In-hospital mortality following pancreaticoduodenectomy (PD) has decreased significantly over the last four decades, from 25%, to currently accepted rates of <5% in high-volume centres worldwide.1–4 Unfortunately, despite careful patient selection and advances in perioperative management, surgical morbidity remains high (40–50%).5 Postoperative pancreatic fistula (POPF) persists as the most challenging and life-threatening complication This study was delivered in part as a poster presentation at the Annual Meeting of the Pancreatic Society of Great Britain and Ireland (PSGBI), 14–16 November 2012, Loch Lomond, Scotland.

HPB 2013, ••, ••–••

despite the application of numerous technical preventative strategies.6–9 The occurrence of POPF has significant cost implications associated with a prolonged postoperative stay, multiple investigations, interventions and admissions to critical care environments.10–12 Of greatest significance is the fact that complications associated with POPF frequently delay or prevent the delivery of adjuvant chemotherapy, potentially impacting longterm survival.13 Subsequently, the timely and appropriate determination of factors14–19 predictive of risk for POPF is vital if the morbidity associated with pancreatic resection is to be minimized. Risk for POPF continues to fuel the controversy surrounding the placement and timing of removal of operatively placed

© 2013 International Hepato-Pancreato-Biliary Association

2

pancreatic drains. Prospective randomized evidence suggesting that drain placement confers no objective benefit and may even increase morbidity20 has been tempered by the early closure of a recent randomized trial (NCT01441492) examining the hypothesis that pancreatic resection without routine intraoperative drainage is associated with decreased postoperative morbidity.21 Certainly, for the majority of pancreatic surgeons, intraoperative drain insertion remains a standard component of PD. Drains serve firstly to identify and control reactionary haemorrhage or bile leak in the initial 24 h and, secondly, as a theoretical mechanism to maintain control in the event of POPF formation. However, the risk for infection and the potential damage that may be induced by mechanical pressure, erosion or suction remain concerning. It has been demonstrated that prolonged drainage is associated with increases in the rate of complications, length of hospital stay and economic resource utilization.22 Consequently, for surgeons who continue to place drains intraoperatively, early removal is desirable in patients in whom this can be achieved safely. Evidence suggests this strategy may avoid or attenuate complications, reducing in-hospital stay and readmissions, and ultimately achieve the provision of cost-effective, high-quality health care.22 Drain fluid amylase content is integral to the diagnosis of POPF and has been employed as an indicator of risk for fistula as early as the first postoperative day (PoD).17,23–26 However, drain fluid amylase content is challenging to assess in the initial 24 h as a result of haemolysis and low drain volume. In response to this, several studies have investigated the relationship between the derangement of routine and easily measured serum and urine biochemical markers, and risk for POPF.27–29 Of note, a rise in serum amylase in the immediate postoperative period has previously been associated with an increased risk for POPF.22,30 Consequently, the current study sought to assess the utility of serum amylase on the night of surgery (PoD 0) to predict the occurrence of POPF in a cohort of 185 patients undergoing PD with routine intraoperative drain placement,in an effort to identify patients suitable for early drain removal.

Materials and methods Patients All patients underwent surgical resection in the West of Scotland Pancreatic Unit, Glasgow Royal Infirmary, Glasgow, UK, during a 5-year period (January 2008 to March 2013). Patients were submitted to either classical or pylorus-preserving PD (PPPD), performed by three surgeons (CRC, CJM and EJD). This analysis included patients undergoing resection for either benign or malignant disease. Resectability for malignant disease was classified in accordance with previous trial inclusion criteria31 and has been outlined in a prior publication.32 Localized lesions adjacent to the superior mesenteric vein or portal vein required the resection of the involved segment and vascular reconstruction. Prospective data collection Demographic, pathological and intraoperative data were recorded in a prospectively maintained database populated from a combi-

HPB 2013, ••, ••–••

HPB

nation of electronic patient records, preoperative imaging and anaesthetic charts. Preoperative clinical data included age, gender, body mass index (BMI), and the results of serum biochemical investigations including serum bilirubin, urea and amylase. The serum amylase measured throughout the study period was total amylase. The POSSUM (physiological and operative severity score for the enumeration of mortality and morbidity) index served as an objective measure of comorbidity and preoperative physiology.33 Preoperative computed tomography images were analysed to calculate pancreatic duct diameter at the line of transection of the pancreas anterior to the portal vein. Intraoperative data included reconstruction technique, texture of the pancreatic remnant and estimated blood loss. Blood loss data were compiled from a combination of anaesthetic charts and perioperative blood transfusion data. Specimens were dichotomized according to whether their pathology was associated with hard pancreatic parenchyma [pancreatic ductal adenocarcinoma (PDAC) and chronic pancreatitis] or with soft or normal pancreatic parenchyma (ampullary carcinoma, duodenal carcinoma, cholangiocarcinoma, neuroendocrine tumours and other lesions). Outcome data included length of stay in a critical care environment [defined as either an intensive care unit (ICU) or a surgical high-dependency unit (SHDU)] and length of postoperative hospital stay. All postoperative complications were prospectively recorded and graded in a detailed weekly consensus discussion by the three operating pancreatic surgeons according to the International Study Group on Pancreatic Fistula (ISGPF) and International Study Group of Pancreatic Surgery (ISGPS) classifications23,34 and the Clavien–Dindo classification.35 Complications of ISGPS Grades B and C and Clavien–Dindo Grades III–V were considered clinically significant. Mortality was recorded at 30-day and 90-day time-points.

Operative procedure All patients received i.v. prophylactic antibiotics and a somatostatin analogue (octreotide 100 μg; Sun Pharmaceuticals UK Ltd, Leeds, UK) subcutaneously at the induction of anaesthesia. Octreotide was continued in the postoperative period for a total of 15 doses. The operative steps of PD and histological analysis were performed as described previously.32 Pancreatic transection was performed with a scalpel, ultrasonic scalpel or diathermy depending on the consultant’s preference. The pancreatic remnant was anastomosed using a dual-layer, duct-tomucosa technique to either the jejunum [pancreaticojejunostomy (PJ)] or the stomach [pancreaticogastrostomy (PG)] according to the surgeon’s preference. Pancreatic stents were not routinely used. Absorbable synthetic monofilament sutures (Biosyn™; Covidien, Inc., Mansfield, MA, USA) were used in both techniques, with 4–0 sutures for the PJ and 3–0 sutures for the PG. At the end of the procedure, one or two non-suction drains (24-Fr, BLAKE®; Ethicon, Inc., Somerville, NJ, USA) were routinely placed adjacent to the pancreatic and biliary anastomoses.

© 2013 International Hepato-Pancreato-Biliary Association

HPB

Perioperative management In the immediate postoperative period all patients were admitted to the SHDU. Perioperative care for all patients was directed by the operating surgeon. Serum haematology and biochemistry laboratory measurements were performed daily from the preoperative day until discharge. Specifically, a blood sample was obtained for serum investigations (including amylase) on admission to the SHDU at least 4 h after the construction of the pancreatic anastomosis (PoD 0). Drain fluid volumes were recorded and drain fluid was analysed for amylase content on PoDs 1 and 3 and thereafter until removal. Octreotide (100 μg given subcutaneously every 8 h) was administered for 5 days and proton pump inhibitors continued indefinitely. The criteria for drain removal remained constant over the study period. Drains were removed when: (i) the volume of drainage amounted to <50 ml over 24 h; (ii) the drain fluid amylase content was not diagnostic of POPF, and (iii) no other clinical contraindication to drain removal (bile or chyle leak) existed. During the study period, PoD 0 serum amylase assessment did not influence drain removal. The management of suspected fistulae included the provision of antibiotics, image-guided percutaneous drainage, and surgical exploration with extensive peripancreatic drainage or completion pancreatectomy. Definitions of outcome measures The upper normal limit for serum amylase at this institute was 100 IU/l. Pancreatitis was defined by a serum amylase level of ≥300 IU/l (three times the upper limit of normal). The normal range for drain fluid amylase was less than three times the serum amylase activity. A drain fluid amylase three times the serum amylase activity on or after PoD 3 was diagnostic of POPF in accordance with the current ISGPF definition.23 In short, an ISGPF Grade A POPF was a transient fistula requiring no additional treatment and no prolongation of hospital stay; a Grade B POPF required additional medication or treatment and led to a prolonged hospital stay, and a Grade C POPF was a potentially life-threatening event, often associated with post-pancreatectomy haemorrhage (PPH) or sepsis requiring re-laparotomy. Grades B and C POPF were defined as clinically significant. Statistical analysis Continuous data are presented as the median (range). Categorical variables were compared using the chi-squared test. The Mann– Whitney U-test was used to compare continuous variables. To identify the optimum threshold value of PoD 0 serum amylase for predicting POPF, receiver operating characteristic (ROC) analysis was performed in a random sample of the cohort (n = 45).36 The area under the curve (AUC) in ROC analysis is a measure of the diagnostic accuracy of a test: an AUC value of >0.50 indicates the ability of a test to significantly differentiate between positive and negative outcomes with regard to the classification variable (in this case, POPF of ISGPF Grades B and C). A diagnostic test with an AUC of >0.75 is considered to have high diagnostic accuracy,

HPB 2013, ••, ••–••

3

with at least 75% of patients with the condition of interest classified correctly.26 The utility of the threshold identified using ROC analysis was subsequently confirmed in a validation group (n = 140) that included the remainder of the study cohort. Univariate binary logistic regression analysis with calculation of odds ratios (ORs) and 95% confidence intervals (CIs) was used to explore the association between perioperative clinicopathological factors, including PoD 0 serum amylase and risk for POPF. Multivariate binary logistic regression analysis was performed on variables showing a significant association on univariate analysis (P < 0.05). Backward stepwise regression was used starting with a saturated model and variables with a P-value of >0.1 were excluded at each step until no more variables could be excluded. List-wise deletion was performed in patients for whom data were missing. Cox proportional hazards regression analysis was used to analyse the relationship between perioperative risk factors and length of hospital stay. A P-value of < 0.05 was considered to indicate statistical significance. All statistical analysis was performed using IBM spss Version 19.0 (IBM Corp., Armonk, NY, USA).

Results Clinicopathological characteristics of the entire patient cohort A total of 185 patients who underwent PD between January 2008 and March 2013 were included in the analysis. The demographic, operative, pathological and treatment characteristics of the cohort are summarized in Table 1. Eighteen (9.7%) patients required en bloc vein resection. The median estimated blood loss was 1450 ml.

Postoperative complications The overall complication rate was 61.1% (n = 113). The rate of complications of Clavien–Dindo Grade IIIa and above was 34.6% (n = 64). A total of 64 (34.6%) patients developed POPF; 21 (11.4%) patients developed Grade A, 27 (14.6%) developed Grade B and 16 (8.6%) developed Grade C POPF. The rate of clinically significant (ISGPF Grade B or C) POPF was 23.2%. The rate of clinically significant PPH (ISGPS Grade B or C) was 12.4% (n = 23). A total of 44 (23.8%) patients required re-intervention in the form of percutaneous drainage of intraabdominal collections or mesenteric angiography and embolization for PPH; of these, 22 (11.9%) patients required a re-laparotomy to both control the source of sepsis and manage the haemorrhage. The rate of 90-day surgical mortality was 4.9% (n = 9). Four (2.2%) deaths occurred as a consequence of PPH and associated complications, three (1.6%) as a result of infected intraabdominal collections, one (0.5%) subsequent to multi-organ failure following severe post-PD pancreatitis, and one (0.5%) following a postoperative myocardial infarction. Eight of the nine patients who died developed Grade C POPF as defined by the ISGPF.

© 2013 International Hepato-Pancreato-Biliary Association

HPB

4

Table 1 Demographic, operative and pathological characteristics and postoperative outcomes in the study test and validation cohorts Characteristic

Age, years

Total cohort (n = 185) n (%)

Test cohort (n = 45, 24.3%) n (%)

≤60

81 (43.8%)

16 (35.6%)

>60

104 (56.2%)

29 (64.4%)

75 (53.6%)

59 (31.9%)

15 (33.3%)

44 (31.4%)

Validation cohort (n = 140, 75.7%) n (%) 65 (46.4%)

Gender

Female Male

126 (68.1%)

30 (66.7%)

96 (68.6%)

Body mass index, kg/m2

<25

78 (44.1%)

18 (40.0%)

60 (42.9%)

≥25

99 (55.9%)

23 (51.1%)

76 (54.3%)

Smoking

No

101(62.3%)

28 (62.2%)

73 (52.1%)

Yes

61 (37.7%)

12 (26.7%)

49 (35.0%)

≤14

6 (3.2%)

POSSUM score

0

>14

167 (90.3%)

43 (95.6%)

124 (88.6%)

≤22

93 (50.3%)

22 (48.9%)

71 (50.7%)

>22

92 (49.7%)

23 (51.1%)

69 (49.3%)

Procedure

Pylorus-preserving PD

42 (22.7%)

11 (24.4%)

32 (22.9%)

143 (77.3%)

34 (75.6%)

108 (77.1%)

Pancreatic anastomosis

Pancreaticogastrostomy

33 (17.8%)

10 (22.2%)

23 (16.4%)

Pancreaticojejunostomy

152 (82.2%)

35 (77.8%)

116 (82.9%)

Pancreatic texture

Hard

70 (37.8%)

18 (40.0%)

52 (37.1%)

71 (38.4%)

17 (37.8%)

54 (38.6%)

133 (72.3%)

32 (71.1%)

101 (72.1%)

Soft Pancreatic duct diameter, mm

>3 ≤3

Estimated blood loss, ml

≤1450 >1450

Postoperative day 1 urea, mmol/l

≤7.5 >7.5

Postoperative pancreatic fistulab

Post-pancreatectomy haemorrhageb

Cardiac complicationsc

Respiratory complicationsc

51 (27.6%)

12 (26.7%)

39 (27.9%)

102 (55.1%)

24 (53.3%)

78 (55.7%)

83 (44.9%)

21 (46.7%)

62 (44.3%)

108 (58.4%)

23 (51.1%)

85 (60.7%)

77 (41.6%)

22 (48.9%)

55 (39.3%)

121 (65.4%)

26 (57.8%)

95 (67.9%)

Grade A

21 (11.4%)

7 (15.6%)

14 (10.0%)

Grade B

27 (14.6%)

6 (13.3%)

21 (15.0%)

Grade C

16 (8.6%)

6 (13.3%)

10 (7.1%)

No

No

160 (86.5%)

37 (82.2%)

123 (87.9%)

Grade A

4 (2.2%)

1 (2.2%)

3 (2.1%)

Grade B

7 (3.8%)

2 (4.4%)

5 (3.6%)

Grade C

14 (7.6%)

5 (11.1%)

9 (6.4%)

Grades 0–II

183 (98.9%)

44 (97.8%)

139 (99.3%)

Grades III–V

2 (1.1%)

1 (2.2%)

1 (0.7%)

Grades 0–II

178 (96.2%)

44 (97.8%)

134 (95.7%)

Grades III–V

7 (3.8%)

1 (2.2%)

6 (4.3%)

Intra-abdominal collectionc

Grades 0–II

155 (83.8%)

38 (84.4%)

117 (983.6%)

Admissions to critical care

1 episode

Reoperation

No Yes

22 (11.9%)

7 (15.6%)

16 (11.4%)

Pathology

PDAC/chronic pancreatitis

79 (42.7%)

15 (33.3%)

64 (45.7%)

Grades III–V

>1 episode

Other Length of stay, days, median (range)d

Critical care Overall

90-day mortality

30 (16.2%)

7 (15.6%)

23 (16.4%)

148 (80.0%)

35 (77.8%)

113 (80.7%)

37 (20.0%)

10 (22.2%)

27 (19.3%)

163 (88.1%)

38 (84.4%)

124 (88.6%)

106 (57.3%)

0.202

0.812

0.981

0.251

6 (4.3%)

Preoperative bilirubin, mmol/l

Classical PD

P-valuea

0.153

0.832

0.827

0.390

0.808

0.940

0.781

0.257

0.533

0.308

0.396

0.529

0.890

0.669

0.467

30 (66.7%)

76 (54.3%)

0.145

6 (2–60)

6.0 (2.0–52.0)

7.0 (3.0–60.0)

0.213

15 (6–127)

16.0 (6.0–68.0)

15.0 (6.0–127.0)

0.909

9 (4.9%)

4 (8.9%)

5 (3.6%)

0.150

aChi-squared

test. Grade refers to International Study Group of Pancreatic Surgery (ISGPS) post-pancreatectomy complications. Grade refers to Clavien–Dindo classification of generic complications. dMann–Whitney U-test. Body mass index data available for 177 patients; smoking status available for 162 patients; POSSUM score available for 173 patients; pancreatic remnant texture data available for 140 patients; pancreatic duct diameter data available for 184 patients. POSSUM, physiological and operative severity score for the enumeration of mortality and morbidity; PDAC, pancreatic ductal adenocarcinoma; Other, ampullary adenocarcinoma, cholangiocarcinoma, duodenal carcinoma, neuroendocrine tumour. b c

HPB 2013, ••, ••–••

© 2013 International Hepato-Pancreato-Biliary Association

HPB

5

PoD 0 Serum amylase clinically significant POPF

N

Area Under Curve

P-value

95% Cl

Threshold serum amylase value (IU/L)

Sensitivity

Specificity

ROC Analysis

45

0.793

0.003

0.66–0.92

127.0

91.7

72.7

Chi-squared test

(a)

N

P-value

Sensitivity

Specificity

Positive predictive value

Negative predictive value

140

<0.001

71.0

65.1

36.7

88.8

(b)

Figure 1 Distribution of postoperative day (PoD) 0 serum amylase in patients who did and did not develop clinically significant postoperative

pancreatic fistula (POPF). Median PoD 0 serum amylase was 92 IU/l (range: 19–906 IU/l) in patients with no POPF or Grade A POPF, and 217 IU/l (range: 19–1833 IU/l) in patients with Grade B or C POPF (Mann–Whitney U-test, P < 0.001). (b) Receiver operating characteristic (ROC) analysis in the test cohort revealed a significant association between PoD 0 serum amylase and clinically significant POPF [area under the curve (AUC) 0.793, 95% confidence interval (CI) 0.66–0.92; P = 0.003]. The optimal threshold of PoD 0 serum amylase for predicting clinically relevant POPF was 127.0 IU/l, with sensitivity of 91.7% and specificity of 72.7%. ISGPF, International Study Group on Pancreatic Fistula

Predicting clinically significant POPF using serum amylase Table 1 illustrates the similarities between the test and validation cohorts with respect to demographic, operative and pathological characteristics and postoperative outcomes. The distribution of PoD 0 serum amylase in patients who went on to develop clinically significant POPF and those who did not is shown in Fig. 1a. Analysis by ROC (Fig. 1b) performed on the test cohort revealed a significant association between PoD 0 serum amylase and clinically significant POPF (AUC = 0.793, 95% CI 0.66–0.92; P = 0.003). The optimal threshold of PoD 0 serum amylase for predicting clinically relevant POPF was 127.0 IU/l, with sensitivity of 91.7% and specificity of 72.7%. This value was corrected to 130 IU/l for clinical use and validated in the subsequent 140 patients (chi-squared test). A PoD 0 serum amylase of ≥130 IU/l was strongly associated with the occurrence of clinically significant POPF (P < 0.001; sensitivity 71.0%, specificity 65.1%), with a negative predictive value of 88.8%. When ROC analysis was performed on the validation cohort (n = 140), or the entire cohort (n = 185), the threshold value of PoD 0 serum amylase was similar

HPB 2013, ••, ••–••

to that found when using the test group, and remained statistically significant with only minor variations in the AUC. Risk factors for POPF Univariate and multivariate binary logistic regression analyses of the associations between clinicopathological characteristics, intraoperative factors, PoD 0 serum amylase (as a dichotomized variable) and clinically significant POPF (Grade B or C) in the entire cohort (n = 185) are shown in Table 2. In a subset analysis examining the influence of pancreatic parenchyma on the level of PoD 0 serum amylase and clinically significant POPF, patients with a hard pancreatic remnant (n = 70) had a lower frequency of clinically significant POPF (4.3%) compared with those with a soft gland (39.4%). Serum amylase on PoD 0 was elevated (≥130 IU/l) in 21.4% of patients with a hard pancreatic remnant compared with 64.8% of patients with a soft pancreatic remnant. Of 71 patients with soft pancreatic parenchyma, 25 (35.2%) had a PoD 0 serum amylase of <130 IU/l. Only three of these patients developed a clinically significant POPF. Serum amylase of <130 IU/l on PoD 0 had a negative predictive

© 2013 International Hepato-Pancreato-Biliary Association

HPB

6

Table 2 The relationship between clinicopathological characteristics and clinically significant postoperative pancreatic fistula (ISGPS Grades B and C) in patients undergoing pancreaticoduodenectomy (n = 185): binary logistic regression analysis Univariate analysis Variable

n

OR

95% CI

Multivariate analysis P-value

OR

95% CI

P-value

4.67

1.50–14.60

0.008

3.09

1.02–9.35

0.045

9.50

2.33–38.78

0.002

7.44

2.24–24.70

0.001

Age ≤60 years

81

>60 years

104

2.12

1.02–4.40

0.043

1.74

0.79–3.82

0.169

2.96

1.34–6.52

0.007

0.79

0.36–1.71

0.545

0.61

0.11–3.45

0.576

0.75

0.38–1.49

0.407

4.16–50.80

<0.001

2.48

1.20–5.11

0.014

1.75

0.72–4.27

0.221

1.82

0.66–5.10

0.252

0.97

0.49–1.92

0.919

6.67

3.03–14.68

<0.001

1.66

0.83–3.29

0.149

3.12

1.43–6.81

0.004

Gender Female Male

59 126

Body mass index <25 kg/m2

78

≥25 kg/m2

99

Smoking No

101

Yes

61

POSSUM score <14

6

≥14

167

Preoperative bilirubin ≤22 mmol/l

93

>22 mmol/l

92

Pancreatic texture Hard

70

Soft

71

14.5

Pancreatic duct diameter >3 mm

133

≤3 mm

51

Procedure Pylorus-preserving PD Classical PD

43 142

Pancreatic anastomosis Pancreaticogastrostomy

33

Pancreaticojejunostomy

151

Estimated blood loss ≤1450 ml

102

>1450 ml

83

PoD 0 serum amylase <130 IU/l

105

≥130 IU/l

80

PoD 1 urea <7.5 mmol/l

108

≥7.5 mmol/l

77

Pathology PDAC/chronic pancreatitis Other

79 106

OR, odds ratio; 95% CI, 95% confidence interval; POSSUM, physiological and operative severity score for the enumeration of mortality and morbidity; PD, pancreaticoduodenectomy; PDAC, pancreatic ductal adenocarcinoma; Other, ampullary adenocarcinoma, cholangiocarcinoma, duodenal carcinoma, neuroendocrine tumour; PoD, postoperative day. Body mass index data available for 177 patients; smoking status available for 162 patients; POSSUM score available for 173 patients; pancreatic remnant texture data available for 140 patients; pancreatic duct diameter data available for 184 patients.

HPB 2013, ••, ••–••

© 2013 International Hepato-Pancreato-Biliary Association

HPB

7

Table 3 The relationship between postoperative day 0 serum amylase of ≥130 IU/l, non-fistulous complications and reoperation rates (n = 185)

Variable

Postoperative day 0 serum amylase <130 IU/l (n = 105, 56.8%) n (%)

≥130 IU/l (n = 80, 43.2%) n (%)

P-valuea

Postoperative pancreatic fistula (ISGPF) No

84 (80.0%)

37 (46.3%)

Grade A

11 (10.5%)

10 (12.5%)

Grade B

6 (5.7%)

21 (26.3%)

Grade C

4 (3.8%)

12 (15.0%)

<0.001

Post-pancreatectomy haemorrhage (ISGPS)b No

94 (89.5%)

66 (82.5%)

Grade A

2 (1.9%)

2 (2.5%)

Grade B

3 (2.9%)

4 (5.0%)

Grade C

6 (5.7%)

8 (10.0%)

0.173

Cardiac complicationsc Grades 0–II Grades III–V

105 (100%) 0

Respiratory complications

72 (97.5%) 2 (2.5%)

0.104

c

Grades 0–II

101 (96.2%)

77 (96.2%)

Grades III–V

4 (3.8%)

3 (3.8%)

0.983

Intra-abdominal abscessc Grades 0–II

96 (91.3%)

59 (74.2%)

Grades III–V

9 (8.7%)

21 (25.8%)

0.001

Admissions to critical care 1 episode

93 (88.6%)

55 (68.8%)

>1 episode

12 (11.4%)

25 (31.2%)

No

97 (92.4%)

65 (81.3%)

Yes

8 (7.6%)

15 (18.7%)

0.001

Reoperation 0.023

a

Chi-squared test. Grade refers to International Study Group of Pancreatic Surgery (ISGPS) classification of post-pancreatectomy complications. c Grade refers to Clavien–Dindo classification of generic complications. b

value of 88.0% for clinically significant POPF in patients with soft pancreatic parenchyma (Table S1, online). Serum amylase, complications and length of stay The relationship between PoD 0 serum amylase of ≥130 IU/l and clinically significant postoperative complications is demonstrated in Table 3. The median cumulative critical care environment (ICU or SHDU) length of stay was 6 days (range: 2–60 days) and the median postoperative hospital stay was 15 days (range: 6–127 days). The relationships between clinicopathological factors, operative factors and length of stay were considered for those 176 patients alive at discharge (Table S2, online). Predictors of postoperative hyperamylasemia Logistic regression was used to determine clinicopathological and operative covariates significantly associated with postoperative

HPB 2013, ••, ••–••

hyperamylasemia (PoD 0 serum amylase of ≥130 IU/l) (Table S3, online). A BMI of ≥25 kg/m2, a soft pancreatic remnant, a pancreatic duct diameter of <3 mm and non-PDAC or chronic pancreatitis pathology were significant by univariate analysis. After adjusting for these factors in a backward stepwise fashion, only a soft pancreatic remnant (OR = 5.30, 95% CI 2.32–12.13; P < 0.001) and non-PDAC or chronic pancreatitis pathology (OR = 2.78, 95% CI 1.22–6.38; P = 0.015) were found to be independently associated with raised PoD 0 serum amylase.

Discussion For the majority of pancreatic surgeons, intraoperative drainage of the pancreatic anastomosis is a standard component of PD, despite the weight of published evidence against it.20,22,37–39 Practice at the study institution favours the routine intraoperative placement of a drain and selective early removal. In its effort to identify patients in whom early drain removal may be appropriate, this study is the first to document serum amylase on the night of surgery (PoD 0 serum amylase) as an objective marker of patients with a low risk for POPF in whom early drain removal and rapid postoperative progress may be predicted. Evaluation of the study cohort shows that the rate of occurrence of at least one complication was high at 61.1%. This undoubtedly represents detailed, rigorous, prospective recording of postoperative events (rather than poor surgical technique) and is similar to the figure of 54% reported in 553 patients undergoing routine drainage at Memorial Sloan–Kettering Cancer Center after surgeries that included distal pancreatectomies and PDs.39 The 23.2% rate of clinically significant POPF in the current cohort is likewise broadly similar to that reported in the literature.40 The reported rate of clinically significant PPH is also high, but, as noted by Welsch et al., who documented an overall rate of PPH of 29.1%, the rigorous application of ISGPS criteria can potentially result in false positive reports of adverse events.41 Previously published evidence suggests that independent risk factors for POPF include: (i) a soft pancreatic remnant; (ii) a small pancreatic duct (≤3 mm in diameter); (iii) non-PDAC or chronic pancreatitis pathology, and (iv) increased intraoperative blood loss.6,16,19,42–45 However, these risk factors are dependent on subjective assessment by the operating surgeon; in particular, gland texture (reflecting both pancreas fat content and fibrosis) and pancreatic duct diameter remain challenging to quantify in a standardized fashion despite their prominence in models of POPF risk prediction and the proposal of a variety of assessment techniques.46 Factors related to an elevated PoD 0 serum amylase in this study included a soft pancreatic remnant, a pancreatic duct diameter of <3 mm, BMI of ≥25 kg/m2 and non-PDAC or chronic pancreatitis pathology; of these, the soft texture of the pancreatic remnant was the dominant determinant (Table S2). The present authors therefore suggest that PoD 0 serum amylase is an aggregate measure of established risk factors for POPF47 and could be utilized in the

© 2013 International Hepato-Pancreato-Biliary Association

HPB

8

immediate postoperative period as an objective confirmation of intraoperatively derived risk factors for POPF. Notably, even in patients with a soft pancreatic parenchyma, low PoD 0 serum amylase was associated with a very low risk for the occurrence of clinically significant POPF (Table S1), which suggests this simple assessment provides risk stratification data beyond those of gland texture and pancreatic duct size. A potential explanation for the utility of serum amylase is that an elevated PoD 0 level represents a marker of surgical trauma to a functioning pancreas, rather than a fistula in evolution. The inflammatory insult of the subsequent postoperative localized necrotizing pancreatitis (despite a ‘sub-diagnostic’ hyperamylasemia) may exacerbate the ongoing systemic inflammatory response to major abdominal surgery,48 leading to a cascade of systemic and local effects, including but not limited to POPF.49,50 A risk score is only of clinical utility if it carries the potential to alter clinical management. The high negative predictive value (88.8%) of a low PoD 0 serum amylase, in conjunction with lowrisk intraoperative factors, might identify patients at low risk for: (i) clinically significant POPF; (ii) clinically significant infected intra-abdominal collections, and (iii) a prolonged length of hospital stay. Such patients therefore might have their drains removed on PoD 1, potentially limiting drain-associated morbidity. By contrast, patients with a PoD 0 serum amylase level of ≥130 IU/l might benefit from prolonged drainage. This study provides justification for drain removal on the first postoperative day on the basis of a PoD 0 serum amylase level of <130 IU/l. This drain management strategy will identify reactionary haemorrhage or bile leak, avoid the morbidity associated with prolonged and unnecessary drainage, and retain control in patients at high risk for POPF.

POPF. Postoperative day 0 serum amylase identifies patients at low risk for POPF and a complicated postoperative course, who would potentially be suitable for selective, early drain removal, thereby avoiding the potential morbidity associated with prolonged drainage. Acknowledgements The authors thank the following members of the West of Scotland Pancreatic Unit: Dr Cindy Chew and Dr Hedvig Karteszi for diagnostic imaging, clinical nurse specialist Sr Elspeth Cowan and audit secretary Diane Stewart for follow-up data, and the staff of the surgical high-dependency unit. Conflicts of interest None declared. References 1. DeOliveira ML, Winter JM, Schafer M, Cunningham SC, Cameron JL, Yeo CJ et al. (2006) Assessment of complications after pancreatic surgery: a novel grading system applied to 633 patients undergoing pancreaticoduodenectomy. Ann Surg 244:931–937. 2. Yeo CJ, Cameron JL, Sohn TA, Lillemoe KD, Pitt HA, Talamini MA et al. (1997) Six hundred fifty consecutive pancreaticoduodenectomies in the 1990s: pathology, complications, and outcomes. Ann Surg 226:248–257. 3. Gouma DJ, van Geenen RC, van Gulik TM, de Haan RJ, de Wit LT, Busch OR et al. (2000) Rates of complications and death after pancreaticoduodenectomy: risk factors and the impact of hospital volume. Ann Surg 232:786–795. 4. Bentrem DJ, Yeh JJ, Brennan MF, Kiran R, Pastores SM, Halpern NA et al. (2005) Predictors of intensive care unit admission and related outcome for patients after pancreaticoduodenectomy. J Gastrointest Surg 9:1307–1312. 5. Bassi C, Butturini G, Molinari E, Mascetta G, Salvia R, Falconi M et al. (2004) Pancreatic fistula rate after pancreatic resection. The importance of definitions. Dig Surg 21:54–59. 6. Bartoli FG, Arnone GB, Ravera G, Bachi V. (1991) Pancreatic fistula and

Limitations Despite the retrospective nature of this non-randomized analysis in a relatively small cohort operated in a single institution, the test and validation cohort strategy utilized in this study provides strong evidence for the utility of PoD 0 serum amylase to predict POPF. Although these data generated through the analysis of a routine and robust serum biochemical test are impressive, this preliminary investigation has not demonstrated that acting upon a PoD 0 serum amylase threshold will impact on POPF rate. Further validation is necessary and is currently ongoing. Ideally, a prospective trial of drain removal in the context of serum amylase on PoD 0 will be necessary to generate Level I evidence that can ultimately guide drain management in patients undergoing PD.

relative mortality in malignant disease after pancreaticoduodenectomy. Review and statistical meta-analysis regarding 15 years of literature. Anticancer Res 11:1831–1848. 7. Callery MP, Pratt WB, Vollmer CM. (2009) Prevention and management of pancreatic fistula. J Gastrointest Surg 13:163–173. 8. Bassi C, Falconi M, Molinari E, Salvia R, Butturini G, Sartori N et al. (2005) Reconstruction by pancreaticojejunostomy versus pancreaticogastrostomy following pancreatectomy: results of a comparative study. Ann Surg 242:767–771. 9. Suc B, Msika S, Fingerhut A, Fourtanier G, Hay J-M, Holmières F et al. (2003) Temporary fibrin glue occlusion of the main pancreatic duct in the prevention of intra-abdominal complications after pancreatic resection: prospective randomized trial. Ann Surg 237:57–65. 10. Pratt WB, Maithel SK, Vanounou T, Huang ZS, Callery MP, Vollmer CM. (2007) Clinical and economic validation of the International Study Group of Pancreatic Fistula (ISGPF) classification scheme. Ann Surg 245:443–

Conclusions The results of the current study suggest that an elevated PoD 0 level of serum amylase represents a robust and objective predictor of POPF. This simple and routine measurement provides confirmation of subjective intraoperatively derived risk factors for

HPB 2013, ••, ••–••

451. 11. Enestvedt CK, Diggs BS, Cassera MA, Hammill C, Hansen PD, Wolf RF. (2012) Complications nearly double the cost of care after pancreaticoduodenectomy. Am J Surg 204:332–338. 12. Daskalaki D, Butturini G, Molinari E, Crippa S, Pederzoli P, Bassi C. (2011) A grading system can predict clinical and economic outcomes of

© 2013 International Hepato-Pancreato-Biliary Association

HPB

9

pancreatic fistula after pancreaticoduodenectomy: results in 755 con-

28. Relles DM, Richards NG, Bloom JP, Kennedy EP, Sauter PK, Leiby BE et al. (2013) Serum blood urea nitrogen and serum albumin on the first

secutive patients. Langenbecks Arch Surg 396:91–98. 13. Petermann D, Demartines N, Severe SM. (2013) Postoperative complications adversely affect longterm survival after R1 resection for pancre-

postoperative day predict pancreatic fistula and major complications after pancreaticoduodenectomy. J Gastrointest Surg 17:326–331. 29. Kawai M, Tani M, Hirono S, Ina S, Miyazawa M, Yamaue H. (2009) How do

atic head adenocarcinoma. World J Surg 37:1901–1908. 14. Schmidt CM, Choi J, Powell ES, Yiannoutsos CT, Zyromski NJ, Nakeeb

we predict the clinically relevant pancreatic fistula after pancreatico-

A et al. (2009) Pancreatic fistula following pancreaticoduodenectomy:

duodenectomy? An analysis in 244 consecutive patients. World J Surg 33:2670–2678.

clinical predictors and patient outcomes. HPB Surg 2009:404520. 15. Kawai M, Kondo S, Yamaue H, Wada K, Sano K, Motoi F et al. (2011)

30. Winter JM, Cameron JL, Yeo CJ, Alao B, Lillemoe KD, Campbell KA et al.

Predictive risk factors for clinically relevant pancreatic fistula analysed in

(2007) Biochemical markers predict morbidity and mortality after

1239 patients with pancreaticoduodenectomy: multicentre data collec-

pancreaticoduodenectomy. J Am Coll Surg 204:1029–1036.

tion as a project study of pancreatic surgery by the Japanese Society of

31. Neoptolemos JP, Stocken DD, Bassi C, Ghaneh P, Cunningham D,

Hepato-Biliary-Pancreatic Surgery. J Hepatobiliary Pancreat Sci 18:601–

Goldstein D et al. (2010) Adjuvant chemotherapy with fluorouracil plus

608.

folinic acid vs. gemcitabine following pancreatic cancer resection: a

16. Akamatsu N, Sugawara Y, Komagome M, Shin N, Cho N, Ishida T et al.

randomized controlled trial. JAMA 304:1073–1081.

after

32. Jamieson NB, Foulis AK, Oien KA, Going JJ, Glen P, Dickson EJ et al.

pancreaticoduodenectomy: the significance of the ratio of the main

(2010) Positive mobilization margins alone do not influence survival fol-

pancreatic duct to the pancreas body as a predictor of leakage.

lowing pancreatico-duodenectomy for pancreatic ductal adenocarci-

(2010)

Risk

factors

for

postoperative

pancreatic

fistula

noma. Ann Surg 251:1003–1010.

J Hepatobiliary Pancreat Sci 17:322–328. 17. Tsujie M, Nakamori S, Miyamoto A, Yasui M, Ikenaga M, Hirao M et al. (2012) Risk factors of pancreatic fistula after pancreaticoduodenectomy – patients with low drain amylase level on postoperative day 1 are safe from

33. Copeland GP, Jones D, Walters M. (1991) POSSUM: a scoring system for surgical audit. Br J Surg 78:355–360. 34. Wente MN, Veit JA, Bassi C, Dervenis C, Fingerhut A, Gouma DJ et al. (2007) Post-pancreatectomy haemorrhage (PPH): an International Study

developing pancreatic fistula. Hepatogastroenterology 59:2657–2660. 18. Okabayashi T, Kobayashi M, Nishimori I, Sugimoto T, Onishi S, Hanazaki

Group of Pancreatic Surgery (ISGPS) definition. Surgery 142:20–25.

K. (2007) Risk factors, predictors and prevention of pancreatic fistula

35. Dindo D, Demartines N, Clavien P-A. (2004) Classification of surgical

formation after pancreatoduodenectomy. J Hepatobiliary Pancreat Surg

complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann Surg 240:205–213.

14:557–563. 19. Pratt WB, Callery MP, Vollmer CM. (2008) Risk prediction for develop-

36. Zweig MH, Campbell G. (1993) Receiver-operating characteristic (ROC)

ment of pancreatic fistula using the ISGPF classification scheme. World

plots: a fundamental evaluation tool in clinical medicine. Clin Chem 39:561–577.

J Surg 32:419–428. 20. Conlon KC, Labow D, Leung D, Smith A, Jarnagin W, Coit DG et al. (2001)

37. Petrowsky H, Demartines N, Rousson V, Clavien PA. (2004) Evidence-

Prospective randomized clinical trial of the value of intraperitoneal drain-

based value of prophylactic drainage in gastrointestinal surgery: a systematic review and meta-analyses. Ann Surg 240:1074–1084.

age after pancreatic resection. Ann Surg 234:487–493. 21. Fisher WE. (2012) Pancreas Resection With and Without Drains. ClinicalTrials.gov2012.

Available

at

http://www.clinicaltrials.gov/ct2/

38. Kawai M, Tani M, Terasawa H, Ina S, Hirono S, Nishioka R et al. (2006) Early removal of prophylactic drains reduces the risk of intra-abdominal

5

infections in patients with pancreatic head resection: prospective study

22. Bassi C, Molinari E, Malleo G, Crippa S, Butturini G, Salvia R et al. (2010)

39. Correa-Gallego C, Brennan MF, D'Angelica M, Fong Y, Dematteo RP,

Early versus late drain removal after standard pancreatic resections:

Kingham TP et al. (2013) Operative drainage following pancreatic resec-

results of a prospective randomized trial. Ann Surg 252:207–214.

tion: analysis of 1122 patients resected over 5 years at a single institution.

show/NCT01441492?term=nct01441492&rank=1

(last

accessed

October 2013).

23. Bassi C, Dervenis C, Butturini G, Fingerhut A, Yeo C, Izbicki J et al. (2005) Postoperative pancreatic fistula: an international study group (ISGPF) definition. Surgery 138:8–13. 24. Molinari E, Bassi C, Salvia R, Butturini G, Crippa S, Talamini G et al. (2007) Amylase value in drains after pancreatic resection as predictive factor of postoperative pancreatic fistula: results of a prospective study in 137 patients. Ann Surg 246:281–287. 25. Sutcliffe RP, Battula N, Haque A, Ali A, Srinivasan P, Atkinson SW et al. (2012) Utility of drain fluid amylase measurement on the first postoperative day after pancreaticoduodenectomy. World J Surg 36:879–883.

for 104 consecutive patients. Ann Surg 244:1–7.

Ann Surg. doi: 10.1097/SLA.0b013e3182813806. 40. Reid-Lombardo KM, Farnell MB, Crippa S, Barnett M, Maupin G, Bassi C et al. (2007) Pancreatic anastomotic leakage after pancreaticoduodenectomy in 1507 patients: a report from the Pancreatic Anastomotic Leak Study Group. J Gastrointest Surg 11:1451–1458. 41. Welsch T, Eisele H, Zschäbitz S, Hinz U, Büchler MW, Wente MN. (2011) Critical appraisal of the International Study Group of Pancreatic Surgery (ISGPS) consensus definition of postoperative haemorrhage after pancreatoduodenectomy. Langenbecks Arch Surg 396:783–791. 42. Yeo CJ, Cameron JL, Lillemoe KD, Sauter PK, Coleman J, Sohn TA et al.

26. Robertson EA, Zweig MH. (1981) Use of receiver operating characteristic

(2000) Does prophylactic octreotide decrease the rates of pancreatic

curves to evaluate the clinical performance of analytical systems. Clin

fistula and other complications after pancreaticoduodenectomy? Results

Chem 27:1569–1574.

of a prospective randomized placebo-controlled trial. Ann Surg 232:419–

27. Uemura K, Murakami Y, Sudo T, Hashimoto Y, Nakashima A, Yamaoka E

429.

et al. (2012) Elevation of urine trypsinogen 2 is an independent risk factor

43. Lin JW, Cameron JL, Yeo CJ, Riall TS, Lillemoe KD. (2004) Risk factors

for pancreatic fistula after pancreaticoduodenectomy. Pancreas 41:876–

and outcomes in post-pancreaticoduodenectomy pancreaticocutaneous

881.

fistula. J Gastrointest Surg 8:951–959.

HPB 2013, ••, ••–••

© 2013 International Hepato-Pancreato-Biliary Association

HPB

10

44. Marcus SG, Cohen H, Ranson JH. (1995) Optimal management of the pancreatic remnant after pancreaticoduodenectomy. Ann Surg 221:635– 645. 45. Choe YM, Lee KY, Oh CA, Lee JB, Choi SK, Hur YS et al. (2008) Risk factors affecting pancreatic fistulas after pancreaticoduodenectomy.

emptying after pancreaticoduodenectomy. J Gastrointest Surg 10:1131– 1139. 50. Räty S, Sand J, Nordback I. (2007) Detection of postoperative pancreatitis after pancreatic surgery by urine trypsinogen strip test. Br J Surg 94:64–69.

World J Gastroenterol 14:6970–6974. 46. Gaujoux S, Cortes A, Couvelard A, Noullet S, Clavel L, Rebours V et al. (2010) Fatty pancreas and increased body mass index are risk

Supporting information

factors of pancreatic fistula after pancreaticoduodenectomy. Surgery

Additional supporting information may be found in the online version of this

148:15–23.

article at the publisher's website.

47. Callery MP, Pratt WB, Kent TS, Chaikof EL, Vollmer CM. (2013) A pro-

Table S1. Relationships between pancreatic parenchyma, postoperative day

spectively validated clinical risk score accurately predicts pancreatic

0 serum amylase and clinically significant postoperative pancreatic fistula.

fistula after pancreatoduodenectomy. J Am Coll Surg 216:1–14.

Table S2. Relationships between clinicopathological characteristics and

48. Marik PE, Flemmer M. (2012) The immune response to surgery and

length of hospital stay following pancreaticoduodenectomy (n = 176): Cox

trauma: implications for treatment. J Trauma Acute Care Surg 73:801–

regression analysis.

808.

Table S3. Relationships between postoperative day 0 serum amylase of

49. Räty S, Sand J, Lantto E, Nordback I. (2006) Postoperative acute pancreatitis as a major determinant of postoperative delayed gastric

HPB 2013, ••, ••–••

≥130 IU/l and clinicopathological characteristics in patients undergoing pancreaticoduodenectomy (n = 185): binary logistic regression analysis.

© 2013 International Hepato-Pancreato-Biliary Association