A total of 471 consecutive patients underwent PD in our center. Fifty seven patients developed a POPF of any type. 21 patients had a fistula type A , 22 patients (4.7%) had a fistula type B and the remaining 14 patients (3%) had a POPF type C. Cirrhotic liver(P=0.05), BMI >25 kg/m2 (P=0.0001), soft pancreas (0.04), pancreatic duct diameter <3 mm (0.0001), pancreatic duct located <3 mm from the posterior border (P=0.02) were significantly associated with POPF. with the multivariate analysis both BMI and pancreatic duct diameter were demonstrated to be independent factors. The hospital mortality in this series was 11 patients (2.3%), the development of POPF type C was associated with a significantly increased mortality ( 7/14 patients). POD 1 and 5 a drain amylase level > 4000 IU/L, WBC, pancreatic duct diameter, pancreatic texture were predictors for clinical evident POPF .
conclusion: Cirrhotic liver, BMI , soft pancreas , pancreatic duct diameter <3 mm , pancreatic duct near posterior border are risk factors for development of POPF. POD1 and 5 a drain amylase level, WBC, pancreatic duct diameter, pancreatic texture may be predictors for POPF B,C
Get your grade
or your money back
using our Essay Writing Service!
Pancreaticoduodenectomy (PD) is a complex surgical procedure. It has been established as a standard surgical operation for malignant and benign diseases in pancreatic head and periampulary regions (1-3). Recently, the operative mortality rate after PD has dramatically decreased to less than 5%, while the incidence of postoperative morbidity remains high, from 40% to 50% (1-5). In the majority of cases, morbidity and mortality after PD are related to surgical management of the pancreatic stump (3-6)
Pancreatic leakage remains the most important cause of morbidity, and also contributes significantly to prolonged hospitalization, increased health care costs and mortality. It remains a challenge at high volume centers for pancreatic surgery (4,5). There is marked variability in the incidence of pancreatic anastomotic leakage after PD among different series, ranging from 5 to 30% (5-9). This wide range could be attributed to the lack of a universally accepted definition of leak. Intra-abdominal abscess, intra-abdominal bleeding, and sepsis are common sequelae of pancreatic leakage, which have been associated with high mortality rate of 40% or more (7, 8).
Recent studies have suggested that many factors influence PF after PD, including age, sex, preoperative jaundice, operative time, intraoperative blood loss, type of pancreatic reconstruction, anastomotic technique, consistency of pancreatic stump, pancreatic duct diameter, use of somatostatin and surgeon experience (6-11). This problem has been studied in many well designed trials addressed surgical techniques (9-11), modified drainage regimens (10-13), or administration of somatostatin (13). The best technique in pancreatic anastomosis is still debated (10-13).
The risk factors for leakage could not always predict the severity and extent of leak and could not precisely distinguish clinically relevant pancreatic fistula from transient pancreatic fistula. The aim of the present study was to analyze perioperative risk factors for pancreatic leakage after PD and evaluate the factors that may predict the extent and severity of leak.
Patients and methods
We retrospectively studied all patients who underwent PD for malignant and benign diseases in pancreatic head and periampulary region in our Gastroenterology Surgical Center, Mansoura University, Egypt, from January 2001 to June 2012. The medical records of patients were reviewed. We use a computerized pancreatic surgery sheet since 2000 which had all preoperative, intraoperative and postoperative variables for each patient (routine practice). Informed consent for the surgical procedures was obtained from each patient. This study was approved by local ethical committee.
Preoperative diagnostic workup included clinical assessment (age, sex, symptoms and signs), laboratory investigations ( complete blood count, liver functions , creatinine, serum amylase, and tumor markers as CEA and CA19-9), radiological investigations (abdominal ultrasound, magnetic resonance cholangiopancreatography MRCP, and abdominal computerized tomography). Preoperative biliary drainage was performed by endoscopic retrograde cholangiopancreatography (ERCP) in selected patients, with serum levels of total bilirubin greater than 10 mg/dl or when biliary obstruction was associated with hepatic dysfunction (transaminase: more than threefold than the normal i.e. more than 120 IU/ml) were detected.
Standard Whipple type operation was performed in 455 patients (96.7%) while the remaining 16 patients (3.4%) underwent a pylorus preserving PD (PPPD). Pancreatic reconstruction was done by either pancreaticogastrostomy (PG) or pancreaticojejunostomy (PJ) based on surgeon preference. Biliary drainage was achieved by end to side hepaticojejunostomy (retrocolic). However, gastric drainage was achieved by gastrojejunostomy (GJ) (antecolic or retrocolic) (manual or using stapler) in standard Whipple operation or dudenojujunostomy in PPPD (end to side or end to end).
Always on Time
Marked to Standard
All patients were managed in the intensive care unit for at least one day before transfer to the ward. All patients received prophylactic antibiotics intraoperatively and for 4 days postoperatively. Prophylactic octreotide was given subcutaneously and continued postoperatively for 4 days, in patients considered high risk for pancreatic fistula.
Outputs from operatively placed drains were recorded daily. The drain was removed in all enrolled patients if no bile leak, pancreatic leak or pus. Outputs from nasogastric tube were recorded daily, and it was removed if the patients passed flatus, no distention, or the daily output less than 500 ml. The nasogastric tube was reinserted after two episodes of vomiting. The patients resumed oral feeding started by a fluid diet, followed by a regular diet once the bowel movement restarted and could be tolerated oral feeding.
The amylase of serum and drainage fluid was measured on postoperative day (POD)1, and POD 5. Liver functions were measured on POD1, and POD6. Abdominal ultrasound was done routinely for all patients, and repeated if we suspect intraabdominal collection. US guided tubal drainage was done if there is abdominal collection.
Postoperative pancreatic fistula was defined as proposed by International Study Group of Pancreatic Fistula (ISGPF) as any measurable volume of fluid on or after POD 3 with amylase content greater than 3 times the serum amylase activity, and classified into grades A, B, C. (14-16)
No fistula group of patients lacks both elevated amylase levels and any clinical sequelae of fistula. Briefly, grading of POPF is dependent on the clinical course: a grade A POPF is transient and asymptomatic fistula (no clinical impact), does not need specific treatment and the abdominal drain is removed within 3 weeks. Grade B POPF is symptomatic, clinically apparent that requires diagnostic evaluation and specific medical treatment or prolonged drainage longer than 3 weeks. Grade C POPF is severe, clinically significant fistula that requires major changes in clinical management or deviation from the normal clinical pathway i.e. requires invasive therapy. Grade B and C were considered to constitute clinically relevant postoperative pancreatic fistula (POPF) (14-16)
Biliary leak was defined as the presence of bile in the drainage fluid that persists to POD 4. Delayed gastric emptying was defined as output from a nasogastric tube of greater than 500 ml per day that persisted beyond POD 10, the failure to maintain oral intake by POD 14, or reinsertion of a nasogastric tube (14, 15).
All patients had a baseline history and physical examination records. Preoperative, intraoperative and postoperative data were collected. Preoperative variables included patient demographics (age, sex, and medical history), patients symptoms (jaundice, abdominal pain, vomiting, diarrhea, loss of weight, diabetes), physical signs (jaundice, abdominal mass, and body weight), laboratory tests (preoperative total bilirubin, alkaline phosphatase, SGPT, albumin, amylase, complete blood picture, tumor markers CEA, CA19-9), preoperative image studies, and preoperative biliary drainage by ERCP.
Intraoperative variables included liver status (normal or cirrhotic), tumor size, pancreatic duct diameter, shape of pancreatic stump and relation of the duct to its borders, consistency of the pancreas (soft, firm), techniques of pancreatic reconstruction (continuous, interrupted or both), type of sutures used (absorbable, non absorbable or both), operative time, blood loss and blood transfusion.
Postoperative variables included postoperative complications (pancreatic leak, bile leak, delayed gastric emptying, hemorrhage, cardiopulmonary complications, collection, bleeding PG, GJ, liver insult, and wound infection). Data of the drain including its output, its nature, and POD of its removal, the amylase of serum and drainage fluid, liver function at day 1 and at day 6, day of starting oral, length of postoperative stay, re-exploration (time and causes), hospital mortality (day and cause), all pathologic specimens were reviewed to show tumor size, tumor type, tumor differentiation, lymph node status, surgical safety margins
Statistical analysis of the data in this study was performed using SPSS software, version 17. For continuous variables, descriptive statistics were calculated and were reported as mean ± standard deviation (SD). Categorical variables were described using frequency distributions. Independent sample t- test was used to detect differences in the means of continuous variables and Chi-square test was used in cases with low expected frequencies. P values ï€¼ 0.05 were considered to be significant. Variables with P<0.05 were entered into a logistic regression model to determine independent risk factors of postoperative PF. The independent risk factors of the variables were expressed as odds ratios (OR) with their 95% confidence intervals (CI). The measurement of drain amylase level on POD 1 and POD 5 has a major impact by enabling the development of PF to be predicted in the early period after PD. In fact, amylase level in drains > 4000 U/L have reported to be a significant predictive factor for the incidence of all grades PF after PD (17,18). Therefore, the optimal cut-off levels of the drain amylase level on POD 1 and POD 5 for differentiation between no PF group and PF group were sought by receiver operating characteristics (ROC) curves, which were generated by calculating sensitivities and specificities of the drain amylase level on POD 1 and POD 5 at different predetermined cut off points. Line graphs were used for graphical visualization.
Characteristics of patients and operative data:
This Essay is
a Student's Work
This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.Examples of our work
Patients characteristic are listed in Table (1). A total of 471 consecutive patients (278 (59%) men and 193 (41%) women) underwent PD in our Gastroenterology Surgical Center, Mansoura University, Egypt, from January 2001 to June 2012. The mean age of patients was 52.58 ± 10.82 years. The most common initial symptoms included jaundice in 431 (91%) patients, abdominal pain in 339 (72%) and weight loss in 175 (37.2%). Preoperative biliary stenting was done in 250 patients (53.1%).
Of the 471 patients, only 16 patients (3.4%) underwent PPPD while the majority 455 patients (96.7%) underwent classic PD. The mean operative time was 5.15± 1.04 hours (range, 2.5-9 hours), and the mean operative blood loss was 533.4 ± 427.36 ml (range,50-3000 ml).
In all patients, intraoperative data consist of consistency of pancreatic parenchyma were documented. Of these patients, 307 (65.2 %) had soft parenchyma and 164 (34.6 %) had firm parenchyma (table 1).
Postoperatively, the mean hospital stay was 11.15 ± 8.2 days (4-71 days), and the mean ICU stay was 1.38 ± 1.65 days (range, 1-50 days). The mean time to resume oral intake was 7.05 ± 5.5 days (range, 0- 52 days). The drain was removed after 10.16 ± 7.68 days postoperatively (4-71days) (Table 2).
As regards postoperative complications, fifty seven patients (12.1%) developed a POPF, 21 patients (4.5%) had a fistula type A, 22 patients (4.7%) had a fistula type B and the remaining 14 patients (3%) had a POPF type C (Table 2)..
Other postoperative complications included delayed gastric emptying in 57 patients (12.1%), intra-abdominal collection in 44 patients (9.3%), wound infection in 33 patients (7%), biliary leakage in 27 patients (5.7%), pulmonary complication in 25 patients (5.3%), internal haemorrhage occurred in 13 patients (2.8%) (4/13 due to erosion of gastroduodenal artery secondary to PF), bleeding gastrojejunostomy in 13 patients (2.8%), bleeding PG in 8 patients (1.7%), liver insult in 7 patients (1.5%), pancreatitis 7 patients (1.5%), pulmonary embolism in 3 patients (0.6%), many of them occurred in combination with others (Table 2).
Ultrasound guided tubal drainage was required in 44 patients with intra-abdominal collection. Thirty three patients required re-exploration because of internal haemorrhage (13 patients, 4/13 due to erosion of gastroduodenal artery), bleeding gastrojejunostomy (10 patients), bleeding PG (8 patients) or debridement and drainage (2 patients). Completion spleno-pancreatectomy was required in one patients had PF and complicated by internal haemorrhage due to erosion of gastroduodenal artery (Table 2).
The hospital mortality in this series was 11 patients (2.3%), and the mortality associated with pancreatic fistula was 8 patients (8/57). The causes of death were liver cell failure as a result accompanying liver cirrhosis (one patient), pulmonary embolism (3 patients), septic shock as a sequence of PF (6 patients) and secondary haemorrhage as sequence of PF (one patient) (Table 2).
Postoperative pathological types were listed in (Table 3).
Risk factors for POPF
General risk factors that were evaluated are shown in (table 4). The incidence of pancreatic fistula was 14/74 (18.9%) in patients with cirrhotic liver compared to 43/397 (10.8%) patients having non-cirrhotic liver (P=0.05). The incidence of PF was 28/112 (25%) in patients with BMI greater than 25 kg/m2, and was 29/359 (8.07%) in those with BMI less than or equal 25 kg/m2 (P= 0.0001).
Patients with soft pancreatic parenchyma or a pancreatic duct diameter < 3mm had a significantly higher incidence of PF. The incidence of PF was 38/133 (28.6%) in patients with pancreatic ducts less than or equal 3 mm and was 19/388 (4.9%) in those with ducts more than 3 mm (P=0.0001). The PF was 13/164 (7.9%) in patients with a firm pancreas, and was 44/307 (14.3%) in those with a soft pancreas (P=0.04). Patients with pancreatic duct close to the posterior edge of pancreatic stump were likely to develop PF, the incidence of PF was 35/224 (15.6%) in patients with pancreatic duct close to the posterior edge less than or equal 3 mm and was 22/247 (8.9%) in those with ducts away from the posterior edge more than 3 mm (P=0.02) (Table 4).
The mean days of removal of drain in patients without PF (8.46 ± 3.9 range; 4-33 days) was significantly eariler than in patients with PF ( 21 ± 11.5 range; 4-71 days) Table (4).
Univariate analysis demonstrated five factors to be significantly associated with pancreatic fistula (BMI, cirrhotic liver status, parenchymal consistency, pancreatic duct diameter, location of pancreatic duct from posterior edge). These five risk factors of PF identified in univariate analysis were further analyzed in multivariate analysis. Both pancreatic duct diameter less than or equal 3 mm and BMI >25 kg/m2 were demonstrated to be independent risk factors (Table 5).
Predictive factors for severity of PF
The mean level of amylase in drain on POD 1, was 2416.3 ± 2027.2 U/L in cases with PF compared with 562.97± 1045.69 U/L in cases without complications (P=0.0001) and on POD 5 was 10525.9 ± 11931.5 U/L in cases with PF compared with 455.09 ± 744.79 U/L in cases without complications (P=0.0001). The mean level of amylase in drains tended to decrease from POD 1 to POD 5 in patients without PF but it tended to increase from POD 1 to POD 5 in patients with PF(Table 6).
We recommended cut off level based on an amylase elevation in the drainage fluid greater than 3 times the upper normal level of serum amylase, which is 200U/L in our hospital.
Considering the sensitivity and specificity of the drain amylase on POD 1, an area under the ROC curve of 0.797 was obtained (p<0.0001; 95% confidence interval: 0.72-0.87) (Fig 1). If Drain amylase level > 1000 U/L on POD 1 was suggested to be the best cut-off for prediction of the clinically relevant pancreatic fistula. The sensitivity, and specificity of drain amylase on POD 1 > 1000 U/L were 71.9% and 86.5% respectively. If Drain amylase level > 4000 U/L on POD 1 was suggested to be the best cut-off for prediction of the clinically relevant pancreatic fistula. The sensitivity, and specificity of drain amylase on POD 1 > 4000 U/L were 28.1% and 97.2% respectively.
With regard to the sensitivity and specificity of the drain amylase on POD 5, an area under the ROC curve of 0.96 was obtained (p<0.0001; 95% confidence interval: 0.93-0.99) (Fig 2). Drain amylase level > 4000 U/L on POD 5 was suggested to be the best cut-off for prediction of the clinically relevant pancreatic fistula. The sensitivity, and specificity of drain amylase on POD 5 > 4000 U/L were 73.7% and 99.3% respectively.
With development of PF , male sex, BMI >25 kg/m2, cirrhotic liver, small pancreatic duct diameter less than or equal 3 mm, pancreatic duct located close to the posterior border within a distance of 3mm, and soft pancreatic texture were a predictive factor for the severity of PF(Table 6).
Postoperative changes in leukocyte counts, and serum albumin are presented in Table 6. Leukocyte counts on POD 5 were significantly decreased compared to POD 1 in PF type A group. In contrast to the grade B and C group leukocyte counts on POD 5 increased significantly compared to POD 1 level. The serum albumin level on POD 5 decreased significantly in grade B and C groups compared to POD 1 level.
Patients with POPF type C were significantly associated with vascular complications likes erosion of gastroduodenal artery leading to internal haemorrhage in 4/14 (28.6%), delayed gastric emptying in 11/14 (78.6%) and pleural effusion in 3/14 (21.4%). Reoperation was required in 8/14 patient (57.1%). Mortality attributed to surgical complications after POPF type C was 7/14 (50%) (Table 6).
Outcome and management of PF
Patients were primarily objected to conservative management without interventional procedure (administration of antibiotics, subcutaneous octreoide, enteral or total parentral nutritional support). All patients had PF type A were successfully managed by conservative treatment. Ultrasound guided tubal drainage was required in 26 patients having intra-abdominal collection (19 patients had PF type B and 7 patients had PF type C. Ten patients required re-exploration the causes were internal haemorrhage in 5 patients, (4 patients due to erosion of gastroduodenal artery and one due to secondary haemorrhage ), debridement and drainage (2 patients) and bleeding PG (3 patients). Completion spleno-pancreatectomy was required in one patients had PF and complicated by internal haemorrhage due to erosion of gastroduodenal artery (Table 6).
No patients with a POPF type A developed other surgical complications and there was no mortality in that group. One patient with POPF type B died. In contrast, the development of POPF type C was associated with a significantly increased mortality (seven out of 14 patients) (Table 6).
Postoperative pancreatic fistula (POPF) after PD remains a challenge even at high volume centers. It remains the major contributor to morbidity after PD (5-9, 19). Many efforts have been made to mitigate this problem including surgical techniques (13, 20-24), administration of somatostatin, and use of adhesive sealants (13).
In this study, we found that PF in a series of 471 consecutive PD was fifty seven patients (12.1%) which appears to be comparable to the PF rate of 5-14 % reported in other specialized center (5-7, 25-30).
The risk of PF formation appears to be multifactorial involving preoperative, intraoperative, and postoperative factors. Many studies reported that male sex was a significant predictor of PF (30, 31). Shmidt et al (30) reported that male sex only achieved significance as a univariate predictor of PF formation in the overall group of patients undergoing PD. In our study, no significant difference between male and female. Some studies found that old age was a significant risk factor of PF, this result was not found as a risk factor in our study.
In our study, BMI, cirrhotic liver status, pancreatic parenchymal consistency, pancreatic duct diameter, location of the pancreatic duct within a distance of 3mm from the posterior border of the pancreas were shown to be significantly associated with developing PF. Multivariate analysis revealed that the pancreatic duct diameter less than or equal 3 mm and BMI >25 kg/m2 were significant independent risk factors. In many series, patients undergoing PD have been categorized into two groups based on character of pancreatic remnant: patients with a soft, fragile, small pancreatic duct, or pancreatic duct located close to the posterior border within a distance of 3mm, who were considered at high risk for PF, and patient with firm, fibrotic, dilated pancreatic duct, or pancreatic duct located far from posterior border, who are at low risk (28, 31-36). Our results approved these reported data.
In our study, Preoperative biliary drainage did not influence the incidence of postoperative complications, and although it can be performed safely in jaundiced patients it should not be used routinely. Preoperative biliary drainage was performed by ERCP in selected patients, with serum levels of total bilirubin greater than 10 mg/dl or when biliary obstruction was associated with hepatic dysfunction (transaminase: more than threefold than the normal i.e. more than 120 IU/ml). There was no significant difference in overall PF between patients with and without preoperative biliary drainage. Preoperative biliary drainage was introduced in an attempt to improve the general condition and thus reduce postoperative morbidity and mortality. Early studies showed a reduction in morbidity (37-38). However, recent studies found that preoperative biliary stenting has been associated with high incidence of PF and other complications (30, 39-40). Therefore, the overall conclusion not to routinely perform preoperative biliary drainage seems evident. Whether biliary drainage should always be performed in jaundiced patients' remains controversial (41-43).
This study showed that the type of anastomosis had no significant impact on the incidence of PF. This finding is matched with the result of Bassi et al (11). Yeo et al (35) found that the incidence of pancreatic fistula was nearly similar for the PG (12.3%) and PJ (11.1%) groups with no significant difference. McKay et al (44) concluded in meta-analysis study that current literature suggests that the safer means of pancreatic reconstruction after PD is PG, but much of the evidence comes from observational cohort study data.
It remains unclear what risk factors can precisely predict which type of POPF (clinical relevant (POPF type B and C) or transient pancreatic fistula (POPF type A) will occur when diagnosing pancreatic fistula on POD3 by ISGPF criteria. We have evaluated the factors that may predict the extent and severity of leak. In our analysis, we found that soft consistency of pancreas, elevation of amylase in drainage fluid on POD 1 and 5, elevation of WBC on POD 1 and 5, small pancreatic duct diameter, its relation to posterior border of pancreas, cirrhotic liver, and obesity significantly increase the risk for development of a type B and C fistula
It is important to predict development of a clinically relevant pancreatic fistula in the early period after PD. In many studies, soft pancreatic parenchyma has been widely recognized as the most significant risk factor of PF (45-48). Kawai M et al found that soft parenchyma was not predictive of pancreatic leak type B and C, although it was an independent risk factor regarding incidence of pancreatic fistula (15).
In the literatures there are few studies that have tried to evaluate the predictive value of amylase in drains with the risk of developing PF, despite the fact that amylase values significantly affect postoperative management (17,49-50). Sutcliffe RP et al (51) reported that the median drain amylase level on POD1 in patients with PF (6205; range 357-23391) was significantly higher than in patients without a PF (69; range 5-2180; P=0.01), and found that no patients with a PF had drain fluid amylase level on POD1 < 350 U/L, compared to 48/61 (79%) without a PF. Using 350 U/L as a cut off, a low drain amylase on POD1 excluded a PF with sensitivity, specificity, positive and negative predictive values 100, 79, 41, and 100% respectively ( 51 ). Molinari et al concluded that a drainage amylase value on POD 1 greater than 5,000 U/l was a significant predictive factor of PF with a sensitivity, specificity, positive and negative predictive values 93, 84, 59, and 98 % respectively. (17). The retrospective study by Shinchi et al (50) on 207 PD defined PF as an output >30 ml/24h with amylase values on POD5 that were more than 5 times the serum value. Hashimoto N and Ohyanagi H (52) reported that a mean level of amylase in drain on POD 1, 10878 ± 14800 U/L in cases with PF compared with 1482 ± 1615 U/L in cases without complications. Some studies reported that amylase level in drainage fluid after PD has no clinical impact (19, 49). Kawai M et al reported that there was no significant difference in amylase level in drainage fluid on POD 1 and POD 4 between transient pancreatic fistula and clinically significant pancreatic fistula, although the increase was significantly greater in cases of pancreatic fistula as compared to those with no pancreatic fistula. Therefore, measuring daily levels of amylase in drainage fluid may not reflect the severity of pancreatic fistula (15). In the other hand, Reido-Lombardo et al have suggested that the ability to detect clinical PF by drain data alone is imperfect (19). In our study we found that, the mean level of amylase in drain on POD 1, was 2416.3 ± 2027.2 U/L in cases with PF compared with 562.97± 1045.69 U/L in cases without complications (P=0.0001) and on POD 5 was 10525.9 ± 11931.5 U/L in cases with PF compared with 455.09 ± 744.79 U/L in cases without complications (P=0.0001).The level of amylase in drain on POD1 and POD 5 has predictive value for the appearance of PF, and in particular when the level of amylase is greater than 4000 U/L.
Kawai M et al found that white blood counts (WBC) on POD 4 were significantly decreased compared to POD 1 in both the no pancreatic fistula group and the grade A group. However, in the POPF type B and C WBC on POD 4 did not decrease significantly compared to POD 1 levels (15). In our study, Leukocyte counts on POD 5 were significantly decreased compared to POD 1 in PF type A group. In contrast to the grade B and C group, leukocyte counts on POD 5 increased significantly compared to POD 1 level
Our study proved that patients with POPF type C are significantly associated with vascular complications likes erosion of gastroduodenal artery leading to internal haemorrhage in 4/14 (28.6%). Reoperation was required in 8/14 patient (57.1%). Mortality attributed to surgical complications after POPF C was 7/14 (50%). Pratt et al (14) reported that POPF C is associated with higher rates of complications, surgical or radiological interventions, ICU and overall hospital stay, and overall cost. Frymerman AS et al (45) found that patients who developed POPF had significantly more vascular but not other surgical complications than patients without POPF. Patients with POPF A had no vascular or surgical complications. Twenty one of the 29 patients with POPF C had surgical complications (17 vascular complications). Mortality attributed to surgical complications after POPF C was 5/29. A soft pancreatic consistency (OR 8.5; p < 0.008) and a high drain lipase activity on postoperative day 3 (OR 4.4; p = 0,065) were predictors for the development of POPF C.
The role of surgically placed prophylactic intra-abdominal drain after pancreatic resection and their effect on morbidity rate and optimal timing for their removal remains controversial (53-57). In our study, the mean days of removal of drain in patients without PF (8.46 ± 3.9 range; 4-33 days) was significantly eariler than in patients with PF ( 21 ± 11.5 range; 4-71 days). Bassi et al after a randomized controlled study on 114 patients who underwent pancreatic resection reported that in patients at low risk of PF, intraabdominal drain can safely removed on POD3 after standard pancreatic resection. A delayed drain removal is associated with a higher rate of postoperative PF (P=0.0001) with increased hospital stay and costs (53) . Kawai M et al concluded that early removal of intra-abdominal drain on POD 4 reducing intra-abdominal infection and was significantly decrease the rate of PF (54) . Although early drain removal is considered desirable, in some cases this may be followed by intra-abdominal collection or sepsis. Yeo et al and Yamaguchi M et al reported that prophylactic drains after PD allow monitoring of the occurrence of intraabdominal bleeding, as well as the detection and drainage of PF (55,56). Most of high volume centers insert a prophylactic drains and usually removed around POD 7 (3, 17, 51,55, 57).
There are several limitations to our study because of the retrospective nature of data collection and the surgical procedures such as PG or PJ or use of pancreatic stent were not randomized but depended on surgeon preference. Therefore, further studies are necessary to prospectively validate these predictive risk factors to confirm the possible relationship between these factors and development of PF.
In conclusion, This study demonstrates five factors to be significantly associated with pancreatic fistula after PD, BMI >25 kg/m2, cirrhotic liver status, soft parenchymal consistency, small pancreatic duct diameter, and location of pancreatic duct within 3 mm distance from posterior edge. Amylase value > 4000 U/L in drain, serum albumin <3 g/dl and leukocyte counts >10,000 on POD1 and POD 5 can predict clinically relevant PF (grade B and C) before it becomes severe. Management of PF depends on the degree of PF; therefore, the identification of these predictive risk factors can provide useful information to assist the surgeon in making a decision about pancreatic reconstruction technique intraoperative and to tailor the postoperative management for patients who are at an increased risk of developing PF. To prevent development of PF we introduce somatostatin analogue treatment and antibiotic therapy for high risk group and delay removal of the drain with close follow up by abdominal ultrasound. Intraoperative placed drain is left in place until leakage stop. However, in cases of severe PF further management depends on rapid decision for re-exploration, especially in cases of vascular complication or uncontrolled sepsis. Surgeons should continue to investigate to reduce the incidence of PF after PD.