Abstract
Objective
The aim of this study was to evaluate the predictive value of the first postoperative day (POD1) drain fluid amylase in predicting pancreatic fistula formation following pancreaticoduodenectomy (PD).
Material and Methods
One-hundred and eighty-five prospective patients undergoing PD between April 2014 and April 2018 were studied retrospectively. Cut-off point to predict the development of POPF was determined by median values for drain fluid amylase of 1883 U/L. Patients were classified into two groups according to POD1 drain fluid amylase values: <1883 U/L (Group 1) and ≥1883 U/L (Group 2). Differences between the groups with clinically relevant POPF and without POPF were evaluated.
Results
The incidence of POPF was 17.2%. POD1 amylase level was the strongest predictor of POPF, with levels of higher than 1883 U/L demonstrating the best accuracy (87.5%), sensitivity (78.1%), specificity (89.5%), positive predictive value (60.9%), and negative predictive value (95.1%). One-hundred and forty-four patients (77.8%) had a POD1 drain amylase level of less than 1883 U/L, and POPF developed in only seven (3.7%) cases, whereas in patients with POD1 drain amylase level of 1883 U/L or higher (n= 41), the POPF rate was 31.4% [OR: 22.24, 95% CI (7.930–62.396), p<0.001].
Conclusion
The cut-off point of POD1 drain fluid amylase level (1883 U/L) might predict the clinically relevant POPF with adequate sensitivity and specificity rates in patients undergoing pancreatic resection.
Introduction
Despite the implementation of advancing techniques into postoperative management of the patients, complication rates after pancreatic resection remain high. Postoperative pancreatic fistula (POPF) is a common complication that threatens the patient’s life after pancreaticoduodenectomy (PD). POPF can cause adverse consequences, including the development of intra-abdominal abscesses and intra-abdominal or intramucosal hemorrhage (1, 2). For that reason, identifying high-risk patients for POPF in the perioperative period is critical for appropriate patient management planning.
The cut-off values were different among studies. Differences in the definition of POPF, differences in the surgical procedure, and retrospective studies prevent the establishment of the optimal cut-off value. From the viewpoint of the mechanism of POPF, the differences in surgical procedures do not affect fistula formation. A large-scale prospective trial is warranted to establish an optimal cut-off value that applies to any surgical procedure (3). However, there is not any consensus on POD1 drain fluid amylase cut-off value. Several studies have suggested that drainage fluid analyses contribute to predicting POPF, although there has long been a debate on the preventive and therapeutic implementation of abdominal drains after performing a pancreatic resection (4-10). The controversial results may be explained by the fact that any cohort of patients undergoing PD may not be homogeneous to determine the diagnostic and therapeutic implications of prophylactic abdominal drainage, in the absence of objective criteria to be used to determine the individual risks for developing POPF. Recently, the levels of amylase in the drainage fluid have received considerable attention, with researchers asking whether they can serve to predict PF or to decide the optimum time for drain removal.
In several studies, predictive values (sensitivity, specificity, and positive and negative predictive values) were investigated and reported as strongly favorable for POPF (11, 12). For this reason, we aimed to determine the optimal predictive value of POD1 drain fluid amylase level in estimating POPF after PD.
Material and Methods
This study was performed after obtaining the permission of Gazi University Faculty of Medicine Clinical Research Ethics Committee (Decision Number: 425). Patients who had elective pancreaticoduodenectomy in the General Surgery Clinic of Gazi University Medical Faculty during the period from April 2014 to April 2018 participated in the study. All patients included in the study were operated by two experienced surgeons, performing more than 50 operations per year. Demographic data of the study patients and their intraoperative, pathological, and postoperative data were logged by the surgeons.
The standard PD as described in the literature was performed on all patients (13). In addition, the lymph nodes between the coeliac trunk and the superior mesenteric artery (triangular area) were retrieved, and the proximal jejunum was pulled up for P-J through the Treitz hole that occurred after duodenectomy. Pyloric preservation was not performed on any patients. An internal stent at the appropriate size was placed between the pancreatic duct and the jejunal opening in all patients, and external drainage was not performed. Two abdominal drains were placed around the P-J and hepaticojejunostomy. Blumgart anastomosis was performed on all patients as described; three or four transpancreatic U-sutures were placed approximately 1 cm distal to the transected edge of the pancreas going from front to back with suture. The seromuscular back wall of the jejunum was utilized, nearer the mesenteric edge, and 10 to 15 mm longitudinal seromuscular stitches were placed. The suturing then reverted to back to front through the pancreas completing the U. A duct-to-mucosa anastomosis was constructed with four to six stitches of polydioxanone 5-6/0 after creating a small opening (about 2-3 mm) in the jejunal loop. The transpancreatic U-sutures were then passed through the seromuscular layer of the anterior wall of the jejunal loop in the direction of the short axis. Each of the U method sutures was placed at a distance of 5 to 10 mm from the last. This procedure completely covered the pancreatic stump with jejunal serosa and protected the knots from cutting through the pancreatic tissue.
Resection and reconstruction of the superior mesenteric vein and portal vein were accepted as vessel resection. To prevent the development of stress ulcerations, proton pump inhibitors were administered to all patients. The parenchymal texture of the pancreas was determined as soft or hard by the surgeon by manually palpating the pancreatic remnant. After the operation, one drain was placed adjacent to the pancreaticojejunostomy and hepaticojejunostomy. The drain fluid amylase (DFA) levels on POD 1-, 3-, and -5 were recorded daily after the operation.
Postoperative pancreatic fistula was defined and classified according to the International Study Group for Pancreatic Surgery-2016 (14). The diagnosis of POPF was made if the amylase level of the drain fluid was three times higher than the serum amylase level on the 3rd postoperative day. This was regarded as biochemical leakage (BL). Grade B fistula was defined as fistula requiring any changes in the treatment (need for antibiotics, enteral or parenteral nutrition, percutaneous, endoscopic, or angiographic interventions) or fistula with abdominal drainage lasting longer than 21 days. Fistula-related organ failure, reoperation, or death was accepted as Grade C fistula. Grade B and C fistulas were defined as clinically relevant POPF (CR-POPF).
First, univariate and multivariate analyses were used to retrospectively investigate the relationship between patient and tumor characteristics, intraoperative, and postoperative features. Cut-off points to predict the development of POPF were determined by median values. Patients were classified into two groups according to POD1 drain fluid amylase level; those with a value <1883 U/L were named the low amylase group, and those with ≥1883 U/L were named the high amylase group. Patient and tumor characteristics were recorded, including age, sex, body mass index (BMI), pathological diagnosis (malignant or benign), operation time (min), estimated blood loss (mL), vascular resection performed/none, soft or hard pancreas parenchyma, main pancreatic duct size (mm), postoperative day one, two, and three drain fluid amylase levels, the median value of POD1 drain amylase value level (≥1883 U/L vs. <1883 U/L), Clavien-Dindo complication classification, and the re-operation rate.
The main pancreatic duct diameter was measured perioperatively by the surgeon who performed the operation. The morbidity rate in the postoperative period was determined by including all complications developing just after the surgery and throughout the postoperative period until the patient was discharged from the hospital. The Clavien-Dindo classification was used for complications (15). Delayed gastric emptying in the postoperative period was diagnosed by the International Study Group on Pancreatic Surgery.
Statistical Analysis
All data were summarized as mean ± standard deviation (SD). For univariate analysis, non-paired t-test and χ2 test were used, and for multivariate analyses, logistic regression analysis was used. P-values <0.05 were considered statistically significant. Cut-off points to predict the development of POPF were determined by median values. Analysis of data and the assessment of sensitivity, specificity, predictive values, and accuracy were performed with diagnostic formulas. SPSS program was used for statistical analysis.
Results
Patient Characteristics
One-hundred and eighty-five patients undergoing PD were included in the study. Median age was 62 years (range= 24-85 years), and there were 104 male (56.2%) patients. Mean BMI of the patients was 26.3 ± 17.8 m2 /kg. Among the patients, 86 (46.5%) had comorbid factors; and 70 (37.8%) patients had ASA III/IV scores. Histopathological diagnosis was benign in 34 (18.4%) patients, and it was reported to be malignant in 81.6% of the patients. Ninety-nine (53.5%) patients were identified as having hard pancreatic parenchyma, and 86 (46.5%) patients had soft parenchyma. The main pancreatic duct diameter was <4 mm in 62 patients (33.5%), and 41 patients (22.2%) needed a blood transfusion during the operation. The postoperative pancreatic fistula rate was 17.3% (32). While the rate of POPF development was 25% (8) in patients with hard pancreatic parenchyma, it was 75% (24) in patients with soft pancreatic parenchyma. The postoperative pancreatic fistula was identified in 24 patients with soft and eight patients with hard pancreatic parenchyma. The main pancreatic duct diameter was less than 4 mm in 68% of patients with POPF (Table 1).
Using the median drain fluid amylase level (<1883 U/L vs. ≥1883 U/L) as a cut-off value.
Clinicopathological characteristics of the patients, intraoperative and postoperative features are presented in Tables 2, 3, and 4. Patients were divided into two groups: Patients with POD1 drain fluid amylase level lower than 1883 U/L (Group 1) and patients with POD1 drain fluid amylase ≥1883 U/L (Group 2). In 21.9% (n= 7) of the patients with POPF, POD1 drain fluid amylase levels were lower than 1883 U/L. Postoperative day one drain fluid amylase levels were higher than 1883 U/L in 78.1% (n= 25) of the patients who developed fistulas on POD1 (Table 2). There were more patients with hard pancreatic parenchyma (59.7% vs. 31.7%, p= 0.002) and ≥4 mm duct diameter (73% vs. 41%, p< 0.001) in Group 1 than Group 2. Patients with soft parenchyma were 46.5% (OR= 3.689, CI= 1.048-12.987) and more likely to form POPF compared to those patients with hard pancreatic parenchyma (p= 0.045). Patients with a small duct diameter (<4 mm) were 33.5% (OR= 3.732, CI= 1.333-10.447) and more likely to form POPF compared to those patients with a larger duct diameter (p= 0.012). It was determined that the need for blood transfusion was higher in Group 2 (36.6% vs. 18.1%, p= 0.012), and the results are shown in Table 3. Postoperative day one in Group 2 was related to a higher rate of overall morbidity that was significantly different (Table 4). There were no statistically significant differences between the groups in terms of demographic patients characteristics, ASA score, comorbidity, histopathological diagnosis, tumor location, vascular resection, estimated blood loss, internal stent replacement, operation time, or length of hospital stay (Tables 2, 3, and 4).
Predictors of PF
The multivariate analysis revealed that a POD1 drain fluid amylase level higher than 1883 U/L (OR= 22.24, 95% CI= 7.93- 62.39), p= 0.0001), soft parenchyma (OR= 3.689, 95% CI= 1.04- 12.98, p= 0.045), and <4 mm main pancreatic duct size (OR= 3.732, 95% CI= 1.33-10.44, p= 0.012) were the independent factors for POPF development, which is shown in Table 5.
Postoperative day one drain fluid amylase level was found to be the powerful predictor of POPF. The best accuracy (87.5%), sensitivity (78.1%), specificity (89.5%), positive predictive value (PPV) (60.9%), and negative predictive value (NPV) (95.1%) were found for POD1 drain fluid amylase level with 1883 U/L or higher (Table 6). The postoperative day one drain fluid amylase level was <1883 U/L in 144 patients (77.8%), and POPF was seen in seven patients (3.7%). The POPF incidence rate was 31.4% among 41 patients with ≥1883 U/L POD1 drain amylase level [OR= 22.24, 95% CI (7.930-62.396), p< 0.001].
Discussion
Postoperative pancreatic fistula (POPF) formation following pancreatic surgery remains a challenge even in pancreatic centers with a high volume of patients. The reported incidence of POPF after PD ranges between 2% and 46% (16). Therefore, predicting the patients who will develop POPF might facilitate the management of this challenging complication and improve surgical outcomes. There are no clearly established tools to identify high-risk patients for developing POPF. The potential risk factors, such as soft parenchymal texture, narrow duct, intraoperative blood loss, and high BMI, have been suggested to be associated with POPF. But their predictive values are not adequate.
However, the early prediction of POPF remains controversial despite the availability of several various methods, proposed to estimate the development of POPF along with the associated risk factors. The precision of fistula according to the ISGPF in 2005 suggested that the drain fluid amylase level measured on postoperative day three or later and postoperative day one drain fluid amylase levels for diagnosis of POPF were not discussed in the ISGPF consensus meeting. This topic is controversial because there are conflicting results on using postoperative day one drain fluid amylase levels to predict the pancreatic fistula (10, 17-22).
In our study, the cut-off value of postoperative day one drain fluid amylase level (1883 U/L) had 78.6% sensitivity, 89.5% specificity, and 87.5% accuracy rates for predicting POPF. A common point of the studies investigating this issue is that drain fluid amylase is very valuable in predicting the clinically relevant POPF on the POD1. However, in almost all studies, there is no consensus on what the cut-off value should be. When the cut-off value is 2365 U/L, 78% sensitivity, 80% specificity, 66% PPV, and 88% NPV have been reported (23). In another study, when the cut-off value was accepted as 350 U/L, the authors reported 79% specificity, 100% sensitivity, 41% PPV, and 100% NPV. Similarly, studies have reported that the cut-off values 4000 U/L and 5000 U/L are optimal for predicting the POPF (24, 25). Our results are consistent with these trials, and we recommend that POD1 drain fluid amylase levels can be used in identifying high-risk patients to predict pancreatic fistula after PD.
Yang Ji et al. reported that eight studies included in their metaanalysis provided published evidence indicating the predictive value of POD1 drain fluid amylase for developing POPF. Ansorge et al. investigated several clinical parameters for their predictive values for estimating the development of POPF and concluded that a POD1 drain fluid amylase level of 1322 U/L was the most clinically relevant parameter for predicting POPF (20).
Similarly, several studies have researched the significance of drain amylase levels for predicting POPF (8, 11, 26). El Nakeeb et al. investigated risk factors for POPF and recommended that a 1000 U/L cut-off value is best for predicting a clinically relevant pancreatic fistula (11). Facy et al. evaluated the levels of both lipase and amylase in the drain fluid in predicting POPF and found that the 500 U/L cut-off value of drain fluid amylase level was a good indicator for clinically relevant POPF (8). Kawai et al. retrospectively analyzed 1239 patients undergoing PD and demonstrated that a POD1 drain fluid amylase >4000 IU/L could predict POPF, based on a ROC curve analysis (18). The authors reported that this cut-off point was associated with a sensitivity of 62.2%, specificity of 89%, and accuracy of 84.8%. Another study reported that higher than 5000 U/L POD1 drain fluid amylase level is a significant predictive factor for POPF following distal pancreatectomy and PD (18). Similarly, another study reported that POD1 drain fluid amylase level higher than 5000 U/L is the best predictive marker for POPF, and they reported a sensitivity of 93% for PD and 100% for distal pancreatectomy (27). In another study that investigated whether POD1 drain fluid amylase could be used to estimate POPF following pancreatectomy, the authors reported that a low cut-off level of 100 U/L was associated with high sensitivity and NPV. Therefore, they recommended that early drain removal would be safe when drain amylase is lower than 100 U/L on the first postoperative day (19).
In our opinion, the wide range of proposed cut-off values depends on whether the drain amylase value is used to determine pancreatic fistula or patients who will not have a fistula. High cut-off values accurately predict those patients that will have a fistula, and low cut-off values predict those patients who will not have a fistula. Small differences in NPV for each cut-off value indicate that clinically relevant consequences of early drain removal should be reviewed from various aspects for the purposes of guiding management, rather than defining one POD1 drain amylase value as a cut-off. Therefore, the cutoff value may be different for each clinic.
In our opinion, it is important to determine the fistula or the patients who do not have a fistula. We believe that the cut-off value should then be determined according to the patient’s fistula status. The comfort level of the surgeon should also be determined in selecting the cut-off point based on the small differences in NPV.
Conclusion
The cut-off point of POD1 drain fluid amylase level (1883 U/L) can help predict the clinically relevant POPF with adequate sensitivity and specificity rates in patients undergoing pancreatic resection. Multi-center, high-volume trials are now required to further investigate these findings.
