Which is the best predictor of clinically relevant pancreatic fistula after pancreatectomy: drain fluid concentration or total amount of amylase?

Abstract Aim Drain fluid amylase concentration (DFAC) has been reported as a predictor of clinically relevant postoperative pancreatic fistula (CR‐POPF) after pancreatectomy. However, the clinical significance of measuring the total drain fluid amylase amount (DFAA) considering the daily drainage volume of CR‐POPF remains unclear. Methods Data from 216 consecutive patients who underwent pancreaticoduodenectomy (PD) (n = 126) or distal pancreatectomy (DP) (n = 90) between August 2014 and November 2020 were reviewed. All drains were closed but not suctioned. DFAA was calculated by multiplying the DFAC and daily drainage fluid volume. DFAC and DFAA were recorded on d 1 and 3 after pancreatectomy. The cutoff value of CR‐POPF was determined using the receiver operating characteristic curve. Results CR‐POPF was found in 75 patients (35%) (PD: 30%, DP: 41%, P = .111); the mortality rate was zero. The cutoff value of DFAC‐day 1 was 1757 U/L (sensitivity [SE]: 84%, specificity [SP]: 62%, and accuracy [AC]: 69%). The cutoff value of DFAA‐day 1 was 139 U (SE: 71%, SP: 72%, and AC: 71%). The cutoff value of DFAC‐day 3 was 1044 U/L (SE: 73%, SP: 79%, and AC: 78%). The cutoff value of DFAA‐day 3 was 21 U (SE: 68%, SP: 72%, and AC: 70%). Multivariate analysis indicated that a nondilated pancreatic duct and high DFAC‐day 3 were independently associated with CR‐POPF after PD, indicating that a prolonged operative duration, massive blood loss, and high DFAC‐day 3 are independently associated with CR‐POPF after DP. Conclusion DFAC is more reliable than DFAA for predicting CR‐POPF after both PD and DP.


| INTRODUC TI ON
Postoperative pancreatic fistula (POPF) remains one of the most common complications after pancreatic surgery, such as pancreaticoduodenectomy (PD) and distal pancreatectomy (DP). Despite modifications in surgical techniques and perioperative patient care to prevent POPF, the incidence of POPF has been reported to range from 3%-50%, even at high-volume centers. [1][2][3][4] The International Study Group of Pancreatic Surgery (ISGPS) has developed and validated a universally applicable definition for POPF. 5,6 According to the classification, grade B or C POPF is re- Many studies have demonstrated risk factors for CR-POPF, such as pancreatic texture, pancreatic duct size, body mass index (BMI), and massive intraoperative blood loss. 7,8 Recently, several studies have reported that the drain fluid amylase concentration (DFAC) on the 1st and 3rd d after pancreatectomy can be a reliable predictor of CR-POPF. [9][10][11][12][13][14][15][16][17][18] DFAC can reveal the optimal timing of drain removal after pancreatectomy by data-driven decisions. However, the value of the drain fluid amylase amount (DFAA) when considering the daily drainage volume has not been investigated, so it is unclear whether DFAC or DFAA is the more reliable predictor of CR-POPF after pancreatectomy. Therefore, this study aimed to investigate the practical significance of DFAA as a predictor of CR-POPF following pancreatectomy.

| PATIENTS AND ME THODS
Between August 2014 and November 2020, 216 consecutive patients who underwent PD (n = 126) or DP (n = 90) at our institution were enrolled in the study. Patients undergoing hepatopancreaticoduodenectomy (n = 12), central pancreatectomy (n = 4), total pancreatectomy (n = 18), or partial pancreatectomy (n = 1) were excluded. The demographics, clinical characteristics, operative details, and postoperative outcomes of patients with and without CR-POPF were retrospectively analyzed. All patients provided written informed consent before undergoing therapy. This study was approved by the Institutional Review Board of our institution (No. 2020-119) and was performed in accordance with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.

| Surgical procedure
After review by a multidisciplinary board, all pancreatic disease cases were assessed by pancreatic surgeons to determine resectability and the most appropriate surgical procedure. The subtotal stomachpreserving method was the standard procedure for PD. In patients with malignant disease, lymph nodes are dissected at the hepatoduodenal ligament, around the common hepatic artery, around the superior mesenteric artery (SMA), and around the pancreatic head.
Transection of the pancreatic parenchyma was performed with an electric scalpel. A modified Child method, with duct-to-mucosa pancreatojejunostomy, was chosen for organ reconstruction in all cases. The modified Blumgart mattress suture was the procedure of choice for pancreatic remnant reconstruction when feasible. 19,20 In all cases, a 4-Fr polyethylene tube was placed through the pancreatojejunal anastomotic site as an external stent. Three silastic flexible drains were routinely placed adjacent to the anastomosis and at both the cranial and caudal sites of the pancreatojejunostomy and choledochojejunostomy.
In the case of DP, transection of the pancreatic parenchyma was performed with a stapler from 2017 on. A silastic flexible drain was placed at the pancreatic stump. However, in the case of radical antegrade modular pancreatosplenectomy (RAMPS) for pancreatic ductal adenocarcinoma (PDAC), an additional drain was placed at the left subphrenic space.
Energy devices, such as LigaSure (TM) (Covidien, Japan) was used only during laparoscopic surgery. All drains were closed but not suctioned. Prophylactic octreotide to prevent POPF was not administered in either PD or DP.

| Definition of DFAC and DFAA
Amylase levels in the drainage fluid were routinely measured on postoperative d (PODs) 1 and 3 (ie, DFAC-day 1 and DFAC-day 3).
If there were multiple drains, the highest DFAC-day 1 value was defined as analysis drain. DFAA-day 1 was calculated by multiplying DFAC-day 1 and the 24-h drainage volume from the morning of POD 1. In the same way, DFAA-day 3 was calculated by multiplying DFACday 3 and the 24-h drainage volume from the morning of POD 3.
The drain was removed on POD 4 or 5 if the drainage fluid was clear and pancreatic fistula and bacterial contamination were absent. In the case that bacterial contamination was detected based on a bacterial culture test of drain fluid, or suspected by nonserous (turbid) fluid of the drain, drains were replaced on POD 7.

| Postoperative complications
Any complications that developed within 90 d after the operation were included. No patients were lost to follow-up. CR-POPF included grade B or C POPF based on the definition of ISGPS. 6 Bile leakage was defined according to International Study Group of Liver Surgery (ISGLS) criteria. 21 Intraabdominal bleeding and delayed gastric emptying (DGE) were also defined by ISGPS criteria. 22,23 Surgical mortality was defined as perioperative death within the first 90 d following surgery.

| Statistical analysis
The continuous data are expressed as the medians (ranges). The statistical analyses were performed using chi square tests, Mann-Whitney U-tests, or Fisher's exact probability tests, as appropriate. The predictive ability of DFAC and DFAA for the occurrence of CR-POPF was assessed by calculating the area under the receiver operating characteristic (ROC) curve. The variables identified as potentially significant by univariate analysis were selected for multivariate analysis with a logistic regression model to identify the independent predictors of CR-POPF. All P values were 2-sided, and P < .05 was considered to indicate a statistically significant difference. All statistical calculations were performed using the IBM SPSS Statistics 27 software package (IBM Japan).

| Patient characteristics and surgical outcomes
The patient demographics and clinical characteristics are shown in Table 1. CR-POPF was found in 75 patients (35%). Among the CR-POPF patients, grade C POPF occurred in only two patients (3%) who underwent PD, and the remaining 73 patients (97%) had grade B POPF. The CR-POPF group had a significantly higher BMI than the no CR-POPF group (P < .001). The CR-POPF group had a significantly lower incidence of PDAC than the no CR-POPF group (P < .001). The median age, sex, preoperative albumin concentration, and incidence of comorbidities did not differ between the two groups.
The laparoscopic approach was performed in 24 cases during DP.
Among the 126 cases of PD, the incidence of patients who underwent the modified Blumgart method was 79% (100 cases). Table 2 shows the surgical outcomes after pancreatectomy. The type of pancreatectomy, operative duration, and total blood loss volume did not differ between the two groups. There were significant differences between the two groups in terms of the pancreatic duct size <3 mm (53% in the no CR-POPF group vs 85% in the CR-POPF group) and soft pancreatic texture rate (62% in the no CR-POPF group vs 80% in the CR-POPF group). The median DFAC-day 1, DFAA-day 1, DFACday 3, and DFAA-day 3 were significantly higher in the CR-POPF group than in the no CR-POPF group. On the other hand, the median drainage volumes on POD 1 and POD 3 were significantly lower in the CR-POPF group than in the no CR-POPF group ( Figure S1).
There were significant differences between the two groups in terms of intraabdominal bleeding (0% in the no CR-POPF group vs 5% in the CR-POPF group) and median postoperative hospital stays (

| Cutoff values of DFAC/DFAA-day 1 and day 3 for predicting CR-POPF
The ROC curves for generating cutoff values of DFAC/DEAA-day 1 and day 3 for rates of CR-POPF for all patient groups are shown in Figure 1 and Table 3. The cutoff value of DFAC-day 1 was 1757 U/L, with a sensitivity of 84%, specificity of 62%, and accuracy of 69%.
The cutoff value of DFAA-day 1 was 139 U, with a sensitivity of 71%, specificity of 72%, and accuracy of 71%. The cutoff value of DFACday 3 was 1044 U/L, with a sensitivity of 73%, specificity of 79%, and accuracy of 78%. The cutoff value of DFAA-day 3 was 21 U, with a sensitivity of 68%, specificity of 72%, and accuracy of 70%.
Altogether, the results indicated that the most reliable predictor of CR-POPF after pancreatectomy was DFAC-day 3, which had the highest area under the ROC curve (AUC) value (0.843; optimal cutoff value 1044 U/L).

| Subgroup analysis of DFAC/DFAA according to the type of pancreatectomy and predictors of CR-POPF
Subgroup analysis of CR-POPF according to the type of pancreatectomy was performed. The median operative duration was 424 min, and total blood loss was 495 mL following PD. The median operative duration was 245 min, and total blood loss was 178 mL following DP. CR-POPF was found in 38 patients (30%) after PD and 37 patients (41%) after DP (P = .111). The ROC curves for generating cutoff values of DFAC/DEAA-day 1 and day 3 for rates of CR-POPF after PD are shown in Figure S2 and Table 4.   Table 5).
The ROC curves for generating cutoff values of DFAC/DEAAday 1 and day 3 for rates of CR-POPF after DP are shown in Figure S3 and Table 6. The most reliable predictor of CR-POPF after DP was DFAC-day 3 (AUC = 0.819; optimal cutoff value 3506 U/L).
The multivariate analysis results indicated that an operative duration independently associated with CR-POPF after DP (Table 7).

| D ISCUSS I ON
Pancreatic surgery has become safer, especially in high-volume centers, due to advances in technology and perioperative management.
Nevertheless, CR-POPF is the most common complication and subsequently triggers other infectious complications, which is concerning for both patients and surgeons. after pancreatectomy in a large sample. They showed that the type of drain is not associated with increased CR-POPF or other postoperative outcomes. In the future, large-series multicenter studies evaluating the clinical impact of DFAC will help to compensate for the limitations of this study.
In conclusion, the results of our study indicate that DFAC is more reliable than DFAA as a predictor of CR-POPF after pancreatectomy.
Routine postoperative assessment of DFAC could provide more meticulous follow-up after pancreatectomy. DFAC can indicate the optimal timing of drain removal after pancreatectomy. However, the cutoff value of DFAC may vary slightly between institutions.
Therefore, we propose an early drain removal strategy based on institution-specific DFAC values, along with consideration of other risk factors for CR-POPF.