Dual‐center randomized clinical trial exploring the optimal duration of antimicrobial prophylaxis in patients undergoing pancreaticoduodenectomy following biliary drainage

Abstract Objectives The aim of this dual‐center randomized controlled trial was to determine the optimal duration of antimicrobial prophylaxis in patients treated with pancreaticoduodenectomy (PD) who underwent preoperative biliary drainage (PBD) but were without cholangitis. Background Some reports showed that PBD in patients undergoing pancreatectomy increased the rate of perioperative complications. However, no clinical trial has evaluated the optimal duration of antimicrobial prophylaxis with a focus on patients who underwent PD following PBD. Methods A total of 82 patients who underwent PD between March 2012 and December 2016 were randomly assigned to either a 1‐day group (n = 40), in which cefozopran (CZOP) as antimicrobial prophylaxis was given only on the day of surgery, or a 5‐day group (n = 42), in which CZOP was given for 5 consecutive days beginning on the day of surgery. We evaluated the incidence of infectious and other complications after PD. Results Outcomes were significantly better in the 1‐day group compared with the 5‐day group (P < 0.05) in terms of the incidence of overall infectious complications (15% vs 36%, respectively), intra‐abdominal abscess (3% vs 21%, respectively), clinically relevant postoperative pancreatic fistula (8% vs 24%, respectively), and Clavien‐Dindo grade III‐V complications (10% vs 31%, respectively). Duration of postoperative hospital stay was significantly shorter in the 1‐day group (10 days vs 15 days, P = 0.018). Anaerobic bacteria and methicillin‐resistant cocci were isolated from the drainage fluid only among patients in the 5‐day group. Conclusion Single‐day prophylactic use of CZOP is appropriate for patients who undergo PD following PBD without preoperative cholangitis.


| INTRODUCTION
Pancreaticoduodenectomy (PD) is a complex surgical procedure carried out for the treatment of periampullary diseases. Although the perioperative mortality rate of PD has remained relatively low, at approximately 1%-2% in high-volume centers, high morbidity of roughly 50% remains a problem. [1][2][3] Obstructive jaundice is the most common symptom in patients with periampullary diseases. Preoperative biliary drainage (PBD) is widely carried out for relieving biliary obstruction, but it is known to cause bile contamination in approximately 80% of patients. 4,5 However, PBD is still considered indispensable before major hepatobiliary and pancreatic surgeries, as several previous retrospective studies have reported that it reduced morbidity and mortality after PD through improving liver function. 6,7 In contrast, a randomized controlled trial (RCT) concluded that routine PBD in patients undergoing pancreatectomy for pancreatic head tumors significantly increased the rate of perioperative complications. 8 First-or second-generation cephem has been used commonly as prophylactic antibiotic in our institutes as well as in Western countries. 9,10 Furthermore, several guidelines recommend discontinuation of antimicrobial prophylaxis within 24 hours of surgery, even for major hepato-pancreato-biliary surgery. 11,12 In contrast, Sudo et al 4  and routine use of broad-spectrum antibiotics as therapeutic antimicrobial therapy has been implicated. 15,16 Thus, hard evidence as to the duration and selection of antibiotics for use as antimicrobial prophylaxis has been lacking for patients with an intermediate risk for infection, such as those who undergo PD after receiving PBD but have no signs of cholangitis preoperatively. In the current study, cefozopran (CZOP) which covers the Enterococcus and Enterobacter species was chosen for controlling infection in patients with PBD.
We herein report results from a RCT comparing antimicrobial prophylaxis with 1-day versus 5-day administration of CZOP, which is a fourth-generation cephalosporin, in patients without cholangitis who underwent PD after PBD, with the incidence of postoperative infectious complications as a primary endpoint.

| Patients
Patients with periampullary disease who were 20 years or older and who underwent PBD followed by PD at Kansai Medical University Hospital and Nagoya University Hospital were eligible for this RCT.
We classified the patients who underwent PD into three risk categories according to biliary contamination or infection, as shown in  15,16 In this study, we always clamped the hepatic duct by clamp forceps to prevent spilling bile juice after resecting the common bile duct. Peritoneal lavage was routinely conducted using 5000 mL each of normal saline at the time when the pancreas head was removed and at the end of the operation, respectively. All patients provided written informed consent before study enrolment. This study was registered in UMIN-CTR (UMIN000007277) and was approved by the ethics committee of each institution (H110276: Kansai Medical University Hospital, 2013-0224: Nagoya University Hospital). An ethical problem assumed in this study was that the incidence of infectious complications in the non-effective group might be increased, and that prolonged antibiotic prophylaxis might be associated with increased acquired antimicrobial resistance.

| Randomization and antimicrobial prophylaxis dosage procedures
Patients were randomly assigned to either the 1-day group or the 5day group before surgery. A central randomization system at Kansai Medical University was applied, and patients were stratified by institution.
In all patients, 1 g CZOP was given i.v. for 30 minutes immediately after induction, and additional doses were given once every 3 hours during the operative procedure. In addition, 1 g postoperative CZOP was given on the day of surgery for patients in the 1-day YAMAMOTO ET AL. group, whereas further doses of the postoperative CZOP were given every 12 hours for another 4 consecutive days in the 5-day group (additional 2 g CZOP per day for 4 days). We diagnosed intra-abdominal abscess by computed tomography (CT) or ultrasonography (US). If the drainage tube was already removed from the patient when we diagnosed the intra-abdominal abscess, we inserted a percutaneous drainage tube. If the patient had a drainage tube in place, we exchanged it for a new tube because the tubes are vulnerable to obstruction by plugs consisting of fibrin and other materials. We investigated the drainage fluid for microbiological testing from the patients who were given a percutaneous drainage tube or who had been placed with a drainage tube for a long period. If the intra-abdominal abscess was located in the center of the body where puncture was considered hazardous, we gave remedial antibiotics before considering surgical drainage.

| Data collection
Clinical data were collected prospectively for all patients and included patient demographics, pathological examination, perioperative clinical information, and complications.

| Statistical analysis
This study was designed to investigate that the infectious complication rate of the 1-day group was not inferior to that of the 5-day group. Based on previous studies, 18,19 the percentage of patients who developed infectious complications was expected to be 20%, and the threshold value was set at 40%. We planned to enrol at least 40 patients in each group without an accurate sample size calculation as a preparatory study.
Patient characteristics and perioperative and postoperative factors between the two groups were compared with Fisher's exact test and the Mann-Whitney U test as appropriate. Statistical significance was defined as P < 0.05. Statistical analyses were carried out with the JMP statistical program version 13 (SAS Inc., Cary, NC, USA).

| RESULTS
Between March 2012 and December 2016, 89 patients from the two institutions were pre-registered as shown in the CONSORT diagram ( Figure 2). Of these, a total of seven patients were excluded because of unresectable tumor (n = 4) and conversion to total pancreatectomy (n = 3). A total of 82 patients who underwent PD were randomized into the 1-day group (n = 40) and the 5-day group (n = 42).

| Clinical characteristics
Patient characteristics are summarized in Table 1 in the 1-day group and in 10 patients (24%) in the 5-day group. The majority of patients in both groups (85%) had a plastic stent inserted, whereas a metal stent had been the choice for the remaining 15%, with no significant difference in preference between groups. Median time from PBD to PD was 29 days in the 1-day group and 38 days in the 5-day group, which was not statistically significant.

| Surgical characteristics
Details on surgical parameters are also shown in Table 1. There were no significant differences in operation time and blood loss between the groups. Diameter of the main pancreatic duct and thickness of the pancreatic parenchyma were also similar. Additionally, the two groups were well matched for reconstruction method of pancreaticojejunostomy (PJ, modified Kakita method 21 or modified Blumgart method 22 ) and stent placement at PJ (external stent or internal stent or no stent). Table 2 shows the postoperative complications in both groups. Incidence of overall infectious complications, including CR-POPF, intraabdominal abscess, cholangitis and wound infection, was significantly lower in the 1-day group compared with the 5-day group (15% vs 36%, respectively; P = 0.029). Of the four types of complications, there were significant differences in the incidence of intra-abdominal abscess (3% vs 21%, respectively; P = 0.005) and CR-POPF (8% vs 24%, respectively; P = 0.038), but no difference in the incidence of postoperative cholangitis or wound infection.

| Postoperative complications
Patients who had CR-POPF had required percutaneous drainage, exchange of drainage tube, or remedial antibiotic. In the 1-day group, three patients had CR-POPF. Of these, percutaneous drainage was conducted in two patients and exchange of drainage tube in one patient, accompanied by giving of remedial antibiotics. In the 5-day group, 10 patients had CR-POPF. Of these, percutaneous drainage was conducted in five patients, exchange of drainage tube in four patients and persistent drainage with remedial antibiotics was needed in one patient. The incidence of POPF of all grades and overall morbidity were significantly lower in the 1-day group compared with the 5-day group (P < 0.05). In addition, the severe complication rate (Clavien-Dindo III to V) was significantly lower in the 1-day group compared with the 5-day group (10% vs 31%, respectively; P = 0.017). Allergic reaction was not found in any patient who participated in this study.

| Postoperative outcomes
As shown in Table 2, method of drain management did not differ significantly between groups, and drains were typically removed by postoperative day 3. Percutaneous drainage after removal was required in two patients (5%) in the 1-day group and in five patients (12%) in the 5-day group, which did not reach statistical significance.

| Microorganisms detected by abdominal drainage fluid culture
Potentially pathogenic microorganisms were identified from the intra-abdominal drainage fluids of three patients with intra-abdominal abscess or clinically relevant POPF in the 1-day group and in 13 patients in the 5-day group ( Table 3). The most commonly isolated microorganisms were Enterococcus species and Enterobacter species. Anaerobic bacteria and methicillin-resistant (MR) cocci were isolated exclusively from the drainage fluid of the 5-day group.

| Risk factors for infectious complications
Risk factors for infectious complications are shown in Table 4.
Univariate analyses by original parameters showed that non-pancreatic disease, soft pancreatic parenchyma and the 5-day group were significantly associated with infectious complications. BMI was found not to be a risk factor for infectious complications in this analysis. Multivariate analysis was performed using factors with significant differences in univariate analyses. Duration of CZOP dosage (5 days) was only independent risk factor for infectious complications.  prophylaxis is sufficient for patients undergoing hepatectomy with extrahepatic bile duct resection, which is one of the most invasive types of surgery, compared with 4-day administration. 24 This study was intended as a preparatory study designed to show that the infectious complication rate of the 1-day group was not inferior to that of the 5-day group. Unexpectedly, giving postoperative antimicrobial prophylaxis for 5 days after PD was shown to significantly T A B L E 1 Patient characteristics and surgical factors  | 447 species and Klebsiella species, 4,26 consistent with our findings (Table 3). Additionally, two previous reports that evaluated bacteria from ascitic fluid after PD reported that most of the isolated bacteria were Enterococcus and Enterobacter species. 27,28 Thus, broad-spectrum antimicrobials targeting these microorganisms are recommended for patients who will undergo PD following PBD. In the current RCT, we used CZOP as antimicrobial prophylaxis covering
In the 1-day group, which had better clinical outcomes, the incidence of overall infectious complications was 15% and that of intra-abdominal infections was only 8%, which were lower than those reported in previous studies. 29,30 An important finding from this study was that anaerobic bacteria and MR cocci including Staphylococcus aureus and Staphylococcus epidermis were isolated exclusively from the 5-day group, and not from the 1-day group. These results imply that CZOP might be an appropriate choice as antimicrobial prophylaxis in limited patients (middle-or high-risk groups who undergo PD), provided that prolonged usage in the absence of clear signs of infection is avoided.
Harmful effects of prolonged prophylactic use of antimicrobial agents have been documented after other types of major surgery.
Anaerobic bacteria were reportedly isolated from patients who developed postoperative infections following radical cystectomy and urinary diversion using the small intestine, when cephalosporin was continued for 1 week as antimicrobial prophylaxis. 31 In cardiac surgery, prolonged prophylaxis with antimicrobials has been associated with an increased risk of acquired antimicrobial resistance. 32 Moreover, it is well known that long-term postoperative antimicrobial prophylaxis has led to outbreaks of postoperative methicillin-resistant Staphylococcus aureus (MRSA) infection. 33 In addition, the incidence of patients with signs of infection in whom causative pathogens cannot be detected by conventional methodology has been increasing.
Recently, an analysis using the 16s ribosomal RNA gene showed higher sensitivity in detecting anaerobic bacteria, 34 but, at the same time, suggested the possibility that the consequences of inadequate prophylaxis antimicrobials are more serious than they seem.
Development of CR-POPF is the most important risk factor for intra-abdominal infections. 35  were not increased by limited use of antimicrobial as shown in Table 2. These findings indicate that the prolonged use of antimicrobials could mar all of the painstaking efforts of the surgical team to reduce the morbidity and mortality associated with pancreatic surgery.
In the current study, the two participating institutes carried out mutual site visits and had taken several steps to standardize the surgical technique and perioperative management, with particular focus on the evaluation of CR-POPF and placement, removal and exchange of the drainage tubes. Our result that the short administration of antibiotics results in superior outcomes in PD was unexpected, but is actually true in other less complex surgical procedures. Perhaps an optimal duration of antibiotics does not solely depend on the complexity of surgery. However, a larger study with a greater number of participating institutes will be needed to confirm our findings.
We are aware of some limitations of this study. First, this was a randomized but unblinded study with a small sample size. The current study was intended as a preparatory study designed to explore whether the issue of the duration of prophylaxis antimicrobials needs further and more extensive exploration in larger studies. Second, there was a significant difference in the baseline BMI of patients between the two groups. Obesity was admittedly one of the risk factors for clinically relevant POPF and postoperative infectious complications. Stratification by BMI was not carried out in this study because, again, the number of enrolled patients was too small for such considerations. In addition, although differences in the incidence of other potential risk factors for POPF, such as pancreatic texture, surgical procedure, fistula risk score and drain management were not statistically significant, one could not completely deny the possibility that some differences in background characteristics could have biased our results.
In conclusion, 1-day administration of broad-spectrum antimicrobial prophylaxis is appropriate as antimicrobial prophylaxis in patients who undergo PD after PBD and are without preoperative cholangitis, and long-term administration of antimicrobial prophylaxis in the absence of apparent signs of infection should be avoided.