Comparing cefazolin/
metronidazole, 
piperacillin‐tazobactam, or c
efoxitin as surgical antibiotic prophylaxis in patients undergoing pancreaticoduodenectomy: A retrospective cohort study

Preoperative antibiotic options for pancreaticoduodenectomy (PD) include cefoxitin (CX), piperacillin‐tazobactam (PT), or combined cefazolin and metronidazole (CM). Recent studies suggest the superiority of PT over CX, but evidence for CM is unclear.

Mortality rates following pancreaticoduodenectomy (PD) have fallen over the years to their current benchmark of less than 5%. 1,2Still, postoperative complications remain high, with estimates of 40%−58%. 1,2Common postoperative complications include delayed gastric emptying (15%−20%), and pancreatic fistula (10%−20%), with the most common complication being surgical site infections (SSI; 11%−48%). 3SSIs are also the most costly complications as they increase hospital stays, emergency department visits, and readmission following initial discharge. 4In extreme cases, SSIs can result in death, with a 2-to 11-fold increase in mortality associated with the development of SSI. 5 Up to 77% of postoperative mortalities in patients with SSI are attributed to the infection itself. 6,7Is following most abdominal surgeries have decreased in recent years, except following PD where postoperative SSIs rates have remained unchanged.2 This could be attributed to preoperative biliary stenting (PBS), malnutrition, neoadjuvant therapy administration, and the prolonged duration of the PD procedure itself.8,9 Therefore, there is a need and opportunity to evaluate and optimize perioperative and postoperative procedures that can reduce this burdensome complication.
Perioperative antibiotics can effectively reduce rates of postoperative SSI. 10 The most common organisms involved in SSI following PD are site-dependent and include Streptococcus pneumoniae, Enterococcus faecalis, Staphylococcus aureus, and Escherichia coli. 11In accordance with the Surgical Care Improvement Project (SCIP), the current standard of care in most centers is cefoxitin (CX)   or cefazolin with metronidazole (CM). 10A recent randomized trial in which our center was a participating site suggested the superiority of the broad-spectrum antibiotic piperacillin-tazobactam (PT) over CX in reducing SSI following PD. 12,13However, evidence for the use of PT versus CM is unclear.

| Outcomes
Outcome data was derived from the ACS-NSQIP database, which is nationally validated, risk-adjusted, with postoperative outcomes adjudicated by trained data abstractors that record outcomes using standardized definitions up to 30 days from the index operation. 14,15e primary outcome was a composite of SSI, which included superficial incisional infections, deep incisional infections, and organ space infections. 5The secondary outcome was a composite of other infectious complications, which included urinary tract infections (UTI), pneumonia, septic shock, sepsis without shock, and Clostridioides difficile infections (CDI).A tertiary outcome was the development of clinically significant POPF. 16

| Statistical analysis plan
Descriptive statistics were used to present patient characteristics.
Categorical variables were reported as frequencies and percentages, while continuous variables were reported as medians (interquartile ranges).It was determined a priori that p < 0.05 would be considered significant.Results were reported as odds ratio (OR) with 95% confidence intervals (CI).Baseline demographic characteristics were compared between the CM, PT, or CX group using independent sample t-tests and χ 2 tests (or Fischer's exact when applicable) for continuous and categorical variables, respectively.
To compare antibiotic selection in the development of the composite SSI, composite infection, and POPF outcomes, a univariable logistic regression analysis was used.Consistent with existing literature, the presumed confounding factor was PBS, as this would impact preoperative antibiotic selection and SSI development. 12,17erefore, a multivariable logistic regression analysis was then used to account for PBS as a potential confounding factor.Multivariable logistic regression analysis was not performed for the CX versus PT comparison since 100% of the patients had been enrolled in the randomized controlled trial. 12llinearity between antibiotic regimen and PBS was assessed by computing the variance inflation factor (VIF). 18 Since each patient could have experienced more than one individual outcome under the composite outcome, robust standard errors were calculated.This was to account for the model assumption that all observations must be T A B L E 1 Baseline characteristics of included patients.| 1415 independent of one another.Under the assumption that the 8% missing data was missing completely at random, complete-case analysis was performed. 19| RESULTS

| Patient demographics
Table 1 displays the demographic data for the 44 and 46 patients included in the CM and PT groups, respectively.There were no differences in baseline characteristics between patients who received CM or PT.There were also no baseline differences across the CM, PT, and CX groups.All patients who had received PT and CX had been enrolled in the trial by D'Angelica et al. 12 None of the patients who had received CM were enrolled in the trial.
The VIF for both variables was 1 across all models, meaning minimal collinearity between the predictor variables.
There were also no significant differences in composite infection (OR: 0.78, 95% CI: 0.34−1.78,p = 0.55) nor clinically significant POPF (OR: 1.58, 95% CI: 0.59−4.25,p = 0.36) outcomes between CM and PT (Table 3).Nonsignificance was maintained following the adjustment for PBS.This was expected considering the increasing antimicrobial resistance of anaerobes to second-generation cephalosporins, like CX, that had initially been deemed to be effective. 20However, in contrast to the D'Angelica trial, the association was not maintained for incisional and organ space SSIs, and there were no statistically significant differences in the development of other infectious outcomes across the antibiotic groups at our center.Considering the low number of events and large CIs, this was likely due to the lack of statistical power to detect any differences, should they exist, at our center. 12ere were also too few occurrences of deep incisional SSIs to draw meaningful conclusions, although it should be noted that in the trial, there were no differences in deep incisional SSIs. 12Again, considering the low number of events, we are limited in our statistical power to draw meaningful conclusions from these results.

| DISCUSSION
Although current evidence suggests the superiority of PT over CX, there is insufficient evidence regarding its superiority to CM.This is an important knowledge gap as CX is only one of the SCIP recommended antibiotic for surgical prophylaxis. 10At JHCC, the preoperative antibiotic regimen of choice is often cefazolin with the addition of metronidazole to cover intestinal anaerobes for patients undergoing PD, even after the D'Angelica trial results were published. 12Compared to CX, CM combined provides greater coverage for Gram-positive species and anaerobes. 21This translated into significant differences in the reduction of SSI with the use of CM over CX in other surgical contexts. 22,23This may explain the associations seen in this study where PT was found to be superior to CX but no different to CM in SSI development.
Still, CM is considered to be a more narrow spectrum antibiotic compared to PT, which has broader Gram negative coverage. 10,24In the substudy of the D'Angelica trial reported by Ellis et al., the presence of Enterococci species (n = 68, 29.4%) was associated with increased SSI in the CX group (OR: 2.33, 95% CI: 1.06−5.15)but not the PT group (OR: 0.55, 95% CI: 0.17−1.84),which may be explained by the coverage of enterococci by PT but not by CX. 25 However, prophylactic coverage for these species may not be as crucial for patients undergoing PD, depending on the institution.Current evidence demonstrates that the predominance of pathogens in SSI among PD patients may be a function of antibiotic selection for prophylaxis and therefore vary by institution. 11For example, sites that used CX had a greater predominance of Enterococci in SSI.
For instance, there were some baseline imbalances between the CM and PT group with respect to the history of diabetes (p = 0.05), hypertension (p = 0.06), and intraoperative drain insertion (p = 0.06).Although these differences were not statistically significant, this should be considered while interpreting the presented data.Altogether, this analysis should be interpreted as exploratory rather than confirmatory in the comparison of PT versus CM.Furthermore, this was a single-center study, thereby limiting the generalizability of the results.Considering the observational design of the study, it is also subject to selection bias.Specifically, we did not include patients that received cefazolin only and so those who received cefazolin and flagyl may have had patient-specific factors that led clinicians to broadening their prophylactic antibiotic regimen.
Currently, there is insufficient evidence to suggest the superiority of PT over CM for patients undergoing PD.PT is a broadspectrum antibiotic and its overuse for prophylaxis may confer greater harm than benefit, as a driving force towards antimicrobial resistance in the community. 27In the spirit of antibiotic stewardship, where the overarching goal is to choose the narrower spectrum antibiotic (i.e., CM) in the absence of evidence for the superiority of broader spectrum antibiotics (i.e., PT), CM remains the preferred choice for perioperative prophylaxis in PD at our center. 28The nodifference between PT and CM demonstrated in the present analysis should be further explored through a dedicated randomized controlled trial.These studies can also consider the collection of SSI cultures when they occur so as to understand the common species that lead to these infectious complications.

| CONCLUSION
This single-center retrospective cohort study confirms the superiority of PT over CX in the development of SSI and suggests there may be no difference compared to CM.There were no differences seen in other infectious complications, including POPF, or composite infections, including CDI, UTI, pneumonia, sepsis, or septic shock.This study supports the use of CM for antibiotic prophylaxis in patients undergoing PD in the appropriate setting.
Therefore, we aimed to compare three different prophylactic antibiotic regimens (CM, CX, PT) in the rates of postoperative SSIs, postoperative pancreatic fistula (POPF), and other infectious outcomes.Given the most common bacterial pathogens and local resistance patterns involved in SSI following PD, we hypothesized that the combination of CM (the current standard of care at many Canadian institutions, including ours) is an equally effective regimen to avoid SSI compared to PT. Considering the involvement of our center in the D'Angelica et al. trial, we expected superiority of PT over CX. 12 2 | MATERIALS AND METHODS 2.1 | Patient selection The sample was derived from a data set contribution from an academic institution, the Juravinski Hospital Cancer Centre (JHCC), to the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP), a North American Surgical Administrative Database.All patients undergoing open PD at JHCC from November 2017 to December 2021 and received either preoperative CM, PT, or CX were considered for inclusion in the analysis.The timeframe of interest corresponded to the duration of the D'Angelica trial; thus, all patients who had received PT or CX had been enrolled in the trial by D'Angelica et al. 12 Exclusion criteria included those who received cefazolin only, and those who had an allergy to penicillin.The study was approved by the Hamilton Integrated Research Ethics Board, and the need for informed consent was waived.

Table 2 .
In the CX group, two patients developed superficial SSI, two developed deep incisional SSI, and 12 patients developed an organ space infection.In total, 14 patients who received CX developed any type of SSI.In the PT group, one patient developed a superficial SSI, 0.99−9.0,p = 0.05).There were too few isolated deep incisional infection outcomes to draw meaningful statistical conclusions.There were also no differences in composite infection (OR: 0.92, 95% CI: 0.39−2.12,p = 0.83) nor clinically significant POPF (OR: 1.69, 95% CI: 0.56−5.06,p = 0.35) between CX and PT use.Effect estimates are summarized in

Table 3 .
Infectious outcomes according to antibiotic selection.
developed organ space SSI (Table2).In total, 11 of the patients who received CM developed any sort of SSI.In the PT group, two patients developed a superficial SSI, and six patients developed organ space SSI.Again, in total, seven patients in the PT group developed any sort of SSI.There were no significant differences in the development of composite SSI between the PT versus CM groups (OR: 1.33 95% CI:T A B L E 1 (Continued)Abbreviations: CDI, Clostridium difficile infection; POPF, postoperative pancreatic fistula; SSI, surgical site infection; UTI, urinary tract infection.
26though CM does not provide Enterococcal coverage, CX is not commonly used for surgical prophylaxis in Canadian institutions, including ours.Therefore, Enterococci cultured in SSI may be less common such that PT does not confer notable benefits compared to CM.Furthermore, based on a retrospective review of unpublished data at our center, only around 10% of cultured bile among those undergoing PD demonstrated E. faecalis.Although a positive culture in bile does not necessarily suggest the same species would be found in the wound, this, in conjunction with lower rates of CX use and the greater coverage for Gram-positive species and anaerobes by CM over CX, may explain why SSI rates were reduced in the PT group versus CX but were no different to the CM group in our center.26Anotable limitation is in the sample size of patients included in this study.The small sample size resulted in wide CIs but also translated into fewer outcome events making it difficult to adjust the analyses for potential confounding factors.Considering the well-established impact of PBS in the development of infectious outcomes, this was included in a multivariable analysis.However, other patient factors could not be adjusted for so as not to risk the stability of the model.increasing the risk of residual confounding.