Introduction

Patients undergoing pancreaticoduodenectomy frequently require biliary stenting due to obstruction of the common bile duct from the malignant mass. Myriad studies have investigated the impact of preoperative biliary stenting or drainage on postoperative complications. In patients undergoing pancreaticoduodenectomy, there is evidence that preoperative biliary stenting or drainage is associated with higher rates of postoperative complications, including surgical site infections (SSIs).1,2,3,4,5,6,7 The mechanism for many of these associations is unclear, but a clear potential causal relationship exists for SSIs: spillage of bile colonized with enteric organisms as a result of the stenting procedure. As such, intraoperative bile culture (IOBC) has been employed, in the hopes of guiding empiric therapy for postoperative SSIs based on the bile culture results. The data supporting this practice are mixed, with reports of IOBC’s sensitivity for predictive postoperative SSI organisms ranging from 30 to 81%.1,2,4,8

Further confounding this topic is the question of stent type, as some believe that metal stents may confer a higher infectious risk than plastic stents. However, the data supporting this belief are anecdotal. Prior studies have shown that stenting does alter the biliary microbiome and increases the proportion of patients colonized with multidrug resistant (MDR) organisms which may not be covered by standard perioperative antibiotic prophylaxis.5,9,10 However, no studies have looked at the influence of stent type with the biliary microbiome or antibiotic resistance patterns. Similarly, few studies have investigated postoperative infectious outcomes by stent type with no difference demonstrated; however, they have been small in scale and predominantly focused on stent-related outcomes (e.g., patency, pancreatitis) rather than the surgical population.11,12,13 Additionally, prior studies have investigated the biliary microbiome in patients with hepato-pancreato-biliary malignancies, and some have characterized institution-level microbiome differences, but none have looked at the potential contribution of stent type to the microbiome in this population.14,15

Regardless of data, dogma surrounding stent type and infection risk has persisted. Given increased SSI risk accompanying patients with stents, postoperative antibiotic prophylaxis has been suggested by many authors and is routinely employed at some institutions. Evidence for this practice is poor, but studies have demonstrated a reduction in postoperative SSIs with this practice.8,16,17,18 Some institutions even tailor postoperative prophylactic antibiotics based on stent type and the belief that metal stents may impart higher infectious risk than plastic stents, for which no published data exists to our knowledge. Notably, the practice of extended postoperative prophylaxis is in stark opposition to quality measures as set forth by the Surgical Care Improvement Project which specify that prophylactic antibiotics ought to be discontinued within 24 h of surgery end time due to general lack of evidence for efficacy and concerns of nurturing antibiotic resistant organisms.19

Given the uncertainty in available evidence, we sought to evaluate the utility of IOBC and the interplay between stent type, the biliary microbiome, and SSI risk following pancreaticoduodenectomy in our high-volume National Cancer Institute (NCI) Comprehensive Cancer Center. We hypothesized that stent presence influences the biliary microbiome and increases infection risk, but that stent type (plastic versus metal) does not correlate with infection risk or changes in the microbiome. Additionally, we hypothesized that due to the numerous potential sources of organ space and incisional SSI from pancreaticoduodenectomy, IOBC results would poorly predict both incision and organ space SSI organisms.

Methods

Patient Population and Data Source

Patients undergoing pancreaticoduodenectomy or total pancreatectomy from January 2008 to April 2020 were identified through our institutional National Surgical Quality Improvement Project (NSQIP) and cancer databases. Standard NSQIP datapoints were captured including procedure-targeted pancreatectomy variables. Chart review was utilized to capture patient stent type (plastic vs metal) as well as IOBC and all postoperative culture data, including culture positivity, speciation, and antibiotic sensitivities. Intraoperative antibiotic prophylaxis was not standardized throughout the study period, and patients received either a second- or third-generation cephalosporin or an extended spectrum agent such as piperacillin/tazobactam per preference of the attending surgeon. Starting on July 2017, patients with plastic and metal stents received 3 or 5 days, respectively, of intravenous ciprofloxacin 500 mg every 12 h and metronidazole 750 mg every 8 h postoperatively starting on postoperative day zero. Patients received extended postoperative antibiotic prophylaxis constituted less than one-third of all study patients. No postoperative antibiotic prophylaxis was given prior to July 2017. Investigating the effectiveness of antibiotic prophylaxis regimens is outside the intended scope of the present manuscript.

Bile Culture, Speciation, and Sensitivity Evaluation

Cultures were taken by swabbing the transected bile duct, following stent removal when present. IOBC was not protocolized and was performed at the discretion of the attending surgeon; ultimately 92% (n = 343) of all patients with a stent (n = 373) and 73% (n = 140) of those without a stent (n = 191) received IOBC.

All samples were subject to direct Gram stain, aerobic culture (Chocolate Agar, MacConkey Agar, and Blood Agar with 5% Sheep Blood) for 3 days and anaerobic culture (Brucella with Hemin/Vitamin K, Bacteroides Bile Esculin Agar) for 5 days per our institutional standard. Speciation of cultured organisms was performed utilizing the VITEK MS device20 (bioMérieux, Marcy-l'Étoile, France), VITEK 2 system21 (bioMérieux, Marcy-l'Étoile, France), and API 20E bacterial identification system22 (bioMérieux, Marcy-l'Étoile, France) according to standard practice. For antibiotic susceptibility testing, organisms were primarily evaluated using VITEK AST cards (GP75 and GN81, bioMérieux, Marcy-l'Étoile, France). Organisms not in the VITEK database were evaluated by Microscan Neg MIC panel type 46 (Beckman Coulter, Brea, CA). Bacillus species, Corynebacterium, and anaerobes were not routinely tested for sensitivities except when requested by the clinical care team. Clinical and Laboratory Standards Institute (CLSI) guidelines were followed for all susceptibility testing, and manufacturer’s instructions were followed for all commercial culture, speciation, and susceptibility testing. For analysis, resistance patterns were grouped into single-agent resistance (either beta-lactam regardless of resistance spectrum or non-beta-lactam resistance). MDR organisms were defined as those with resistance to two separate classes of antibiotics (e.g., beta lactams and fluoroquinolones).

Statistical Evaluation

Descriptive statistics of clinicopathologic variables were tabulated. Numerical data were reported as medians with interquartile ranges (IQR). Chi-squared and analysis of variance (ANOVA) testing were utilized as appropriate with Tukey’s honestly significant difference post hoc testing for ANOVA P values < 0.05. NSQIP-defined SSIs were analyzed with binomial logistic regression generating odds ratios (OR) with 95% confidence intervals (CI). Organ space SSIs, incisional SSIs (combining both deep and superficial NSQIP SSI categories), and all SSIs were each analyzed as separate outcome variables. All clinicopathologic variables with P < 0.2 on univariable analysis were included in initial multivariable models. Final multivariable models were arrived at by single backwards elimination of nonsignificant variables until further elimination would reduce model fit with P < 0.05. Due to significant co-linearity between patients with biliary stents and positive IOBC, the stent variable was utilized for multivariable modeling in cases where both variables would otherwise have been included. All statistical operations were performed using SPSS version 26 (IBM Corp, Armonk, NY).

Results

Clinicopathologic Characteristics

Of the 655 patients undergoing pancreaticoduodenectomy during the study period, 483 (74%) had IOBC and were included in the study; the median patient age was 67 years. One-hundred-eighty-nine (39%) patients had plastic stents, 154 (32%) had metal stents, and 140 (29%) were not stented; four patients had a stent of unknown composition and were excluded from the analysis. Stent groups differed in several clinicopathologic characteristics (Table 1). Among patients with available data on wound protector use (N = 220), use did not differ between patients with (n = 117, 80%) and without biliary stents (n = 60, 82%, P = 0.78).

Table 1 Clinicopathologic and treatment characteristics of patients undergoing pancreaticoduodenectomy by stent type
Full size table

Three-hundred-twenty-nine (96%) patients with stents had bactibilia, compared to 18 (13%) without stents (P < 0.001) (Fig. 1). Eighty-five percent of cultures were polymicrobial in patients with stents, compared 5% in patients without stents (P < 0.001). The proportion of positive cultures and biliary microbiome in patients with metal versus plastic stents was nearly identical, however, significantly differed from patients without a biliary stent. The most common organisms identified included Enterococcus, Streptococcus, and Klebsiella species. In patients with positive bile cultures, only the proportion with Klebsiella and Enterococcus differed significantly between patients with and without stents (P = 0.01 and 0.002, respectively). Identified genus or species did not significantly differ between patients with plastic stents compared to metal stents (all P > 0.4).

Fig. 1
figure 1

Heat map of intraoperative bile culture results in patients undergoing pancreaticoduodenectomy by stent type

Full size image

Analysis of Surgical Site Infections

Overall, 159 (33%) experienced an SSI, most commonly incisional (n = 92, 22%). On univariable analysis, the presence of any stent was associated with higher odds of incisional SSI (OR 3.39, 95% CI 1.70–6.79, P < 0.001). Similarly, bactibilia (OR 3.69, P < 0.001), plastic stents (OR 4.06, P < 0.001), and metal stents (OR 2.65, P = 0.01) were each significantly associated with incisional SSI (Table 2). Additionally, incisional SSI risk significantly decreased over time. Due to high co-linearity with the stent variable, positive bile culture was not included in multivariable modeling. Diabetes, hypoalbuminemia, and ASA class 3–4 were not associated with incisional SSI but met the threshold for inclusion in the initial multivariable model. On single backwards elimination, the only variables independently associated with higher incisional SSI risk were later years of operation and either plastic or metal stents. Notably, when positive bile culture was split into mono- and polymicrobial cultures, both monomicrobial (OR 3.1, 95% CI 1.15–8.33, P = 0.03) and polymicrobial (OR 3.8, 95% CI 1.83–7.89, P < 0.001) cultures were associated with increased odds of incisional SSI. Gland texture, pancreatic duct size, BMI > 30 kg/m2, age, sex, histology, operative duration, and vascular reconstruction were not associated with odds of superficial SSI (all P > 0.4).

Table 2 Univariable regression of factors associated with incisional surgical site infection following pancreaticoduodenectomy
Full size table

Plastic stents and metal stents were not significantly associated with organ space SSIs (Table 3) nor was positive bile culture (OR 0.78, 95% CI 0.47–1.29, P = 0.33). On multivariable analysis, PDAC or chronic pancreatitis histology was independently associated with lower odds of organ space SSIs, while male gender and undergoing a vascular reconstruction were independently associated with higher odds. Pancreatic duct size, gland texture, BMI >30 kg/m2, diabetes, and operative duration met significance cutoffs for inclusion in the initial multivariable model but were removed on single backwards elimination. Hypoalbuminemia, age, and ASA status were not associated with organ space SSIs on univariable analysis.

Table 3 Uni- and multivariable regression of factors associated with organ space surgical site infection following pancreaticoduodenectomy
Full size table

On multivariable analysis of factors associated with any SSI, plastic stents, metal stents, and male gender were independently associated with higher odds, while later years of operation and PDAC/chronic pancreatitis histology were independently associated with lower odds of any SSI (Table 4). Pancreatic duct size met criteria for inclusion in the initial multivariable model but was eliminated in the final model. Gland texture, diabetes, ASA class, and vascular resection were not associated with odds of any SSI on univariable analysis and were not included in the initial multivariable model.

Table 4 Uni- and multivariable regression of factors associated with any surgical site infection following pancreaticoduodenectomy
Full size table

In a separate analysis, we investigated the association of biliary candidiasis at operation with infectious outcomes given recent works suggesting an association with SSIs.23,24 Compared to patients with negative bile culture, isolation of any candida species was not associated with incisional SSI (OR 1.86, 95% CI 0.59–5.87, P = 0.29), while patients with other organisms on IOBC were at higher risk (OR 3.99, 95% CI 1.93–8.27, P < 0.001). Similarly, candidiasis was not associated with organ space SSIs (OR 0.53, 95% CI 0.19–1.48, P = 0.23).

Prediction of Causative Organisms in Postoperative Infections with Intraoperative Bile Culture

Seventy-three patients with a NSQIP-defined SSI had speciated SSI cultures, and 39 (53%) were polymicrobial. At least one organism identified by IOBC was present in just over one-half of cases (n = 42, 58%), while at least one organism not identified by IOBC was present in two-thirds of cases (n = 49, 67%). When analyzed by SSI type, 20 patients had speciated incisional SSI cultures. IOBC performed slightly better in this setting, correctly identifying at least one SSI organism in 70% (n = 14) of cultures while missing at least one causative organism in 45% (n = 9). In the 53 speciated cultures from organ space SSIs, IOBC identified at least one organism in 53% (n = 28) while missing at least one organism in 75% (n = 40). Notably, all cultured bacteria were sensitive to standard empiric broad-spectrum regimens for SSIs such as vancomycin and piperacillin/tazobactam. No patients developed positive blood or sputum cultures with an organism identified from IOBC.

Discussion

The findings of our study challenge longstanding dogma regarding the association of stent type with infection risk as well as the utility of IOBC in patients with biliary stents undergoing pancreaticoduodenectomy. It has long been known that stent presence has a near-perfect correlation with bactibilia and a strong association with incisional SSIs; however, stent type has not previously been rigorously investigated as an independent risk factor. Despite this, some institutional protocols treat metal stents as being higher risk than those with plastic stents. The present results clearly show that metal stents do not pose significantly higher risk for either incisional or organ space SSIs than plastic stents nor are there differences in the biliary microbiome or antibiotic resistance patterns. Finally, we found that IOBC is poorly predictive of organisms cultured from SSIs, even for incisional SSIs, and is not suitable for guiding empiric antibiotic therapy. We therefore do not recommend IOBC in patients with or without biliary stents. We additionally recommend empiric broad-spectrum antibiotic therapy informed by institutional antibiograms, with subsequent narrowing guided by cultures for all surgical site infections following pancreaticoduodenectomy. Empiric therapy based on reassuring IOBC results has the potential to cause harm from under-treatment and is not recommended as a substitute for empiric broad-spectrum therapy in light of the present findings. IOBC may be considered in practice environments with prevalent MDR organisms on institutional antibiograms or in patients with a history of colonization by MDR organisms that would not respond to standard broad-spectrum antibiotic regimens; however, the utility of IOBC in this setting has not been well-studied.

Our results are largely consistent with those of prior studies evaluating the utility of IOBC. Sudo et al. reported that only 30% of postoperative abdominal infections had identical organisms to those from IOBC, however, did not comment on the clinical utility of IOBC in light of this low number.8 Similarly, Sandini et al. recently showed that the presence of MDR organisms on IOBC only occurred in one-third of postoperative infections and suggested that the use of IOBC may merit re-evaluation.7 In the study by Howard et al., only 42% of cultured infectious complications had identical organisms on IOBC.1 In contrast, the large multi-institutional study by Fong et al. concluded that IOBC “should be routinely obtained during PD to predict microorganisms isolated in wound cultures and allow for tailored antibiotic therapy in the setting of pending or unavailable wound culture data”.14 However, the authors did not report the percent of postoperative SSIs containing an organism identified on IOBC nor what proportion of cultures contained an organism not present on IOBC. Without these crucial data, it is difficult to evaluate the clinical utility of IOBC in their study. Finally, biliary candidiasis has recently been associated with SSIs by Kato et al..23,24 We were unable to replicate these results, possibly due to differences in culture methodology or SSI definitions.23,24 Given the variation in utility of IOBC in the current literature, additional study of the rationale for IOBC is needed. Overall, the available data suggest that IOBC results do not sufficiently correlate with causative organisms of postoperative SSIs or other infections to warrant routine clinical use.

Our study has several limitations, chiefly its retrospective nature, though it is founded in a prospectively maintained institutional NSQIP database. Additionally, NSQIP-defined variables may be insufficient for accounting for the full spectrum of variables associated with SSIs. For example, later year of operation was independently associated with lower odds of incision and any SSI but not organ space SSI. We suspect that progressively higher utilization of wound protectors likely accounts for this association, given that it was most pronounced for incisional SSI risk; temporal differences in perioperative practices such as intraoperative antibiotic prophylaxis, glove, and instrument changes prior to closure may also have contributed. These data points were not captured for the majority of the study period, and analyzing their efficacy in reducing SSIs is outside the scope of the present manuscript.

Additional unknown patient-level factors may influence SSI risk, as our multivariable analyses showed that male gender was a risk factor for organ space SSIs despite a lack of a biologic explanation for this; this may indicate that male gender acted as a surrogate variable in our analyses. Notably, prior studies have reported that males have higher risk of SSIs in abdominal operations, and no definitive explanation is known.25 Also, results may differ when using alternative SSI definitions or culture methodologies. Regarding the biliary microbiome, our findings are only reflective of the spectrum of potentially pathogenic organisms investigated with standard-of-care culture techniques, and the biliary microbiome may differ with respect to non-pathogenic or difficult to culture organisms. Finally, while a minority of patients with stents in our study received postoperative antibiotic prophylaxis, some stented patients undergoing pancreaticoduodenectomy did not have IOBC and were excluded from the study cohort. Our population was therefore not suited to investigate the efficacy of perioperative antibiotic prophylaxis regimens, wound protector use, or other perioperative practices in reducing the risk of SSI. These practices are deserving of additional study.

Conclusions

In conclusion, IOBC does not have utility in predicting surgical site infection risk or guiding empiric therapy in patients undergoing pancreaticoduodenectomy and is not recommended. Intraoperative bile culture may be considered selectively in institutions or patients with a high risk of MDR organisms; however, the evidence supporting even this application remains poor. We recommend the use of adjunctive methods to decrease infection risk, including the use of double ring wound protectors.26 Additional study is needed to identify an independent beneficial effect of postoperative antibiotic prophylaxis on SSIs in pancreaticoduodenectomy patients, to better ensure the surgical community is practicing evidence-based perioperative care.