Discussion

In this retrospective study, we assessed risk factors and microbial patterns of SSIs developing after pancreatoduodenectomies at a single institution over 10 years. From data extracted from the NSQIP database, the overall SSI rate was 20.3%, similar to previously reported rates of 20%–40%. ^{Reference Ceppa, Pitt and House16} Stabilized SSI rates in recent years may be related to increased availability and standardization of microbial testing. Moreover, the hospital system in this study transitioned to a new electronic medical record system in 2011, so variability in provider SSI reporting and familiarity with EMR documentation over time may have also contributed to perceived increase in SSI rate over time. ¹⁷

At the start of this study in 2012, our hospital pancreatoduodenectomy protocol called for broad-spectrum as the preoperative antibiotic of choice. In late 2016, a change to first-generation cephalosporins was implemented and later switched back to broad-spectrum in early 2018. During this interval, we observed a surge in the SSI rate from 29.0% in 2016 to 37.1% in 2017 and subsequently down to 23% in 2018. Given missing data on preoperative antibiotics, we could not correlate antimicrobial use with SSI rates. Minor discrepancies between percentages can occur because patients may steer away from protocol-assigned preoperative antibiotics due to patient-specific contraindications (ie, allergies or chronic kidney disease stage 4 or 5 for ertapenem). ^{Reference Kara, Sahin, Ayaz, Yildirim, Sahutoglu and Tufekci18} Despite changes in SSI rate, the incidence of multidrug-resistant organisms during this interval did not change. The antibiogram format allows visualization of changes in resistance patterns over time (2012–2016 vs 2017–2021). We were able to observe decreases in antimicrobial resistance patterns between year ranges. Interestingly, Klebsiella oxytoca, Pseudomonas, E. coli, and Citrobacter spp increased in sensitivities.

Unmodifiable risk factors, such as age, sex, and race, were not linked to post-Whipple SSIs in this study. The data set did not explore socioeconomic status, access to healthcare, or degree of health literacy. These social determinants of health may contribute to patient outcomes and offer important information because our healthcare system serves rural communities in the state of Alabama. ¹⁹ Regarding modifiable risk factors, smoking is strongly correlated with delayed wound healing in the literature, and obesity has been associated with infection, specifically of skin and soft tissue. ^{Reference Nolan, Martin, Thompson, Schroeder, Hanson and Warner20,Reference Huttunen and Syrjänen21} Interestingly, we did not identify associations between increased BMI or smoking within 1 year and the development of SSIs. Additional patient comorbidities have also been cited as SSI risk factors, such as diabetes mellitus, hypertension, immunosuppression, malnourishment, and bleeding disorders. ^{Reference Azoury, Farrow and Hu8} Only the presence of bleeding disorder demonstrated a higher SSI rate. This finding may be attributed to standardized preoperative management of the other comorbidities studied compared to bleeding disorders. Guidelines recommending target A1C, blood pressure range, or albumin levels have been well established. Previous reports describe similar increased SSI risks in thoracic and orthopedic cases among patients with bleeding disorders; however, no documentation in abdominal surgeries has been cited. ^{Reference Aeschbacher, Nguyen, Dorn, Kocher and Lutz22,Reference Lai, Song and Guo23} No differences in corresponding preoperative platelet, INR, or PTT values were found between SSI and no SSI cohorts. Nonetheless, low BUN, albumin, and hematocrit were associated with increased SSIs. Collectively, these values illustrate a state of malnutrition either from hypermetabolic states associated with malignancy or decreased oral intake. This discrepancy between laboratory and clinical malnutrition may be due to error in documentation or delayed symptom recognition. Furthermore, the documentation of a comorbidity does not indicate whether patient comorbidities were well managed versus poorly managed and symptomatic, which may be better understood through more objective measures such as preoperative laboratory values. For cancer patients who underwent chemotherapy or radiation therapy within 90 days prior to surgery, no increases in SSI rate were established, reassuring providers that patients may proceed with neoadjuvant therapy to improve cancer outcomes without compromising postoperative infection risk.

Overwhelmingly, the most common indication for undergoing pancreatoduodenectomy is malignancy. Our investigation demonstrates surgical indication of cancer as a negative prognostic factor. This finding has not been previously reported and may warrant further study. Patients with pancreatic head or biliary tract masses often present with preoperative biliary obstruction and may require biliary stenting (metal, plastic, or both). Patients with biliary obstruction did exhibit higher SSI rates; unexpectedly, patients with documented preoperative biliary stents had lower SSI rates in an unadjusted analysis. Although it has been recognized that endoscopic stenting harbors an increased risk of infection by exposing the biliary tree to duodenal bacteria, we hypothesize that confounding variables may be present and are worth investigating. ^{Reference Gavazzi, Ridolfi and Capretti24,Reference Lamarca, Rigby, McNamara, Hubner and Valle25}

Controversy remains over the choice of preoperative antibiotic therapy as the effectiveness of antibiotic classes is debated, and emerging evidence suggests that institution-specific and/or targeted therapy is more effective in lowering SSI risk than any single antibiotic choice. ^{Reference Barreto, Singh and Perwaiz26–Reference Pham, Chen, Nahm, Lam, Pang and Richardson29} Intrabdominal surgery most commonly uses early-generation cephalosporins and broad-spectrum antibiotics to cover for gram-positives and/or enterococci and coliform bacteria. Moreover, coverage varies by regional resistance patterns, though antibiotic choices may be limited by patient-specific characteristics (ie, drug reactions, renal insufficiency, or liver disease). There is also concern that preoperative antibiotics may select for bacterial growth. Although we did not detect differences between antibiotic class and SSI rate, bacterial culture and speciation demonstrated a notable discrepancy. Specifically, the use of first-, second-, and third-generation cephalosporins later developed SSIs composed primarily of Enterococcus spp and E. coli. In contrast, the use of broad-spectrum antibiotics led to Klebsiella spp–dominated cultures, followed by Enterococcus spp. The rate of polymicrobial infections did not differ according to antibiotic class, suggesting low concern for developing resistance to preoperative antibiotics over time.

We evaluated duration of surgery and wound classification as potential risk factors for developing SSI. As expected, surgery duration was predictive of increased SSI risk; therefore, systemic and team-based approaches to reducing operative length may benefit both patient outcomes and hospital efficiency. ^{Reference Cheng, Chen, Soleas, Ferko, Cameron and Hinoul30} The degree of intraoperative wound contamination has also been suggestive of SSIs in prior studies. ^{Reference Onyekwelu, Yakkanti, Protzer, Pinkston, Tucker and Seligson31} Wounds were categorized as clean, clean-contaminated, contaminated, or dirty-infected. Pancreatic surgeries often involve dividing the bile duct and proximal small bowel, therefore considered clean-contaminated. ^{Reference Speicher, Nussbaum and Scarborough32} The vast majority (89%) of our pancreatoduodenectomies reported clean-contaminated wounds with a 20.1% SSI rate, comparable to our overall 20.3% rate. Unsurprisingly, 5 of 5 dirty or infected wounds developed an SSI.

Almost half of patients undergoing Whipple procedures develop postoperative complications, with delayed gastric emptying and pancreatic leak being the most common. ^{Reference Saraee, Vahedian-Ardakani and Saraee1,Reference Changazi, Ahmed and Bhatti3,Reference Clancy and Ashley4,Reference Karim, Abdulla, Abdulkarim and Rahim6} Pancreatic leaks themselves have been linked to both delayed gastric emptying and SSIs and therefore represents an important prognostic factor to study. Biochemical leaks may be approximated by drain amylase levels monitored in the postoperative course and indicate insufficient pancreatojejunostomy anastomosis. As such, leaked bile, containing autodigestive enzymes, that may seep into surrounding tissues and promote necrosis and bacterial colonization. Our study identified an increase in SSI incidence in patients reporting pancreatic leak, consistent with literature documenting pancreatic leak as an independent risk factor. ^{Reference Suragul, Rungsakulkij and Vassanasiri33} Those with increased extra-abdominal postoperative complications related to anesthesia/induction (pneumonia, increased ventilator requirement) or kidney function (renal insufficiency, dialysis requirements) had increased SSI rates. Confounding variables contributing to renal insufficiency include microbial infection or adverse effects of antimicrobial treatment itself. No preoperative differences in creatinine were noted; postoperative creatinine values were not documented. The SSI patient cohort was admitted for 1 week longer on average, further increasing patient morbidity risk.

As predicted, SSI cultures primarily grew gram-negative isolates (E. coli) and gram-positive isolates (Enterococcus spp), commonly known as culprits of hepatobiliary infection. ^{Reference Melzer, Toner, Lacey, Bettany and Rait34} Although >99% of gastrointestinal flora are anaerobic, we found that anaerobes only comprised 3% of cultured bacteria. ^{Reference Vedantam and Hecht35} Therefore, anaerobes are either sufficiently covered by preoperative antibiotics or are difficult to isolate. However, yeasts (specifically Candida) make up a considerable percentage of SSIs cultured from our institution, possibly related to antibiotic-induced yeast infections within bowels and superficial skin where they are colonized. More than half of cultures were polymicrobial; polymicrobial infections are common in abdominal surgeries yet do not have inferior outcomes compared to monomicrobial infections. ^{Reference Frisan36–Reference Shah, Edwards and Dietch38} Distributions of bacterial grown from year ranges were comparable, suggesting no significant microbial surge or acquired antibiotic resistance (Fig. 2b and 2c).

This study had several limitations.

Given a patient population of 645, our study utilized patient specific data reported in the NSQIP database. Missing values in this database limited our statistical analyses. Nearly half of our patients (322 of 645) lacked data on preoperative antibiotic class, which restricted our ability to perform logistic regression analyses to evaluable whether choice of preoperative antibiotics was predictive of SSI. Moreover, completion of antibiograms was hindered by unstandardized laboratory testing because bacteria were often tested against different antibacterial agents. For instance, Pseudomonas grew from 4 SSI cultures between 2017 and 2021: 2 were resistant to meropenem, and the other 2 were suppressed per laboratory protocol. With data missing from the other 2 cultures, we were unable to calculate a resistance percentage for the antibiogram depiction.

In conclusion,

we identified surgery duration as a prognostic indicator of SSIs in patients undergoing the Whipple procedure. Additionally, we identified a strong correlation between cancer as a surgical indication and decreased SSI rates. This finding has not been reported in previous epidemiologic studies, and future analyses may be warranted. We argue that choice of antimicrobials may be tailored to specific patient needs without compromising SSI risk. Maintaining current Whipple-protocol preoperative antibiotics at our institution while making systematic efforts to decrease operative times may help improve patient outcomes after pancreatoduodenectomy.