Effectiveness of Antibiotic Prophylaxis in Non-emergency Cholecystectomy Using Data from a Population-Based Cohort Study

The CholeS study-enrolled patients undergoing cholecystectomy for benign gallbladder diseases in acute UK and Irish hospitals participating in this study between 1 March and 1 May 2014 were included. This secondary analysis investigated the group of patients undergoing non-emergency cholecystectomy for benign gallbladder diseases. Patients undergoing non-emergency cholecystectomy as delayed operations (defined as a scheduled cholecystectomy following an emergency admission with gallbladder disease) or elective operations (defined as a planned elective admission for cholecystectomy referred by their family doctor and added to the routine surgical waiting list from the outpatient department with no prior emergency admission with gallbladder disease) were included. Patients undergoing an emergency cholecystectomy were thus excluded. Open, laparoscopic and laparoscopic converted to open surgeries were included. Patients who had a cholecystectomy for known gallbladder cancer or as a part of another surgical procedure, e.g. pancreaticoduodenectomy, bariatric, anti-reflux or transplant operations or an emergency cholecystectomy (defined as a cholecystectomy during an acute admission) were excluded.

The primary outcome of interest was superficial SSI within 30 days. Secondary outcomes included a deep SSI, all-cause readmissions, all-cause complications, all re-interventions and post-operative administration of antibiotics within 30 days. These have been defined previously [9]. Briefly, the following definitions were used:

To standardise data quality, a quality assurance programme was developed [8]. This included a detailed study protocol, pilot phase and a requirement for a minimum of 95% data completeness at submission. Case ascertainment and data accuracy were further validated by independent investigators at selected hospitals, who checked data correctness from 2077 (23.3%) patients against original medical records. These independent investigators were not involved in the original data collection. Case ascertainment and accuracy of collected data were above 95.2 and 99.2%, respectively.

The main explanatory variable was the antibiotic prophylaxis, defined as antibiotics administered at induction or during the operation. Other pre- and peri-operative characteristics were considered as potential explanatory variables influencing the primary outcomes. A full list including definitions has been published previously [8]. Briefly, patient characteristics included here were: age, sex, American Society of Anaesthesiologists (ASA) fitness grade (1, normal healthy patient; 2, mild systemic disease; >2, severe systemic disease/severe systemic disease that is a constant threat to life/moribund patient who is not expected to survive without the operation), body mass index (BMI in kg/m²; <24.9, 25.0–29.9, 30.0–34.9, >35.0), indication [biliary colic, cholecystitis, pancreatitis, common bile duct (CBD) stones and others]. Surgical factors considered were grade of operating surgeon (non-consultant grade; consultant), operative method (laparoscopic, converted to open), operative difficulty (as defined by the Nassar scale of difficulty for cholecystectomy from 1 to 4 [13]) and intra-operative events (bile spilt, stones spilt, bleeding, bowel Injury, common bile duct injury (CBD) injury, cholangiogram, CBD explored). In addition, a composite variable of ‘high-risk’ group for developing an SSI was generated, including patients aged over 60 years, undergoing an intra-operative cholangiogram, CBD exploration, conversion to open, bile spilt, stones spilt, CBD injury or bowel injury during surgery. These factors have been identified in guidelines to represent patients at potential risk of developing a SSI [7]. Low-risk surgery was considered as patients undergoing surgery with none of the above risk factors.

To adjust for factors found to be associated with antibiotic prophylaxis, multivariable binary logistic regression models were produced for each outcome. Antibiotic prophylaxis was entered into the model as a factor, and all of the other potentially confounding factors were considered for inclusion, with a forward stepwise entry procedure used to select independent predictors of outcome.

The data were also analysed using a matched approach. Patients who received antibiotic prophylaxis were exact matched to those who did not on all of the confounding factors considered, with age treated as categorical. The outcomes were then compared across the resulting pairs of patients using McNemar’s test, and odds ratios (OR) with 95% confidence interval (CI) were produced, in order to compare the results to those of the multivariable analysis for consistency.

Finally, to test whether the effect of antibiotic prophylaxis differed by patient risk, an additional set of multivariable binary logistic regression models were produced. The patient risk group, antibiotic prophylaxis and the interaction term were entered into the model. One model included the whole cohort of patients, so confounding factors were added to the model using a forward stepwise approach, to give adjusted odds ratios. The analysis was also repeated for the subgroup of matched patients, which did not require this adjustment. The significance of the interaction terms was from these models which were interpreted as testing the difference in the effectiveness of antibiotic prophylaxis in the low- and high-risk groups.

All analyses were performed using IBM SPSS 22 (IBM Corp. Armonk, NY). Patients with missing data were excluded on a per-analysis basis, and p < 0.05 was deemed to be indicative of statistical significant throughout.