Prophylactic Antibiotics for Elective Laparoscopic Cholecystectomy

Abstract

Surgical site infections with elective laparoscopic cholecystectomy are less frequent and less severe, leading some to suggest that prophylactic antibiotics are no longer indicated. We compared the incidence of surgical site infections prior to and following an institutional practice change of withholding prophylactic antibiotics for elective laparoscopic cholecystectomy. Between May 7, 2013, and March 11, 2015, no prophylactic antibiotics were given to patients selected for elective cholecystectomy by two surgeons at a single center. The only patients excluded were those who received antibiotics prior to surgery for any reason. All others, including those at high risk for infection, were included. The incidence and severity of infections were compared to historical controls treated with prophylaxis by the same two surgeons from November 6, 2011, to January 13, 2013. There were 268 patients in the study group and 119 patients in the control group. Infection occurred in 3.0% in the study group compared to 0.9% in the controls (p = 0.29). All infections were mild except one. Based on these data, the routine use of prophylactic antibiotics for elective laparoscopic cholecystectomy is not supported.

Introduction:

During the 1960s through the 1980s, the era immediately before the common use of laparoscopic cholecystectomy (LC), surgical site infection (SSI) rates between 3% and 47% were reported for open biliary tract surgery. Data from a meta-analysis of 42 studies (4129 cases) showed that prophylactic antibiotics (PA) reduced infection rates by 9% (15% without and 6% with antibiotics).¹ Since both the frequency and severity of infections with open cholecystectomy were compelling, these data weighed heavily in favor of PA.

As LC replaced open cholecystectomy from the late 1980s through the early 1990s, observers began to notice that infections were less frequent and less severe. Over time, several articles have shown that PA is probably not as beneficial for elective LC as it was for open biliary tract surgery.^2–20 Paralleling the apparent loss of benefit, there is a newfound, heightened awareness of the complications of perioperative period including anaphylaxis, secondary infections like C. difficile colitis, and cost.

Some reports evaluating the use of PA for elective LC are prospectively randomized, but almost all are understandably under-powered, given the estimated number of cases required for the low incidence of variables being studied. All prior studies are helpful but most have deficiencies, especially in the areas of intention-to-treat analysis, exclusion of patients at high risk for SSI, and statistical power.^2–20 Our study provides data that address the intention-to-treat issue by including all cases, even those with high risk for SSI, and by strict adherence to a protocol that prohibits perioperative antibiotics except for established SSI or other nosocomial infections. As evidenced by the lack of standardized practices and clear-cut guidelines for PA for elective LC, there is an ongoing need for studies that add to the weight of evidence for or against PA for elective LC. We hypothesized that withholding PA for elective LC would have no significant effect on SSI rate. Thus, the purpose of our study was to compare the incidence of SSI for elective LC prior to and following a change institutional practice of withholding PA.

Methods:

This report is an IRB approved, retrospective, quality improvement study designed to test the efficacy of PA for elective LC at a single tertiary academic institution. It was initially designed to measure quality of care at a single institution by studying the incidence of SSI in elective LC with the universal use of PA. It included all elective cases that were not already on antibiotics performed by two surgeons. The cases performed between 11/6/2011 and 1/13/2013 were reviewed. The incidence of SSI was low, which led to the conclusion that PA may have more risk than benefit. Therefore, the study was continued from 5/7/2013 to 3/1½015 with the goal of establishing the incidence of SSI in a study group (SG) with no PA for comparison with this PA cohort, which was chosen as a secondary database to serve as a control group (CG). Going forward, the goal was to address concerns about the available data raised by authors of consensus reports and meta-analyses: 1) shortcomings in intention-to-treat analysis, 2) insufficiently powered reviews, 3) risk of case selection bias. All SG cases received no PA. The selection and exclusion criteria remained the same as for CG. Excluding only cases already given antibiotics before surgery for any reason, all other cases of elective LC were included, even those with co-morbidities presumed to increase the risk of SSI. No intraoperative antibiotics were given for any reason, and no postoperative antibiotics were given in the absence of documented infection after surgery. Elective LC was defined as outpatient LC performed for indications such as past documented or suspected cholecystitis, biliary colic, and cholelithiasis. Based on the SSI rate of the historical control (1½011 to ½013), it was calculated that a sample size of 268 patients was required to reach a power of 0.8 for inferential comparison of SSI rates; thus the study was continued until this point.

Data collected included age, sex, procedure, preoperative risk factors for SSI (diabetes, BMI>30, age>60, steroid therapy, and immunosuppression), intraoperative risk factors for SSI (unsuspected acute cholecystitis, bile or stone spillage, and conversion to open), indications for exclusion because of preoperative antibiotics (gallstone pancreatitis, clinical evidence of acute cholecystitis, surgical or endoscopic manipulation, PA given in error, and choledocolithiasis), SSI occurrence, and treatment interventions for SSI.

The incidence of SSI in CG, the PA cohort, was compared to that of SG, the no PA cohort. The two groups were compared for age, sex, co-morbidities thought to be risk factors for infection, adverse intraoperative events, and outcome, defined as mortality, bile duct injury, or morbidity related to LC. Student’s t-test and Fischer’s exact test were used to compare the data. A p value of less than 0.05 was considered statistically significant.

Results:

After removing 25 patients meeting exclusion criteria from the CG database, there were 119 patients for inclusion. After removing 70 patients meeting exclusion criteria from the SG database, there were 268 patients for inclusion.

Groups CG and SG were similar in all respects. 82.4 % (98/119) of CG patients were female compared to 86.6 % (232/268) of patients in SG (p = 0.28). The average age of patients in CG was 44.9 compared to 42.4 in SG (p = 0.08). Comparing the presence of each preoperative and intraoperative risk factor for SSI, there was no difference between CG and SG except in the case of diabetes (p = 0.01) (Table 1). Notwithstanding 17 patients excluded from SG because of inappropriate PA, the percent of patients excluded for CG and SG was proportionate: 25/144=17.4% for CG and 53/338=15.7% for SG (p = 0.40).

Table 1.

Comparison of groups with (CG) and without (SG) prophylactic antibiotics n = number of patients with comorbidity or adverse event in subset. % = n as the numerator and the total number of patients in the group as the denominator.

	Controls CG (# - 119) (n/%)	Study Group SG (# - 268) (n/%)	P value
CO-MORBIDITY
Diabetes	6/5	37/14	0.01
BMI>30	73/61	157/59	0.61
Steroid therapy	0/0	3/1	0.25
ASA>2	23/19	75/28	0.07
Immunosuppression	5/4	12/4	0.90
Age>60	7/6	16/6	0.97
ADVERSE INTRAOPERATIVE EVENT
Unexpected acute cholecystitis	15/12	30/11	0.69
Bile spillage	40/34	96/36	0.67
Conversion to open	1/1	2/1	0.92

All 17 inappropriately administered PA were breaches in protocol ordered by residents. A further review of these found no rationale in 11/17, immunosuppressed comorbidity in 3/17 (sickle cell disease = 2; myelodysplastic syndrome = 1), indwelling abdominal catheter from a ventriculoperitoneal shunt in 1 patient, and possible untreated recent episode of acute cholecystitis in 1 patient.

The proportion of cases with at least one preoperative risk factor for SSI was similar for both groups: 90/119=75.6% for CG and 213/268=79.5% for SG (p = 0.40). There were 56 intraoperative events (some cases had multiple events) predisposing to SSI in CG and 128 in SG: 56/119 = 47.1% in CG and 128/268 = 47.8% in SG (p = 0.90).

SSI included one patient in the CG (0.9%) and eight patients in the SG (3.0%) (p = 0.29). There was one deep infection in SG. All other infections were superficial. There were no common bile duct injuries, mortality, or need for re-operation.

Six of the superficial SSI were umbilical port site infections, one superficial SSI was a Kocher incision of a LC converted to open, and one was an organ site infection. All superficial SSI were diagnosed by clinical exam and treated with antibiotics, and 2/6 umbilical port site superficial SSI had an incision and drainage upon initial diagnosis. All superficial SSI resolved with antibiotic treatment. The one organ site SSI was in a patient on postoperative day 14 who had an infected fluid collection in the gallbladder fossa, diagnosed by computed tomography of the abdomen and pelvis. This was treated with percutaneous drainage by interventional radiology and 7 days of intravenous antibiotics. Cultures grew E. coli and Klebsiella.

The incidence of SSI for each preoperative and intraoperative risk factor is listed in Table 2. Although the SSI rate in SG was noticeably higher in the categories of unsuspected acute cholecystitis and immunosuppression, there was no statistical difference between groups in any category.

Table 2.

Comparing SSI in subsets with (CG) and without (SG) prophylactic antibiotics n = the number of cases with SSI in the subset. t = total number of cases in the subset. % = n/t. Since many cases fell into more than one subset, the sums of cases and SSIs are skewed.

Subsets	Controls (CG) subset SSI n/t=%	Study Group (SG) subset SSI n/t=%	P value
PREOP CO-MORBIDITIES
None	0/29=0	0/55=0	NA
Diabetes	0/6=0	1/37=2.7	0.68
BMI>30	1/73=1.4	5/157=3.1	0.44
Steroid therapy	0/0=0	0/3=0	NA
ASA>2	0/23=0	2/75=2.6	0.42
Immunosuppression	0/5=0	1/12=8.3	0.52
Age>60	0/7=0	0/16=0	NA
INTRAOP ADVERSE EVENTS
Unexpected acute cholecystitis	0/15=0	2/30=6.7	0.31
Bile spillage	0/40=0	2/96=2.1	0.36
Conversion to open	0/1=0	1/2=50	0.38

Discussion:

Researchers have probed the efficacy of PA for elective LC from the beginning of minimally invasive surgery for gallbladder disease.^2–20 Since the incidence of SSI is low, failing to acquire enough cases has left almost all prospective randomized trials underpowered. This has led to meta-analyses to validate results.^21,22 In addition to small numbers, these reviews have identified two additional concerns: 1) Twelve of 19 prospective, randomized studies fail in the important area of intention-to-treat analysis; 2) Case selection bias is possible since almost all of the studies excluded patients at high risk for SSI. By including patients with co-morbidities that might increase the frequency and severity of SSI, reaching a power of 0.80, and rigidly addressing intention-to-treat analysis, our study fills some gaps in the previously available information.

By comparison to the randomized trials, our study is a retrospective non-randomized review. Only one of the 19 randomized controlled studies has a larger no-antibiotic cohort.¹⁸ Given our goals, randomization would have little if any value. We used historical controls as a secondary database, enabling us to complete the study in a relatively short time period without compromising intention-to-treat analysis.²³ Although 17 otherwise eligible patients were among the patients excluded from the SG cohort immediately because they received PA in error, all remaining eligible patients were included, and there were no protocol violations: 1) No patient who underwent elective LC without PA was given a belated intraoperative or perioperative antibiotic for bile spillage, unexpected acute cholecystitis, conversion to open cholecystectomy, or any other reason; 2) No patient received a postoperative antibiotic without a documented established infection. The shortcomings of insufficient power, exclusion of patients at high risk for SSI, and flawed intention-to-treat analysis were mitigated.

As noted, our no-antibiotic cohort, SG, included all patients, even those presumed to be at high risk for SSI because of co-morbid conditions that predispose to infection. 18.7% of the patients in our study had at least two risk factors for SSI. Even with these high-risk cases, our results show no statistical advantage for PA compared to no PA. Even though SAGES guidelines specifically state that PA is not required for low-risk patients, they hedge on patients at high risk. The guidelines include a recommended antibiotic regimen, implying ongoing uncertainty regarding case selection criteria.²⁴ By adding the patients with preoperative, high-risk co-morbidity variables, and omitting preemptive perioperative antibiotic therapy for adverse intraoperative events, we provide new opportunities for subgroup analysis with ongoing data collection. With continued data collection, risk/benefit for all of the preoperative and intraoperative variables can be statistically validated. At this point, numbers are too small for meaningful multivariate analysis, which is the noteworthy limitation of this study. More accumulated data may identify patients with risk factors for unexpected acute cholecystitis and comorbid conditions that may benefit from PA.

We and others have shown that there is no benefit to PA for elective LC for low-risk patients. In addition to lack of benefit, PA occasionally causes serious complications. The incidence of anaphylaxis with cephalosporins is one per thousand.²⁵ Although less well studied, other antibiotics, including quinolones, are implicated as well.²⁶ In one representative center, the incidence and severity of C. difficile infection secondary to perioperative PA are increasing and has reached 14.7 cases per 1000 operations.²⁷ Cost is now under intense scrutiny. For example, at another institution where approximately 140 cholecystectomies are performed annually, researchers estimate that ceasing the use of PA for LC would result in an institutional savings of $30,060 annually.¹⁰

In spite of data currently available, approximately 36% of surgeons continue to use PA for elective LC for low-risk patients.²⁸ This is not surprising, given the confusion in the literature. Only 64% of surgeons have ceased PA for low-risk LC, and many of them still give perioperative antibiotics after intraoperative events like bile spillage, unsuspected acute cholecystitis, and conversion to open cholecystectomy. Given these divergent practice patterns, the establishment of definitive guidelines will require better clarification of the risk/benefit ratio of PA for patients at both low and high risk for SSI.

This study confirms that antibiotic prophylaxis for elective cholecystectomy is not beneficial, even though patients with co-morbidities that are known to predispose to SSI were included, and intention-to-treat analysis was impeccable. Including patients at high risk for SSI adds validity to our study by helping to eliminate bias from case selection and intention-to-treat analysis. However, questions regarding the efficacy of PA in high-risk patients remain unanswered, because the subgroups are small. Since data from available reports are not standardized, meta-analyses have fallen short as well. Ongoing case studies from a single database -- either at one institution or through collaboration -- are required to provide meaningful univariate and multivariate analysis. For now, surgeons should cease using prophylactic antibiotics for elective cholecystectomy in low-risk patients and should feel more comfortable individualizing care of high-risk patients. To categorically eliminate PA for high-risk patients must await additional data.