Meta-analysis: antibiotic prophylaxis in elective laparoscopic cholecystectomy

Authors


Prof. Z. Hu, Professor of Surgery, Department of General Surgery, Changzheng Hospital, Second Military Medical University, No.415, Fengyang Road, Shanghai, 200003, China.
E-mail: czhuzq@yahoo.com.cn

Summary

Background  Current guidelines do not support routine antibiotic prophylaxis during elective laparoscopic cholecystectomy. However, routine antibiotic prophylaxis for elective laparoscopic cholecystectomy is still popular in many clinical settings.

Aim  To evaluate the role of antibiotic prophylaxis in elective laparoscopic cholecystectomy.

Methods  Electronic databases and manual bibliographical searches (updated to April 2008) were conducted. A meta-analysis of all trials evaluating antibiotic prophylaxis in elective laparoscopic cholecystectomy was performed.

Results  Fifteen trials were included, involving 2961 patients. After pooling all the trials, 48 wound infections occurred (48/2961, 1.62%), 22 in antibiotic prophylaxis group (22/1494, 1.47%) and 26 in control group (26/1467, 1.77%). The pooled odds ratio (OR) was 0.79 (95%CI: 0.44, 1.41). Four major infections occurred (4/2961, 0.14%), 3 in antibiotic prophylaxis group (3/1494, 0.20%), and one in control group (1/1467, 0.07%). The pooled OR was 2.51 (95%CI: 0.35, 17.84). Fifteen distant infections occurred (15/2961, 0.51%), six in antibiotic prophylaxis group (6/1494, 0.40%) and nine in control group (9/1467, 0.61%). The pooled OR was 0.53 (95%CI: 0.19, 1.50). Sensitivity analyses also failed to support antibiotic prophylaxis’s preventive effect.

Conclusions  Considering the absent role of antibiotic prophylaxis in reducing the infectious complications, we suggest that antibiotic prophylaxis is unnecessary and should not be routinely used in low-risk elective laparoscopic cholecystectomy patients.

Introduction

Laparoscopic cholecystectomy (LC) has been the standard procedure for the management of gallstone disease since 1990s.1–4 Compared with open cholecestectomy (OC), LC is associated with lower post-operative infections, which make the role of routine antibiotic prophylaxis (ABP) in this minimally invasive procedure debatable.5,6 In fact, given the extremely low infectious incidences, current guidelines do not support routine ABP in elective LC.7,8 However, routine ABP for elective LC was still popular in many clinical settings, because many surgeons believed that prophylactic antibiotics would reduce the incidences of post-operative infections in elective LC.9–12 McGuckin et al. showed that 79% of patients who had undergone LC procedures had received prophylactic antibiotics preoperatively, 63% had received them post-operatively.13 Is it really necessary to perform ABP in elective LC? Two meta-analyses had been made trying to answer the question.14,15 However, these meta-analyses were conducted 5 years ago with relatively small sample sizes and low powers. As there is still some controversy with regard to the use of ABP in elective LC and several new randomized controlled trials (RCTs) are available, we conducted an up-to-date meta-analysis to provide the current best evidence on this topic.

Materials and methods

Identification and selection of studies

Relevant studies were identified and selected by searching the databases, including:(1) Medline, Embase, Cochrane controlled trials register and Chinese Biological Medicine Disk (updated to April 2008) under the search words ‘laparoscopic cholecystectomy’, ‘antibiotic prophylaxis’, and ‘randomized controlled trial’. (2) A review of reference bibliographies from original research articles and reviews. No language or date limitations were imposed.

The following inclusion criteria were applied: (1) Study design: randomized, placebo or untreated controlled trial, blind or not, including multi-centre trials; (2) Study population: patients undergoing elective LC; (3) Antibiotics should be administrated before the end of LC; (4) each study should contain information on rates of infectious complications. The following exclusion criteria were applied: (1) Studies compared two groups, both submitted to ABP with different dosage or route of the same antibiotic, or with different antibiotics; (2) Comparing ABP with other prophylactic intervention; (3) Tropical antibiotic prophylaxis. In the case of sequential or multiple publications of analyses of the same data or overlapping data sets, the publication that reported data from the largest or most recent study was included.16

Data extraction

Data were extracted from each study by two authors (Zhou and Hu) independently according to a predesigned protocol with the journal, year and country of the study veiled to avoid a possible subjective prejudice. Following data were recorded: (1) Basic characteristics: sample size, mean age, inclusion and exclusion criteria, antibiotics used and duration of follow-up; (2) Details in study design: method of randomization, blinding (single, double or none), withdraw and dropouts; (3) Outcomes: wound infection (defined as the clinical appearance of erythema and pus or serous discharge with or without bacteriological evidence), major infection (defined as intra-abdominal collections or abscesses with clinical or ultrasonic evidence), distant infection (defined as the clinical manifestation of respiratory and urinary infections with or without bacteriological evidence), overall infection (wound, major and distant infections combined). Two authors (Zhou and Hu) recorded data independently and then crosschecked the information.

Quality assessment

We applied the Jadad composite scale to assess the methodological quality of studies included in the meta-analysis17 (Box 1). This is a 5-point quality scale, with low-quality studies scoring ≤2, and high-quality studies scoring ≥3.18,19 Quality assessment was independently performed by two authors (Zhou and Hu). Each included study was given an overall quality score based on the Jadad composite scale. Any disagreement was resolved by discussion and consensus.

Table Box 1..   Criteria for grading quality of randomized controlled trials: the Jadad score
Study received 1 point for each yes or 0 point for each no for each of the following quesitoin:
1. Was the study decribed as randomized such as using the words randomly, random, and radomization?
 (a) An additional point was given if method of randomization was described and it was appropiate (for example, table of random numbers, computer-generated).
 (b) A point was deducted if the method of randomization was inappropriate (for example, patients allocated alternately, by birth date or hospital number).
2. Was the study described as double-blind?
 (a) A point was given if method of blinding was described and it was appropriate (for example, identical placebo).
 (b) An additional point was deducted if method of blinding was inapproprite (for example, comparing placebo tablet with injection).
3. Was there a description of withdrawals and dropouts?
Maximun number of points is 5.

Statistical methods

The odds ratio (OR) was used as the effect size. Summary ORs and their corresponding 95% confidence intervals (CI) were estimated by fixed effect (Mantel-Haenszel) or random effect (DerSimonian and Laird) models.20,21 Tests for heterogeneity were performed with each meta-analysis using the Cochran Q statistic and the I2 test, with P = 0.05 indicating significant heterogeneity. Random effect model was performed when heterogeneity was present. Zero total event trials (trials with zero events in both prophylaxis and control arms) would be excluded from the analyses, because they make no contribution to the magnitude of the treatment effect.22,23 To check for publication bias, a funnel plot was constructed using Egger’s linear regression method.24 Statistical analyses were performed using Revman 4.2 and STATA 9.0 software packages. A P value <0.05 was considered statistically significant.

Results

Characteristics of included studies

Sixty eight records were identified by the search strategy; 48 records were excluded because of the reasons shown in Figure 1. Of the 20 potentially appropriate studies,25–44 three were excluded for comparing two kinds of ABP,25–27 one for comparing ABP with other intervention,28 and one for assessing tropical ABP.29 Finally, 15 RCTs were included in the meta-analysis (Table 1).30–44

Figure 1.

 Flow chart shows the method of inclusion of trials in our analysis.

Table 1.   Baseline characteristics of trials included in meta-analysis
StudyNo. of patients (Pro/Con)AnitbioticControlFollow-up (Day)
  1. Pro, prophylaxis; Con, control; U, untreated; P, placebo; LC, laparoscopic cholecystectomy.

Illig et al.30128/122Cefazolin OR Vancomycin (Before and after LC)U30
Dobay et al.31 29/24Cefotetan (Only before LC)P30
Higgins et al.32277/135Cefotetan OR Cefazolin (Only before LC)P30
Tocchi et al.33 44/40Cefotaxime (Before and after LC)P42
Chen et al.34 57/56Cefazolin OR Ciprofloxacin (Before and after LC)U60
Mahatharadol35 50/50Cefazolin (Only before LC)U30
Shi et al.36 64/55Ceftriaxone OR Metronidazole (Before and after LC)U30
Wang et al.37 84/115Cefmetazole OR Ciprofloxacin (Before and after LC)U60
Koc et al.38 49/43Cefotaxime (Before and after LC)P30
He et al.39 70/68Clavulanic Acid + Amoxicillin
OR Ceftriaxone (Before and after LC)
U30
Hu et al.40 98/98Ceftriaxone (Only before LC)U30
Chang et al.41141/136Cefazolin (Only before LC)P30
Kuthe et al.42 40/53Cefuroxime (Only before LC)P30
Fang et al.43263/372Cefoperazone (Only before LC)U30
Wang et al.44100/100Levofloxacin (Before and after LC)U30

The meta-analysis involved 2961 patients: 1494 were randomized to ABP group and 1467 to control group. All the included studies were single centre trials. 6 studies were placebo controlled and 9 studies were untreated controlled. The mean age ranged from 42 to 54 years and the mean follow-up interval ranged from 30 to 60 days. As for the methods of follow-up, three trials used the method of structured interview and clinical examination, with the surgeons responsible for reporting infectious complications.38,39,41 One trial used the method of telephone contact, with the surgeons responsible for reporting infectious complications.44 Two trials reported the surgeons were responsible for reporting infectious complications, but did not report the exact method for follow-up.31,32 Nine trials did not report the exact method of follow-up, as well as the person who was responsible for reporting infectious complications.30,33–37,40,42,43 Seven studies used antibiotic before LC only, while eight studies used antibiotic before and after LC. Exclusion criteria shared by almost all the studies were acute cholecystitis, pancreatitis, jaundice, pregnancy, known-allergies to antibiotic, concurrent antibiotic therapy, immunosuppression due to other diseases or drugs and use of prosthetic devices. Five trials excluded patients converted to OC,33,37,40,41,43 four trials excluded patients with a history of previous biliary surgery,31,32,38,41 three trials excluded old patients (age > 60 years),32,39,44 and two trials excluded patients with bile spillage.40,43 The methodological quality scores of included trials ranged from 1 to 5 (Table 2).

Table 2.   Jadad quality score of trials included in meta-analysis
StudyRandomized methodDouble blindingWithdrawals dropoutsTotal
Illig et al.301012
Dobay et al.311203
Higgins et al.322215
Tocchi et al.332114
Chen et al.341001
Mahatharadol351012
Shi et al.361001
Wang et al.371001
Koc et al.381214
He et al.391001
Hu et al.401012
Chang et al.412114
Kuthe et al.421102
Fang et al.431012
Wang et al.442002

Results of meta-analyses

As there was no statistical heterogeneity among included studies, we used fixed effect model to pool the results (Figure 2). After pooling all the trials, 48 wound infections occurred (48/2961, 1.62%), 22 in ABP group (22/1494, 1.47%) and 26 in control group (26/1467, 1.77%). The pooled OR was 0.79 (95%CI: 0.44, 1.41). Four major infections occurred (4/2961, 0.14%), three in ABP group (3/1494, 0.20%) and one in control group (1/1467, 0.07%). The pooled OR was 2.51 (95%CI: 0.35, 17.84). In all, 15 distant infections occurred (15/2961, 0.51%), six in ABP group (6/1494, 0.40%) and nine in control group (9/1467, 0.61%). The pooled OR was 0.53 (95%CI: 0.19, 1.50). Sixty seven overall infectious complications occurred (67/2961, 2.26%), 31 in ABP group (31/1494, 2.07%) and 36 in control group (36/1467, 2.45%). The pooled OR was 0.77 (95%CI: 0.47, 1.27). On the basis of these results, there was no statistically significant difference between ABP and control in prevention of infectious complications in elective LC.

Figure 2.

 Meta-analyses of antibiotic prophylaxis in the prevention of infectious complications in elective laparoscopic cholecystectomy. *Zero total event trials (trials with zero events in both prophylaxis and control arms) were excluded from the analyses and were not displayed in the forest plots.

Sensitivity analyses

Sensitivity analyses were performed by excluding low quality studies, including studies using placebo as control and including large size trials (with the sample size >100 per arm). In addition, we pooled the results according to the administrative ways of antibiotics (used only before LC vs. used before and after LC). The results were similar to those of the primary analyses, with details listed in Table 3. The sensitivity analyses also failed to support the positive effect of ABP.

Table 3.   Sensitivity analyses
ItemNumber of included studyPooled effects (95% confidence interval)Test for heterogeneity, chi-square testZ test for pooled effect size
  1. All pooled by fixed effect model. NA, not applicable.

  2. WI, wound infection; MI, major infection; DI, distant infection; OI, overall infection; LC, laparoscopic cholecystectomy.

Excluding low quality studies
 WI40.70 (0.27, 1.84)1.91 (P = 0.59)Z = 0.72 (P = 0.47)
 MI26.94 (0.43, 111.10)0.00 (P = 0.98)Z = 1.37 (P = 0.17)
 DI20.50 (0.11, 2.35)0.48 (P = 0.49)Z = 0.87 (P = 0.38)
 OI40.77 (0.34, 1.70)0.63 (P = 0.89)Z = 0.65 (P = 0.51)
Using placebo as control
 WI50.70 (0.29, 1.70)1.91 (P = 0.75)Z = 0.79 (P = 0.43)
 MI26.94 (0.43, 111.10)0.00 (P = 0.98)Z = 1.37 (P = 0.17)
 DI20.50 (0.11, 2.35)0.48 (P = 0.49)Z = 0.87 (P = 0.38)
 OI50.75 (0.35, 1.61)0.64 (P = 0.96)Z = 0.73 (P = 0.47)
Large size trials
 WI50.79 (0.29, 2.17)3.10 (P = 0.54)Z = 0.46 (P = 0.64)
 MI26.94 (0.43, 111.10)0.00 (P = 0.98)Z = 1.37 (P = 0.17)
 DI30.75 (0.18, 3.10)0.23 (P = 0.89)Z = 0.40 (P = 0.69)
 OI50.79 (0.35, 1.74)2.16 (P = 0.71)Z = 0.60 (P = 0.55)
Antibiotics used only before LC
 WI60.77 (0.30, 1.95)3.07 (P = 0.69)Z = 0.56 (P = 0.58)
 MI17.13 (0.14, 359.64)NAZ = 0.98 (P = 0.33)
 DI10.97 (0.09, 10.92)NAZ = 0.02 (P = 0.98)
 OI60.88 (0.38, 2.05)1.56 (P = 0.91)Z = 0.30 (P = 0.76)
Antibiotics used before and after LC
 WI80.81 (0.38, 1.70)1.44 (P = 0.98)Z = 0.56 (P = 0.57)
 MI31.77 (0.18, 17.06)2.57 (P = 0.28)Z = 0.49 (P = 0.62)
 DI40.47 (0.15, 1.46)0.43 (P = 0.93)Z = 1.31 (P = 0.19)
 OI80.73 (0.39, 1.34)2.02 (P = 0.96)Z = 1.03 (P = 0.30)

Publication bias

We assessed the publication bias based on the results of wound infection and overall infection. No evidence of publication bias existed in the studies included in the meta-analysis, based on the egger’s publication bias plots (Figure 3).

Figure 3.

 Egger’s publication bias plots for wound infection and overall infection. Egger’s method is based on doing a linear regression of the standardized effect size against the precision (1/variance), and detects funnel plot asymmetry by determining whether the intercept deviates significantly from zero in a regression of the standardized effect estimates against their precision. The 95%CIs of the regression line’s y intercept should included zero if there is no evidence of publication bias. The y intercepts (95%CI) for wound infection and overall infection were −0.50 (−2.10, 0.90) and 0.11 (−1.24, 1.46), respectively. Thus, no evidence of publication bias existed in the studies included in the meta-analysis.

Discussion

Several studies have concluded that in OC the use of ABP leads to a significant decrease in infectious complications.45 However, these data cannot be extrapolated to elective LC. Compared with OC, LC was superior with its smaller incision, minimal trauma, shorter hospital stay, better preservation of the immune system and considerable decrease of post-operative infections,5,6 thus making the role of routine ABP in this minimally invasive procedure questionable. Two previous meta-analyses addressed this subject.14,15 However, these meta-analyses were conducted 5 years ago with relatively small sample sizes and low powers. As there is still no definitive conclusion on this issue, it may be particularly germane to do such a systematic review to resolve this problem that is frequently encountered all over the world.

The main finding of our meta-analysis was similar to that of the previous meta-analyses,14,15 namely, ABP had no role in significantly reducing the incidences of post-operative infectious complications in elective LC, including wound infection, major infection, distant infection and infectious complication as a whole. By conducting sensitivity analyses, we found that the results were independent of the design, quality and sample size of original studies. Moreover, we found that additional antibiotic doses after LC also did not reduce the incidences of infection. On the other hand, after pooling all currently available RCTs, we found that rates of infection were extremely low in elective LC (wound infection: 1.62%; major infection: 0.14%; distant infection: 0.51%; overall infection: 2.26%). This was comparable with that of two meta-analyses of LC series.1,2 Given the extremely low incidences of infections and the absent role of ABP in elective LC, our meta-analysis indicated that it was unnecessary to give ABP in elective LC.

However, before drawing a conclusion, there are still lots of things we need to reconsider. First of all, although there was no statistical heterogeneity among included studies, we could not exclude the potential clinical heterogeneity. In fact, there was a wide variability in reported rates of infections after LC among included studies, with the rates of wound infection ranging from 0% to 8.33% and the rates of overall infection ranging from 0% to 14.28%. An important source of clinical heterogeneity may come from the variability in definition of infectious complications. A more strict definition may result in a high rate of infection, while a more compromised definition may lead to a lower rate of infectious complication. Another source of clinical heterogeneity may be ascribed to the differences in exclusion criteria, which may reflect different infectious risks of enrolled patients among included studies. Although all the studies seemed to enrol only patients at low risk of infection, substantial variability existed. Actually, 5 trials excluded patients converted to OC,33,37,40,41,43 four trials excluded patients with a history of previous biliary surgery,31,32,38,41 three trials excluded old patients (age > 60 yr)32,39,44 and two trials excluded patients with bile spillage.40,43 In a previously published meta-analysis, low-risk patients were defined as those who did not have acute cholecystitis, a recent history of acute clolecystitis, common bile duct calculi, jaundice, immune suppression, prosthetic implants.15 As patients converted to OC, with a history of previous biliary surgery, age > 60 yr, and bile spillage were reported to be at higher risk of post-operative infections,46–50 low-risk patients should be also defined as those without any of these risk factors. On the contrary, high-risk patients were defined as those with the risk factors. So far, there is no RCT conducted to assess the role of ABP in high-risk elective LC patients, so the results of our meta-analysis may restrict to low-risk elective LC patients. Last but not least, the different rates of post-operative infections among included studies may be ascribed to the surgical technique and experience of different surgeons, which were often ignored. Perhaps, delicate surgical technique and other non-antibiotic-based physical prophylactic procedures were more important because major infectious complications, such as intra-abdominal abscess and acute pancreatitis, were usually related to technical pitfalls.1,51,52 In fact, several technical modifications were proposed to reduce the post-operative infectious complications.28,41,53 The use of plastic bags for protection of umbilical trocar sites may reduce the rates of wound infection, the use of meticulous irrigation of the operation field when bile spillage and gall-bladder stones is encountered may reduce the rates of intra-abdominal abscesses and the use of a drain in the subhepatic space, if a difficult and ruptured gallbladder is encountered, may minimize the risk of subhepatic fluid accumulation.

To the best of our knowledge, none of the published meta-analyses could avoid the potential clinical heterogeneity, because included trials would certainly have some deference in design, measurement and statistical calculation. However, by performing sensitivity analyses, we got results similar to those of the primary analyses, which confirmed our results, suggesting that large effects due to clinical heterogeneity are unlikely.

Another factor we need to reconsider is the costs of ABP in elective LC. The numbers need to treat (NNT) for wound infection and overall infection were 333 and 526 respectively. This means that one would need to treat 333 patients undergoing elective LC using antibiotics to control one wound infection episode and to treat 526 patients using antibiotics to control one infection episode, which is definitively not cost-effective in clinical practice. Higgins et al. estimated that if no antibiotic was needed in elective LC, this resulted in a saving of $30 060 USD/year.32 These results further suggested that ABP should not be administered to elective LC patients at low risk of infection, because avoidance of the unnecessary administration of antibiotics could lead to significant savings in terms of costs.

None of the 15 studies reported tolerability and/or adverse reaction of ABP. Although ABP seemed to be well-tolerated, we could not exclude the potential side effects. Furthermore, prolonged use of prophylactic antibiotics may induce bacterial resistance and increase further opportunistic nosocomial infections,54,55 thus the unnecessary use of antibiotics, should be discouraged.

Our study also has some limitations. Firstly, 10 out of 15 included studies were low-quality studies, with Jadad score <3. However, by performing sensitivity analyses that only included high quality studies, we got similar results. It is unlikely that this shortcoming affected our findings. Secondly, it is possible that unpublished articles, such as scientific dissertations, may be missed because only published studies can be involved in the meta-analysis. However, by constructing funnel plots by Egger’s linear regression methods, we found no publication bias. Thirdly, although our meta-analysis included all the currently available RCTs on this issue, involving 2961 patients, the relatively low rates of infections recorded after LC in this group of subjects and the extremely low rate difference between the two arms make the number of patients included in the meta-analysis insufficient to avoid a type II error, a significant bias present in many reports from negative RCT published in the surgical literature.56 Based on 0.30% rate difference in wound infection and 0.38% rate difference in overall infection, 27767 patients and 23984 patients per arm are required to rule out a type II error (α = 0.05, power = 0.8; two tailed test) respectively. Finally, all the included studies assessed ABP in low-risk elective LC patients. It is still uncertain whether ABP has a role in high-risk elective LC patients.

In conclusion, the results of our study were in accordance with guidelines for pre-operative antibiotic use, which did not support routine antibiotic prophylaxis in elective laparoscopic cholecystectomy. Considering the extremely low post-operative infectious incidences in elective LC and the absence of role of ABP in reducing the infectious complications, we suggest that ABP is unnecessary and should not be routinely used in low-risk elective LC patients. Future studies are warranted to assess the role of ABP in high-risk elective LC patients.

Acknowledgement

Declaration of personal and funding interests: None.

Ancillary