Prophylactic antibiotics to prevent surgical site infection after breast cancer surgery
Abstract
Background
Surgery has been used as part of breast cancer treatment for centuries; however any surgical procedure has the potential risk of infection. Infection rates for surgical treatment of breast cancer are documented at between 3% and 15%, higher than average for a clean surgical procedure. Pre‐ and perioperative antibiotics have been found to be useful in lowering infection rates in other surgical groups, yet there is no consensus on the use of prophylactic antibiotics for breast cancer surgery.
Objectives
To determine the effects of prophylactic (pre‐ or perioperative) antibiotics on the incidence of surgical site infection (SSI) after breast cancer surgery.
Search methods
For this third update we searched the Cochrane Wounds Group Specialised Register (5 December 2013); the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library); the Database of Abstracts of Reviews of Effects (DARE) (The Cochrane Library); Ovid MEDLINE; Ovid MEDLINE (In‐Process & Other Non‐Indexed Citations); Ovid EMBASE; and EBSCO CINAHL. We applied no language or date restrictions.
Selection criteria
Randomised controlled trials of pre‐ and perioperative antibiotics for patients undergoing surgery for breast cancer were included. Primary outcomes were rates of surgical site infection (SSI) and adverse reactions.
Data collection and analysis
Two review authors independently examined the title and abstracts of all studies identified by the search strategy, then assessed study quality and extracted data from those that met the inclusion criteria.
Main results
A total of eleven studies (2867 participants) were included in the review. Ten studies evaluated preoperative antibiotic compared with no antibiotic or placebo. One study evaluated perioperative antibiotic compared with no antibiotic. Pooling of the results demonstrated that prophylactic antibiotics administered preoperatively significantly reduce the incidence of SSI for patients undergoing breast cancer surgery without reconstruction (pooled risk ratio (RR) 0.67, 95% confidence interval (CI) 0.53 to 0.85). Analysis of the single study comparing perioperative antibiotic with no antibiotic found no statistically significant effect of antibiotics on the incidence of SSI (RR 0.11, 95% CI 0.01 to 1.95). No studies presented separate data for patients who underwent reconstructive surgery at the time of removal of the breast tumour.
Authors' conclusions
Prophylactic antibiotics administered preoperatively reduce the risk of SSI in patients undergoing surgery for breast cancer. Further studies involving patients undergoing immediate breast reconstruction are needed as studies have identified this group as being at higher risk of infection than those who do not undergo immediate breast reconstruction.
PICOs
Plain language summary
Antibiotics to prevent surgical site infection after breast cancer surgery
Breast cancer accounts for one in 10 of all new cancer cases diagnosed and surgical removal of the breast is a common treatment approach. An infection of the surgical wound is often a complication of surgery and taking antibiotics just before the operation significantly reduces the chances of developing an infection. The review is not able to establish which antibiotic is most appropriate. No trials were found which considered the effect of antibiotics when the operation involved immediate breast reconstruction.
Authors' conclusions
Background
Breast cancer accounts for one in 10 of all new cancer cases diagnosed around the world each year (Bray 2004) and is the leading cause of cancer death in women (Pisani 1999). Surgery for removal of breast cancer has been common practice for centuries (Donegan 1995) and this is normally used as part of a multi‐faceted approach to care with the aim of curing the patient of their cancer in early stage tumours or prolonging life for others (NICE 2002). Surgical intervention ranges from removing the breast and associated axillary lymph nodes, to lumpectomy with or without sentinel node biopsy (Harris 2004). Whilst the risk of breast cancer for men is only 1%, treatment for men is very similar to that of women (Harris 2004). As with all surgical procedures, breast cancer surgery runs the risk of complications. One such risk is postoperative surgical site infection (SSI), even though breast cancer surgery is considered a 'clean surgical procedure'. Clean surgical procedures, as defined by Haley 1985, are those which have a low risk of bacterial contamination during the surgery.Some women have immediate breast reconstruction; however this group of patients has a higher risk of SSI (Spauwen 2000).
Despite internationally recognised infection control guidelines (Mangram 1999), the incidence of SSI in those being treated for breast cancer is thought to range between 3% (Lefebvre 2000) and 15% (Witt 2003). This is a higher incidence of infection than the 3.4% SSI rate associated with clean surgical techniques (Vazquez‐Aragon 2003). A recent review (Pittet 2005) found that women who had been treated for breast cancer and who had immediate reconstruction had a SSI rate of between 0% and 53%, whilst non‐cancer patients undergoing the same reconstructive surgery had an average rate of 2.5%. There are several factors that are documented as increasing the risk of infection for surgical patients generally. These include: patient risk factors, e.g. diabetes, obesity or smoking (Haley 1985; Mangram 1999); surgical technique, e.g. aseptic technique (Ritter 1988); and type of surgery, e.g. whether the wound is contaminated (Gruendemann 2001). In addition, surgery for breast cancer has several risk factors that make this patient group more susceptible to infection, including use of chemotherapy prior to surgery (neo‐adjuvant chemotherapy); technique of diagnostic biopsy; re‐operation for recurrence or to achieve better tumour margins; reconstructive surgery with implants and seroma accumulation and drainage (Morris 1988; Tran 2003). Infection may lead to significant morbidity for the patient, delay in adjuvant treatment, such as radiotherapy, and increased cost of care if the patient requires supplementary treatment due to infection (Coello 1993).
Pre‐ and perioperative antibiotics have been shown to reduce the risk of postoperative infection in several patient groups (the term "perioperative" refers to administration between induction of anaesthetic and the patient leaving the recovery room) (Gruendemann 2001; Majoribanks 2004; SIGN 2008a). In colorectal surgery antibiotic prophylaxis has been found to reduce long and short‐term morbidity, decrease length of hospital stay and lower the overall cost of care (SIGN 2008a). However, the use of prophylactic antibiotics in preventing infection is still a controversial issue and their routine use is not common in breast cancer surgery. Some feel that a clean surgical procedure should not require prophylactic antibiotics (Sheridan 1994) and that the use of pre‐ or perioperative antibiotics merely masks the symptoms of infection until after the patient is discharged (Wagman 1990). In addition increased antibiotic use may lead to antibiotic resistance (PHLS 2000) and adverse effects such as clostridium difficile infection that causes gastro‐intestinal problems (SIGN 2008a). In order to clarify the situation, this systematic review evaluated the effectiveness of pre‐ or perioperative antibiotics in reducing the incidence of postoperative infections in patients undergoing breast cancer surgery.
Objectives
To determine the effects of prophylactic antibiotics on SSI after breast cancer surgery.
Methods
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs) and controlled clinical trials (where patients were allocated by quasi‐random methods such as alternation, case records numbers or days of the week).
Types of participants
People with breast cancer undergoing breast surgery with or without immediate re‐construction as part of their treatment.
We included studies that involved mixed patient groups (i.e. cancer and non‐cancer, other surgeries or breast implants not as part of cancer treatment) as long as it was possible to extract separate data for those undergoing surgery primarily to treat breast cancer.
Types of interventions
Any pre‐ or perioperative antibiotics used as prophylaxis where there was no known infection and where the use of antibiotics was the only systematic treatment difference between comparison groups.
We only included trials of one antibiotic compared with another if there was a control or placebo arm, as benefit from prophylactic antibiotics has not yet been established in this patient group.
Definitions of key terms:
-
'Antibiotic regimen' describes the characteristics of the antibiotic treatment, i.e. type of antibiotic, route, dose, number of doses and timing of administration.
-
'Preoperative antibiotic prophylaxis' is antibiotic therapy given within 24 hours prior to surgery, solely for prophylaxis (i.e. not for an infection that is already suspected).
-
'Perioperative antibiotic prophylaxis' is antibiotic therapy administered between commencement of induction of surgery and the patient leaving the recovery room.
Comparisons of interest were as follows.
-
Preoperative antibiotic compared with no antibiotic or placebo.
-
Perioperative antibiotics compared with no antibiotic or placebo.
-
Head to head comparisons of antibiotics.
Types of outcome measures
Primary outcomes
-
Incidence of postsurgical breast surgical site (wound) infection (SSI)*. Where possible, this should be reported as the number of participants in each group with a clinically significant infection. Research demonstrates that 98% of acute SSIs related to non‐implant breast surgery occur within 28 days (Mitchell 1999). However, where there is surgical re‐construction, guidelines recommend that this time is increased to one year post surgery (Mangram 1999). Therefore we included all studies that present data on acute SSI within one year of surgery.
-
Adverse reactions (e.g. anaphylaxis, gastro‐intestinal or skin rash).
*Surgical site infection: ideally this will be defined using outcomes from a validated assessment tool such as ASEPSIS (Wilson 1986) which are based on CDC definitions (Mangram 1999).
Secondary outcomes
-
Death.
-
Delay in adjuvant cancer treatment because of breast wound infection.
-
Time to wound healing.
-
Time to infection.
-
Readmission to hospital.
-
Cost of care (should be a comparison between the treatment and control group).
Search methods for identification of studies
Electronic searches
See Appendix 1 for the search strategy used for the second update of this review.
For the third update of this review we revised the search strategies and searched the following electronic databases over all years:
-
the Cochrane Wounds Group Specialised Register (searched 5 December 2013);
-
the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2013, Issue 11);
-
The Database of Abstracts of Reviews of Effects (DARE) (The Cochrane Library 2013, Issue 11)
-
Ovid MEDLINE (1946 to November Week 3 2013);
-
Ovid EMBASE (1974 to 2013 Week 48);
-
Ovid MEDLINE (In‐Process & Other Non‐Indexed Citations 03 December, 2013);
-
EBSCO CINAHL (1982 to 20 December 2013).
We used the following search strategy in the Cochrane Central Register of Controlled Trials (CENTRAL):
#1 MeSH descriptor: [Surgical Wound Infection] explode all trees 2643
#2 MeSH descriptor: [Surgical Wound Dehiscence] explode all trees 346
#3 (surg* near/5 infect*):ti,ab,kw 4032
#4 (surg* near/5 wound*):ti,ab,kw 4427
#5 (surg* near/5 site*):ti,ab,kw 1016
#6 (surg* near/5 incision*):ti,ab,kw 1043
#7 (surg* near/5 dehisc*):ti,ab,kw 385
#8 (wound* near/5 dehisc*):ti,ab,kw 551
#9 (wound* near/5 infect*):ti,ab,kw 4440
#10 (wound near/5 disruption*):ti,ab,kw 42
#11 (wound next complication*):ti,ab,kw 420
#12 {or #1‐#11} 8341
#13 MeSH descriptor: [Breast Neoplasms] explode all trees and with qualifiers: [Surgery ‐ SU] 1627
#14 ((breast next cancer*) near/5 surg*):ti,ab,kw 756
#15 ((breast next neoplasm*) near/5 surg*):ti,ab,kw 919
#16 ((breast next carcinoma*) near/5 surg*):ti,ab,kw 47
#17 MeSH descriptor: [Mastectomy] explode all trees 1212
#18 MeSH descriptor: [Mammaplasty] explode all trees 218
#19 (mastectomy or mammaplasty):ti,ab,kw 2029
#20 MeSH descriptor: [Breast] explode all trees and with qualifiers: [Surgery ‐ SU] 97
#21 {or #13 #20} 3148
#22 MeSH descriptor: [Anti‐Bacterial Agents] explode all trees 8601
#23 (antibiotic* or clindamycin or cefuroxime or cefuroxim or ceftazidime or ofloxacin or levofloxacin or azithromycin or sulbactam or ampicillin or mezlocillin or oxacillin or vancomycin or tobramycin or ciprofloxacin) 21709
#24 #22 or #23 25029
#25 #12 and #21 and #24 33
The search strategies for Ovid MEDLINE, Ovid EMBASE and EBSCO CINAHL can be found in Appendix 2, Appendix 3 and Appendix 4 respectively. We combined the MEDLINE search with the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE: sensitivity‐ and precision‐maximising version (2008 revision); Ovid format (Lefebvre 2011). We combined the EMBASE and CINAHL searches with the trial filters developed by the Scottish Intercollegiate Guidelines Network (SIGN) (SIGN 2008b). We applied no language or date restrictions.
Searching other resources
In addition, we screened references in all articles found by the above search strategy for further studies. We contacted experts in the field and interest groups to try and obtain access to unpublished or ongoing work. We followed up conference proceedings and grey literature that was considered to be potentially eligible for inclusion by both authors by contacting the study authors for further information.
Data collection and analysis
Selection of studies
Two review authors independently examined the title and abstract of citations identified by the search. We obtained all reports of potentially eligible trials as full‐text articles and two review authors independently applied the inclusion criteria, resolving disagreements by discussion.
Data extraction and management
Two review authors independently extracted trial data using a specifically designed data extraction tool. We extracted data on study risk of bias (as defined below), antibiotic intervention (i.e. drug name, dose route, duration of treatment), setting, source of funding, length of follow‐up and outcomes.
Assessment of risk of bias in included studies
For this update two review authors independently assessed each included study using the Cochrane Collaboration tool for assessing risk of bias (Higgins 2011). This tool addresses six specific domains, namely sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting and other issues (e.g. extreme baseline imbalance) (see Appendix 5 for details of criteria on which the judgement was based). We assessed blinding and completeness of outcome data for each outcome separately. We completed a 'Risk of bias' table for each eligible study. We discussed any disagreement amongst all review authors to achieve a consensus. We presented assessment of risk of bias using a 'Risk of bias' summary figure, which presents all of the judgements in a cross‐tabulation of study by entry. This display of internal validity indicates the weight the reader may give the results of each study.
Assessment of heterogeneity
We assessed heterogeneity between study results using the I2 statistic (Higgins 2003). This examined the percentage of total variation across studies due to heterogeneity rather than chance. We considered values of I2 over 75% to indicate a high level of heterogeneity and would have resulted in a random‐effects model being applied or not pooling results.
Data synthesis
Where possible for each trial we calculated the risk ratio (RR) of infection and 95% confidence interval (95% CI), such that a risk ratio of greater than one indicates a higher risk of infection in the first group named. We reported continuous data (i.e. number of days to infection), where possible, as mean difference (MD) with 95% CI.
Methods of synthesising the studies were dependent on trial quality, design and heterogeneity. We explored both clinical and statistical heterogeneity. In the absence of clinical and statistical heterogeneity we applied a fixed‐effect model to pool data. Where synthesis was inappropriate we have presented a narrative overview.
Subgroup analysis and investigation of heterogeneity
As patients undergoing reoperation, reconstruction with or without implants and patients receiving neo‐adjuvant chemotherapy are documented as having a higher risk of infection (Tran 2003) we conducted a prespecified subgroup analysis of each of these factors where there were sufficient data available. The proposed subgroups were:
-
patients undergoing immediate reconstruction without implants (i.e. TRAM flap);
-
patients undergoing immediate reconstruction with implants (i.e. silicone or saline); and
-
patients who have received chemotherapy (excluding hormone treatment) prior to surgery.
Sensitivity analysis
Since there is evidence that the quality of allocation concealment particularly affects the result of studies (Schulz 1995), we examined the effect of excluding studies judged to have inadequate allocation concealment in a prespecified sensitivity analysis.
Results
Description of studies
Results of the search
We identified two further studies which met the inclusion criteria for this third update (Cabaluna 2012; Gulluoglu 2013) and excluded one study (Nicholas 2007). A further four abstracts which were considered to be multiple publications of the same study were awaiting assessment pending clarification from the study authors, these are now excluded as the study authors have not responded to our requests for further clarification (Kumar 2005). In total eleven studies met the inclusion criteria for this version of the review (Amland 1995; Bold 1998; Cabaluna 2012; Chow 2000; Gulluoglu 2013; Gupta 2000; Hall 2006; Paajanen 2009; Platt 1990; Wagman 1990; Yetim 2010).
Included studies
Participants
Of the eleven studies, eight (Bold 1998; Cabaluna 2012; Chow 2000; Gulluoglu 2013; Gupta 2000; Paajanen 2009; Wagman 1990; Yetim 2010) included women only, one almost entirely women (Hall 2006) and two (Amland 1995; Platt 1990) may have contained male and female breast surgery participants, although this could not be established from the data presented in the report or by contacting the authors. All of these studies included breast cancer patients as one of multiple patient groups being analysed. The studies were conducted between 1990 and 2013. Study sizes ranged between 44 (Yetim 2010) and 618 (Hall 2006). In total 2867 participants were included for meta‐analysis, 1439 in treatment arms and 1428 in control arms. These studies were conducted in the hospital setting, were single‐centre trials and were conducted in eight different countries. Country of origin for studies were: Australia (Hall 2006), Norway (Amland 1995), United States of America (Bold 1998; Platt 1990; Wagman 1990), Japan (Chow 2000), Finland (Paajanen 2009), The Philipines (Cabaluna 2012), Turkey (Gulluoglu 2013; Yetim 2010) and United Kingdom (Gupta 2000). All included studies had been published.
Types of surgery
Types of participants included patients undergoing plastic surgery (Amland 1995), herniorrhaphy or breast surgery (Platt 1990), axillary lymph node dissection for breast cancer (Bold 1998) and primary, non‐reconstructive surgery for breast cancer (Cabaluna 2012; Gulluoglu 2013; Gupta 2000; Hall 2006; Wagman 1990). One study (Chow 2000) was designed to look at inflammatory rather than infective episodes, however discrete data on infection rates were presented and therefore the study was eligible for inclusion. Two studies looked at axillary lymph node dissection as part of breast cancer treatment (Bold 1998; Gulluoglu 2013). One study (Paajanen 2009) looked at core needle biopsy and primary, non‐reconstructive surgery for breast cancer. The four remaining studies (Cabaluna 2012; Gupta 2000; Wagman 1990; Yetim 2010) looked solely at breast cancer patients receiving primary, non‐reconstructive surgery for breast cancer.
Length of follow‐up
Length of follow‐up from surgery ranged from five days (Chow 2000) to six months (Yetim 2010). One study (Gupta 2000) followed up patients between 10 and 14 days post discharge, but did not document the length of hospital stay for these patients.
Source of funding
Three studies (Amland 1995; Bold 1998; Platt 1990) stated that they were sponsored by a pharmaceutical company (Pfizer AS, Smith Kline & Beecham and Smith Kline & French laboratories, respectively). One study (Wagman 1990) was funded by the American Cancer Society and another (Paajanen 2009) by the Finnish cultural foundation. The source of funding was not reported in the other studies.
Antibiotics used
The antibiotics evaluated included:
-
azithromycin, single dose decided according to body weight, taken 8 pm the night before surgery (Amland 1995).
-
oral clarithromycin (500mg) for 10 doses (Chow 2000).
-
intravenous augmentin (1.2g) (Gupta 2000).
-
a single dose of intravenous flucloxacillin (2g) (Hall 2006).
-
cefazolin (six doses) (Wagman 1990).
-
a single dose of intravenous dicloxacillin (1g) (Paajanen 2009).
-
a single dose of cefonicid (1g) (Bold 1998; Platt 1990).
-
a single dose of intravenous cefazolin (1g) (Cabaluna 2012)
-
a single dose of intravenous ampicillin‐sulbactam (1g) (Gulluoglu 2013)
-
collagen plus gentamycin sulphate (200mg) inserted under the surgical wound prior to surgical closure (Yetim 2010).
Five studies (Bold 1998; Cabaluna 2012; Gulluoglu 2013; Platt 1990; Wagman 1990) are very similar in terms of length of follow‐up, choice of antibiotic and type of surgery undertaken. All studies had similar inclusion and exclusion criteria.
Immediate reconstruction with or without implants
No eligible studies evaluating prophylactic antibiotics for reconstructive surgery (with or without implants) were identified. Whilst three studies (Amland 1995; Baker 2000; Franchelli 1994) included patients undergoing reconstructive surgery, we excluded the studies following scrutiny. It was not clear that the patients had undergone surgery as part of breast cancer treatment (Amland 1995; Franchelli 1994) whilst one study was excluded because the research was addressing the needs of dental patients who have existing implants (Baker 2000).
Neo‐adjuvant chemotherapy
Two studies included patients who had received neo‐adjuvant chemotherapy (Bold 1998; Platt 1990).
Excluded studies
We excluded a total of 27 studies for the following reasons: two were reviews, 11 were not RCTs or quasi RCTs, one was a multiple drug comparison excluded as there was no placebo or control arm. One compared different regimens and doses, but had no control or placebo arm. We excluded one study as it could not be obtained from the British Library. Four abstracts were excluded as they may have been multiple publications of the same study and clarification could not be obtained from the authors. Five studies did not provide discrete data for breast cancer patients and two were found to be studies focused on other types of surgery (see Characteristics of excluded studies table).
Risk of bias in included studies
See 'Risk of bias' summary figure: Figure 1. Studies were judged to be at overall unclear or high risk of bias if they were described as unclear or at high risk of bias in the majority of the domains.
Methodological quality summary: review authors' judgements about each methodological quality item for each included study.
Allocation
Sequence generation
Eleven studies were described as RCTs, but only eight adequately generated the randomisation sequence by reporting the use of computer‐generated numbers or sequences of blocks of 10 and were at low risk of bias for this domain (Amland 1995; Bold 1998; Cabaluna 2012; Chow 2000; Gulluoglu 2013; Gupta 2000; Hall 2006; Wagman 1990). Three studies were classified as unclear as the authors failed to report the method by which randomisation sequence was generated (Paajanen 2009; Platt 1990; Yetim 2010).
Allocation concealment
Adequate allocation concealment was described for eight studies (Bold 1998; Cabaluna 2012; Gulluoglu 2013; Gupta 2000; Hall 2006; Paajanen 2009; Platt 1990; Wagman 1990) and they were therefore at low risk of bias. Three of these studies used the hospital pharmacy to generate the allocation for participants (Bold 1998; Platt 1990; Wagman 1990). One study stated that consecutive patients were allocated to group by a computer program (Chow 2000) however the method of allocation was not described and two studies used sealed opaque sequentially numbered envelopes (Gupta 2000; Hall 2006). One study reported the use of both hospital pharmacy as well as sealed, opaque, sequentially numbered envelopes (Paajanen 2009). In the remaining three studies the method of allocation concealment was not described (Amland 1995; Chow 2000; Yetim 2010) and therefore they are classified as at unclear risk of bias.
Blinding
Blinding (participants and treatment providers ‐ all outcomes)
Adequate blinding of participants and treatment providers was clearly reported in seven trials and therefore these were at low risk of bias (Amland 1995; Bold 1998; Cabaluna 2012; Gupta 2000; Paajanen 2009; Platt 1990; Wagman 1990). Three trials were classified as having inadequate blinding of both participants and treatment providers mainly because the control groups were not blinded as they were not given any treatment. (Chow 2000; Gulluoglu 2013; Yetim 2010). Whilst blinding was not specifically reported by Hall 2006 the antibiotic was administered after the induction of anaesthesia therefore it is possible that blinding was adequate but as there was no statement by the study authors we judged this to be unclear.
Blinding (outcome assessors ‐ all outcomes)
Nine studies described adequate blinding of outcome assessors and these were at low risk of measurement bias. All antibiotic compared with placebo studies stated that the key physician was unaware of patient allocation until data collection was complete (Amland 1995; Cabaluna 2012; Chow 2000; Gulluoglu 2013; Gupta 2000; Hall 2006; Paajanen 2009; Platt 1990; Wagman 1990). In one study it remained unclear if the outcome assessors were adequately blinded (Bold 1998) and in another (Yetim 2010) it was judged that the nature of the collagen implants under the wound site would unblind the outcome assessors.
Incomplete outcome data
In ten studies we judged the loss to follow‐up to be low, with similar numbers of participants lost in both control and treatment groups and valid reasons given (Amland 1995; Bold 1998; Cabaluna 2012; Chow 2000; Gulluoglu 2013; Gupta 2000; Hall 2006; Paajanen 2009; Platt 1990; Wagman 1990). In one study (Yetim 2010) the study was judged to be unclear for this domain because the authors stated that patients would be followed up for six months post surgery but only reported data at seven days.
We judged six studies to have undertaken an ITT analysis either because they explicitly reported this or because there were no drop outs from the study and the numbers of participants in the groups analysed at the final follow up of the study were the same as those randomised at the outset (Amland 1995; Cabaluna 2012; Gulluoglu 2013; Gupta 2000; Hall 2006; Paajanen 2009). Intention‐to‐treat analysis was not reported in the other five studies (Bold 1998; Chow 2000; Platt 1990; Wagman 1990; Yetim 2010).
Selective reporting
The study protocols were not available but all the important outcome measures stated in the methods section are reported in the results and therefore we judged this domain to be at low risk of bias for all studies.
Other potential sources of bias
We judged seven trials to be at low risk of bias for this domain because there was no imbalance in the baseline characteristics and the studies appeared free from other forms of bias (Cabaluna 2012; Chow 2000; Gupta 2000; Hall 2006; Paajanen 2009; Wagman 1990; Yetim 2010). In three remaining studies (Amland 1995; Bold 1998; Platt 1990) there was some funding reported from pharmaceutical companies but it was unclear the extent of the industry involvement and we have adopted a cautious approach by judging there to be a high risk of bias. One study (Gulluoglu 2013) highlighted a difference in the baseline characteristics, stating that patients in the control group had significantly more frequent open surgical biopsies than those in the prophylaxis group, as a result this study was judged at a high risk of bias for this domain.
Effects of interventions
Preoperative antibiotics compared with placebo or no antibiotic (ten trials, 2823 participants)
Seven studies (Amland 1995; Bold 1998; Cabaluna 2012; Gupta 2000; Paajanen 2009; Platt 1990; Wagman 1990) compared preoperative antibiotics with placebo. Three studies (Chow 2000; Hall 2006; Gulluoglu 2013) compared preoperative antibiotics with no treatment.
Incidence of postoperative wound infection
All ten trials recorded incidence of wound infection as an outcome. Results are presented as risk ratio (RR) where the risk ratio is the risk of infection in the intervention group divided by the risk of infection in the control group. A risk ratio of less than one indicates fewer infections in the intervention group. Two studies compared cefonicid with placebo (Bold 1998; Platt 1990), one compared azithromycin with placebo (Amland 1995), one compared augmentin with placebo (Gupta 2000), two compared cefazolin with placebo (Cabaluna 2012; Wagman 1990), one compared flucloxacillin with no treatment (Hall 2006), one compared ampicillin‐sublactam with no treatment (Gulluoglu 2013), one compared dicloxacillin with placebo (Paajanen 2009) and one compared clarithromycin with no treatment (Chow 2000). One study (Chow 2000) reported no infections in either group but in the remaining nine trials there were fewer infections in the groups treated with antibiotics, this was statistically significant in one study (Gulluoglu 2013) which considered overweight and obese breast cancer patients. In the other eight trials the results were not statistically significant (Analysis 1.1).
In addition pooling the two studies which compared cefonicid with placebo (Bold 1998; Platt 1990) showed a statistically significant reduction in infection associated with preoperative antibiotics (RR 0.56, 95% confidence interval (CI) 0.33 to 0.95) (Analysis 1.2). Pooling the two studies which compared cefazolin with placebo (Cabaluna 2012; Wagman 1990) showed no significant reduction in infection (RR 0.82, 95% CI 0.47 to 1.42)
We pooled all the trials using a fixed‐effect model as there was no evidence of heterogeneity (I2 = 0%). The pooled risk ratio shows that giving preoperative antibiotics significantly reduces the risk of wound infection after breast cancer surgery (RR 0.67, 95% CI 0.53 to 0.85) (Analysis 1.1). We carried out a sensitivity analysis to exclude one study (Chow 2000), as this study had short follow‐up, only compared antibiotics with no antibiotic and reported inflammation rather than infection as its primary outcome. Sensitivity analysis demonstrated no effect from removing Chow from the pooled analysis.
One study (Bold 1998) documented infection rates in those who received neo‐adjuvant chemotherapy; there was no statistically significant difference between the groups treated with cefonicid compared with the placebo group (RR 0.21, 95% CI, 0.01 to 4.12) (Analysis 1.4). Another study provided details of the number of patients who had previously received chemotherapy (Platt 1990) but did not report separate data on infection rates for these patients.
Since there is evidence that the quality of allocation concealment influences study results (Schulz 1995) we examined the effect of excluding studies judged to have inadequate allocation concealment in a prespecified sensitivity analysis. We judged two studies (Amland 1995; Chow 2000) to have unclear allocation concealment. Removing these studies from the meta‐analysis resulted in a pooled RR of 0.67 (95% CI 0.52 to 0.85) which was still significantly in favour of prophylactic antibiotics.
Cost of care
Two studies (Bold 1998; Gulluoglu 2013) reported the cost of care (Analysis 1.5). In one study (Bold 1998) the cost did not include the cost of operation or associated stay in hospital, but calculated the cost of any additional care or medications (i.e. antibiotic prophylaxis, postoperative antibiotics or wound care). They found that the average cost per patient was USD 49.80 in the antibiotic prophylaxis group and USD 364.87 in the control group. The majority of this cost difference was accounted for in patients readmitted to hospital for wound complications. In the other study (Gulluoglu 2013) little detail is given regarding what the cost included. The report states it calculated "SSI‐related treatment cost". The study reported that the control group had a significantly higher cost (USD 20.26) when compared with the treatment group (USD 8.48).
Adverse reactions to treatment
Six studies (Bold 1998; Gupta 2000; Hall 2006; Paajanen 2009; Platt 1990; Wagman 1990) reported adverse events (please refer to other data tables for adverse effects from antibiotics under antibiotic versus none or placebo) (Analysis 1.6). Seven studies reported there were no adverse events (Bold 1998; Cabaluna 2012; Gulluoglu 2013; Hall 2006; Paajanen 2009; Platt 1990; Wagman 1990) and one study (Gupta 2000) reported 41 adverse events (23%) in the treatment group and 33 (18%) in the control group, but no details were reported on type of adverse events. Although we contacted authors for clarification about the nature of these events, they did not reply. The remaining three studies made no mention of adverse events in the study report.
Death
No studies presented information on deaths.
Time to wound healing
No studies presented information on time to wound healing.
Delay in adjuvant cancer treatment caused by SSI
No studies presented information on delays in adjuvant cancer treatments due to SSI.
Time to onset of infection
Four studies reported time to onset of infection (Analysis 1.7), however they all provided the mean time to onset of infection and not a range and therefore we have not combined this in a meta‐analysis. Two studies (Gupta 2000; Platt 1990) documented similar mean times to onset of infection: 12 and 11 days in the intervention group and 11 and 10 days in the control group respectively. Wagman 1990 documented mean time of onset of infection of 17.7 days in the intervention group and 9.6 days in the control group. Gulluoglu 2013 states that the time to onset of infection was similar in both the control and intervention group. The study provides a table of data with a range of onset of infection times. The majority of infections (62%) were detected between 3 and 7 days postoperatively.
Readmission to hospital
Three studies (Bold 1998; Gulluoglu 2013; Platt 1990) reported readmission rates following treatment. In one study (Gulluoglu 2013) no patients required readmission to hospital. In the other two studies, due to heterogeneity (I2 = 70.8%) we did not pool results. One study (Bold 1998) reported statistically significantly lower readmission rates in those treated with prophylactic antibiotics (RR 0.11, 95% CI 0.01 to 0.88) (Analysis 1.8) and a shorter duration of readmission (placebo group 5.9 days, prophylaxis group 3.0 days); the other study found no reduction in readmission rates (RR 1.0, 95% CI 0.29 to 3.42) (Analysis 1.8). As such no conclusions can be drawn on this outcome.
Perioperative antibiotics compared with placebo or no antibiotic (one trial, 44 participants)
One study (Yetim 2010) compared perioperative antibiotics with no antibiotic.
Incidence of postoperative wound infection
This small study at overall high risk of bias presented wound infection as an outcome. The study compared gentamycin‐infused collagen (Gentacoll) inserted perioperatively with no antibiotic. There were no infections in the antibiotic‐treated group compared with four infections in the control group. Whilst the study author stated this to be significantly better in favour of the antibiotic group this was not replicated in our analysis (RR 0.11, 95% CI 0.01 to 1.95) (Analysis 2.1).
Cost of care
The study did not report the cost of care.
Adverse reactions to treatment
The study did not report any adverse reactions to treatment.
Deaths
The study did not report any information on deaths.
Delay in adjuvant cancer treatment caused by SSI
The study did not report any information on delays in adjuvant cancer treatment caused by SSI.
Time to onset of infection
The study did not report any information on the time to onset of infection.
Readmission to hospital
The study reports that two patients in the control group had to be readmitted for parenteral antibiotics as a result of wound infection. No patients in the antibiotic group were readmitted.
Discussion
This review found that preoperative antibiotics significantly reduce the risk of SSI in people undergoing surgery for breast cancer when compared with placebo or no treatment. Of the nine studies that reported data on adverse events only one found an increase of events in the intervention group, however detailed information about the nature of the adverse events was not given and adverse events were generally poorly reported across the included studies. In addition data for some of the outcomes, including deaths, delays in adjuvant cancer treatments, cost and readmissions were reported by few of the included studies. We found one study that evaluated perioperative antibiotics compared with no antibiotic; this small study found that perioperative antibiotics did not significantly reduce the incidence of SSI. We found no studies evaluating antibiotics for breast reconstruction at the time of the initial surgery.
We found one systematic review and meta analysis on the effects of antibiotic prophylaxis for breast cancer surgery which also concluded that prophylactic antibiotics reduce SSIs (Tejirian 2006). Another systematic review (Sajid 2012) also considered non breast cancer patients and concluded that prophylactic antibiotics reduce the incidence of SSIs.Two non‐systematic reviews (D'Amico 2001; Hall 2000) did not draw any firm conclusions. Similar systematic reviews in other types of clean surgery are scarce and have produced varied results (Gillespie 2010; Sanchez‐Manuel 2007).
We found only eleven studies with a total of 2867 participants; not many considering the number of people affected globally by breast cancer. Whilst it is encouraging that a statistically significant result was found it is possible that the numbers are not adequate to evaluate fully the risks and benefits of antibiotic prophylaxis for breast cancer surgery. In addition, although we found some trials that included people having immediate breast reconstruction we excluded them as we were unable to obtain discrete data specifically for breast cancer patients.
Whilst all efforts were made to obtain unpublished data, all the included studies had been published, therefore there is potential for publication bias. Testing for publication bias was not done due to the small number of studies obtained.
Although there was no statistical heterogeneity only four studies compared the same antibiotic using the same regimen (Bold 1998; Cabaluna 2012; Platt 1990; Wagman 1990), therefore we were unable to make conclusions about the most effective antibiotic and regimen. Other recent research has, however, recommended that antibiotic prophylaxis should generally be administered as a single dose preoperatively in order to maximise benefit and minimise adverse effects from treatment (SIGN 2008a).
In general the included trials were at low risk of bias for the main domains of sequence generation and allocation concealment. Three studies had unclear allocation concealment (Amland 1995; Chow 2000; Yetim 2010) and excluding these studies from the analysis made little difference to the result. One study (Chow 2000) had a follow‐up of only five days. As the average time to onset of infection in the other included studies ranged between 11 to 17.7 days it may have been appropriate to specify in the protocol a minimum length for follow‐up. However, excluding data from this study made no difference to the overall outcomes. However, we judged one study (Yetim 2010) which compared perioperative antibiotics with no antibiotic to be at high risk of bias overall due to a failure of blinding and insufficient information given regarding selection bias.
Overall, there are sufficient data from this review to suggest that antibiotic prophylaxis reduces surgical site infections in those undergoing non‐reconstructive breast cancer surgery. However further research would be required to establish the best protocols for practice.
Methodological quality summary: review authors' judgements about each methodological quality item for each included study.
Comparison 1 Preoperative antibiotics versus none or placebo, Outcome 1 Wound infections.
Comparison 1 Preoperative antibiotics versus none or placebo, Outcome 2 Wound infection cefonicid.
Comparison 1 Preoperative antibiotics versus none or placebo, Outcome 3 Wound infection cefazolin.
Comparison 1 Preoperative antibiotics versus none or placebo, Outcome 4 Infection rates in those who received neo‐adjuvant chemo.
| Study | Antibiotic | Placebo | Cost calculation |
| Bold 1998 | Total cost in the treatment group: USD 4382.57 | Total cost in the placebo group: USD 32,838.16 | Treatment costs were calculated from: cost of prophylaxis administration, charges for outpatient treatment and charges for inpatient treatment. |
| Gulluoglu 2013 | Average SSI related treatment cost: USD 8.48 | Average SSI related treatment cost: USD 20.26 | Little information is given regarding what is included in the SSI (surgical site infection) treatment cost. |
Comparison 1 Preoperative antibiotics versus none or placebo, Outcome 5 Cost of care.
| Study | Antibiotic | Control |
| Preoperative antibiotics versus placebo | ||
| Amland 1995 | Side effects considered by the investigator to be related to treatment were recorded in 4 of the 171 patients receiving the antibiotic (2.3%) | Side effects considered by the investigator to be related to treatment were present in 5 of the control group (3.0%) |
| Bold 1998 | Stated as: "no patient suffered a complication related to the antibiotic administration" | None recorded |
| Gupta 2000 | 41 adverse events noted, details not provided as to whether these were per patient or per event | 33 adverse events noted, details not provided as to whether these were per patient or per event |
| Paajanen 2009 | None recorded | None recorded |
| Platt 1990 | None recorded | None recorded |
| Wagman 1990 | Stated as: "no untoward reactions" | Stated as: "no untoward reactions" |
| Preoperative antibiotics versus none | ||
| Chow 2000 | No adverse events recorded | No adverse events recorded |
| Hall 2006 | Stated as 'no side effects observed' from the flucloxacillin | None stated |
Comparison 1 Preoperative antibiotics versus none or placebo, Outcome 6 Adverse effects from antibiotics.
| Study | Antibiotic | Control |
| Preoperative antibiotic versus placebo | ||
| Gulluoglu 2013 | 0 infections between 0 and 2 days, 6 (67%) infections detected between 3 and 7 days, 2 (22%) infections detected between 8 and 14 days, 1 (11%) infection detected between 15 and 30 days, no infections detected beyond 30 days | 1 (4%) infection detected between 0 and 2 days,15 (60%) infections detected between 3 and 7 days, 7 (28%) infections detected between 8 and 14 days, 2 (8%) infections detected between 15 and 30 days, no infections detected beyond 30 days |
| Gupta 2000 | Mean time to onset of infection 12 days | Mean time to onset of infection 11 days |
| Platt 1990 | Mean time to onset of infection 11 days | Mean time to onset of infection 10 days |
| Wagman 1990 | Mean time to onset of infection 17.7 days | |
Comparison 1 Preoperative antibiotics versus none or placebo, Outcome 7 Time to onset of infection.
Comparison 1 Preoperative antibiotics versus none or placebo, Outcome 8 Readmission to hospital.
Comparison 2 Perioperative antibiotics compared with no antibiotic, Outcome 1 Wound infection.
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Wound infections Show forest plot | 10 | 2823 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.67 [0.53, 0.85] |
| 1.1 Preoperative antibiotic versus placebo | 7 | 1784 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.74 [0.56, 0.97] |
| 1.2 Preoperative antibiotic versus none | 3 | 1039 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.48 [0.28, 0.81] |
| 2 Wound infection cefonicid Show forest plot | 2 | 747 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.56 [0.33, 0.95] |
| 3 Wound infection cefazolin Show forest plot | 2 | 336 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.82 [0.47, 1.42] |
| 4 Infection rates in those who received neo‐adjuvant chemo Show forest plot | 1 | 47 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.21 [0.01, 4.12] |
| 5 Cost of care Show forest plot | Other data | No numeric data | ||
| 6 Adverse effects from antibiotics Show forest plot | Other data | No numeric data | ||
| 6.1 Preoperative antibiotics versus placebo | Other data | No numeric data | ||
| 6.2 Preoperative antibiotics versus none | Other data | No numeric data | ||
| 7 Time to onset of infection Show forest plot | Other data | No numeric data | ||
| 7.1 Preoperative antibiotic versus placebo | Other data | No numeric data | ||
| 8 Readmission to hospital Show forest plot | 2 | Risk Ratio (M‐H, Random, 95% CI) | Subtotals only | |
| 8.1 Preoperative antibiotics versus placebo | 2 | 784 | Risk Ratio (M‐H, Random, 95% CI) | 0.39 [0.04, 3.49] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Wound infection Show forest plot | 1 | Risk Ratio (M‐H, Fixed, 95% CI) | Totals not selected | |
