Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs), published or unpublished, with allocation to interventions at the individual (Patient-RCT) or at the group level (Cluster-RCT), testing the efficacy and safety of antibiotic prophylaxis for the prevention of burn wound infections. Quasi-randomised studies will be excluded.
Types of participants
Studies involving people of any age or gender, with any type of burn injury to the epidermis, dermis, subcutaneous tissues, vessels, nerve, tendons, or bone admitted to any unit in the hospital setting or treated in an outpatient setting.
We will include studies regardless of the severity of the burn (determined by either clinical evaluation or objective assessment, or both) or the method of burn injury (e.g., chemical, scald, or flame burns). We will not exclude studies depending on the presence of inhalation injury as a co-morbidity. We will exclude studies that include mixed population, i.e. people with already infected wounds in addition to those without an infection unless the data can be presented separately.
Types of interventions
We will assess antibiotic prophylaxis compared with placebo, no treatment, usual care or an alternative intervention (for example: non pharmacological - isolation of the burn patient, surgical excision; or pharmacological, such as another antibiotic regimen (trials comparing different antibiotics or different antibiotic dosages, routes of administration, timings or duration of administration)).
Prophylaxis is defined as the administration of antibiotics to patients without a documented infection, regardless of the signs of systemic inflammation, with the intention of preventing infection in the wound and invasive infection, including systemic antibiotics given intravenously, orally, or by intramuscular injection, oral nonabsorbables antibiotics, or topical antibiotics (dressings, ointments, or by inhalation). Antibiotics can be given at any moment after admission ("general") or can be specifically connected to a surgical procedure ("perioperative"). We will not consider a minimum duration of the intervention or of the follow-up as an inclusion criteria.
Studies evaluating antibiotic impregnated catheters will be excluded, as will evaluations of ointments or dressings that contain antimicrobials (iodine, chlorhexidine), non-antibiotics or antifungals. Dressings for superficial burns with partial thickness have been evaluated in a previous review by Wasiak 2008, but the principal objective of that study was not the evaluation of antibiotic prophylaxis.
Types of outcome measures
Burn wound infection: Studies must report an objective measure of burn wound infection in order to be included in the review. Diagnosis should rely on clinical examination (burn wound appearance) and culture data, if possible. However, burn wound infections diagnosed only by clinical examination will be also eligible.
Infections in the burned patient: such as pneumonia, urinary tract infection, bacteremia or blood infection (sepsis), or central venous catheter-associated bloodstream infections. We will admit any metric for quantifying infections, such as incidence rate or incidence density rate.
Infection-related mortality: for example, mortality due to burn wound infection, sepsis, or another infection complication.
Adverse events: as considered by the study investigators to be related to the antibiotic prophylaxis, such as toxicity, allergies, antibiotic-associated diarrhoea due to the overgrowth of toxigenic strains of Clostridium difficile, etc.
Objective measures of wound healing rate: such as time to complete healing; proportion of wounds completely healed within a trial period; proportion of participants with completely healed wounds; or proportion of wounds partly healed in a specified time period.
Antibiotic resistance: defined as clinical infection or colonization caused by bacteria resistant to one or more of the antibiotics included in the prophylactic regimen (proportion or rate of isolates of a specific pathogen ).
All-cause mortality: We will try to analyse this outcome according to longest common time point of assessment among the included studies.
Length of hospital stay (LOS).
Studies will be eligible for inclusion even if they only report secondary outcomes, as these outcomes are relevant to patients.
Search methods for identification of studies
We will search the following electronic databases for potentially relevant studies (without any language or date of publication restriction):
Cochrane Wounds Group Specialised Register
The Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library latest Issue).
Ovid MEDLINE - 1950 to present
Ovid EMBASE - 1980 to present
EBSCO CINAHL - 1982 to present
The Cochrane Central Register of Controlled Trials (CENTRAL) will be searched using the following exploded MeSH headings and keywords:
#1 MeSH descriptor Burns explode all trees
#2 (burn or burns or burned or scald*):ti,ab,kw
#3 (thermal NEXT injur*):ti,ab,kw
#4 (#1 OR #2 OR #3)
#5 MeSH descriptor Anti-Bacterial Agents explode all trees
#6 MeSH descriptor Anti-Infective Agents, Local explode all trees
#7 (antibiotic* or amoxicillin or ampicillin* or bacitracin or cephalothin or cefazolin or cefotaxime or cefoperazone or ceftazidime or ceftriaxone or cefuroxime or chloramphenicol or ciprofloxacin or clarithromycin or clindamycin or cloxacillin or colistin or colymycin or erythromycin or flucloxacillin or furazolidone or "fusidic acid" or gentamicin or gramicidin or imipenem or "mafenide acetate" or mupirocin or natamycin or neomycin or nitrofurazone or oxacillin or penicillin or piperacillin or polymyxin or rifam* or "silver nitrate" or "silver sulfadiazine" or "sulfacetamide sodium" or tobramycin or amphotericin or tazocin or teicoplanin or tetracylcin or (trimethopri* NEXT sulfamethoxazole) or vancomycin):ti,ab,kw
#8 (#5 OR #6 OR #7)
#9 (#4 AND #8)
This strategy will be adapted to search Ovid MEDLINE, Ovid EMBASE and EBSCO CINAHL. The Ovid MEDLINE search will be combined with the Cochrane Highly Sensitive Search Strategy for identifying randomized trials in MEDLINE: sensitivity- and precision-maximizing version (2008 revision) (Lefebvre 2009). The EMBASE and CINAHL searches will be combined with the trial filters developed by the Scottish Intercollegiate Guidelines Network (SIGN 2009).
Websites such as www.controlled-trials.com or www.clinicaltrials.gov will be searched in order to find additional ongoing clinical trials.
Searching other resources
The references of all identified studies will be searched in order to find any further relevant trials.
Data collection and analysis
Selection of studies
At least two review authors (LB, CJ) will independently screen all titles and abstracts of studies identified by the search strategy to assess for eligibility. If a reliable decision cannot be made based on this information, the full text version will be obtained for further assessment.
The full text versions of all potentially eligible studies will be retrieved and at least one pair of review authors will assess the eligibility of each study against the inclusion criteria. We will detail in the table of excluded studies all studies that appear initially to meet our inclusion criteria, but on closer examination fail to, and we will document the reasons for exclusion. The two review authors will resolve any disagreement by discussion, if there is no consensus; they will consult a third review author.
Data extraction and management
Two review authors (LB, CJ) will independently extract data from included studies using a predefined data form. We will extract details of participants, setting, methods, intervention and control, outcome data, risk of bias and results. Any discrepancy will be resolved by consensus. If there is no consensus a third review author or the editorial base of the Cochrane Wounds Group will settle the discrepancies. When information regarding any of the above is unclear, we will attempt to contact authors of the original trial reports to provide further details.
Assessment of risk of bias in included studies
At least two review authors will independently assess the risk of bias of each included study. We will use the risk of bias tool designed by the Cochrane Collaboration (Higgins 2009a).
We will consider the following domains:
Sequence generation - we will describe for each included study the methods used to generate the allocation sequence in sufficient detail to allow an assessment of whether it should produce comparable groups.
Allocation concealment - we will describe for each included study the method used to conceal the allocation sequence in sufficient detail and determine whether intervention allocation could have been foreseen in advance of, or during recruitment, or changed after assignment.
Blinding - we will describe for each included study all the methods used, if any, to blind study participants and personnel from knowledge of which intervention a participant received. Where blinding is not possible, we will assess whether the lack of blinding was likely to have introduced bias. Blinding will be assessed separately for different outcomes or classes of outcomes. We will also describe if the outcome assessors are blind to intervention. Note: In some situations there may be partial blinding e.g. where outcomes are self-reported by unblinded participants but they are recorded by blinded personnel without knowledge of group assignment.
Incomplete outcome data - we will assess the completeness of outcome data for each main outcome, including attrition and exclusions from the analysis. We will state whether attrition and exclusions were reported, the numbers in each intervention group (compared with total randomized participants), reasons for attrition/exclusions where reported, and any re inclusions in analyses performed by the review authors.
Selective outcome reporting - where possible we will assess whether study reports are free of any suggestion of selective outcome reporting.
Additional sources of bias - for cluster-randomised trials, we will assess these additional sources of bias: recruitment bias; baseline imbalance in either clusters or individuals (patient); loss of clusters and incorrect analysis (Higgins 2009b, Section 16.3.2). For the trials where the unit of randomisation is the patient we will also assess the existence of additional sources of bias: Were baseline characteristics similar?
Assessments will be made for each main outcome (or class of outcomes). We will label each criterion as 'Yes' (meaning low risk of bias), 'No' (high risk of bias) or 'Unclear'. See Appendix 1 for details of criteria on which the judgements will be based. We will try to obtain this information from the trial reports but, if there is not enough information to make a judgement, we will write to the trial authors for clarification. Disagreements will be resolved by discussion and consensus, and by consulting a third review author (JLA or IS if necessary).
Two figures will be also included in the review: a ‘Risk of bias graph figure’ and a ‘Risk of bias summary figure’. We will assess the overall risk of bias for each outcome (or class of similar outcomes) within each study. Each outcome (or class of outcomes) will be defined as having a ‘low risk of bias’ only if it is at low risk of bias for all the domains; at ‘high risk of bias’ if it demonstrates high risk of bias for one or more of the domains; or at ‘unclear risk of bias’ if it demonstrates unclear risk of bias for at least one key domain without any of the other domains being described as ‘high risk of bias’.
Finally, we will incorporate the results of the risk of bias assessment into the review through systematic narrative description and commentary and we will explore the effect of the risk of bias in the meta-analysis by carrying out sensitivity analysis as detailed below.
Measures of treatment effect
We will report study results, organised by type of intervention and study design. Analyses will be stratified by antibiotic mode and intervention: systemic antibiotic prophylaxis (general); systemic antibiotic prophylaxis (perioperative); non-absorbable antibiotics, and topical antibiotics. Regimens including both systemic and non-absorbable or topical antibiotics will be included in the systemic category. All outcome effects will be shown with their associated 95% confidence intervals (CI). We will report, when possible, the risk ratio (RR) for dichotomous data (for example, incidence of patients with infection); for continuous data (for example, length of hospital stay) we will report mean difference (MD) and for time-to-event date (for example, time to healing) we will report hazard ratios (HR).
Dealing with missing data
Missing outcome data will be assessed for the included studies and reported. We will contact the authors of the primary studies for missing data and clarification of issues. If we do not obtain this data, it will be clearly documented on the data extraction form and the narrative of the review. We will explore the impact in the overall treatment effect of including studies with high levels of missing data (>20%) using a sensitivity analysis.
As far as possible, we will carry out analyses on an intention-to-treat basis for all dichotomous outcomes; i.e. we will attempt to include all participants randomised to each group in the analyses, irrespective of what happened subsequently, assuming that missing participants experienced a negative outcome. We will explore the impact of this assumption using a sensitivity analysis.
Assessment of heterogeneity
Where possible we will display graphically the results of clinically and methodologically comparable studies and will assess heterogeneity visually. We will also use the I² statistic (Higgins 2003) which describes the percentage total variation across studies that is due to heterogeneity rather than chance. We will judge the importance of the observed value of I² depending on the magnitude and direction of effects and the strength of evidence for heterogeneity (moderate to high heterogeneity will be defined as I² ≥ 50%) (Deeks 2009).
Assessment of reporting biases
If sufficient studies are found, we will assess publication bias by means of a funnel plot for each outcome (a simple scatter plot of the intervention effect estimates from individual studies against some measure of each study’s size or precision (Sterne 2009)). Funnel plot asymmetry will be assessed statistically. If there is evidence of asymmetry, publication bias will be considered as only one of a number of possible explanations.
The outcome measures from the individual trials will be combined in a meta-analysis to provide a pooled effect estimate if there are enough studies and if these studies are sufficiently similar.
Cluster-randomised trials will be combined with individually randomised trials in the same meta-analysis only if unit of analysis errors are not detected. If the data analysis is determined to have been performed incorrectly and sufficient information is available, an 'approximately correct analysis' will be performed for each cluster-RCT. If it is not possible, the results of the study will be reported as point estimates of the intervention effect without presentation of any statistical analysis (P values) or confidence intervals and they will not be included in the meta-analysis (Higgins 2009b). 'Unit of analysis error' is defined as taking place when in some studies individuals are assigned in groups, rather than in an individual manner (i.e., by healthcare centre, hospital, or community). When this is done, often the assignment unit is different from the analysis unit, i.e., if subjects are assigned by group but analysed individually (Whiting-O'Keefe 1984).
We will use a random-effects model to pool data, although we will assess in the sensitivity analysis the influence of a fixed-effect model. In the event that relevant statistical heterogeneity is detected (I² ≥50%) or if the meta-analysis is inappropriate for any other reason, we will present a narrative analysis of eligible studies, providing a descriptive presentation of the results, grouped by intervention and study design, with supporting tables.
We will perform the analyses using Review Manager 5 (RevMan 2008) a statistical package provided by the Cochrane Collaboration and we will present the results with 95% confidence intervals (CI).
Subgroup analysis and investigation of heterogeneity
Analyses will be conducted separately for the following groups:
Patient age: children (ages between 0 and 18 years) compared with adults (>18 years).
Severity of the burn: burns involving less than 20% total burned body surface area (TBSA) versus burns involving greater than 20% TBSA.
If there are sufficient studies the following sensitivity analyses will be undertaken:
We will repeat the meta-analysis to assess the effect of including only studies with allocation to interventions at the group level ('cluster designs').
We will assess the effect of including studies with high or unclear risk of bias (as defined above), excluding these trials in the sensitivity analysis.
We will assess the effect of missing data, excluding studies with high levels of missing data (>20%) in the sensitivity analysis.
We will assess the impact of imputation of missing data performing a sensitivity analysis without imputation (i.e. analyzing available data on the included papers).
We will assess the effect of the statistical model chosen, using a fixed-effect model in the sensitivity analysis.
We will do sensitivity analyses restricted to studies including patients with specific co morbidities (i.e. diabetes mellitus, respiratory or kidney failure).