Types of studies

Randomized controlled clinical trials (RCTs) of any design conducted in people undergoing orthognathic surgery. Quasi‐randomized and non‐randomized trials, observational studies, narrative reviews, commentaries and letters to editors will be excluded.

Types of participants

People of any age undergoing orthognathic surgery in any setting.

Types of interventions

Intervention: any type of antibiotic (penicillin and its derivatives, cephalosporins, etc.), with any regimen, or mode of administration (short‐term or long term; oral, endovenous or intramuscular; pre‐ or peri‐operative regimen).

Comparison: placebo, or another antibiotic, or another regimen of antibiotic.

Types of outcome measures

Electronic searches

We will search the following electronic databases to identify reports of relevant randomised clinical trials:

• The Cochrane Wounds Group Specialised Register;

• The Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library) (Latest issue);

• Ovid MEDLINE (1948 to present);

• Ovid EMBASE (1974 to present);

• EBSCO CINAHL (1982 to present)

We will use the following search strategy in The Cochrane Central Register of Controlled Trials (CENTRAL):

#1 MeSH descriptor Orthognathic Surgery explode all trees
#2 MeSH descriptor Orthognathic Surgical Procedures explode all trees
#3 MeSH descriptor Osteotomy, Le Fort explode all trees
#4 (orthognathic NEAR/5 surg*):ti,ab,kw
#5 ((maxillary NEXT osteotom*) or "Le Fort" or (mandibular NEAR/5 osteotom*) or (vertical NEXT ramus NEXT osteotom*) or genioplast*):ti,ab,kw
#6 (#1 OR #2 OR #3 OR #4 OR #5)
#7 MeSH descriptor Antibiotic Prophylaxis explode all trees
#8 MeSH descriptor Anti‐Bacterial Agents explode all trees
#9 (antibiotic* or cephalosporin* or cefazolin or cefuroxime or amoxicillin* or amoxycillin* or clindamicin or clindamycin or penicillin* or levofloxacin):ti,ab,kw
#10 (#7 OR #8 OR #9)
#11 (#6 AND #10)

We will adapt this strategy to search Ovid MEDLINE, Ovid EMBASE and EBSCO CINAHL. We will combine the Ovid MEDLINE search with the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE: sensitivity‐ and precision‐maximising version (2008 revision) (Lefebvre 2011). We will combine the EMBASE search with the Ovid EMBASE filter developed by the UK Cochrane Centre (Lefebvre 2011). We will combine the CINAHL searches with the trial filters developed by the Scottish Intercollegiate Guidelines Network (SIGN 2011). We will not restrict studies with respect to language, date of publication or study setting. If an article in a language other than English is identified, we will make an effort to have all relevant data extracted by a translator.

We will search the following ongoing trials databases:

• Current Controlled Trials (http://www.controlled‐trials.com/);

• ClinicalTrials.gov (http://www.clinicaltrials.gov/);

• WHO International Clinical Trials Registry Platform (ICTRP) (http://www.who.int/ictrp/en/)

Searching other resources

We will examine reference lists of relevant articles that were identified by the electronic searches for other pertinent articles to include in the review. We will also search in Scholar Google in order to detect unpublished or grey literature.

To account for any delay in indexing in the electronic databases mentioned above, we will also search the last six months of the following journals:

•  Journal of Oral and Maxillofacial Surgery;
•  International Journal of Oral and Maxillofacial Surgery;
•  British Journal of Oral and Maxillofacial Surgery;
•  Journal of Craniofacial Surgery;
•  Head & Neck: Journal for the Sciences & Specialties of the Head and Neck.

Finally, we will handsearch the last five years of the online abstract indexes of the conference proceedings of the annual meetings of the American Association of Oral and Maxillofacial Surgeons and the International Association for Dental Research.

Selection of studies

In a first screening, the title and the abstract of all potentially relevant articles will be listed and evaluated using a pre‐established selection criteria form. This process will be done independently by two review authors who will follow instructions especially designed for this stage and will be widely inclusive. All the articles selected for full text screening by either review author will be retrieved. The full text of all articles that potentially meet the eligibility criteria will be assessed independently by two review authors. Disagreements will be solved by consensus, and, if no consensus is reached, a third review author will act as arbiter. We plan to include a study flow diagram, as recommended by the PRISMA statement (Liberati 2009), to illustrate the results of all searching activity and the process of screening and selecting studies for inclusion in the review.

Data extraction and management

Two review authors will extract the data from all the selected studies independently, using a Microsoft® Excel Office for Mac 2011 standardized form created for this purpose. Discrepancies between the data will be reviewed by the two review authors and, if needed, a third review author will act as arbiter.

Assessment of risk of bias in included studies

The risk of bias of the included studies will be evaluated using the Cochrane Risk of Bias tool (Higgins 2011a). All the domains of this tool will be used (sequence generation, allocation concealment, blinding of participants, personnel and outcome, incomplete outcome data, selective outcome reporting, and other sources of bias), and the instructions published in the Cochrane Handbook will be followed for assessing each domain and for doing the evaluation of the overall risk of bias. The evaluation will be performed by two review authors independently, based on the full text of the trials. Any disagreements between the review authors will be discussed and consensus will be reached in order to classify the articles as having low, high or unclear risk of bias. Since it has been shown that blinding to author and/or affiliation of the study is not associated with the overall results of the evaluations (Moher 1999), the review authors will not be blinded to these characteristics of the trials.

Measures of treatment effect

The measures of treatment effect that will be used for evaluating the outcome in each trial will be:

• SSI: rates of infection at the surgical site will be analysed as dichotomous, risk ratio (RR) and its 95% confidence interval (CI) will be calculated. Trials should define how the presence of SSI was evaluated, and two review authors with clinical expertise will determine whether this definition is in accordance with the CDC definition.

• Systemic infection: it will be analysed as dichotomous, risk ratio (RR) and its 95% confidence interval (CI) will be calculated.

• Length of hospital stay (days): will be analysed as continuous, mean difference (MD) with 95% CI. If the data are presented as median and likely skewed we will not consider pooling.

• Adverse events: where trials report adverse events in sufficient detail (e.g. the number of participants who experienced at least one adverse event) we will analyse these data dichotomously. Where it is unclear whether the denominator is the total number of adverse events, or the number of participants, we will report these data narratively.

We will contact authors of trials that do not report outcome data appropriately. If we cannot gather enough data from the authors to pool the results; the results will be reported in tables and text.

Unit of analysis issues

No unit of analysis issues are anticipated. If cluster RCTs are included and enough data is available, we will pool the results of these trials with the ones from the other trials using the Adjusted Mantel‐Haenszel test (Donald 1987).

Dealing with missing data

If data for the outcomes of interest are missing, we will contact trial authors in order to obtain the information. A standard form will be designed to request trial authors to provide the specific data needed for the review, and the fields required will be highlighted in each case.

Each trial will be analysed with regard to missing data, and if data are judged to be missing at random (e.g., missing data does not seem to have an association with the intervention or any other patient characteristic) only the available data will be used. For dichotomous data, if data are judged not to be missing at random, we will perform a worst‐case scenario analysis by assuming that missing people from groups receiving antibiotic prophylaxis had the outcome of interest and missing people from groups receiving placebo or no treatment did not have the outcome of interest . For continuous data, if data are judged not to be missing at random, imputation of the missing data and accounting for the fact that these were imputed with uncertainty (e.g. multiple imputation, simple imputation methods with adjustment to the standard error) will be performed, as recommended in the Cochrane Handbook (Higgins 2011b)

Assessment of heterogeneity

The chi‐square test will be used to determine the presence of statistical heterogeneity, using a level of significance of 0.1. Quantification of inconsistency across the studies will be done using the I² statistic, and its interpretation will be based on The Cochrane Collaboration recommendations (Deeks 2011): i.e. an I² between 0% and 40% might be considered as unimportant heterogeneity amongst the trials; 30% to 60% might represent moderate heterogeneity; 50% to 90% might represent substantial heterogeneity; and 75% to 100% might represent considerable heterogeneity. Clinical heterogeneity will be assessed qualitatively considering patients, setting and intervention characteristics with the help of experts. Methodological heterogeneity will be evaluated using the domains of the risk of bias tool (Higgins 2011a). Exploration of heterogeneity will be based on subgroup analyses (detailed below).

Assessment of reporting biases

Efforts will be made to detect reporting biases, if possible, in accordance with the recommendations of The Cochrane Handbook for Systematic Reviews of Intervention (Sterne 2011). Outcome‐reporting biases will be explored by looking for published protocols of the trials included in the systematic review. If there are more than 10 included studies, publication bias will be explored using funnel plots for all outcomes, and tested using the Egger test (Egger 1997).

Data synthesis

We will present a narrative overview of the included trials. Where appropriate, we will present meta‐analyses of outcome data using RevMan 5.1. The decision to pool data in a meta‐analysis will depend on the availability of outcome data and assessment of between‐trial heterogeneity. For comparisons where there is no apparent clinical heterogeneity and the I² value is less than, or equal to, 40%, we will apply a fixed‐effect model. Where there is no apparent clinical heterogeneity and the I² value is greater than 40%, we will apply a random‐effects model.

The main findings will be presented using a Summary of Findings table (Rosenbaum 2010).

Subgroup analysis and investigation of heterogeneity

There are two factors that we anticipate could cause heterogeneity across the results of the trials and these will be explored through subgroup analyses. The number of osteotomies is the predictor of greatest interest. Based on previous findings (Chow 2007), the a priori hypothesis for this factor is that trials in which patients had a lower number of osteotomies will show a higher treatment effect in favour of antibiotic prophylaxis than trials in which patients underwent a higher number of osteotomies. Other sources of heterogeneity that will be investigated, if possible, include: mode of administration of the antibiotic, and length of surgery.

Sensitivity analysis

In a sensitivity analysis, we will exclude trials with high risk of bias. Since the allocation concealment is the most critical risk of bias domain for this review, a trial will be classified as high risk of bias if either the allocation concealment domain or two or more other domains are judged to have high risk of bias. The same criteria will be applied to classify a trial as having unclear risk of bias. Otherwise, the trial will be classified as having low risk of bias.

If there is need for data imputation, we will perform a sensitivity analysis using only the data available and compare its results with the results from the meta‐analysis with imputed data.

In addition, if a meta‐analysis shows a trial with an effect importantly different to the others (outlier), and this trial has clinical features that make it different from the others, a sensitivity analysis will be performed excluding it from the meta‐analysis.