Antibiotic prophylaxis for preventing meningitis in patients with basilar skull fractures
Abstract
Background
Basilar skull fractures predispose patients to meningitis because of the possible direct contact of bacteria in the paranasal sinuses, nasopharynx or middle ear with the central nervous system (CNS). Cerebrospinal fluid (CSF) leakage has been associated with a greater risk of contracting meningitis. Antibiotics are often given prophylactically, although their role in preventing bacterial meningitis has not been established.
Objectives
To evaluate the effectiveness of prophylactic antibiotics for preventing meningitis in patients with basilar skull fractures.
Search methods
We searched CENTRAL (2014, Issue 5), MEDLINE (1966 to June week 1, 2014), EMBASE (1974 to June 2014) and LILACS (1982 to June 2014). We also performed an electronic search of meeting proceedings from the American Association of Neurological Surgeons (1997 to September 2005) and handsearched the abstracts of meeting proceedings of the European Association of Neurosurgical Societies (1995, 1999 and 2003).
Selection criteria
Randomised controlled trials (RCTs) comparing any antibiotic versus placebo or no intervention. We also identified non‐RCTs to perform a separate meta‐analysis in order to compare results.
Data collection and analysis
Three review authors independently screened and selected trials, assessed risk of bias and extracted data. We sought clarification with trial authors when needed. We pooled risk ratios (RRs) for dichotomous data with their 95% confidence intervals (CIs) using a random‐effects model. We assessed the overall quality of evidence using the GRADE (Grades of Recommendation, Assessment, Development and Evaluation) approach.
Main results
In this update we did not identify any new trials for inclusion. We included five RCTs with 208 participants in the review and meta‐analysis. We also identified 17 non‐RCTs comparing different types of antibiotic prophylaxis with placebo or no intervention in patients with basilar skull fractures. Most trials presented insufficient methodological detail. All studies included meningitis in their primary outcome. When we evaluated the five included RCTs, there were no significant differences between antibiotic prophylaxis groups and control groups in terms of reduction of the frequency of meningitis, all‐cause mortality, meningitis‐related mortality and need for surgical correction in patients with CSF leakage. There were no reported adverse effects of antibiotic administration, although one of the five RCTs reported an induced change in the posterior nasopharyngeal flora towards potentially more pathogenic organisms resistant to the antibiotic regimen used in prophylaxis. We performed a subgroup analysis to evaluate the primary outcome in patients with and without CSF leakage. We also completed a meta‐analysis of all the identified controlled non‐RCTs (enrolling a total of 2168 patients), which produced results consistent with the randomised data from the included studies.
Using the GRADE approach, we assessed the quality of trials as moderate.
Authors' conclusions
Currently available evidence from RCTs does not support prophylactic antibiotic use in patients with basilar skull fractures, whether there is evidence of CSF leakage or not. Until more research is available, the effectiveness of antibiotics in patients with basilar skull fractures cannot be determined because studies published to date are flawed by biases. Large, appropriately designed RCTs are needed.
PICOs
Plain language summary
Antibiotics to prevent infection of the brain coverings (meningitis) in patients with basilar skull fracture
Review question
Is it beneficial for patients with basilar skull fractures to receive a course of intravenous antibiotics?
Background
Basilar skull fracture (7% to 15.8% of all skull fractures) places the central nervous system in contact with bacteria from the nose and throat and may be associated with cerebrospinal fluid leakage (occurring in 2% to 20.8% of patients). Blood or watery discharge from the nose or ears, bruising behind the ear or around the eyes, hearing loss, inability to perceive odours or facial asymmetry may lead physicians to the diagnosis of basilar skull fracture. Patients with a basilar skull fracture may develop meningitis and some doctors give antibiotics in an attempt to reduce this risk.
Study characteristics
This review examined five randomised controlled trials, comprising a total of 208 participants with basilar skull fracture, which compared those who received preventive antibiotic therapy with those who did not receive antibiotics, to establish how many participants developed meningitis. The evidence is current to June 2014.
Key results
The available data did not support the use of prophylactic antibiotics, as there is no proven benefit of such therapy. There was a possible adverse effect of increased susceptibility to infection with more pathogenic (disease‐causing) organisms. We suggest that research is needed to address this question, as there are too few studies available on this subject and they have overall design shortcomings and small combined numbers of participants studied.
Quality of the evidence
We ranked the evidence as being of moderate quality because it is based on randomised data, although with some methodological limitations in design that caused us to downgrade the quality of the trials.
Authors' conclusions
Summary of findings
| Antibiotic prophylaxis compared with placebo for preventing meningitis in BSF | ||||
| Patient or population: patients with a recent BSF independent of the presence or severity of CSF leakage Settings: in‐hospital care Intervention: antibiotic prophylaxis Comparison: placebo | ||||
| Outcomes | Relative effect | No. of participants | Quality of the evidence | Comments |
| Frequency of meningitis | OR 0.69 (0.29 to 1.61) | 208 (4) | ⊕⊕⊕⊝ | |
| All‐cause mortality | OR 1.68 (0.41 to 6.95) | 208 (4) | ⊕⊕⊕⊝ | |
| Meningitis‐related mortality | OR 1.03 (0.14 to 7.40) | 208 (4) | ⊕⊕⊕⊝ | |
| Need for surgical correction in patients with CSF leakage | Not estimable | 109 (1) | ⊕⊕⊝⊝ | |
| Non‐CNS infection | OR 0.61 (0.15 to 2.46) | 52 (1) | ⊕⊕⊕⊕ | |
| BSF: basilar skull fracture; CI: confidence interval;CNS: central nervous system; CSF: cerebrospinal fluid; OR: odds ratio | ||||
| GRADE Working Group grades of evidence | ||||
| 1Downgraded for limitations in study design as two of the four studies had unclear blinding of allocation and outcome assessment leading to possible selection and detection bias; also unclear risk of reporting bias in one study. 2 Downgraded this single study, with concurrent limitations in study design (non‐blinded outcome assessment, leading to potential detection bias). | ||||
Background
Description of the condition
The estimated incidence of basilar skull fracture from non‐penetrating head trauma varies between 7% and 15.8% of all skull fractures, with associated cerebrospinal fluid (CSF) leakage occurring in 2% to 20.8% of patients (Buchanan 2004). Clinical signs that may lead a physician to suspect a basilar skull fracture include CSF otorrhoea or rhinorrhoea, bilateral periorbital ecchymosis, Battle's sign, peripheral facial nerve palsy, haemotympanum or tympanic membrane perforation with blood in the external auditory canal, hearing loss, evidence of vestibular dysfunction and anosmia. High‐resolution bone computed tomographic (CT) scans have dramatically improved the radiological diagnosis of this type of fracture. Basilar skull fractures are of special significance because the dura mater may be torn adjacent to the fracture site, placing the central nervous system (CNS) in contact with bacteria from the paranasal sinuses, nasopharynx or middle ear. If the dura mater is torn, CSF leakage could occur. Basilar skull fractures will predispose the patient to meningitis. A greater associated risk has been reported when CSF leakage exists, in particular if it persists for more than seven days (Leech 1973).
Description of the intervention
The role of prophylactic antibiotics for preventing bacterial meningitis in patients with basilar skull fractures is controversial. Growing concern about the emergence of resistant organisms argues against their use. In addition, there are reports of a higher incidence of meningitis in patients with basilar skull fractures who have received prophylactic antibiotics (Choi 1996).
How the intervention might work
Chemoprophylaxis with antibiotics in basilar skull fractures may reduce the incidence of meningitis.
Why it is important to do this review
A meta‐analysis showed a statistically significant reduction in the incidence of meningitis with prophylactic antibiotic therapy for patients with post‐traumatic CSF leakage (Brodie 1997). Another meta‐analysis concluded that antibiotic prophylaxis after a basilar skull fracture does not appear to decrease the risk of meningitis, independent of whether or not CSF leakage has occurred (Villalobos 1998). These studies did not include an extensive review of the literature; both searched papers only until 1995 and 1996 respectively, and their conclusions were based mainly on retrospective and observational studies.
The inadequacies of these reviews and their conflicting conclusions led us to decide to search for and analyse evidence for the use of prophylactic antibiotics for preventing bacterial meningitis in patients with a basilar skull fracture.
Objectives
To evaluate the effectiveness of prophylactic antibiotics for preventing meningitis in patients with basilar skull fractures.
Methods
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs) comparing any antibiotic versus placebo or no intervention. We also identified non‐RCTs to perform a separate meta‐analysis in order to compare results.
Types of participants
Patients of any age with a recent basilar skull fracture, independent of the presence and severity of CSF leakage.
Types of interventions
Any antibiotic administered at the time of primary treatment of the basilar skull fracture compared with placebo or no antibiotic. We excluded trials comparing different antibiotics, different antibiotic dosages, different routes of administration, or differences in the timing or duration of administration.
Types of outcome measures
Primary outcomes
-
Frequency of meningitis: suspected clinically (fever, neck stiffness, deterioration of neurological status, headache) and confirmed by lumbar puncture (CSF analysis including biochemistry, Gram stains or bacteriological cultures (or both)).
Secondary outcomes
-
All‐cause mortality/meningitis‐related mortality.
-
Need for surgical correction in patients with CSF leakage.
-
Non‐CNS infection.
Search methods for identification of studies
Electronic searches
For this update, we searched the Cochrane Central Register of Controlled Trials (CENTRAL 2014, Issue 5) (accessed 18 June 2014), which contains the Cochrane Acute Respiratory Infections (ARI) Group's Specialised Register, MEDLINE (January 2011 to June week 1, 2014), EMBASE (January 2011 to June 2014) and LILACS (2011 to June 2014). Previous searches are described in Appendix 1.
We used the search strategy described in Appendix 2 to search MEDLINE and CENTRAL. We used no filter for the MEDLINE search as we wished to identify both randomised and non‐randomised studies and the number of search results was manageable without a filter. We adapted the search strategy for EMBASE (Appendix 3) and LILACS (Appendix 4). We applied no language or publication restrictions.
Searching other resources
We searched the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) and ClinicalTrials.gov for completed and ongoing trials (27 June 2014). We screened titles, keywords and abstracts of the citations downloaded from the electronic searches and obtained full copies of reports of potentially suitable trials for further assessment. The search strategy also included a search of the reference lists of identified trials and basilar skull fracture review articles, and personal communication with other researchers in the field. We handsearched abstracts of meeting proceedings from the European Association of Neurosurgical Societies (1995, 1999 and 2003). We were unable to search the latter source in this 2014 update (see Differences between protocol and review).
We contacted researchers active in the field, for information regarding unpublished trials. We also contacted authors of published trials for further information and unpublished data. We did not apply any language restrictions.
Data collection and analysis
Selection of studies
Three review authors (BR, JC, LP) independently assessed the studies identified by the search strategy, to identify potentially suitable trials for the review according to the criteria outlined above. We resolved disagreements by discussion with the fourth author (CS).
Data extraction and management
Three authors (BR, JC, LP) independently assessed the full papers for type of participants, type and dose of antibiotic used, methodological quality, number of patients excluded or lost to follow‐up and the outcome measures stated in the protocol. We recorded extracted data on a data collection form. We resolved disagreements by discussion.
Assessment of risk of bias in included studies
We investigated sources of bias. Two authors (LP, BR) independently assessed the global quality of included trials using the 'Risk of bias' assessment tool, as outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We resolved disagreements by discussion.
Unit of analysis issues
We performed statistical analysis using Review Manager software (RevMan 2014). We calculated the significance of any differences between ORs using a standard method (Egger 2001).
Dealing with missing data
We contacted the original trial authors whenever relevant missing data were detected.
Assessment of heterogeneity
We performed statistical analyses using the statistical software provided by the Cochrane Collaboration (RevMan 2014). We investigated statistical heterogeneity between trial results using the I2 statistic (Higgins 2002).
Assessment of reporting biases
We would have assessed publication bias according to the recommendations on testing for funnel plot asymmetry if there had been sufficient numbers of trials (more than 10) in any meta‐analysis (Sterne 2011). We would have examined possible causes if asymmetry had been identified.
Data synthesis
We reported the results of meta‐analysis as odds ratios (ORs) and 95% confidence intervals (CIs) for dichotomous outcomes.
Subgroup analysis and investigation of heterogeneity
We previously planned to investigate heterogeneity by undertaking a subgroup analysis of patients with or without CSF leakage in the event of uncovering significant heterogeneity.
Sensitivity analysis
We performed a meta‐analysis of all the controlled, non‐randomised studies identified in order to evaluate the consistency of the results of the main meta‐analysis.
Summarising and interpreting results
We used the GRADE (Grades of Recommendation, Assessment, Development and Evaluation) approach to interpret findings (Schünemann 2011) and used GRADE profiler (GRADEpro 2014) to import data from RevMan 2014 to create summary of findings Table for the main comparison. This table provides information on the quality of evidence from studies, the magnitude of effects of the interventions examined, and the sum of available data on all important outcomes from each study included in the comparison. The GRADE approach (Schünemann 2011) considers 'quality' to be a judgement of the extent to which we can be confident that the estimates of effect are correct. We downgraded evidence from randomised controlled studies that was initially graded as 'high quality', by one or two levels on each of five domains after full consideration of any limitations in the design of the studies, indirectness (or applicability) of the evidence, inconsistency, imprecision of the effect estimates, and the possibility of publication bias.
A GRADE quality level of 'high' reflects confidence that the true effect lies close to the estimate of the effect of a given outcome. A judgement of 'moderate' quality indicates that the true effect is likely to be close to the estimate of the effect, but acknowledges that it could be substantially different. Evidence of 'low' and 'very low' quality limits our confidence in the effect estimate (Balshem 2011).
We selected the following outcomes for summary of findings Table for the main comparison.
-
Frequency of meningitis.
-
All‐cause mortality.
-
Meningitis‐related mortality.
-
Need for surgical correction of CSF leakage.
-
Non‐CNS infection.
Results
Description of studies
Results of the search
In this 2014 update we retrieved a total of 794 records. The 2011 update retrieved a total of 168 new records (when duplicates were removed from searches of MEDLINE (28 records), EMBASE (118 records), CENTRAL (14 records) and LILACS (eight records)) (Figure 1). This 2014 update yielded the addition of six new excluded studies.
Study flow diagram.
This review identified five RCTs comparing prophylactic antibiotics in basilar skull fracture with placebo or no antibiotics (Demetriades 1992; Eftekhar 2004; Hoff 1976; Ignelzi 1975a; Klastersky 1976).
Included studies
Meningitis in patients with basilar skull fractures
We found five RCTs with available data comparing prophylactic antibiotics in basilar skull fractures with placebo or no antibiotics (Demetriades 1992; Eftekhar 2004; Hoff 1976; Ignelzi 1975a; Klastersky 1976) (see Characteristics of included studies). All these studies were single‐centre, conducted in South Africa, Iran, USA or Belgium, and were published between 1975 and 2004. All had a parallel design and were stated by the trial authors to be randomised, although the method of randomisation was not clearly described in any trial report.
All trials included participants with a clinical or radiological diagnosis of basilar skull fracture. Entry criteria did not differ considerably. Exceptions were the Hoff 1976 trial, in which CSF leakage was an exclusion criterion, and the Klastersky 1976 trial, in which the participants had to have evidence of CSF leakage to be included.
The primary outcome for all trials included the occurrence of meningitis. In three trials the primary outcome was a composite outcome that also included extracranial infection (wound sepsis, pneumonia, urinary tract infection), bacterial colonisation of bronchial secretions or urine, change in the posterior nasopharyngeal flora, or death from brain damage (Demetriades 1992; Ignelzi 1975a; Klastersky 1976). Criteria for these diagnoses were based on clinical grounds and further investigations and prophylactic medication were commenced as soon as the diagnosis of basilar skull fracture was made in all trials. None of the studies reported data on outcomes of safety and tolerability of prophylactic antibiotics.
Ignelzi 1975a performed a small controlled trial with 10 participants with basilar skull fractures that was included in a report of a larger retrospective study. The presence of CSF fistulae in these participants was not described and the participants were randomised to one group that received prophylactic ampicillin or cephalothin 1 g six‐hourly for 10 days or another group that did not. The Klastersky 1976 study performed a double‐blind controlled trial that enrolled 52 participants and compared five mega units of penicillin G given intravenously six‐hourly for a mean duration of 7.7 days; the placebo was given under identical conditions.
Hoff 1976 enrolled 160 participants assigned randomly and blindly to one of three groups: no antibiotic (group 1), 1.2 million units of intravenous penicillin daily for three days (group 2), or 20 million units of intravenous penicillin daily for three days (group 3). No cases of meningitis were found but the numbers of participants enrolled in each group were not provided. Although we have contacted the trial author, further information was no longer available.
Demetriades 1992 randomised 37 participants to three groups: no antibiotic (group A), 1 g intravenous ceftriaxone daily for three days (group B), or combined ampicillin (1 g intravenous six‐hourly)/sulphadiazine (0.5 g intravenous six‐hourly) (group C).
Eftekhar 2004 studied 109 participants with acute traumatic pneumocephalus verified by a CT scan, who were followed until occurrence of meningitis or at least for five days post‐trauma. They randomised the participants to one of two groups: the prophylactic antibiotic treatment given (PAT+) group, in which ceftriaxone was administered at a dose of 1 g twice a day for five days; and the prophylactic antibiotic treatment not given (PAT‐) group, in which ceftriaxone was not administered.
Overall, 368 participants were enrolled in these five studies. Two of them enrolled 73% of these participants (Eftekhar 2004; Hoff 1976). Since we could not access the number of participants included in each group of the Hoff 1976 trial, we could not include it in the meta‐analysis. We therefore analysed a total of 208 participants from four RCTs: 109 participants in the treatment group and 99 in the control group.
In three trials participants were well matched between the treatment and control arms for demographics, clinical status at admission and presence of rhinorrhoea or otorrhoea (Demetriades 1992; Eftekhar 2004; Klastersky 1976). The other trials did not describe the characteristics of the population included in each group (Hoff 1976; Ignelzi 1975a).
Two studies provided sufficient descriptions for withdrawals and dropouts to determine the number of participants in each treatment group entering and completing the trial (Demetriades 1992; Klastersky 1976).
Meningitis in participants with basilar skull fracture concerning the presence of CSF leakage
There was only one study in which the presence of CSF leakage was not specified (Ignelzi 1975a). CSF leakage was an exclusion criteria in the Hoff 1976 study. Traumatic rhinorrhoea or otorrhoea had to be present in the participants included in the Klastersky 1976 study. The other two trials included participants with and without CSF leakage (Demetriades 1992; Eftekhar 2004).
Excluded studies
The Characteristics of excluded studies table contains all studies that have been systematically reviewed. We excluded 23 studies. Of these, 15 were retrospective controlled studies (Ash 1992; Choi 1996; Clemenza 1995; Dagi 1983; Einhorn 1978; Eljamel 1993; Frazee 1988; Friedman 2001; Helling 1988; MacGee 1970; McGuirt 1995; Raskind 1965; Steidtmann 1997; Tos 1973; Zrebeet 1986), four were prospective observational studies with an historical control group (Gonzalez 1998; Ibrahim 2012; Ignelzi 1975b; Lauder 2009), three were author statements and opinion (Bellamy 2013; Prosser 2011; Sherif 2012), and one was a randomised controlled trial with both groups receiving antibiotics for orbital blow‐out fractures (Zix 2013). There was no specification about the presence of CSF leakage in seven of these studies (Ash 1992; Gonzalez 1998; Helling 1988; Ignelzi 1975b; Lauder 2009; Prosser 2011; Tos 1973). Eight included only participants with CSF leakage, either otorrhoea or rhinorrhoea (Clemenza 1995; Eljamel 1993; Friedman 2001; Ibrahim 2012; MacGee 1970; McGuirt 1995; Raskind 1965; Sherif 2012), and one excluded participants with CSF leak (Zix 2013).
The remaining six studies included participants with or without CSF leakage (Choi 1996; Dagi 1983; Einhorn 1978; Frazee 1988; Steidtmann 1997; Zrebeet 1986). Overall, 2168 participants were included in these 17 studies, in which 1141 participants were treated with antibiotics and 1027 participants were not.
Risk of bias in included studies
The overall risk of bias is presented graphically in Figure 2 and summarised in Figure 3. (See Characteristics of included studies).
'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.
Allocation
All studies stated that patients were randomised between the treatment and control groups. The precise method of randomisation and details of concealment of allocation were not explained in any trial. We considered the method of allocation to be unclear in all trials.
Blinding
Only one study was double‐blinded throughout, using interventions of identical appearance (antibiotics or placebo) (Klastersky 1976). The other studies were not placebo‐controlled and did not measure outcomes blindly. In addition, only two studies reported the number of and reasons for patients leaving the trials (Demetriades 1992; Klastersky 1976).
Data were analysed on a per‐protocol basis in all trials.
Incomplete outcome data
Missing data precluded several planned analyses in this systematic review. We were able partly to overcome this problem because we had access to further data from the original trial of Eftekhar (Eftekhar 2004), kindly provided by the author himself. We sought further information for some of the other studies, but without success. As previously stated, the number of patients in each group was not accessible in the Hoff 1976 study; although none of the 160 patients enrolled had meningitis, we could not include this trial in the meta‐analysis.
Other potential sources of bias
We identified no other potential sources of bias.
Effects of interventions
See: Summary of findings for the main comparison
We were able to perform a meta‐analysis with five of the randomised controlled trials (RCTs) included in this review. Since we primarily aimed to compare prophylactic antibiotics with no antibiotic or placebo in patients with basilar skull fractures and to identify the influence of cerebrospinal fluid (CSF) leakage on the frequency of meningitis in these patients, we performed a subgroup analysis of patients with and without CSF leakage. We tested statistical heterogeneity between trial results using the I2 statistic and found no evidence of heterogeneity in any of the outcomes measured (I2 statistic = 0%).
The efficacy outcomes did not show significant differences between treatment and control groups in any of the included trials, when considering either the total population or the subgroup of patients with CSF leakage. (See Summary of main results).
Primary outcome
1. Frequency of meningitis
We found no significant differences for this outcome (odds ratio (OR) 0.69; 95% confidence interval (CI) 0.29 to 1.61) (Analysis 1.2). In addition, we found no differences in the subgroups of patients with CSF leakage (OR 0.44; 95% CI 0.09 to 2.15) or without CSF leakage (OR 0.77; 95% CI 0.25 to 2.41) (Analysis 1.1 ‐ 1.1.1 and 1.1.2). Using GRADE, we graded the evidence as being of moderate quality since there was potential selection and detection bias in two of the four RCTs as well as unclear risk of reporting bias in one study (summary of findings Table for the main comparison).
Secondary outcomes
1. All‐cause mortality/meningitis‐related mortality
We accessed relevant data from the five trials. We found no significant differences for all‐cause mortality (OR 1.68; 95% CI 0.41 to 6.95) (Analysis 2.1) or for meningitis‐related mortality (OR 1.03; 95% CI 0.14 to 7.40) (Analysis 3.1). We rated the evidence using GRADE as moderate quality, since there was potential selection and detection bias in two of the four RCTs as well as unclear risk of reporting bias in one study (summary of findings Table for the main comparison).
2. Need for surgical correction in patients with CSF leakage
Only one study provided data for this secondary outcome and no participants in either treatment or control groups underwent surgical correction for CSF leakage in this trial (Eftekhar 2004) (Analysis 4.1). We rated the quality of evidence as low quality according to GRADE, since it pertained to a single study with potential detection bias (summary of findings Table for the main comparison).
3. Non‐central nervous system (CNS) infection
Only one study provided data for this outcome (Klastersky 1976). No significant differences were found (OR 0.61; 95% CI 0.15 to 2.46) (Analysis 5.1). We rated the evidence as high quality, since although extracted from a single RCT, there was little risk of bias and few concerns raised regarding inconsistency, indirectness or imprecision (summary of findings Table for the main comparison).
Meta‐analysis of controlled non‐randomised studies identified
In order to study the consistency of these results we performed a meta‐analysis of all the controlled non‐randomised studies identified and previously described in the Excluded studies section. Globally, these studies enrolled 2168 participants (treatment group 1141; control group 1027). Tests for heterogeneity were not statistically significant (Chi2 test, P value = 0.16; I2 statistic = 26%). Globally, the frequency of meningitis in the treatment group was 6.92% and in the control group 6.52% (P value = 0.65) (random‐effects model OR 1.13; 95% CI 0.67 to 1.88). Individually, only one study showed a significant difference favouring the treatment group (OR 0.47; 95% CI 0.25 to 0.88) (Eljamel 1993). This study contributed most to the results (weight 19.4%), but it did not impact significantly on the direction of the results. Additionally, we performed a subgroup analysis for patients with CSF leakage (529 participants in the treatment group and 260 in the control group) and without CSF leakage (334 participants in the treatment group and 292 in the control group). In five studies the presence of CSF leakage was not specified (278 participants in treatment groups and 475 in control groups) (Ash 1992; Gonzalez 1998; Helling 1988; Ignelzi 1975b; Tos 1973). The OR (random‐effects model) for participants with CSF leakage was 0.61 (95% CI 0.37 to 0.99) and for patients without CSF leakage it was 0.86 (95% CI 0.27 to 2.78). In the subgroup of patients for which no data were available regarding the presence of CSF leakage, the OR was 2.01 (95% CI 0.91 to 4.44).
We found no statistically significant differences for all‐cause mortality in the eight studies in which relevant data were available (Ash 1992; Einhorn 1978; Frazee 1988; Friedman 2001; Ignelzi 1975b; MacGee 1970; McGuirt 1995; Zrebeet 1986). These included 460 participants in the treatment group and 265 in the control group (random‐effects model OR 0.78; 95% CI 0.26 to 2.28). We found no significant differences for meningitis‐related mortality (random‐effects model OR 0.43; 95% CI 0.08 to 2.29).
Discussion
Summary of main results
Curiously, the frequency of meningitis in the Eftekhar 2004 trial was significantly higher than in the other trials. The diagnosis of meningitis was based on cerebrospinal fluid (CSF) analysis in participants with compatible clinical findings and was comparable with the other trials. However, Eftekhar 2004 included only the subset of patients with basilar skull fractures and pneumocephalus that is associated with a dural tear with an open communication with air in the paranasal sinuses, mastoid air cells or petrous temporal regions and the central nervous system (CNS). These participants with pneumocephalus might have had an additional risk factor for developing meningitis that may have been independent of CSF leakage. The same authors developed a trial with the intent to clarify this hypothesis but it was not completed and no data were available upon contacting the author (Eftekhar 2006). Further investigations are necessary to clarify this issue.
Given the current data, it is not possible to recommend the use of prophylactic antibiotics in patients with basilar skull fractures. Our results did not show that the administration of antibiotics had an effect on the frequency of meningitis. No significant difference was found in the subgroup of participants with CSF leakage, although there was a tendency to favour the treatment group. Again, no significant difference was found for all‐cause or meningitis‐related mortality. Although no significant differences were found, the confidence interval (CI) for all outcomes was wide and we could not exclude the possibility that antibiotic prophylaxis is either better or worse than the control. This is partially explained by the relatively small number of participants enrolled and the small number of events recorded.
The global results of the analysis of data extracted from the excluded studies are in agreement with the randomised data. Subgroup analysis within the excluded trials suggests a benefit from antibiotic prophylaxis in patients with CSF leakage. However, treatment interventions caused significantly more meningitis in the subgroup of patients without specification regarding CSF leakage status. These analyses should be read with caution since they are based mostly on retrospective studies and the data are not randomised. Additionally, the types of participants, interventions, diagnoses and outcome measures were significantly different between these studies. This makes the data difficult to interpret. Nevertheless, we thought it would be interesting to compare data from randomised controlled trials (RCTs) with non‐RCTs since non‐randomised studies tend to overestimate treatment effect size, which may be the case here.
According to the frequency of events in the treatment and control groups, and the relative risk for meningitis in the subgroup of patients with CSF leakage (0.64), a sample size of 798 participants was needed in order to show a statistically significant result between the two interventions, with a power of 90% and the probability of a type I error of 5%. This figure is similar when considering the data from the non‐randomised case‐controlled studies for the subgroup of patients with CSF leakage, for which the sample size necessary to show a significant result was 737 patients.
This is the first systematic review to study the effect of prophylactic antibiotics in basilar skull fractures. Based on the analysis of five RCTs, there is insufficient evidence to support or refute the use of antibiotics to prevent meningitis in patients with basilar skull fractures.
Overall completeness and applicability of evidence
There is no support for routine prophylactic antibiotics in all patients with a basilar skull fracture. Further RCTs are needed to assess the benefits and risks clearly.
Quality of the evidence
The quality of the evidence available to evaluate the use of prophylactic antibiotics in basilar skull fractures was indicated by the identification of only five RCTs that we considered suitable for this review. Even these had important methodological shortcomings. In general, the quality of the included trials was poor, as assessed by the 'Risk of bias' tool (Higgins 2011). All trials used a per‐protocol based analysis. For the Eftekhar 2004 trial we had access to unpublished data that allowed us to perform some comparisons. We were able to study a total of 208 participants.
We assessed the quality of evidence using the GRADE method (GRADEpro 2014). We judged the quality of evidence for the effect on frequency of meningitis, all‐cause mortality and meningitis‐related mortality as being of moderate quality due to potential selection, detection and reporting bias in two of the four studies. We assessed the evidence for the effect on need for surgical correction for CSF leakage to be of low quality, since it was extracted from a single study with potential detection bias. We judged the evidence on the effect of antibiotic prophylaxis on the frequency of non‐CNS infections as high quality, as there were few concerns regarding inconsistency, indirectness or imprecision, with low risk of bias of the RCT (See summary of findings Table for the main comparison).
Potential biases in the review process
We attempted to minimise publication bias by checking the reference lists of all related studies for further references and searching multiple databases with a comprehensive search strategy without any language restrictions; we did not identify any ongoing trials in three clinical trial registries. We believe we reduced other sources of bias by having three authors independently conducting the study selection, quality assessment and data extraction.
Agreements and disagreements with other studies or reviews
Despite the commonality of antibiotic prophylaxis in the treatment of basilar skull fractures, surprisingly no systematic review had previously specifically evaluated the efficacy of this treatment. We found two meta‐analyses with conflicting conclusions: Brodie concluded there is a benefit of antibiotic prophylaxis (reducing the incidence of meningitis in patients with post‐traumatic cerebrospinal fluid (CSF) leakage), whilst Villalobos found no decrease in meningitis in patients with basilar skull fractures (with or without CSF leakage) afforded antibiotic prophylaxis (Brodie 1997; Villalobos 1998). The findings of our systematic review are similar to those in the Villalobos meta‐analysis.
'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.
Comparison 1 Comparison of frequency of meningitis with antibiotic prophylaxis versus no antibiotic, Outcome 1 Frequency of meningitis by subgroup.
Comparison 1 Comparison of frequency of meningitis with antibiotic prophylaxis versus no antibiotic, Outcome 2 Frequency of meningitis.
Comparison 2 Comparison of all‐cause mortality with antibiotic prophylaxis versus no antibiotic, Outcome 1 All‐cause mortality.
Comparison 3 Comparison of meningitis related mortality with antibiotic prophylaxis versus no antibiotic, Outcome 1 Meningitis‐related mortality.
Comparison 4 Comparison of the need for surgical correction in patients with CSF leakage with antibiotic prophylaxis versus no antibiotic, Outcome 1 Need for surgical correction in patients with CSF leakage.
Comparison 5 Comparison of frequency of non‐CNS infection with antibiotic prophylaxis versus no antibiotic, Outcome 1 Non‐CNS infection.
| Antibiotic prophylaxis compared with placebo for preventing meningitis in BSF | ||||
| Patient or population: patients with a recent BSF independent of the presence or severity of CSF leakage Settings: in‐hospital care Intervention: antibiotic prophylaxis Comparison: placebo | ||||
| Outcomes | Relative effect | No. of participants | Quality of the evidence | Comments |
| Frequency of meningitis | OR 0.69 (0.29 to 1.61) | 208 (4) | ⊕⊕⊕⊝ | |
| All‐cause mortality | OR 1.68 (0.41 to 6.95) | 208 (4) | ⊕⊕⊕⊝ | |
| Meningitis‐related mortality | OR 1.03 (0.14 to 7.40) | 208 (4) | ⊕⊕⊕⊝ | |
| Need for surgical correction in patients with CSF leakage | Not estimable | 109 (1) | ⊕⊕⊝⊝ | |
| Non‐CNS infection | OR 0.61 (0.15 to 2.46) | 52 (1) | ⊕⊕⊕⊕ | |
| BSF: basilar skull fracture; CI: confidence interval;CNS: central nervous system; CSF: cerebrospinal fluid; OR: odds ratio | ||||
| GRADE Working Group grades of evidence | ||||
| 1Downgraded for limitations in study design as two of the four studies had unclear blinding of allocation and outcome assessment leading to possible selection and detection bias; also unclear risk of reporting bias in one study. 2 Downgraded this single study, with concurrent limitations in study design (non‐blinded outcome assessment, leading to potential detection bias). | ||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Frequency of meningitis by subgroup Show forest plot | 4 | 208 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.63 [0.25, 1.59] |
| 1.1 CSF leakage (rhinorrhoea or otorrhoea) | 3 | 92 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.44 [0.09, 2.15] |
| 1.2 No CSF leakage | 2 | 106 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.77 [0.25, 2.41] |
| 1.3 Presence of CSF leakage not specified | 1 | 10 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.0 [0.0, 0.0] |
| 2 Frequency of meningitis Show forest plot | 4 | 208 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.69 [0.29, 1.61] |
| 2.1 Frequency of meningitis | 4 | 208 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.69 [0.29, 1.61] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 All‐cause mortality Show forest plot | 4 | 208 | Odds Ratio (M‐H, Fixed, 95% CI) | 1.68 [0.41, 6.95] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Meningitis‐related mortality Show forest plot | 4 | 208 | Odds Ratio (M‐H, Fixed, 95% CI) | 1.03 [0.14, 7.40] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Need for surgical correction in patients with CSF leakage Show forest plot | 1 | 109 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.0 [0.0, 0.0] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Non‐CNS infection Show forest plot | 1 | Odds Ratio (M‐H, Fixed, 95% CI) | Totals not selected | |
| 1.1 CSF leakage (rhinorrhoea or otorrhoea) | 1 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.61 [0.15, 2.46] | |
