Acute bacterial meningitis acquired in the community is associated with significant morbidity and mortality, even when appropriate treatment is administered.1^–3 Although current recommendations regarding most aspects of the treatment of this disease are evidence based, the issue of the appropriate duration of antibiotic treatment has not been subjected to adequate and conclusive research. As a consequence, relevant recommendations are primarily based on accumulated clinical experience as recorded in observational studies.4

Current treatment guidelines of community-acquired bacterial meningitis recommend a 7-day course of antibiotic therapy for Neisseria meningitidis, and a 7–14-day antibiotic course for either Haemophilus influenzae or Streptococcus pneumoniae, with regard to the most common aetiological pathogens.4 However, shortened courses of antibacterial therapy, if equally effective with standard duration therapy, could presumably have the advantage of fewer adverse events, reduced length of hospitalisation along with lower rates of hospitalisation-related complications, diminished cost of therapy, or improved quality of life.

The above presumed benefits of potentially effective shortened courses of antibacterial therapy for community-acquired meningitis would be of particular importance for several developing countries with limited healthcare resources, but a high baseline incidence of the disease, coupled with frequent epidemics.5 6 In such settings, the availability of appropriate drugs is frequently limited, and the provision of healthcare services is uneven and at times inappropriately curtailed.7 In more advanced healthcare systems, shortening the minimum required duration of the treatment for bacterial meningitis would also be of value. This is illustrated by the fact that outpatient parenteral therapy, after the completion of a short course of inhospital therapy, is increasingly being considered as an alternative treatment strategy for select patients.4 8 9

In this context, we sought to evaluate the effectiveness and safety of shortened courses of antibacterial therapy for community-acquired bacterial meningitis, by means of a meta-analysis of relevant randomised controlled trials (RCT).

METHODS

Data sources

The PubMed database (up to 7 November 2007), and the Cochrane Central Register of Controlled Trials (up to 13 November 2007) were searched for studies eligible for inclusion in the meta-analysis. Bibliographies of relevant articles were also hand-searched for the same purpose. The PubMed search strategy used was: (meningitis OR meningoencephalitis OR CNS infections) AND (antibiotics OR antimicrobials OR treatment) AND (short OR long OR day OR days OR course OR duration) AND clinical trials NOT review[pt] NOT case reports[pt]. The Cochrane Central Register of Controlled Trials was searched applying the term “meningitis”.

Study selection criteria

Retrieved studies were included in the meta-analysis if they: represented RCT; involved patients of any age group diagnosed with community-acquired bacterial meningitis; compared treatment with the same antibiotic agents, administered via the same route, in the same total daily dosage, but for a different duration (short-course treatment versus long-course treatment); the duration of short-course treatment did not exceed 7 days; the duration of the long-course treatment was at least 2 days longer than the corresponding short-course treatment. Abstracts presented in scientific conferences were not included in the meta-analysis, neither were studies written in languages other than English, Spanish, French, German, Italian and Greek.

Data extraction

Data extracted from each trial referred to study design, the characteristics of the included population, the type, dosage and duration of study treatments, the type of any concomitantly administered therapy, the number of included patients, the timing of assessment of outcomes, as well as data regarding the effectiveness and safety of compared treatments.

Definitions

Bacterial meningitis was defined by the presence of relevant clinical manifestations, such as fever, headache, altered mental status, or signs of meningeal irritation, accompanied by evidence of a bacterial aetiology of the disease consisting of positive cerebrospinal fluid (CSF) Gram stain, or culture, or latex agglutination test, or positive blood culture in conjunction with CSF abnormalities suggestive of bacterial meningitis (increased white blood cell count with neutrophilic predominance, reduced glucose and increased protein concentration). The intention-to-treat population included patients who were randomly assigned to receive study treatments. The per protocol population referred to patients who satisfied the eligibility criteria for evaluation for a specific outcome, which were used in each included trial.

The primary outcome of the meta-analysis was clinical success, termed as complete recovery or substantial improvement of symptoms and signs of meningitis, of the per protocol patients, at the end-of-therapy evaluation. Secondary outcomes included: all-cause inhospital mortality; persistence of CSF abnormalities at the end of therapy; duration of hospitalisation; total adverse events; patient withdrawals due to adverse events; secondary nosocomial infections; hearing impairment at the late follow-up evaluation; as well as long-term neurological complications at the late follow-up evaluation.

Quality assessment

Assessment of the methodological quality of each included trial was performed by the Jadad criteria.10 These criteria assess the presence and the quality of randomisation and blinding procedures, as well as the report of data regarding patient withdrawals from the study. The maximum score that can be assigned to a trial is 5. A score of at least 3 is considered to denote high methodological quality.

Statistical analysis

Pooled odds ratios (OR), for outcomes presented as dichotomous data and weighted mean differences, for outcomes presented as continuous data, along with corresponding 95% CI, were calculated using the Mantel–Haenszel fixed effect model,11 in case statistical heterogeneity between trials was not significant, or using the DerSimonian–Laird random-effects model12 in the opposite case. The evaluation of statistical heterogeneity between trials was performed by the χ² test.13 A p value of less than 0.10 for this test was defined as denoting the presence of significant between-trials heterogeneity. The presence of publication bias with regard to the sample size of included trials was assessed graphically with the funnel plot method.14 All statistical analyses were performed using the “RevMan analyses v1.0 for Windows” (The Cochrane Collaboration, Copenhagen, Denmark).

RESULTS

Characteristics of included trials

Five RCT were identified as qualifying for inclusion in this meta-analysis, after screening 473 articles retrieved from the search performed at the PubMed database and 569 articles retrieved from the search performed at the Cochrane Central Register of Controlled Trials.15^–19 The flow diagram of the detailed process of selection of trials for inclusion in the meta-analysis is presented in fig 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Flow diagram of the detailed process of selection of trials for inclusion in the meta-analysis.

The main characteristics of the five included RCT (study design, patient inclusion criteria, treatment regimens administered, concomitant therapy, number of randomly assigned patients, timing of the assessment of outcomes, methodological quality) are presented in table 1. The included trials were performed in various parts of the world. Two trials were performed in European countries,16 17 one was performed in North America,15 one in Latin America18 and one in Asia.19 All of the included RCT had an open-label study design. Three of the included RCT were assigned a Jadad score of 3,16 17 19 which is the maximum score that can be assigned to open-label RCT, whereas the remaining two were assigned a Jadad score of 2.15 18

Regarding the characteristics of the study population, all of the RCT identified as eligible for inclusion in the meta-analysis were performed on infants and children. In particular, the age of included patients varied between 3 weeks and 16 years. In two of the included RCT, the random assignment of the study population was performed at the end of the short course of therapy and referred to the continuation or not of treatment.18 19 In one of these two RCT, randomisation was limited to children with rapid initial recovery.18 In an additional RCT, patients with a positive repeat CSF culture after a minimum of 15 h of therapy were secondarily excluded.17 Four of the overall five included RCT exclusively randomly assigned patients with N meningitidis, S pneumoniae, or H influenzae as the causative pathogens.16^–19 In the remaining RCT, randomisation was limited to patients with disease caused by S pneumoniae, H influenzae and Streptococcus agalactiae.15

The antibiotic regimen that was used in all of the included RCT involved intravenous ceftriaxone. The duration of administration of ceftriaxone varied between 4 and 7 days in the short-course treatment arms and between 7 and 14 days in the long-course treatment arms of the included RCT. In two of the included RCT, the duration of ceftriaxone treatment in each of the compared treatment arms related to the type of causative pathogen. Data on concomitantly administered therapy were reported in two RCT. In one of these, intravenous dexamethasone was administered to a considerable percentage of patients.18 In the remaining RCT, no concomitant therapy was allowed.17

Clinical outcomes were determined at the end of therapy for each treatment arm, in all of the included RCT. The interval between the beginning of therapy and the determination of clinical outcomes was thus shorter in the short-course treatment arms compared with the long-course treatment arms. In two of the five overall included RCT, specific data regarding the outcome of clinical success were not available.15 18 Instead, data on the resolution of specific signs and symptoms were reported. Among these, resolution of fever was elected as a surrogate marker for clinical success and was included in the respective analysis. The timing of the late follow-up assessment varied among the included trials, the shorter corresponding to the day of discharge and the longer to 6 months after discharge.

Findings of the meta-analysis

The data extracted from each included RCT regarding the outcomes of this meta-analysis are presented in table 2.

Clinical success

No difference was found regarding clinical success at the end of therapy in patients with community-acquired bacterial meningitis treated with short-course compared with those treated with long-course antibiotic regimens (five RCT,15^–19 383 patients, fixed effect model (FEM), OR 1.24, 95% CI 0.73 to 2.11, fig 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Meta-analysis of the end-of-therapy clinical success in patients with bacterial meningitis treated with short-course versus long-course antibiotic regimens. Horizontal lines, 95% CI; squares, point-estimates; size of the squares, weight of the study in the meta-analysis; diamond, pooled odds ratio (OR) plus 95% CI; vertical line, “no difference” line between compared treatments; area to the right of the vertical line indicates higher clinical success for short-course regimens.

Mortality

Two deaths were reported in the trials included in the meta-analysis. One was attributed to hospital-acquired sepsis,19 and the other was attributed to bacteraemia as a result of meningitis. The latter occurred before the planned discontinuation of study treatment.15 A meaningful meta-analysis of these data could not be performed.

Persistence of CSF abnormalities

Relevant data were reported in one of the five included RCT.16 Five patients had persistent CSF pleocytosis at the time of completion of short-course therapy, whereas one patient had CSF pleocytosis at the completion of long-course therapy.

Duration of hospitalisation

The duration of hospitalisation was significantly shorter in the short-course treatment arms of two RCT,15 19 (137 patients, FEM, weighted mean difference −2.17 days, 95% CI −3.85 to −0.50). In the remaining RCT, either the minimum duration of hospitalisation was predefined by protocol (two RCT),17 18 or specific relevant data were not reported.16

Total adverse events

No difference was found regarding total adverse events in patients treated with short-course compared with long-course antibiotic regimens (two RCT,15 16 122 patients, FEM, OR 1.29, 95% CI 0.57 to 2.91).

Withdrawals due to adverse events were either not reported or were not observed in the included RCT.

Secondary nosocomial infections

No difference was found regarding the occurrence of secondary nosocomial infections in patients treated with short-course compared with long-course antibiotic regimens (two RCT,15 19 139 patients, random effects model, OR 0.45, 95% CI 0.05 to 3.71).

Hearing impairment

No difference was found regarding hearing impairment at the late follow-up evaluation in patients treated with short-course compared with long-course antibiotic regimens (four RCT,15 16 18 19 241 patients, FEM, OR 0.59, 95% CI 0.28 to 1.23, fig 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Meta-analysis of long-term neurological complications in patients with bacterial meningitis treated with short-course versus long-course antibiotic regimens. Horizontal lines, 95% CI; squares, point-estimates; size of the squares, weight of the study in the meta-analysis; diamond, pooled odds ratio (OR) plus 95% CI; vertical line, “no difference” line between compared treatments; area to the right of the vertical line indicates more neurological complications for short-course regimens.

Long-term neurological complications

No difference was found regarding neurological complications at the late follow-up evaluation in patients treated with short-course compared with long-course antibiotic regimens (five RCT,15^–19 367 patients, FEM, OR 0.60, 95% CI 0.29 to 1.27, fig 4).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Meta-analysis of long-term hearing impairment in patients with bacterial meningitis treated with short-course versus long-course antibiotic regimens. Horizontal lines, 95% CI; squares, point-estimates; size of the squares, weight of the study in the meta-analysis; diamond, pooled odds ratio (OR) plus 95% CI; vertical line, “no difference” line between compared treatments; area to the right of the vertical line indicates higher rates of hearing impairment for short-course regimens.

DISCUSSION

In this meta-analysis, which identified as eligible for inclusion only RCT performed in children, no difference could be demonstrated regarding the primary outcome of clinical success with a shorter course of antibiotic treatment for community-acquired bacterial meningitis compared with treatment for a longer duration. No difference could also be found between the compared treatment strategies regarding adverse events, secondary nosocomial infections, long-term neurological complications and hearing impairment. The total duration of hospitalisation, in trials in which this was not defined by protocol, was significantly shorter by approximately 2 days in patients who were allocated to short-course treatment. However, the amount or quality of the available data are not sufficient strongly to support the hypothesis that short-course therapy is equally effective compared with therapy of longer duration.

The findings of several trials, with a different methodology than those selected for inclusion in the meta-analysis, support the concept that shortened antibiotic therapy may be adequately effective for the treatment of bacterial meningitis. However, the majority of relevant data refer to meningococcal meningitis. In particular, a randomised, open label trial on epidemic meningococcal meningitis demonstrated the non-inferiority of treatment consisting of a single dose of ceftriaxone compared with standard treatment with a single dose of long-acting oily chloramphenicol.20 Moreover, a 2-day regimen of ceftriaxone was found comparable in clinical and microbiological terms to a 6-day regimen of penicillin G in a small randomised trial, although patients with severe disease in the ceftriaxone arm required further treatment.21 Small non-randomised trials have also shown favourable outcomes with a 4–5-day treatment of meningococcal meningitis using appropriate antibacterial agents.22 23 Similar findings have been demonstrated in trials of bacterial meningitis, not limited to meningococcal disease, which have evaluated antibacterial treatment administered for 4–7 days.24^–26

The presumed lack of inferiority of shorter courses of antibacterial therapy for meningitis, on the basis of the findings of this meta-analysis and the cumulative data from other supportive studies, could be attributed to the rapid bactericidal activity of third-generation cephalosporins, such as ceftriaxone and cefotaxime, which constitute the mainstay of empirical therapy for this entity. These agents have been shown to reach CSF concentrations that are many times higher than the minimum inhibitory concentration for most causative pathogens.24 27 Accordingly, sterilisation of the CSF can be documented during the first or the second day of therapy in most cases.21 24 28 In fact, in the RCT included in our meta-analysis that provided relevant data, early repeat CSF cultures were negative.15 17 Besides, it should be noted that inflammatory CSF changes,23 25 as well as certain clinical manifestations,23 29 might persist after the eradication of causative pathogens from the CSF and, therefore, these parameters might not be accurate markers of the effectiveness of therapy.

The findings of this meta-analysis should not be interpreted without the consideration of several potential limitations. First, this meta-analysis shares the common shortcomings of this type of study that refers to the presence of differences between the included trials regarding the characteristics of the included population and the treatments administered, as well as the criteria for the assessment of outcomes. Nevertheless, the statistical between-studies heterogeneity regarding most of the outcomes analysed was of a very low degree (figs 2–4). Another important limitation of this meta-analysis is the relatively small total sample size of the included RCT (minimum 52, maximum 100 per protocol patients). Consequently, the confidence intervals regarding the estimates of this meta-analysis are rather wide, particularly for specific outcomes for which data were not provided in all of the included studies. Of importance, the lowest bound of the 95% CI regarding the OR for clinical success with short-course antibiotic treatment is 0.73, which corresponds to an absolute reduction in the clinical success rate of approximately 5%. In our view, such a difference, if true, has clinical relevance regarding a critical disease such as bacterial meningitis.

It should also be mentioned that the open-label design of the included RCT might have allowed several types of bias to interfere with the assessment of outcomes. This may be reflected by the fact that the included RCT tended to show better patient outcomes in terms of hearing impairment or neurological complications for short-course compared with long-course antibiotic treatment, a finding that is rather counterintuitive. In addition, the two studies with the poorest methodological quality by the Jadad criteria (score of 2) tended to favour short-course treatment in terms of clinical success, as shown in fig 2.15 18 We performed, in this regard, a post-hoc sensitivity analysis, but did not find a significant difference between the latter trials and those with higher methodological quality (Jadad score of 3), regarding the assessment of clinical success (χ² test for subgroup differences, p = 0.14). Bias may have also affected the selection of the per protocol population that was evaluated for clinical success in each of the included RCT. We did not use, however, the intention-to-treat population as the denominator in the assessment of this outcome, because the differences between compared treatments would appear more conservative.

A major consideration in the extrapolation of the findings of this meta-analysis is that patients with severe disease or patients with potentially resistant pathogens may not represent appropriate candidates for shortened courses of antibacterial therapy. These two categories of patients have been rather underrepresented among the trials included in this meta-analysis. In particular, three out of five RCT excluded patients with specific characteristics of increased risk.16^–18 In addition, the relatively low overall mortality observed in the population of the RCT included in this meta-analysis indicates that mainly patients with favourable initial prognosis were included.

The findings of this meta-analysis may also not be applicable if dexamethasone is used as adjunct to antimicrobial therapy. Such a treatment strategy was not adequately evaluated in the included RCT. The potential suppression of immune responses caused by dexamethasone may warrant a relatively long duration of antimicrobial therapy for optimal microbiological outcomes. The same may also hold true for other conditions associated with impaired immunity, such as HIV infection or malnutrition,30 which can often be observed in children with meningitis in certain developing countries. However, the nutritional status (assessed as the percentage of expected body weight) did not correlate with patient outcome in one of the included trials, which provided relevant data.19 Furthermore, antimicrobial agents other than ceftriaxone, which was used in all of the studies included in this review, may differ substantially in their pharmacokinetic and pharmacodynamic profile. A longer duration of administration may therefore be required for some of these agents.

The various causative pathogens of bacterial meningitis may have substantially different patterns of disease pathogenesis and susceptibility to antibiotics. Due to a lack of specific relevant data, we could not separately assess the appropriate duration of therapy for each of the main aetiological pathogens of bacterial meningitis. Rather, we evaluated antimicrobial therapy of shorter duration than commonly used,30 in comparison with therapy of longer duration. It should be mentioned that findings similar to those presented herein referring to acute meningitis have also been demonstrated for other types of infectious diseases, such as acute bacterial sinusitis and community-acquired pneumonia, using the same methodology.31 32 Last but not least, findings of studies performed in developed countries may not directly relate to the actual clinical setting in developing countries, in which the issue of a shortened duration of therapy for bacterial meningitis is crucial.

Despite the aforementioned limitations, the value of this meta-analysis is that by pooling all the available data the issue of the appropriate duration of the antibiotic treatment of bacterial meningitis can be approached with greater confidence. The findings of this meta-analysis imply that paediatric patients without adverse prognostic factors (such as an important delay in the institution of appropriate therapy, severe clinical condition at presentation, or persistence of positive CSF culture), who show rapid initial clinical improvement, could be considered as candidates for an early discontinuation of antibiotic treatment. However, due to the lack of conclusive evidence, the approach on this issue should be rather conservative, given the hazardous nature of the disease. Of note, a multicentre randomised placebo controlled trial comparing 5 to 10 days of ceftriaxone treatment for bacterial meningitis in children (International Standard Randomised Controlled Trial Number Register ISRCTN38717320) may provide more insight into this issue.

In conclusion, the pooled available data did not show a difference in terms of the clinical effectiveness and safety of shortened (7 days or less) than standard antibiotic treatment for community-acquired bacterial meningitis, in infants and children, particularly for those without severe disease or potentially resistant causative pathogens. However, definitive conclusions cannot be drawn as the relatively limited available data cannot preclude that a difference in clinical terms between long and short-course antibiotic therapy for acute meningitis truly exists. Rather, the findings of this meta-analysis support further research on shortened antibiotic therapy for bacterial meningitis in appropriately selected patients.