Osmotic therapies added to antibiotics for acute bacterial meningitis

Abstract Background Every day children and adults die from acute community‐acquired bacterial meningitis, particularly in low‐income countries, and survivors risk deafness, epilepsy and neurological disabilities. Osmotic therapies may attract extra‐vascular fluid and reduce cerebral oedema, and thus reduce death and improve neurological outcomes. This is an update of a Cochrane Review first published in 2013. Objectives To evaluate the effects of osmotic therapies added to antibiotics for acute bacterial meningitis in children and adults on mortality, deafness and neurological disability. Search methods We searched CENTRAL (2017, Issue 1), MEDLINE (1950 to 17 February 2017), Embase (1974 to 17 February 2017), CINAHL (1981 to 17 February 2017), LILACS (1982 to 17 February 2017) and registers of ongoing clinical trials (ClinicalTrials.com, WHO ICTRP) (21 February 2017). We also searched conference abstracts and contacted researchers in the field (up to 12 December 2015). Selection criteria Randomised controlled trials testing any osmotic therapy in adults or children with acute bacterial meningitis. Data collection and analysis Two review authors independently screened the search results and selected trials for inclusion. Results are presented using risk ratios (RR) and 95% confidence intervals (CI) and grouped according to whether the participants received steroids or not. We used the GRADE approach to assess the certainty of the evidence. Main results We included five trials with 1451 participants. Four trials evaluated glycerol against placebo, and one evaluated glycerol against 50% dextrose; in addition three trials evaluated dexamethasone and one trial evaluated acetaminophen (paracetamol) in a factorial design. Stratified analysis shows no effect modification with steroids; we present aggregate effect estimates. Compared to placebo, glycerol probably has little or no effect on death in people with bacterial meningitis (RR 1.08, 95% CI 0.90 to 1.30; 5 studies, 1272 participants; moderate‐certainty evidence), but may reduce neurological disability (RR 0.73, 95% CI 0.53 to 1.00; 5 studies, 1270 participants; low‐certainty evidence). Glycerol may have little or no effect on seizures during treatment for meningitis (RR 1.08, 95% CI 0.90 to 1.30; 4 studies, 1090 participants; low‐certainty evidence). Glycerol may reduce the risk of subsequent deafness (RR 0.64, 95% CI 0.44 to 0.93; 5 studies, 922 participants; low to moderate‐certainty evidence). Glycerol probably has little or no effect on gastrointestinal bleeding (RR 0.93, 95% CI 0.39 to 2.19; 3 studies, 607 participants; moderate‐certainty evidence). The evidence on nausea, vomiting and diarrhoea is uncertain (RR 1.09, 95% CI 0.81 to 1.47; 2 studies, 851 participants; very low‐certainty evidence). Authors' conclusions Glycerol was the only osmotic therapy evaluated, and data from trials to date have not demonstrated an effect on death. Glycerol may reduce neurological deficiency and deafness.


A B S T R A C T Background
Every day children and adults die from acute community-acquired bacterial meningitis, particularly in low-income countries, and survivors risk deafness, epilepsy and neurological disabilities. Osmotic therapies may attract extra-vascular fluid and reduce cerebral oedema, and thus reduce death and improve neurological outcomes. This is an update of a Cochrane Review first published in 2013.

Objectives
To evaluate the e ects of osmotic therapies added to antibiotics for acute bacterial meningitis in children and adults on mortality, deafness and neurological disability.

Search methods
We

Selection criteria
Randomised controlled trials testing any osmotic therapy in adults or children with acute bacterial meningitis.

Data collection and analysis
Two review authors independently screened the search results and selected trials for inclusion. Results are presented using risk ratios (RR) and 95% confidence intervals (CI) and grouped according to whether the participants received steroids or not. We used the GRADE approach to assess the certainty of the evidence.

Main results
We included five trials with 1451 participants. Four trials evaluated glycerol against placebo, and one evaluated glycerol against 50% dextrose; in addition three trials evaluated dexamethasone and one trial evaluated acetaminophen (paracetamol) in a factorial design. Stratified analysis shows no e ect modification with steroids; we present aggregate e ect estimates. Cochrane Database of Systematic Reviews

B A C K G R O U N D Description of the condition
Community-acquired acute bacterial meningitis is a devastating infection with associated rates of death and disability that have changed little over the last 10 to 15 years. In high-income countries, 5% to 30% of adult patients die, rising to 50% to 60% in low-income countries, despite highly e ective antibiotics against the causative pathogens ( Increasing understanding of the pathways of cerebral inflammation in meningitis has led several investigators to try treatments that aim to reduce brain oedema and inflammation and improve brain perfusion. The intervention most extensively tested in clinical trials has been corticosteroids. A Cochrane Review shows a mortality benefit in adults in Europe with meningitis due to S pneumoniae and an overall reduction in deafness in adults and children (Brouwer 2015). Another systematic review of individual patient data from five randomised studies suggests that the e ect of dexamethasone on outcomes for bacterial meningitis in these countries is limited to reducing the incidence of hearing loss in survivors (van de Beek 2010). A long-held concern exists over excessive fluids contributing to brain oedema; a further Cochrane Review suggests that judicious fluid resuscitation guided by the clinical condition is appropriate to maximise brain perfusion without contributing to brain oedema (Maconochie 2016).

Description of the intervention
Osmotic therapies work by increasing the concentration of the blood. They exert an osmotic pressure across a semi-permeable membrane (such as a cell wall or blood vessel lining in the brain), which draws water from the brain into the blood and reduces pressure in the brain. This is theoretically advantageous if brain swelling is causing reduction in brain function.
Osmotic therapies have long been used in acute brain trauma (BTF 2000), and their use has been postulated in other forms of acute brain injury, particularly stroke (Bereczki 2007; Yu 1992; Yu 1993) and cerebral malaria (Namutangula 2007; Okoromah 2011). Mannitol and hypertonic saline are the most commonly used osmotic therapies (Wakai 2013), but glycerol, sorbitol and sodium lactate have also been investigated (Righetti 2004;Stoll 1998). Details of all these therapies are reported in Table 1. Glycerol has been studied in animals with meningitis, where no e ect was noted.
Conclusions from these studies are limited by the applicability of animal models of meningitis, where set doses of pathogenic bacteria are introduced directly into the animal's central nervous system, to the complex host pathogen interactions in human disease (Blaser 2010; Schmidt 1998). The excellent safety profile of glycerol in previous studies (Righetti 2004), combined with its low cost and easy administration and availability, has led investigators to look for its e icacy as an adjuvant treatment in acute bacterial meningitis in both adults and children, particularly in low-income countries.

How the intervention might work
All osmotic therapies have slightly di erent and poorly understood mechanisms of action. The osmotic drug's mechanism of action causes dehydration of central nervous system (CNS) cells, lowering intracranial pressure (ICP). However this e ect may only be temporary and lead to a rebound phenomenon where cells subsequently draw in too much water, increasing the oedema. Mannitol has this mechanism of action but acts primarily by erythrocyte deformity through increases in intravascular water, allowing increased tissue oxygenation in the CNS. Mannitol produces a large diuresis through this e ect, which causes a reflex cerebral vasoconstriction, temporarily reducing ICP. However, there is a significant risk of subsequent rebound raised ICP and mannitol is now used sparingly due to this concern. The main mechanism of action of glycerol in humans is unknown but there are some data to suggest that the addition of glycerol in meningitis could potentially improve cerebral blood flow and metabolism (Mathew 1972;Meyer 1972). Glycerol also has a mild e ect on serum osmolality (Singhi 2008).
Hypertonic saline and sodium lactate appear to have direct osmotic actions on cells and they do not cause diuresis. These drugs may therefore be better than mannitol in reducing ICP (Ichai 2009). Osmotic diuretics such as mannitol and sorbitol could potentially also have a clinical benefit in meningitis through reduction in ICP but may risk volume depletion in the febrile patient. All osmotic therapies ideally require an intact blood brain barrier to exert their e ects. Bacterial meningitis causes disruption of the barrier due to intense inflammation in the subarachnoid space and therefore it cannot be assumed that osmotic therapies would be beneficial. Table 1 gives details of all the properties of currently available osmotic therapies. Cochrane Database of Systematic Reviews

O B J E C T I V E S
To evaluate the e ects of osmotic therapies added to antibiotics for acute bacterial meningitis in children and adults on mortality, deafness and neurological disability.

M E T H O D S
Criteria for considering studies for this review

Types of studies
Randomised controlled trials (RCTs).

Types of participants
Adults and children diagnosed with acute community-acquired bacterial meningitis, as defined by the trial authors, on the basis of cerebrospinal fluid (CSF) culture, white cell count, biochemical composition and clinical presentation.

Types of interventions
Intervention: osmotic therapy, including at least one of the following: orally administered glycerol, intravenous (IV) hypertonic saline, sodium lactate and osmotic diuretics including IV mannitol and sorbitol.
Control: standard IV therapy or matched placebo.
All participants received broad-spectrum intravenous antibiotic treatment.

Secondary outcomes
1. Residual neurological deficit at the end of the follow-up period, including focal neurological deficit, epilepsy and deafness. Deafness was defined as hearing loss greater than 40 decibels bilaterally. 2. Epilepsy/seizures. 3. Deafness (hearing loss greater than 40 decibels bilaterally). 4. Adverse e ects. We used the search terms described in Appendix 1 to search MEDLINE and CENTRAL. We combined the MEDLINE search strategy with the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE: sensitivity-maximising version (2008 revision); Ovid format (Lefebvre 2011). We adapted the search strategy to search Embase (Appendix 2), CINAHL (Appendix 3) and LILACS (Appendix 4).

Searching other resources
We searched the following clinical trials registers on 21 February 2017.
For previous versions of this review we also searched conference abstracts and contacted researchers in the field (to 12 December 2015).

Selection of studies
One author (EW) screened all search results (title and abstract) and selected relevant studies according to the review inclusion criteria. Two authors (EW, KA) screened all selected studies by reading the published full text to ensure each study met the inclusion criteria. The same two authors then agreed which studies were to be included in the review. We emailed trial authors to clarify duplication and study numbers.

Data extraction and management
Two review authors (EW, KA) independently extracted all data from the selected studies using a data extraction form. We discussed all trial data, which were then included only when the data matched those extracted by both review authors. We contacted one trial author regarding duplication and we excluded one study from the analysis as a result. No further discrepancies arose during data extraction. We entered data for analysis using RevMan 5.3 so ware (Review Manager 2014).

Assessment of risk of bias in included studies
The data extraction form included a 'Risk of bias' collection tool. Two review authors (EW, KA) independently judged the potential risk of bias for each included study as low, uncertain or high for the following parameters (Higgins 2011). Both review authors then discussed and agreed the final judgements. One review author (EW) synthesised these judgements into a standard 'Risk of bias'

Measures of treatment e ect
The primary outcome of this review was binary and the studies included were all RCTs, therefore we used the risk ratio (RR) as the most appropriate statistical tool to express the results of the Cochrane Database of Systematic Reviews treatment e ect in a meta-analysis. We displayed the results as forest plots.
All included studies had outcomes defined by the trial authors using standardised measurements. We counted hearing loss of greater than 40 decibels (dB) as significant where measured. If a formal neurological score was used to define neurological disability we used this. However, where only a description was given, we counted a described deficit that results in the participant not being able to work or attend school as significant. As the number of studies was small we were not able to analyse mortality by continental geographical area and resource setting as secondary outcomes, as planned in the protocol.
Due to the small number of studies retrieved, we were unable to group results for both primary and secondary outcomes by the follow-up period: acute phase, less than three months since inclusion in the study and longer-term up to one year of follow-up.

Unit of analysis issues
We did not anticipate any cluster-randomised trials on this topic. However, within the trials included, a four parallel-arm design was employed. We separated data into groups comparing the intervention alone with placebo, and the intervention plus a second intervention with the second intervention alone. These results are expressed in Analysis 1.1 and Analysis 1.4.

Dealing with missing data
We found some relevant data to be missing from Kilpi 1995, Sankar 2007 and Molyneux 2014. We contacted the authors for clarification or additional data. Molyneux provided information and data; we did not receive responses from Kilpi or Sankar.

Assessment of heterogeneity
We intended to use the I statistic and to explore explanations for heterogeneity by subgroup analysis as outlined in the protocol, but the data were insu icient.

Assessment of reporting biases
We assessed each study for reporting bias. Where it was suspected that selected results had been presented, we contacted the authors for clarification (see Dealing with missing data).

Data synthesis
We entered all extracted data into RevMan 5.3 (Review Manager 2014) and performed all analyses using this so ware. We expressed all results using forest plots. We used a fixed-e ect model for analysis and found minimal heterogeneity between the studies. We repeated the analyses using a random-e ects model where heterogeneity was detected. We present the results from the fixede ect model. Where disagreement in e ect size was determined between the fixed-e ect and random-e ects models, we present data from both models.

GRADE and 'Summary of findings' table
We created Summary of findings for the main comparison using the following outcomes: death, neurological disability, seizures, hearing loss and adverse e ects. We used the five GRADE considerations (study limitations, consistency of e ect, imprecision, indirectness and publication bias) to assess the quality of evidence as it related to the studies that contributed data to the meta-analyses for the prespecified outcomes (Atkins 2004). We used the methods and recommendations described in Section 8.5 and Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) using GRADEpro GDT so ware (GRADEpro GDT 2014). We justified all decisions to downgrade or upgrade the quality of studies in footnotes, and made comments to aid readers' understanding of the review where necessary.

Results of the search
We obtained 31 records from the 2017 update search; two duplicates were excluded. We assessed 29 records and could exclude 19 titles and abstracts. We obtained five full texts and excluded four. We included one new study in this update (Molyneux 2014).
We screened a total of 752 abstracts following the initial search in November 2010. Further records were screened following update searches in November 2012 (35 records from electronic databases), November 2014 (24 records) and February 2017 (24 records from electronic databases and five records from trials databases). This resulted in 840 screened abstracts in total over the history of this review including updates. See Figure 1.

Library
Trusted evidence. Informed decisions. Better health.
Cochrane Database of Systematic Reviews All included studies tested glycerol compared to matched placebo, with some studies including a dexamethasone arm and one study an acetaminophen (paracetamol) arm.

Interventions
All included studies used oral glycerol as the primary intervention.
The potential mechanism of action of glycerol is detailed in

Outcomes
Death was the primary outcome in all included studies.
In Peltola 2007, we noted di erent results in tables 2 and 3. As there appeared to be exclusions in table 3, we used the data from table 2, which appeared to be intention-to-treat.

Excluded studies
We excluded 13 studies (14 records). We found that 11 studies, which each used or mentioned the use of osmotic therapies, were not randomised controlled trials (RCTs) and these were excluded.
Reasons for exclusion were as follows: • five studies were not randomised trials; • four were review articles; • one was a systematic review (we screened the reference list and found no new studies to include in our review); and • one was a letter replying to an editorial comment.
We also excluded one study that included children with acute central nervous system infections and raised intracranial pressure (ICP) randomised to receive cerebral perfusion pressuretargeted therapy or intracranial pressure-targeted therapy (Kumar 2014); and one study that was a registered trial record (CTRI/2015/04/005668). The trial registry stated that it had been suspended and this was confirmed with the trialists. See Figure 1 for a flow diagram of the study selection process.

Studies awaiting classification
There are currently no studies awaiting classification.

Ongoing studies
We did not identify any ongoing studies.

Risk of bias in included studies
Risk of bias was mostly low; 70% of our judgements were of low risk of bias (see Figure 2). See Figure 3 for our judgements for each risk of bias item for each included study. Cochrane Database of Systematic Reviews

Figure 3. 'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study Allocation
The risk of bias was low for random sequence generation across all studies. Allocation concealment was adequately described for all but one study (Molyneux 2014), which we judged at unclear risk of bias (Figure 2 and Figure 3). We judged Peltola 2007 at unclear risk of allocation bias due to changes in the protocol that occurred during the study, a change from two dexamethasone to one placebo to one dexamethasone to one placebo, as reported by a meta-analysis of individual patient data testing dexamethasone compared to placebo for bacterial meningitis (van de Beek 2010).

Blinding
The risk of bias was low for blinding across four studies. We judged Kilpi 1995 at high risk of performance and detection bias, as no details of any concealment were given, so we assumed that the allocations were not blinded ( Figure 2). The review authors requested clarification from the authors of Kilpi 1995 but no response has been received.

Incomplete outcome data
Four studies reported complete data and we judged them to have a low risk of attrition bias. Data on two participants were missing from Kilpi 1995 and we judged this study to have a high risk of attrition bias.

Selective reporting
We judged Ajdukiewicz 2011 and Peltola 2007 to have a low risk of reporting bias as all data appeared to be presented clearly and completely. Kilpi 1995 presented selected data as there was significant attrition bias, so we judged it to have a high risk of reporting bias. We judged Sankar 2007 to have an unclear risk of reporting bias as neither adverse e ects nor time of stopping treatment were presented. Kilpi 1995 did not respond to our request for data on all enrolled trial participants.

Library
Trusted evidence. Informed decisions. Better health.

Other potential sources of bias
No trials were sponsored by pharmaceutical companies, nor were the authors declared to have conflicts of interest. Peltola 2007 was partly funded by a pharmaceutical company, which supplied the dexamethasone for the trial but not the glycerol, so we did not judge this to have a significant bias e ect on this analysis. Kilpi 1995 was also partially funded by a pharmaceutical company and we judged the risk of bias as unclear.

E ects of interventions
See: Summary of findings for the main comparison Glycerol for acute bacterial meningitis We included five trials, all evaluating glycerol. Four of the trials had four arms, which also compared glycerol plus dexamethasone with dexamethasone alone or glycerol plus paracetamol and paracetamol alone.
We carried out the initial analysis comparing participants who received glycerol or placebo only, or glycerol with paracetamol or placebo with paracetamol, labelled 'no steroids'. We carried out a subgroup analysis with the remaining trial participants who received either glycerol plus dexamethasone or dexamethasone plus placebo, labelled 'with steroids'. All trial participants received the antibiotic ce riaxone, so no antibiotic subgroup analysis was necessary. Due to the small number of included studies, a subgroup analysis of paediatric data was not possible.

All-cause mortality
In the adult study, there were more deaths in the glycerol group and this led to the study being stopped by the data monitoring committee (risk ratio (RR) 1.30, 95% confidence interval (CI) 1.04 to 1.62) (Ajdukiewicz 2011). None of the other studies detected harm with glycerol and the meta-analysis did not detect an e ect on mortality (RR 1.08, 95% CI 0.90 to 1.30, 1272 participants, 5 trials, I = 17%, Analysis 1.1, moderate-certainty evidence). The stratified analysis found no significant di erence whether dexamethasone was administered or not.

Residual neurological deficit at the end of the follow-up period
Overall, a slight reduction (54/644 cases) in neurological disability was reported in the glycerol group compared with the placebo group (77/626) (RR 0.73, 95% CI 0.53 to 1.00, 1270 participants, 5 trials, I = 50%, Analysis 1.2, low-certainty evidence). The e ect size was further reduced using the random-e ects model (RR 0.70, 95% CI 0.38 to 1.27). Little or no di erence was detected in the subgroup of participants who received steroids (RR 0.82, 95% CI 0.38 to 1.77, 419 participants, 3 trials, I = 25%).

Epilepsy/seizures
Convulsions on admission and during treatment were reported in all studies but none reported data for persistent epileptic seizures post discharge. In the adult study, the risk of seizures was higher with glycerol (RR 1.62, 95% CI 1.18 to 2.23) (Ajdukiewicz 2011). However, this was not found in the other studies and the metaanalysis did not detect a di erence (RR 1.08, 95% CI 0.90 to 1.30; 1090 participants, 4 trials, I = 54%, Analysis 1.3, low-certainty evidence).

Deafness
Fewer surviving participants given glycerol were reported as deaf at four to eight weeks of follow-up compared to placebo (RR 0.64, 95% CI 0.44 to 0.93; 5 trials, 922 participants, I = 7%, Analysis 1.4, moderate-certainty evidence). Using the random-e ects model, the estimate of the e ect size of glycerol on deafness was slightly lower (RR 0.67, 95% CI 0.44 to 1.01).

Adverse e ects
Neither glycerol nor dexamethasone were associated with significant adverse e ects in the included studies but systematic recording of adverse events was not reported. Only Ajdukiewicz 2011 reported on serious adverse events (SAEs). One SAE was reported each in glycerol and placebo arm participants, both considered possibly due to the study drug but the researchers reported that the most likely diagnosis for both participants (HIVpositive adults in Malawi) was a major cerebrovascular event secondary to meningitis.
Common adverse e ects were nausea and vomiting; there were also small numbers of cases of gastrointestinal bleeding.

Nausea, vomiting, diarrhoea
Two studies reported on nausea, vomiting or diarrhoea, with 221/426 events in the glycerol groups and 200/425 in the placebo groups. The meta-analysis did not detect a di erence (RR 1.09, 95% CI 0.81 to 1.47; 2 trials, 851 participants, I = 79%, Analysis 1.5, very low-certainty evidence) but heterogeneity was high. Peltola 2007, a study conducted with children in South America, reported more adverse events in the glycerol without steroids group (80/148) than in the placebo group (53/148) (RR 1.51, 95% CI 1.16 to 1.96; 296 participants).
Two studies reported results that could not be added to the metaanalysis. Sankar 2007, a study conducted with children in India, reported that the incidence of vomiting in the glycerol and nonglycerol groups was "similar", and Kilpi 1995, a trial with children conducted in Finland, reported a higher incidence of vomiting on days 2 and 3 in the glycerol and glycerol with steroid groups (day 2: 38%, day 3: 23%) than in the steroid and placebo groups (day 2: 14%, day 3: 4%) and that vomiting led to discontinuation of glycerol treatment in three cases.

Summary of main results
We included five trials evaluating glycerol in acute bacterial meningitis. Other osmotic diuretics, such as mannitol and hypertonic saline, have not yet been tested. Cochrane Database of Systematic Reviews the five included trials, glycerol was evaluated in a complex trial design including dexamethasone or acetaminophen. The review and meta-analysis did not detect an overall e ect of glycerol on mortality from acute bacterial meningitis in children and adults. However, in the only trial in adults, glycerol was associated with increased mortality. We assessed the quality of the evidence using GRADE criteria as low (GRADEpro GDT 2014; Summary of findings for the main comparison).

Glycerol was tested in adults and children with acute bacterial meningitis in a variety of di erent clinical settings and in four of
The meta-analysis of low-quality evidence suggested that glycerol may reduce hearing loss (Summary of findings for the main comparison).
The small numbers seen overall in the studies in children were not su icient to fully exclude the impact of dexamethasone, particularly on neurological disabilities and deafness in children, as this has been shown to be e ective elsewhere (van de Beek 2010).
The overall number of study participants in this review was small and a significant degree of bias was found to be present in Kilpi 1995. Analysis was mainly weighted on Ajdukiewicz 2011 and Peltola 2007, two large studies that were both well conducted, but limited in their population demographics and follow-up data. Data from Peltola 2007 have been subject to systematic reviews investigating the e ect of dexamethasone, and some methodological concerns were raised regarding the randomisation schedule (van de Beek 2010). As a result we have assigned this study an unclear risk of allocation bias.
Each study was undertaken in a very di erent environment and the population for each has its own particular issues. The HIV prevalence in Ajdukiewicz 2011 was 83.5% and the impact of this on mortality and other outcomes has not been measured and may be significant. Ajdukiewicz 2011 and Molyneux 2014 were conducted in a severely resource-limited environment in Malawi, with no access to advanced resuscitation or intensive care units (UNDP 2016). All other included studies were carried out in hospitals with intensive care units and paediatric specialist teams, which is not necessarily representative of most hospitals in low-income countries. This may introduce a degree of confounding, particularly regarding lower mortality rates in children.
Peltola 2007 was conducted at multiple sites and excluded participants who had received parenteral antibiotics but not oral antibiotics before the first dose of glycerol or dexamethasone or both glycerol and dexamethasone. The authors of Peltola 2007 did not include these data in the analysis, so it is unclear if prior antibiotic treatment had an e ect on outcomes, particularly deafness.
The doses and duration of glycerol used varied across the included studies, introducing further inconsistencies among studies (see Table 2). We were unable to control for this e ect in the analysis, which may have introduced further heterogeneity (Brouwer 2011; Saez-Llorens 2007). Prolonged use of osmotic agents, such as the four-day courses of glycerol used in Ajdukiewicz 2011, have been suggested to be harmful. Peltola 2007 and Sankar 2007 both used two-day courses due to this concern. However, most seizures and deaths in Ajdukiewicz 2011 occurred in the first two days, and therefore an association between mortality and glycerol duration is unlikely.
Di erent agents were used as placebo comparators in the studies. Ajdukiewicz 2011 used 50% dextrose, Peltola 2007, Sankar 2007 and Molyneux 2014 used carboxymethylcellulose, and Kilpi 1995 did not use a placebo agent. It may be argued that the placebo agents used were not wholly inert and may exert an independent osmotic action. All trial authors designed control agents that had a similar taste and texture to glycerol for concealment purposes, and whether any of the substances used exerted an independent osmotic action is untested. However, the higher mortality reported by Ajdukiewicz 2011 in the glycerol group suggests that glycerol had an action beyond any osmotic e ect exerted by the dextrose placebo, particularly as the glycerol was diluted in dextrose for some participants (Brouwer 2011).
The slight reduction in hearing loss observed suggests that glycerol may be acting to reduce oedema or improve cerebral blood flow in particular areas of the brain, either the nucleus or length of the vestibular-cochlear nerve (which is encased in a bony canal). There is some evidence to suggest that glycerol is required for bacterial metabolic pathways in the central nervous system (CNS) (Mahdi 2012). Genetic susceptibility to hearing loss following meningitis has been suggested and the presence of glycerol may attenuate the production of free radicals that may a ect CNS damage leading to hearing loss (van Well 2012). We selected greater than 40 dB as the cut-o for hearing loss to capture all clinically significant deficits; the e ect of glycerol on more severe hearing loss was not evaluated. Currently, there are no clear data showing the mechanistic e ects of glycerol on either hearing or mortality in humans and more research is needed. Experimental animal work has shown no e ect of glycerol in a bacterial meningitis model (Blaser 2010). The cause of increased mortality with glycerol in adults is unclear. Risk stratification of patients in that trial by disease severity showed that glycerol exerted harmful e ects on those patients with low predicted risk of death on admission (Wall 2017). It is possible that increased mortality from glycerol in these patients with a more intact blood-brain barrier may relate to enhanced virulence of pneumococci in the CNS in the presence of glycerol (Mahdi 2012), or harmful e ects of osmotic shi across the blood-brain barrier.
The use of dexamethasone did not have any impact on the outcomes studied when used with or without glycerol. Other larger reviews have found an impact of dexamethasone on the reduction of hearing loss in children with meningitis (van de Beek 2010). There were too few data available for analysis to inform a robust conclusion about the utility of dexamethasone for treatment of people with bacterial meningitis.

Overall completeness and applicability of evidence
This is an update of a Cochrane Review that examines the evidence for the use of osmotic therapies in acute bacterial meningitis. To date, the evidence is incomplete and unequivocal findings for the use of glycerol for people with meningitis could not be derived. Data from further studies are required, particularly in children, to assess the impact of glycerol on meningitis-induced hearing loss. There is no evidence testing any other osmotic therapy apart from glycerol for meningitis: data from clinical studies are required. The highquality evidence from Ajdukiewicz 2011 demonstrates harm from glycerol in adults with bacterial meningitis in Malawi and no further testing or clinical use of glycerol in adults is currently warranted.

Library
Trusted evidence. Informed decisions. Better health.

Quality of the evidence
We assessed the quality of evidence provided by this review using the GRADE methods (Summary of findings for the main comparison). We generally assessed the evidence as low-or very low-quality, which indicates that further research is very likely to change the estimates of e ect.
The main reasons for downgrading evidence quality were the small size of the trials, the low numbers of events and the substantial di erences between locations, sizes and participant populations studied in the included studies. Much larger trials would be necessary to prove or exclude significant benefits or harms.
We also downgraded the evidence quality for mortality and seizures due to inconsistency. The only trial in adults was stopped early due to small but statistically significant harm (Ajdukiewicz 2011), while four trials in children did not demonstrate statistically significant e ects.

Potential biases in the review process
Dr Katherine Ajdukiewicz is an author of this Cochrane Review and was the principal investigator for one of the included studies. To minimise bias she did not extract any data from her study to include in the analysis or perform any of the analysis.

Agreements and disagreements with other studies or reviews
There are no current systematic reviews examining glycerol or other osmotic agents for use in acute bacterial meningitis.

A U T H O R S ' C O N C L U S I O N S Implications for practice
There is no evidence to support the use of glycerol as adjunctive treatment for acute bacterial meningitis. Glycerol may have a small beneficial e ect on reducing deafness in surviving children but further data are needed. Overall, the evidence quality is low.

Implications for research
Trials testing other osmotic interventions in acute bacterial meningitis may be considered, particularly in children.

A C K N O W L E D G E M E N T S
We wish to thank Sarah Thorning for assistance with the search strategy and support and Dr David Sinclair for his help synthesising the 'Summary of findings'

Methods
Randomised controlled trial Placebo is potentially not completely inactive and 50% glucose may exert a neurological effect in meningitis

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias) Low risk "A randomisation number list in blocks of 12 was produced by an independent statistician using Stata version 9.0" Allocation concealment (selection bias) Low risk "Numbers and allocation were placed into sealed envelopes. Envelopes were opened sequentially by an independent person not involved in the clinical care or assessment of trial participants" Blinding (performance bias and detection bias)

Placebo
All placebo-controlled: carboxymethylcellulose (placebo for glycerol) and cocoa butter base suppository (placebo for paracetamol) Doses: glycerol 6 g/kg/day in 4 daily doses (maximum 2.5 mg/dose) for 2 days Acetaminophen rectal suppository 35 mg/kg first dose followed by 20 mg/kg 6-hourly for 42 hours Outcomes Primary outcome: • Survival to 6 months post discharge with no sequelae Secondary outcomes: • Sequelae that affect daily life (e.g. hemiplegia, deafness, blindness, seizures, global developmental delay) at 6 months • Severe incapacitating sequelae • Death

Notes
Source of funding: the Academy of Finland In the trial registration from 2008, the primary outcomes were: death, severe neurological sequelae and hearing loss; secondary outcomes were: audiological or neurological sequelae

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias) Low risk "...randomisation was computer generated in permuted blocks of 12" Allocation concealment (selection bias)

Unclear risk No information
Blinding (performance bias and detection bias) All outcomes

Low risk
No report in trial. Email from author that the trial was "double blind"

Molyneux 2014
Osmotic therapies added to antibiotics for acute bacterial meningitis (Review)

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias) Unclear risk "Stratified block randomisation took place in blocks of 20" Allocation concealment (selection bias) Low risk "All treatment kits were packaged according to the randomisation lists in Santiago, Chile. Saline and carboxymethylcellulose were the placebo preparations for dexamethasone and glycerol, respectively. The agents were provided in identical ampoules or bottles and were labelled only with a study code" Blinding (performance bias and detection bias) All outcomes Low risk "All treatment kits were packaged according to the randomisation lists in Santiago, Chile. Saline and carboxymethylcellulose were the placebo preparations for dexamethasone and glycerol, respectively. The agents were provided in identical ampoules or bottles and were labelled only with a study code" Incomplete outcome data (attrition bias)