on 26-May-2022 (Thu)

Normal CSF is crystal clear. Turbid CSF is due to hyperleuko- cytosis. The detection threshold for turbid CSF is approximately 200 leukocytes/mm 3

Cytological examination. The first step consists in counting the number of leukocytes and red blood cells (RBC) in CSF. Normal CSF is devoid of cells (< 5 cells/mm 3 ). Bac- teria may be detected during microscopic examination, in the absence of cellular reaction, even though very rarely. This test may only be performed with a minimum of 10 leukocytes/mm 3

In case of traumatic lumbar puncture, it is impossible to rely on the cytological analysis. The meningitis diagnosis is unlikely when the leukocytes/RBC ratio is < 1/700.

In the absence of a dedicated tube and if the total CSF volume is sufficient, 500 L (10 drops) should be stocked by the laboratory at −20 ◦ C before performing any test, for potential microbiological test using gene amplification (PCR)

Turbid CSF usually indicates cellular reaction of at least 200 leukocytes/mm 3 with a predominance of more or less altered neutrophils.

Even after antibiotic administration, bacterial meningitis is associated with a leukocyte count > 1000 cells/mm 3 in 87% of patients and > 100 cells/mm 3 in 99% of patients. Patients presenting with viral meningitis usually have < 100 cells/mm 3 .

There is a correlation between the number of neutrophils and bacterial inoculum: 67% of CSF samples with high cellularity have a bacterial inoculum > 10 3 CFU/ml (P < 0.01) [19].

the cytological analysis of CSF may be normal if the lum- bar puncture is performed early on. A mixed count or even lymphocytic predominance may be observed if the lumbar puncture is performed very early on, especially in case of N. meningitidis [20,21]

viral meningitis is usually associated with lymphocytic predominance. However, polymorphonuclear leukocyte pre- dominance or a mixed count may be observed in case of enteroviral meningitis when the lumbar puncture is performed early on

Direct microbiological examination following Gram staining. This is a simple and rapid examination. Its sensitivity is improved when the CSF is concentrated using centrifugation. Its sensitivity ranges from 60 to 97% for a specificity close to 100% in the absence of antibiotic therapy [22]. It depends on the causative agent and on transportation time to the laboratory, but also on the health professional’s experience in collecting samples. Its sensitivity rapidly drops to 40–60% — or even lower — when antibiotics are initiated before lumbar puncture [23].

A minimum inoculum of 10 5 bacteria/ml is required to be detected by Gram staining. Gram staining sensitivity is 25% when the inoculum is < 10 3 bacteria/ml, 60% when the inoculum is between 10 3 and 10 4 bacteria/ml, and 97% when the inoculum is > 10 5 bacteria/ml [19].

Gram staining is always performed, irrespective of the cytological and biochemical results.

in case of S. pneumoniae suspicion, Etest ® assays should be performed (MIC determination) at least for cefotaxime or ceftriaxone. When the MIC of the tested cephalosporin is > 0.5 mg/l, the MIC of the second cephalosporin should be subsequently determined.

when the Gram staining direct examination is positive, an antimicrobial susceptibility test should be performed to deter- mine the minimum inhibitory concentrations (MIC) (Etest ® method) directly from the samples if the remaining volume of CSF is sufficient and if the quantity of bacteria observed at direct examination indicates sufficient inoculum

Latex agglutination tests are no longer recommended.

Soluble antigen detection. The conventional latex agglutination test should be distinguished from the more recent immunochromatographic test.

A CSF immunochromatographic test is recommended to detect pneumococcal soluble antigens when the clinical signs and symptoms are highly indicative of bacterial meningitis even when the direct examination is negative

Immunochromatographic test (BINAX Now Streptococcus pneumoniae ® test). This rapid (15 minutes) and simple test detects the C-polysaccharide molecules present in the wall of all S. pneumoniae strains, irrespective of the serotypes

On the basis of studies including more than 2000 patients (more than half being children), of whom 156 presented with pneumococcal meningitis, the specificity of the immunochro- matographic test was 100% and its sensitivity 95–99%.

However, cross-reactivity has been reported espe- cially with Streptococcus mitis and Streptococcus oralis. The immunochromatographic test has the advantage of not being impacted by prior antibiotic therapy administration

A positive culture confirms the diagnosis, identifies the causative agent, and determines its susceptibility to antibiotics. This examination may be faulted because of: • antibiotic intake before lumbar puncture performance; • the conditions and time required for transporting the sample to the laboratory, which may not be compatible with the survival of particularly fragile bacteria; • a very low bacterial inoculum

Ceftriaxone MICs should be determined in case of reduced susceptibility to cefotaxime, or cefotaxime MICs should be determined in case of reduced susceptibility to ceftriaxone because the MICs of both of these molecules are sometimes different

Accurate determination of MICs is required for amoxicillin and for one of the two 3GCs (cefotaxime or ceftriaxone). The use of strips impregnated with a predefined concentration gradient of antibiotics is recommended (Etest ® ).

Tests targeting a specific bacterial or viral agent should be distinguished from those aiming at detecting the presence of any bacterial DNA (universal PCR) from a biological speci- men.

Two amplification techniques have been developed and val- idated for the main microorganisms responsible for meningitis (multiplex PCR and real-time PCR)

When bacterial meningitis is highly suspected and when the direct examination is negative, physicians should perform the following before culture result availability whenever possible meningococcal PCR, pneumococcal PCR, and Listeria PCR (when patients have risk factors for the latter infection); • or universal PCR

When the direct examination is positive and the culture is negative at 24 hours, these PCR tests are also recommended

When the suspicion of a bacterial etiology is low, a PCR test is recommended in infants and children to detect an enterovirus (Genexpert ® test). Considering the high sensitivity (86–100%) and specificity (92–100%) of this test, the high prevalence of enteroviruses in pediatric acute meningitis, and the rapid time to results (two hours), a positive enterovirus PCR test rules out the need for a bacterial PCR and allows for discontinuing the antibiotic therapy if initiated [28]. However, several epidemic enteroviruses in infants such as parechoviruses, cannot be iden- tified with this technique and a specific PCR test should be performed

Meningococcal blood PCR performed on EDTA and/or skin biopsy of purpuric lesions in case of purpura fulminans suspicion, allows for confirming the diagnosis when meningo- coccemia is suspected. Blood PCR is, however, useless when performed more than 24 hours after treatment initiation

Blood cultures. At least one pair of blood culture bottles should be inoculated. A second blood culture may be performed in adults just before initiating the antibiotic therapy

Blood cultures are positive in 50–75% of cases even when the CSF culture is negative

An association between CSF bacterial load and blood inoculum has been demonstrated.

Skin biopsy. Patients presenting with purpura skin lesions should have a skin biopsy performed, even more so when the antibiotic therapy has already been initiated or when the CSF direct examination is negative or has not been performed. The causative agent may be isolated from culture of this type of sample in 60 to 80% of cases (most frequently N. meningitidis). N. meningitidis persists in skin lesions for at least 24 hours after antibiotic therapy initiation [29]

CSF glucose levels. CSF glucose levels should be interpreted according to blood glucose levels, which samples should be taken at the same time. Normal CSF glucose levels are usually two thirds (66%) of those of blood glucose. CSF glucose levels of patients presenting with bacterial meningitis are < 40% of those of blood glucose (sensitivity 80%, specificity 98% [30]).

CSF protein level. High CSF protein levels are significantly associated with bacterial meningitis. The CSF pro- tein level threshold associated with bacterial meningitis ranges from 0.5 to 1.2 g/l.

CSF lactate levels. The 2008 consensus confer- ence recommended routine measurement of CSF lactate levels for differentiating bacterial from viral meningitis using the 3.5 mmol/l threshold; CSF lactate levels above that threshold are indicative of bacterial meningitis in adults. Since 2008, two meta-analyses of 25 and 33 studies have been published on that topic in 2010 and 2011, respectively (22 studies similar to both meta-analyses) [31,32]. They demon- strated the excellent diagnostic performance of this biochemical marker, with a threshold ranging from 3.7 to 4 mmol/l to distin- guish bacterial from viral meningitis

The lowest CSF lactate level for bacterial meningitis was 3.2 mmol/l and the highest CSF lactate level for viral meningitis was 3.7 mmol/l [34].

Lactate levels should always be interpreted alongside the CSF cytological and biochemical results. Lactate level measurement is only useful when the direct examination is negative and when the other CSF parameters do not point to a bacterial origin.

Procalcitonin (PCT). In 2008 the jury recom- mended performing serum measurement of PCT levels to help diagnose bacterial meningitis, using 0.5 ng/ml as the discrimi- natory value. Several studies [35–37] and one meta-analysis of nine stud- ies (725 patients) [38] have since confirmed the effectiveness of this marker in discriminating bacterial and viral meningitis, with thresholds ranging between 0.5 and 1 ng/ml.

PCT level measurement is only use- ful when the direct examination is negative and when the other CSF parameters do not indicate a bacterial origin. When sepsis is observed in infants, PCT levels may be normal the first six hours; thus, the bacterial origin cannot be officially ruled out at that stage

The diagnos- tic effectiveness of serum PCT level measurement seems to be even better with the new version of the BRAHMS/Thermofisher diagnostic kit (Kryptor ® ), and the discriminatory value is lower (close to 0.25 ng/ml) [34].

Hoen’s algorithm [43] combines the number of blood leuko- cytes, blood glucose levels, CSF protein levels, and the number of neutrophils in CSF in adults and children

Bacterial Meningitis Score (BMS) is based on the presence of seizures, the number of blood neutrophils (≥ 10,000/mm 3 ), CSF protein levels (≥ 0.8 g/l), the number of neutrophils in CSF (≥ 1000/mm 3 ), and on a positive direct examination using CSF Gram staining [46] in children

Meningitest ® , an improved version of BMS, is based on the presence of purpura, severe presentation in children (irritability, lethargy, prolonged time to skin recoloration), seizures, positive direct examination using CSF Gram stain- ing, CSF protein levels ≥ 0.5 g/l, or PCT levels ≥ 0.5 ng/ml [42] in children

In France and in other countries, brain imaging test — usually a CT-scan — is too often performed before lumbar puncture when confronted with meningitis suspicion [50].

Removing consciousness disorders, recent seizures, and immunodeficiency (even severe) from the list of scenarios requiring brain CT scan to be performed before the lumbar puncture (besides those associated with brain herniation signs) led to performing the lumbar puncture earlier on and to better prognosis (case fatality and severity of sequelae) [52].

The contribution and compliance with indications for imaging tests before lumbar puncture recommended by the IDSA [30] in case of community-acquired bacterial meningitis suspicion were assessed in eight hospitals of Houston (Texas, United States) and its region between 2005 and 2010 [59]. A total of 614 cases of meningitis were included in this assessment. Imaging tests were performed in 549/614 cases (89%). They were not recommended in two-thirds of cases and only led to identifying two cases of minor abnormalities in this population. An imaging test was indicated in 193 cases and revealed major abnormalities requir- ing a change in clinical management in only 15 cases. Indication was based on isolated consciousness disorders in 17% of cases (5.8% overall). Two-thirds of the 119 patients presenting with consciousness disorders had a second contraindication to lum- bar puncture. Median times to lumbar puncture observed in this study were respectively 5.5 hours and 8.25 hours in patients who did not undergo or who underwent an imaging test before the lumbar puncture. This is not compatible with adequate mana- gement of bacterial meningitis, which only accounted for 46 of 614 cases (7.4%).

However, some factors may lead physicians to consider the likelihood of brain herniation such as the following contraindications to lumbar puncture without prior brain imag- ing: sus- and infratentorial focal neurological signs (with or without consciousness disorders or even direct signs of brain herniation) are absolute contraindications to the lumbar punc- ture; however, in case of brain herniation signs, normal imaging results do not allow the lumbar puncture to be performed.

However, the risks of brain herniation following lumbar punc- ture are very low even when performed in the presence of intracranial lesions with a mass effect. The authors of a study of 94 brain abscesses [61] reported that 55 patients (59%) had a lumbar puncture performed while — besides the abscess — mass effect had been observed at CT scan in 65% of them. Only one of these patients died from brain herniation within six hours following the lumbar puncture, but the authors could not establish the causative role of the lumbar puncture in the patient’s death. This figure is similar to those reported in brain abscesses.

Post- critical state may hide neurological abnormalities indicative of focal neurological lesions because of the associated hypotonia and confusion. The neurological examination is not contribu- tory in case of status epilepticus and is a cause of intracranial hypertension. Children aged below 5 years often present with generalized seizures in case of fever, irrespective of the ori- gin. They are therefore not considered contraindications to the lumbar puncture, except for unilateral seizures, which should be considered a sign of localization

Recent seizures may be manifestations of intracra- nial expansion; the lumbar puncture is in that case dangerous. A Dutch study of adult patients presenting with bacterial menin- gitis reported that 5% of the 666 patients experienced one or more seizures before hospital admission. Seventy per cent of these patients had a brain imaging performed before lumbar puncture, with focal lesions observed in 32% of them [62].

Consciousness disorders associated with brain herniation (pupillary abnormalities [fixed unilateral or bilateral mydriasis], dysautonomia [blood pressure and bradycardia, respiratory rate abnormalities], cerebellar tonic seizures, no reaction to stimuli, decortication or decerebration symptoms) are absolute contraindications to lumbar puncture without prior imaging test.

Isolated consciousness disorders are not considered con- traindications to an immediate lumbar puncture

When bacterial meningitis is suspected, brain imaging should be performed BEFORE lumbar puncture in the following situations only (neurological contraindications to the lumbar puncture): • presence of clinical signs potentially indicating intracranial expansion: ◦ signs of localization (Table 1), ◦ focal AND recent seizures (< 4 days); • signs of brain herniation: consciousness disorders AND one or more of the following signs: ◦ pupillary abnormalities (fixed unilateral or bilateral mydri- asis), ◦ dysautonomia (blood pressure and bradycardia, respiratory rate abnormalities), ◦ cerebellar tonic seizures, ◦ no reaction to stimuli, ◦ decortication or decerebration symptoms; • persisting seizures (i.e., generalized motor seizures) prevent- ing lumbar puncture

The antibiotic therapy should be urgently initiated in patients presenting with bacterial meningitis. The immediate and mid- term prognoses depend on how early the antibiotic therapy is initiated [63]

Proulx et al. studied prognostic factors in 123 patients [64]. The median time between arrival at the emergency department and antibiotic therapy initiation was 3.8 hours (IQR 1.4–6.1 hours). The multivariate analy- sis revealed that more than 6 hours between admission and administration of the first dose of antibiotics were significan- tly associated with a higher rate of deaths, with an adjusted odds ratio (OR) of 8.4 (95% CI 1.6–40.9, P < 0.01).

Another study assessed the impact of a delayed antibiotic therapy initiation in 176 patients aged above 16 years and hospitalized for bacterial meningitis, includ- ing 40% caused by S. pneumoniae. Although the time between admission and antibiotic therapy initiation was short (median of 1 hour, mean of 1.2 ± 0.9 hours) and was not considered a risk factor for poor prognosis, the multivariate analysis revealed that a time interval > 24 hours between symptom onset and treat- ment initiation was associated with a higher risk of poor outcome (OR 2.8; 95% CI 1.13–7, P = 0.026) [66].

After adjustment based on age, sex, consciousness disorders, use of corticosteroids, and antibiotic choice, odds ratios for antibi- otic administration within one or two hours following admission were respectively 1.82 (95% CI 1.15–2.89, P < 0.05) and 2.07 (95% CI 1.34–3.20, P < 0.01) during the second period. The authors also reported a decreased case fatality between both periods, from 11.7% to 6.9% (P < 0.05), but the difference was not statistically significant after adjustment based on the above- mentioned criteria.

One of the key findings of this study was that time to treat- ment administration was significantly associated with the risk of death, with a relative increase of 12.6% (95% CI 3.1–23.1%, P < 0.01) per delayed hour of administration of the first dose of antibiotics.

Recommendations: The antibiotic therapy (and the admin- istration of dexamethasone when indicated) should ideally be initiated within the hour following hospital admission, irrespec- tive of the presumed time since meningitis onset. Any delay in antibiotic therapy initiation is associated with a poorer progno- sis

Recommendations: The antibiotic therapy should be initiated before lumbar puncture (but after blood culture) in the three following situations: • purpura fulminans; • patients who cannot be admitted to hospital within 90 minutes [54] ; • contraindication to the immediate lumbar puncture (Question 1)

The lumbar puncture should be performed as soon as possi- ble following correction of abnormalities — whenever possible. Several experts do not recommend performing the lumbar punc- ture in patients presenting with purpura fulminans, even when abnormalities have been corrected.

Bacterial menin- gitis case fatality is highly reduced if the antibiotic therapy is initially adapted to the causative agent in terms of in vitro sus- ceptibility [74–84], and sequelae are less common when CSF sterilization is rapidly obtained [85,86].

Since April 2013 vacci- nation with PCV13 followed by the 23-valent polysaccharid pneumococcal vaccine is recommended in older children and adults at higher risk of pneumococcal invasive infection [91]. High vaccination coverage with PCV13 is currently observed among children aged below 2 years (> 90% since 2010).

Although vaccination coverage for group C meningococcal infections is still too low to induce the required herd immu- nity, these new recommendations — especially those relating to pneumococcal vaccination — led to substantial changes in the epidemiology of bacterial meningitis [93]

The latest data published by GPIP/ACTIV indicates that group B Streptococcus is the leading cause of meningi- tis (57.2%) in infants aged below 2 months, while the leading cause is S. pneumoniae in infants aged between 2 and 12 months (44%). N. meningitidis and S. pneumoniae respectively account for half and a third of meningitis cases in infants aged above 1 year.

The case fatality of pediatric bacterial meningitis remains stable at approximately 7% with variations depending on the pathogen and age, with a higher case fatality in infants aged below 2 months (11%) (Table 2).

S. pneumoniae. A reduced incidence of pneumococ- cal meningitis and of all pneumococcal invasive infections is being observed in France and abroad [97–100]. The incidence of pneumococcal meningitis has indeed significantly decreased in France between 2009 and 2014 in all age groups (Table 3) and especially among children aged < 2 years (66% decrease) [88]

N. meningitidis. The 2015 incidence of meningo- coccal invasive infections was 8.5/100,000 (64 cases) among newborns aged < 1 year. It ranged from 3 to 1/100,000 among children aged between 1 and 4 years (64 cases) and was approxi- mately 0.9/100,000 among children aged between 5 and 14 years (50 cases). Serogroup B is predominant in all age groups and accounts for more than 60% of pediatric cases. The proportion of serogroup C ranges from 22% among newborns aged < 1 year to 36% among children between 5 and 14 years of age. Serogroup Y accounts for < 10% of cases among children and serogroup W remains rare (< 5%) [88,101].

The end of the mandatory BCG vaccination did not lead to an increase in tuberculous meningitis cases, which account for 0.9% of pediatric meningitis cases (Table 2) [102,103] .

Group B streptococcus is responsible for slightly more than 10% of pediatric meningitis cases (excluding new- borns) and the EPIBAC data reported 16 patients aged between 2 and 11 months in 2014 [88]

H. influenzae meningitis accounts for 3.7% of pediatric meningitis cases and the EPIBAC network reported 20 cases in 2014 [88]

Escherichia coli is observed in 4.3% of cases and Listeria meningitis (nine cases reported by EPIBAC in 2014) and group A streptococcal meningitis remain rare in children (Table 2)

Irrespective of the etiology, the incidence of bacterial menin- gitis in adults was 1.74 cases per 100,000 inhabitants aged 15 years or above in 2013–2014

A decreased incidence (−19%, P < 10 −4 ) has therefore been observed compared with that of 2008–2009 (1.93 cases per 100,000 inhabitants aged 15 years or above). This is due to the decreased incidence of pneumococcal and meningococcal meningitis as well as to a stable incidence of other types of bacterial meningitis [88]

The incidence of meningococcal meningitis decreased by 24% (P = 0.007) among adults between 2008–2009 and 2013–2014 [88]. This decreased incidence was mainly observed among young adults aged 15–24 years, and was mainly due to the decreased incidence of group B meningo- coccal infections [101]. No decrease in the incidence of group C meningococcal invasive infections in adults was observed because of the insufficient vaccination coverage for group C meningococcal infections [93]

The introduction of the PCV7 vaccine was associated with a paradoxical increase in adult pneumococcal meningitis between 2001–2002 and 2008–2009 [10], but the incidence of pneumococcal meningitis decreased following the introduc- tion of the PCV13: by 23% (P < 10 −4 ) among individuals aged above 15 years [88]. This decreased incidence was observed in all age groups, except in the 15–24-year age group.

These changes did not substantially change the epidemiology of adult bacterial meningitis. S. pneumoniae and N. meningitidis are still mostly responsible for adult bacterial meningitis. They respectively account for 55% and 24% of cases, i.e. 500 and 220 cases per year in 2013–2014. L. monocytogenes is the third cause of bacterial meningitis

N. meningitidis is most common in adolescents and young adults, accounting for 79% of cases among the 15–24-year age group and for 60% of cases among the 25–29-year age group

S. pneumoniae is responsible for 64% of meningitis cases in adults aged above 30 years.

Etiologies are slightly more diverse among elderly patients: S. pneumoniae accounts for 59% of adult bacterial meningitis cases, L. monocytogenes for 16% of cases, N. meningitidis for 9% of cases, H. influenzae for 9% of cases, and group A and B streptococci for less than 5% of cases

To sum up, despite a decreased incidence of pneumococ- cal meningitis due to the indirect effect of infant vaccination with PCV13, S. pneumoniae remains the leading cause of bac- terial meningitis in adults. N. meningitidis is the leading cause of meningitis in adults aged below 30 years. L. monocytogenes is the second cause of bacterial meningitis in elderly individu- als

The most recent 2016 data reports the following rates of pneumococci responsible for meningitis with reduced susceptibility to beta-lactams: 26% for penicillin, 6.2% for amoxicillin, and 2% for cefotaxime

Irrespective of age, 1.5% of strains had reduced susceptibility to cefotaxime (MIC > 0.5 mg/l) but no strain was resistant to the molecule (all have MICs < 2 mg/l). The highest cefotaxime MIC measured in 2014 was 1 mg/l.

The highest amoxicillin MIC measured in 2014 was 4 mg/l. This was observed in two strains isolated from adults (Fig. 4).

The proportion of pneumococci with reduced susceptibility to amoxicillin or cefotaxime has therefore not really been differ- ent between children and adults since 2012 (Fig. 5). Antibiotic susceptibility of S. pneumoniae strains isolated from patients presenting with meningitis in 2016 is displayed in Table 5

Most strains with reduced susceptibility to beta-lactams were represented by vaccine serotypes (6B, 9V, 14, 19F, 23F, 19A, and 6A) no longer observed in pediatric meningitis (< 2 years) and which incidence has significantly decreased in the rest of the population (Fig. 6) [104,106]

The incidence of pneumococcal meningitis with reduced sus- ceptibility to beta-lactams has now bottomed out since the start of the surveillance in 2001 — in all age groups (Figs. 7 and 8).

N. meningitidis. Susceptibility profiles of N. menin- gitidis strains isolated from patients presenting with invasive meningococcal infections were defined by the national reference center for meningococci for the antibiotics of therapeutic or pro- phylactic interest (penicillin G, amoxicillin, injectable 3GCs, rifampicin, and ciprofloxacin) [108]. In 2015 all strains assessed at the national reference center (n = 322) were susceptible to 3GCs, rifampicin, and ciprofloxacin. However, just like in 2006, 30% of these strains showed reduced susceptibility to penicillin G and amoxicillin

The empirical treatment with a 3GC is therefore highly likely to be effective when administered at the high dosage usually rec- ommended in this indication. However, the use of amoxicillin at high dosage is probably ineffective against strains with reduced susceptibility to penicillin G and amoxicillin (Table 6)

H. influenzae. The susceptibility profiles of H. influenzae strains have not substantially changed since 2005. Eighteen per cent (84/470) of all strains assessed at the Haemophilus national reference center in 2013 (vs 19% in 2005) were resistant to amoxicillin (MICs > 2 mg/l) through penicillinase secretion, and 18% (85/470) (vs 19% in 2005) had a modified PLP3 with amoxicillin MICs between 2 and 16 mg/l and cefotaxime MICs between 0.064 and 1 mg/l.

PLP3 modification (preferred target of cephalosporins) is associated with reduced cross susceptibility to all beta-lactams, although at variable rates depending on the position and number of mutations, and type of beta-lactam. Resistance to amoxicillin is often moderate, with MICs ranging from 1 to 16 mg/l (modal MIC at 2 mg/l). Clavulanic acid does not restore susceptibility to amoxicillin. The activity of first-generation and second-generation cephalosporins is reduced, just like that of carbapenems. 3GCs remain the most active cephalosporins, with MICs rarely exceeding 0.125 mg/l

Nineteen per cent (29/156) of invasive strains assessed at the Haemophilus national reference center in 2013 were resistant to amoxicillin (MIC > 2 mg/l), including only two strains with a modified PLP3. No strain was resistant to injectable 3GCs (MIC ≤ 0.125 mg/l) [109]. Only one strain (0.6%) was resistant to fluoroquinolones, and all were susceptible to rifampicin

However, H. influenzae strains with higher resistance level have been isolated in France since 2013 from patients presenting with pulmonary tract infections, and in 2016 from a patient presenting with meningitis. These strains are resistant to oral cephalosporins (cefpodoxime and cefixime) but also to cefotaxime (MICs ranging from 0.25 to 1 mg/l). They show intermediate susceptibility to meropenem (MIC ≥ 0.5 mg/l), and most of them remain susceptible to ceftriaxone (O. Gaillot, national reference center for Haemophilus infections, private data).

Regarding aminoglycosides, the prognostic analysis of patients enrolled in the MONALISA cohort and presenting with Liste- ria bacteremia and/or neurolisteriosis highlights the association between a better prognosis in terms of case fatality at 3 months and aminoglycoside prescription [110].

As for the susceptibility of other bacteria to antibiotics, it should be reminded that L. monocytogenes is naturally resistant to cephalosporins

Adding aminoglyco- sides to amoxicillin in case of Listeria meningitis suspicion is always recommended, even more so when microbiological documentation is available

Susceptibility of E. coli strains isolated from meningitis patients is not specific. Resistance to aminopenicillins is high (48% to 60% of strains), but resistance to 3GCs and to gentam- icin reported for all invasive strains reached almost 10% in 2013 [111]

The second objective of bacterial meningitis treatment is to avoid sequelae. Sequelae are less common if CSF sterilization is rapidly obtained [85,86]

The proportion of sterilized CSF 12 or 24 hours after treatment initiation depends on the bacterium [115]: pneumococcal meningitis is associated with longer time to CSF sterilization [116]. “Early” sterilization takes approxi- mately the same time for all administered antibiotics, provided the bacterium is susceptible to the antibiotic [115]

Generating synthetic time-series and sequential data is more challenging than tabular data where normally all the information regarding one individual is stored in a single row. In sequential data, information can be spread through many rows, like credit card transactions, and preservation of correlations between rows — the events — and columns — the variables is key. Furthermore, the length of the sequences is variable; some cases may comprise just a few transactions while others may have thousands.

we describe and apply an extended version of a recent powerful method to generate synthetic sequential data — DoppelGANger. It is a framework based on Generative Adversarial Networks (GANs) with some innovations that make possible the generation of synthetic versions of complex sequential datasets.

Considering the bacterial concentrations initially observed after the first lumbar puncture (approximately 5 log/ml for S. pneumoniae and Haemophilus, and 4 log/ml for meningococci) [117] and the need for early CSF sterilization, the operational treatment objective is to reach an in vivo bactericidal level of 0.5 log CFU/ml/h. This data can only be assessed with an analysis of experimental model results

Pharmacodynamic parameters more often focus on the mini- mum bactericidal concentration (MBC) than on the MIC, while efficacy parameters depend on the antibiotic class.

A strong correlation is observed between the three main parameters: concentration ratio in CSF/MBC (CCSF/MBC), area under the curve of CSF concentration (AUCCSF) and MBC ratio, and concentration time fraction in CSF/MBC (TCSF/MBC). On the basis of experimental works, CCSF/MBC ratio ≥ 5 is associated with early sterilization [119,122,123]

For the use of 3GCs in the treatment of experimental pneumococcal menin- gitis “resistant” to 3GCs, it has been demonstrated that the TCSF/MBC ratio had to be at least equal to 95% of time to obtain sterilization [123]; this may also be expressed by the need for a minimum concentration (in CSF) always higher than the MBC. Beta-lactam prescription therefore requires immediately efficient CSF concentrations

Lipophilic antibiotics (quinolones, rifampicin, and to a lesser extent linezolide), antibiotics with low molecular weight (quinolone, rifampicin, fosfomycin), and those with low protein binding are associated with a more rapid and more important diffusion.

Other factors play a role in antibiotic penetration such as inflammation of the meninges, which facilitates the diffusion of beta-lactams in the meninges [119]

Some antibiotic molecules may also be administered via con- tinuous infusion following a loading dose to obtain optimal and rapid diffusion [125]. Continuous infusion helps optimize the probability of antibiotic therapy success in case of systemic infections [127] and increases the meningeal and cerebro- meningeal penetration of antibiotics. Continuous infusion is particularly useful with beta-lactams with a relatively short half- life and with glycopeptides such as vancomycin. However, the benefit of the continuous infusion in case of bacterial meningitis has not been clinically proven

Cefotaxime and ceftriaxone are most frequently used because of their bactericidal properties and their good diffusion in CFS [128] . Ceftriaxone has a slightly better in vitro activity profile against S. pneumoniae than cefotaxime [129]. However, cefo- taxime may be administered at very high doses without inducing adverse events. Ceftriaxone has high protein binding rate, which may theoretically limit its diffusion in CSF [130]

Vancomycin is only used in combination, either via contin- uous infusion following a loading dose or through clustered injections to optimize CSF concentrations. A proportional relation may be observed between serum and meningeal con- centrations [131,132]. Using continuous infusion and even with corticosteroid addition, sufficient CSF levels of vancomycin are obtained when vancomycin blood levels are at least 25–30 mg/l [132]. Tolerability of S. pneumoniae to vancomycin seems rare although more frequent in case of resistance to beta-lactams. It is also associated with poorer treatment response and relapses [133–138]. These findings are also observed in experimental models [135]. No pneumococcal strain tolerant to vancomycin has been observed in France so far

Many clinical and experimental studies demonstrated the limitations of vancomycin in the management of community- acquired pneumococcal meningitis: • ineffectiveness of vancomycin alone in obtaining a cure following pneumococcal meningitis, leading to the con- traindication of vancomycin alone [139]; • no potentialization of 3GC bactericidia in experimental models of pneumococcal meningitis [140,141]; • benefits of vancomycin in experimental models, when adding a 3GC, only if the pneumococcus is resistant to the 3GC (MIC > 2 mg/l); very rare in France (no strain reported in 2014); • reduced penetration of vancomycin in the CSF in case of co-administration of dexamethasone [122] (currently highly recommended). Reduced penetration has been associated with delayed sterilization in experimental meningitis with resistant strain to 3GCs and concomitant administration of corticosteroids [142]; • the risk of nephrotoxicity of vancomycin is higher in adults when administered at high doses [143]. Adding vancomycin should therefore not be suggested in the absence of benefits and in case of toxic risks. This data is not available in children

The authors of a mouse study reported that rifampicin reduced lipoteichoic and teichoic acid concentrations as well as early mortality [144]

Clinical prospective studies are required. However, the authors of a clinical retrospective study reported that adding rifampicin to a 3GC was not associated with better prognosis in the multivariate analysis [6]

Except for the last case patient reported, cephalosporin failure observed in this study was either due to very high MICs (> 2 mg/l) — no longer observed nowadays — or to an insufficient dose.

Goldwater [150] (2005) thus assessed the rate of 3GCs in the CSF of 14 children treated with ceftriaxone 100 mg/kg/day or with cefo- taxime 200 mg/kg/day who had undergone a second lumbar puncture. The lowest rate was 0.58 mg/l and the highest rate was 35 mg/l, with a median rate of approximately 3 mg/l. No failure was observed as strains had low MICs of approximately 0.01 mg/l. The same diversity of rates was observed in a study of 19 patients who received 300 mg/kg/day of cefotaxime asso- ciated with vancomycin, without dexamethasone injection. The median CSF rate of cefotaxime was 4.4 mg/l (1.9–10.7 mg/l)

Compared with studies assessing lower doses of cefo- taxime (200 mg/kg/day), increasing the dose of cefotaxime to 300 mg/kg/day in children does not seem to be associated with a significant increase in CSF concentrations [151].

The CSF levels observed in this study were not high enough to reach the optimal inhibitory quotient for strains with cefotaxime MICs ≥ 0.5 mg/l. Required values for the inhibitory quotient are also not reached for strains with MICs of 1 mg/l. No clinical study is available to assess the potential impact of insufficient concentrations on clinical outcome.

Therefore, the systematic addition of vancomycin is no longer required. It should be restricted to cases where injectable 3GC MICs are > 0.5 mg/l. Obtaining 3GC CSF concentrations five times higher than the MIC is in that case not guaranteed. An early switch with IV amoxicillin (200 mg/kg/day, four times a day) is recommended if the strain is susceptible to the molecule (amoxicillin MIC ≤ 0.5 mg/l).

GPIP also suggested favoring cefotaxime (instead of ceftri- axone) at the initial dosing of 300 mg/kg/day divided into four intravenous infusions because of the very long half-life of ceftri- axone and of its high digestive concentrations. These properties may favor the emergence of Enterobacteriaceae strains resistant to 3GCs through ESBL production [152]. However, the theoret- ical risk in terms of ecological impact should be assessed and compared with the benefit/risk ratio of such choice

On the basis of currently available data and especially epidemiological data, when the optimal dose of 3GC is admin- istered for the treatment of pneumococcal meningitis, the systematic addition of vancomycin is not justified in adults and children

Guidelines for the first-line antibiotic therapy depend on the positivity of the direct examination or PCR tests (Table 8). Cor- ticosteroids may be administered before antibiotics

Patients presenting with meningitis and turbid CSF should be prescribed an empirical antibiotic therapy as soon as the physician per- forming the lumbar puncture notices the turbid CSF

Physicians should NOT wait for the results of the direct examination and the CSF biochemical analysis to initiate antibiotics; while wait- ing for Gram staining results, the choice of empirical antibiotic therapy should comply with treatment guidelines for meningi- tis with a negative direct examination

S. pneumoniae suspicion. A monotherapy with cefo- taxime 300 mg/kg/day is recommended. The administration is performed intravenously either with a continuous infusion or a discontinuous infusion with a minimum of four infusions (75 mg/kg/6 h). The daily dose for the continuous infusion is initiated immediately after the administration of a loading dose of 50 mg/kg over one hour.

Ceftriaxone may be administered at a dosage of 100 mg/kg/day as one or two intravenous infusions. A pharmacokinetic modeling study reported the benefit of cef- triaxone administration as two infusions daily in patients with normal renal function (glomerular filtration rate > 30 ml/min)

Although cefotaxime has better parmacokinetic param- eters (higher protein binding for ceftriaxone) and ceftriaxone shows better microbiological data (MIC usually one dilution lower than cefotaxime), there is no clinical study available to recommend one molecule over the other.

There is no consensus on the maximum dose in adults: some studies report doses of 24 g/day of cefotaxime [146]. Children (< 15 years of age) may receive cefotaxime at the max- imum daily dose of 12 g (4 g maximum for ceftriaxone).

Doses should be adjusted to the renal function. Measuring residual plasma concentrations after 48 hours of treatment may be indi- cated when administering a daily dose > 10 g/day, or in patients aged > 70 years or with a creatinine clearance < 30 ml/min

N. meningitidis suspicion. A monotherapy with cefo- taxime 200 mg/kg/day is recommended. The administration is performed intravenously either with a continuous infusion or a discontinuous infusion with a minimum of four infusions (50 mg/kg/6 h). The daily dose for the continuous infusion is administered concomitantly with the loading dose of 50 mg/kg over one hour. Ceftriaxone may be administered at a dose of 75 mg/kg/day as one or two intravenous infusions.

Listeriosis suspicion. A two-drug combination therapy is recommended with intravenous amoxicillin 200 mg/kg/day divided into 4 to 6 infusions and intra- venous gentamicin 5–6 mg/kg/day as a single daily dose over 30 minutes.

H. influenzae suspicion. A monotherapy with cefo- taxime 200 mg/kg/day is recommended. The administration is performed intravenously either with a continuous infusion or a discontinuous infusion with a minimum of four infusions (50 mg/kg/6 h). The daily dose for the continuous infusion is administered concomitantly with the loading dose of 50 mg/kg over one hour. Ceftriaxone may be administered at a dose of 75 mg/kg/day as one or two intravenous infusions

E. coli suspicion. A monotherapy with cefotaxime 200 mg/kg/day is recommended. The administration is per- formed intravenously either with a continuous infusion or a discontinuous infusion with a minimum of four infusions (50 mg/kg/6 h). The daily dose for the continuous infusion is administered concomitantly with the loading dose of 50 mg/kg over one hour. Ceftriaxone may be administered at a dose of 75 mg/kg/day as one or two intravenous infu- sions.

When ESBL-producing E. coli meningitis is highly suspected (ESBL colonization or positive specimen from another site), cefotaxime or ceftriaxone should be replaced by meropenem 40 mg/kg three times daily as a slow IV infusion (expert advice required)

Negative direct examination and negative PCR tests The distribution of microorganisms is in that case different, as detailed in Table 9 using the results of the COMBAT cohort of adult patients presenting with bacterial meningitis. As the sensitivity of the direct examination is lower in case of Listeria meningitis, the respective proportion of this microorganism is more important in patients presenting with meningitis with a negative direct examination (Table 9)

Evidence for listeriosis: age > 70 years, comorbidities, immunodeficiency, progressive onset of symptoms, rhomben- cephalitis (cranial nerve involvement and/or cerebellar syn- drome)

12.8.2.2. Evidence for listeriosis. A triple therapy with cephalosporin, amoxicillin, and aminoglycoside is recom- mended. Cefotaxime is administered at the dose of 300 mg/kg/day. The administration is performed intravenously either with a contin- uous infusion or a discontinuous infusion with a minimum of four infusions (75 mg/kg/6 h). The daily dose for the continuous infusion is initiated immediately after the administration of a loading dose of 50 mg/kg. Ceftriaxone may be administered at a dose of 100 mg/kg/day as one or two intravenous infusions. Amoxicillin is administered at the dose of 200 mg/kg/day intravenously, divided into 4 to 6 infusions. Gentamicin is administered at the dose of 5 mg/kg/day intra- venously as a single daily dose over 30 minutes in adults and at the dose of 5–8 mg/kg in children

12.8.2.3. Beta-lactam allergy. Severe allergy to beta-lactams is defined as a history of immediate anaphylactic hypersensitivity, including anaphylactic shock. Except for these extremely rare cases, the use of cefotaxime or ceftriaxone is recommended. One should bear in mind that the CSF diffusion of aztreonam is poor. Meropenem could be used because of the associated low risk of cross allergy [154]. However, no data is available to draft guidelines in this rare situation. The infectious disease specialist’s advice should be sought, and the first 24 hours of treatment may be monitored in the continuous monitoring unit (even if the patient’s state does not require it) (Table 10)

As cefotaxime is mainly renally excreted (including 60% as unchanged form and 20% as active metabolites), the dosing regimen should be adjusted to the renal function. As the suscep- tibility of the causative agent is initially unknown, we suggest using high antibiotic doses the first 24 hours and then reducing them by 25% to 75% according to the renal function impairment (25% reduction for a glomerular filtration rate [GFR] between 30 and 60 ml/min, 50% reduction for a GFR between 15 and 30 ml/min, 75% reduction for a GFR below 15 ml/min). High antibiotic doses should not be adjusted in patients receiving continuous hemofiltration

The ceftriaxone dose should also be adjusted as it is equally excreted from the hepato-biliary and renal routes. A full dose may also be administered the first 24 hours (2 injections/24 h). A 50% reduction of the dose may be suggested when the creatinine clearance is < 30 ml/min (single administration) [153]. Plasma (to avoid excessive dosing) and CSF concentrations should be measured in all these situations to check if CSF antibiotic concentrations are at least 10 times higher than the MIC

Dexamethasone is the only adjuvant to the bacterial meningi- tis treatment to have been extensively assessed in clinical studies.

Dexamethasone reduces the inflammation of the subarach- noid spaces and the vasogenic edema induced by meningitis, which are associated with potentially deleterious effects

This anti-inflammatory activity was highest in animal models when dexamethasone was administered before or at the same time as the first dose of antibiotic

A meta-analysis of randomized studies concluded to the benefit of dexamethasone in children as it helped reduce the frequency of severe hearing loss when the causative agent was H. influenzae or S. pneumoniae and when the first injection was administered before or at the same time as the first dose of antibiotics [155]

The most recent Cochrane meta-analysis (2015) included 25 randomized trials (1517 adults and 2511 children) for a total of 4121 bacterial meningitis cases [155]. Irrespective of the microorganism, findings from this meta-analysis revealed that corticosteroids were not associated with reduced case fatality but with morbidity improvement in terms of neurological sequelae reduction and hearing loss reduction. Taking into consideration the type of microorganisms or the country’s socio-economic sta- tus, the subgroup analyses revealed that corticosteroids were associated with reduced case fatality for microbiologically con- firmed pneumococcal meningitis only and in industrialized countries only

In the randomized, double-blind, and placebo-controlled European study of 301 adult patients presenting with bacte- rial meningitis included in this meta-analysis, patients were followed up for an average of 13 years. The impact of corti- costeroids on the reduction of case fatality, observed during the initial follow-up, was sustained up to 20 years after the initial episode [156]

Conversely, the prospective French study MONALISA reported that dexamethasone prescription was an independent risk factor for increased mortality, thus contraindicating its use when listeriosis is confirmed [110]

Time to corticosteroid initiation was assessed in 22 trials included in the 2015 Cochrane meta-analysis [155]. Corti- costeroids were administered before or at the same time as antibiotics in 13 trials (3143 patients) and after antibiotics in nine trials (797 patients). No significant difference was observed in terms of dexamethasone benefit between both administration modalities based on the analysis of the various endpoints

However, corticosteroids should be administered as soon as possible after antibiotic therapy initia- tion when they cannot be administered before or at the same time as antibiotics. No study assessing the maximum time to corti- costeroid administration following antibiotic therapy initiation is available. On the basis of experts’ advice, the 2016 Euro- pean guidelines recommend a maximum time to corticosteroid administration of 4 hours [56]. The 2016 British guidelines rec- ommend a maximum time of 12 hours [55].

Dexamethasone injection is recommended, just before or at the same time as the first injection of antibiotics in case of: • a suspicion of bacterial meningitis without microbiological confirmation, but with an empirical antibiotic treatment deci- sion taken. This scenario is observed when: ◦ the indication for brain imaging delays the lumbar puncture (neurological contraindications to the lumbar puncture) (Question 1), ◦ the lumbar puncture is contraindicated for non- neurological reasons (Question 1), ◦ turbid or even purulent CSF is observed at lumbar puncture, ◦ a negative direct examination of CSF is observed at lum- bar puncture, but other CSF and blood biological findings confirm the bacterial meningitis diagnosis; • initial microbiological diagnosis indicative of: ◦ pneumococcal bacterial meningitis (Binax positive and/or Gram-positive cocci at CSF direct examination), ◦ meningococcal bacterial meningitis (Gram-negative cocci at CSF direct examination). Several experts do not recom- mend dexamethasone in this scenario.

The initial dose of dexamethasone in adults is 10 mg every 6 hours, for 4 days. This treatment is not recommended in immunocompromised patients and in patients presenting with neuroinvasive listeriosis [12].

When the corticosteroid therapy cannot be administered just before or at the same time as the first injection of antibiotics, it should be administered as soon as possible up to 12 hours following antibiotic initiation.

Following definitive identification of the microorganism, dexamethasone should be continued for a total duration of four days in case of documented pneumococcal and Haemophilus meningitis. Whether or not to continue dexamethasone for four days in case of meningococcal meningitis is left to the physician’s choice. There is no indication for continuing dexam- ethasone in case of meningitis caused by other microorganisms

Onset of thrombosis with cerebral ischemia several days or several weeks later, following an initially favorable outcome, is observed in 1% of patients [158]. Most of these patients present with pneumococcal meningitis and have received dexametha- sone. Most experts believe that dexamethasone should therefore be reintroduced when this rebound effect occurs without failure of the antibiotic therapy